Environmental Protection and Disaster Risks: Selected Papers from the 1st International Conference on Environmental Protection and Disaster RISKs ... in Systems, Decision and Control, 361) 3030701891, 9783030701895

Citation preview

Studies in Systems, Decision and Control 361

Nina Dobrinkova Georgi Gadzhev   Editors

Environmental Protection and Disaster Risks Selected Papers from the 1st International Conference on Environmental Protection and Disaster RISKs (EnviroRISKs)

Studies in Systems, Decision and Control Volume 361

Series Editor Janusz Kacprzyk, Systems Research Institute, Polish Academy of Sciences, Warsaw, Poland

The series “Studies in Systems, Decision and Control” (SSDC) covers both new developments and advances, as well as the state of the art, in the various areas of broadly perceived systems, decision making and control–quickly, up to date and with a high quality. The intent is to cover the theory, applications, and perspectives on the state of the art and future developments relevant to systems, decision making, control, complex processes and related areas, as embedded in the fields of engineering, computer science, physics, economics, social and life sciences, as well as the paradigms and methodologies behind them. The series contains monographs, textbooks, lecture notes and edited volumes in systems, decision making and control spanning the areas of Cyber-Physical Systems, Autonomous Systems, Sensor Networks, Control Systems, Energy Systems, Automotive Systems, Biological Systems, Vehicular Networking and Connected Vehicles, Aerospace Systems, Automation, Manufacturing, Smart Grids, Nonlinear Systems, Power Systems, Robotics, Social Systems, Economic Systems and other. Of particular value to both the contributors and the readership are the short publication timeframe and the world-wide distribution and exposure which enable both a wide and rapid dissemination of research output. Indexed by SCOPUS, DBLP, WTI Frankfurt eG, zbMATH, SCImago. All books published in the series are submitted for consideration in Web of Science.

More information about this series at http://www.springer.com/series/13304

Nina Dobrinkova · Georgi Gadzhev Editors

Environmental Protection and Disaster Risks Selected Papers from the 1st International Conference on Environmental Protection and Disaster RISKs (EnviroRISKs)

Editors Nina Dobrinkova Institute of Information and Communication Technologies Bulgarian Academy of Sciences (IICT-BAS) Sofia, Bulgaria

Georgi Gadzhev National Institute of Geophysics, Geodesy and Geography Bulgarian Academy of Sciences (NIGGG-BAS) Sofia, Bulgaria

ISSN 2198-4182 ISSN 2198-4190 (electronic) Studies in Systems, Decision and Control ISBN 978-3-030-70189-5 ISBN 978-3-030-70190-1 (eBook) https://doi.org/10.1007/978-3-030-70190-1 © The Editor(s) (if applicable) and The Author(s), under exclusive license to Springer Nature Switzerland AG 2021 This work is subject to copyright. All rights are solely and exclusively licensed by the Publisher, whether the whole or part of the material is concerned, specifically the rights of translation, reprinting, reuse of illustrations, recitation, broadcasting, reproduction on microfilms or in any other physical way, and transmission or information storage and retrieval, electronic adaptation, computer software, or by similar or dissimilar methodology now known or hereafter developed. The use of general descriptive names, registered names, trademarks, service marks, etc. in this publication does not imply, even in the absence of a specific statement, that such names are exempt from the relevant protective laws and regulations and therefore free for general use. The publisher, the authors and the editors are safe to assume that the advice and information in this book are believed to be true and accurate at the date of publication. Neither the publisher nor the authors or the editors give a warranty, expressed or implied, with respect to the material contained herein or for any errors or omissions that may have been made. The publisher remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. This Springer imprint is published by the registered company Springer Nature Switzerland AG The registered company address is: Gewerbestrasse 11, 6330 Cham, Switzerland

Organization

The 1st International Conference on Environmental Protection and Disaster Risks (ENVIRORISKs 2020) was held as an online participation event between 29 and 30 September 2020 Sofia, Bulgaria.

Conference Co-chairs Kostadin Ganev, Bulgarian Academy of Sciences (Bulgaria) Nina Dobrinkova, Institute of Information and Communication Technologies, Bulgarian Academy of Sciences (Bulgaria) Georgi Gadzhev, National Institute of Geophysics, Geodesy, and Geography, Bulgarian Academy of Sciences (Bulgaria) Nikolay Miloshev, National Institute of Geophysics, Geodesy, and Geography, Bulgarian Academy of Sciences (Bulgaria) Petya Trifonova, National Institute of Geophysics, Geodesy, and Geography, Bulgarian Academy of Sciences (Bulgaria) Stelian Dimitrov, Sofia University (Bulgaria)

Program Committee Alexander Arakelyan, American University in Armenia, Armenia Artemi Cerdà, University of Valencia, Spain Anna Ganeva, Institute of Biodiversity and Ecosystem Research, Bulgarian Academy of Sciences, Bulgaria Aleksandar Petrovski, University “Goce Delcev” Stip and Military Academy “General Mihailo Apostolski” Skopje, North Macedonia Bojko Berov, Geological Institute, Bulgarian Academy of Sciences, Bulgaria Chuck Bushey, International Association of Wildland Fire, USA v

vi

Organization

Christos Dimopoulos, European University Cyprus, Cyprus Constantin Ionescu, National Institute for Research and Development for Earth Physics, Romania Dimitrios Melas, Aristotle University of Thessaloniki, Greece Dimcho Solakov, National Institute of Geophysics, Geodesy and Geography, Bulgarian Academy of Sciences, Bulgaria Evangelos Katsaros, European University Cyprus, Cyprus George Boustras, European University Cyprus, Cyprus George Drakatos, National Observatory of Athens, Greece Georgios Eftychidis, Centre for Security Studies—KEMEA, Greece Geert Seynaeve, European Society for Disaster and Emergency Medicine, Belgium Hrachya Astsatryan, Institute for Informatics and Automation Problems, National Academy of Sciences of the Republic of Armenia, Armenia Harald Pauli, Global Observation Research Initiative in Alpine Environments, Austria Hristo Chervenkov, National Institute of Meteorology and Hydrology, Bulgaria Horst Schwichtenberg, Fraunhofer SCAI, Germany Ilias Gkotsis, Centre for Security Studies—KEMEA, Greece Ivan Georgiev, National Institute of Geophysics, Geodesy, and Geography, Bulgarian Academy of Sciences, Bulgaria Jesús Rodrigo-Comino, University of Valencia, Spain, and Trier University, Germany Kristalina Stoykova, Geological Institute, Bulgarian Academy of Sciences, Bulgaria Nikolai Dobrev, Geological Institute, Bulgarian Academy of Sciences, Bulgaria Reneta Dimitrova, Sofia University and National Institute of Geophysics, Geodesy, and Geography, Bulgarian Academy of Sciences, Bulgaria Roberto San Jose, Technical University of Madrid, Spain Snejana Moncheva, Institute of Oceanology, Bulgarian Academy of Sciences, Bulgaria Stefan Florin Balan, National Institute for Research and Development for Earth Physics, Romania Tania Marinova, National Institute of Meteorology and Hydrology, Bulgaria Todor Gurov, Institute of Information and Communication Technologies, Bulgarian Academy of Sciences, Bulgaria Velichka Milousheva, Institute of Mathematics and Informatics, Bulgarian Academy of Sciences, Bulgaria Viacheslav Berman, Institute of Hydromechanics at National Academy of Sciences of Ukraine, Ukraine Yancho Todorov, VTT Technical Research Centre of Finland, Finland

Preface

Environmental Protection and Disaster Risks topics are challenging fields that scientific world is trying to address as much as it can. Earthquakes, floods, fires, droughts, blizzards, dust storms, natural releases of toxic gases and liquids, diseases, and other environmental variations affect hundreds of millions of people each year. Many disaster events are triggered by human activities. Examples that affect the environment and natural biodiversity are activities such as adding contaminants to air and water, changing land use, reducing and fragmenting the habitat of some species, introducing non-native species, and changing natural fluxes and cycles of energy and materials. The challenges associated with environmental protection today are multifaceted and affected by many interacting factors. Usually, they cover various, often large, spatial scales, unfold on long temporal scales, and have global implications (e.g. carbon dynamics, nutrient cycles, and ocean acidification). Dealing with these problems will require systems thinking and integrating multidisciplinary science. Actions in these directions are taken more and more in the recent years by political bodies, NGOs, and scientific groups trying to find sustainable solutions for the future generations. Every point of view matters when it comes to our global home—The Planet Earth. This volume is a result of discussions done during the International Conference on “Environmental Protection and Disaster Risks”, Sofia, Bulgaria, 2020, held as online participation event in the period 29–30 September 2020. The participants have agreed that the relevance of the conference topic and quality of the contributions have clearly suggested that a more comprehensive collection of extended contributions devoted to the area would be very welcome and would certainly bring value to a wider public in the field of Environmental Protection and Disaster Risks. The topics covered by this volume are: Air pollution, climate and health, natural hazards and risks, water

vii

viii

Preface

resources, human activities and management and informatics, remote sensing, high performance computing and GIS for environmental monitoring and management. Sofia, Bulgaria January 2021

Nina Dobrinkova Georgi Gadzhev Editors of Environmental Protection and Disaster Risks ENVIRORISKs 2020

Contents

Air Pollution, Climate and Health Effects of Satellite Data Assimilation in Air Quality Modelling in Bulgaria . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Dimiter Syrakov, Maria Prodanova, and Emilia Georgieva

3

Degree-Days and Agro-meteorological Indices in CMIP5 RCP8.5 Future Climate—Results for Central and Southeast Europe . . . . . . . . . . . Hristo Chervenkov, Georgi Gadzhev, Vladimir Ivanov, and Kostadin Ganev

19

Assessment of the Joint Quantiles of Temperature and Precipitation in CMIP5 Future Climate Projections over Europe . . . . . . . . . . . . . . . . . . . Hristo Chervenkov, Georgi Gadzhev, Vladimir Ivanov, and Kostadin Ganev

31

Coastal Boundary-Layer Characteristic During Night Time Using a Long-Term Acoustic Remote Sensing Data . . . . . . . . . . . . . . . . . . . . . . . . . Damyan Barantiev and Ekaterina Batchvarova

43

Climatological Study of Extreme Wind Events in a Coastal Area . . . . . . . Damyan Barantiev, Ekaterina Batchvarova, Hristina Kirova, and Orlin Gueorguiev

59

Anthropogenic and Solar Influence on Temperature over Bulgaria . . . . . Yavor Chapanov

75

Chemical Characteristics of Precipitation and Cloud Water at High Elevation Site in Bulgaria . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Elena Hristova, Blagorodka Veleva, Krum Velchev, and Emilia Georgieva

91

Performance of Operational Chemical Transport Models for Particulate Matter Concentrations in Bulgaria . . . . . . . . . . . . . . . . . . . . 107 Hristina Kirova, Nadya Neykova, and Emilia Georgieva

ix

x

Contents

Study of the Extreme Thermal Conditions for the Sofia Region—Preliminary Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 123 Vladimir Ivanov and Reneta Dimitrova Interaction Between Particulate Matter Characteristics and Atmospheric Boundary Height Over Sofia Based on Case Studies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 139 Plamen Savov, Nikolay Kolev, Ekaterina Batchvarova, Hristina Kirova, and Maria Kolarova The Seasonal Recurrence of Air Quality Index for the Period 2008–2019 Over the Territory of Sofia City . . . . . . . . . . . . . . . . . . . . . . . . . . . 161 Georgi Gadzhev Modelling of the Seasonal Sulphur and Nitrogen Depositions over the Balkan Peninsula by CMAQ and EMEP-MSC-W . . . . . . . . . . . . 171 Georgi Gadzhev and Vladimir Ivanov The Use of LES CFD Urban Models and Mesoscale Air Quality Models for Urban Air Quality Simulations . . . . . . . . . . . . . . . . . . . . . . . . . . . 185 R. San Jose, J. L. Pérez, and R. M. Gonzalez-Barras Urban Heat Island and Future Projections: A Study in Thessaloniki, Greece . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 201 Stavros Keppas, Daphne Parliari, Serafeim Kontos, Anastasia Poupkou, Sofia Papadogiannaki, Paraskevi Tzoumaka, Apostolos Kelessis, and Melas Dimitrios Chemical Characteristics of Flue Gas Particulates: An Experimental Investigation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 213 Tsvetelina Petrova, Iliyana Naydenova, Ricardo Ferreira, Yordanka Karakirova, and Mário Costa Natural Hazards and Risks Spatial Variation of Precursory Seismic Quiescence Observed Before Earthquake from 01.04.2010 in the Region of Crete . . . . . . . . . . . . 231 Emil Oynakov, Dimcho Solakov, Irena Aleksandrova, and Yordan Milkov Earthquake Ground Motion Scenarios for the City of Ruse . . . . . . . . . . . . 243 Dimcho Solakov, Stela Simeonova, Plamena Raykova, Boyko Rangelov, and Constantin Ionescu Precipitation Chemistry in Bulgaria During Saharan Dust Outbreaks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 263 Emilia Georgieva, Elena Hristova, and Blagorodka Veleva

Contents

xi

Modelling a Composite Tsunami Scenario for Karpathos Island (Aegean Sea) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 279 Lyuba Dimova, Reneta Raykova, Gianluca Pagnoni, Alberto Armigliato, and Stefano Tinti Seismic Scenario and People Exposure for Blagoevgrad Region, Bulgaria . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 293 Petya Trifonova, Dimcho Solakov, Stela Simeonova, Metodi Metodiev, and Stefan Florin Balan Informatics, Remote Sensing, High Performance Computing and GIS for Environmental Monitoring and Management Forecasting the Propagation of HF Radio Waves Over Bulgaria . . . . . . . . 309 Rumiana Bojilova and Plamen Mukhtarov Wildfire Risk Reduction Based on Landscape Management . . . . . . . . . . . 325 Nina Dobrinkova, Carlos Trindade, Craig Hope, Chuck Bushey, Alexander Held, Ciaran Nugent, Georgios Eftychidis, Adrián Cardil, George Boustras, and Evangelos Katsaros Numerical Weather Prediction for the Bulgarian Antarctic Base Area and Sensitivity to the SST Variable . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 339 Boriana Chtirkova, Elisaveta Peneva, and Gergana Georgieva Water Resources, Human Activities and Management Value Eco-Innovation as a Basis for Clean Production Through Ecodesign in the Bulgarian Food Industry . . . . . . . . . . . . . . . . . . . . . . . . . . . 359 Silviya Topleva, Tsvetko Prokopov, and Donka Taneva Porewater Nutrient and Oxygen Profiles and Sediment-Water Interface Fluxes Under Extreme Organic Loading in Different Sedimentary Habitats in Sozopol Bay (SW Black Sea): A Laboratory Experiment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 371 Stefania Klayn, Dimitar Berov, and Ventzislav Karamfilov Remote Sensing and Modelling of the Mopang Oil Pollution Near the Bulgarian Black Sea Coast . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 383 Irina Gancheva and Elisaveta Peneva

Air Pollution, Climate and Health

Effects of Satellite Data Assimilation in Air Quality Modelling in Bulgaria Dimiter Syrakov , Maria Prodanova , and Emilia Georgieva

Abstract The operational Bulgarian Chemical Weather Forecast System (BgCWFS) was modified and applied for assimilation of satellite retrieved atmospheric chemistry parameters—Aerosol Optical Depth (AOD) and columnar values of NO2 and SO2 . The work outlines the methodology based on calculation of correction factors between model estimated and assimilated satellite derived parameters. Simulations by two versions of the system were performed for two months (August 2017 and February 2019) for all 5 domains of BgCWFS. The first version, mod-run, is without satellite data assimilation, the second one, sat-run assimilates satellite data. The effects of the assimilation is demonstrated for different pollutants analysing the difference between the results of the two versions on particular days in different model domains and as domain mean values for the Balkan Peninsula and for Bulgaria. The domain mean monthly particulate matter concentrations increase by more than 100% in summer and by about 50% in winter. The increase in the domain mean monthly SO2 concentrations is about 110% in summer and 130% in winter. Keywords Chemical transport model · Satellites · Data assimilation · Air quality

D. Syrakov (B) · M. Prodanova · E. Georgieva National Institute of Meteorology and Hydrology, 66, Tsarigradsko Shose Blvd., 1784 Sofia, Bulgaria e-mail: [email protected] M. Prodanova e-mail: [email protected] E. Georgieva e-mail: [email protected] D. Syrakov · M. Prodanova National Institute of Geophysics, Geodesy and Geography—Bulgarian Academy of Sciences, Acad. Georgi Bonchev Str., Bl. 3, 1113 Sofia, Bulgaria © The Author(s), under exclusive license to Springer Nature Switzerland AG 2021 N. Dobrinkova and G. Gadzhev (eds.), Environmental Protection and Disaster Risks, Studies in Systems, Decision and Control 361, https://doi.org/10.1007/978-3-030-70190-1_1

3

4

D. Syrakov et al.

1 Introduction Satellite derived air pollution data are nowadays increasingly used in combination with comprehensive chemical transport models (CTM) for better description of the atmospheric composition and for improved forecast of pollutants concentrations at ground level [1, 8]. In the last few decades, major efforts are put on improving modeled aerosol parameters, as aerosols play significant role in the Earth’s energy budget and climate. Aerosols in the atmosphere are product of complex interactions between sources and chemical transformations at different scales, determining their high variability in space and time [2]. Satellites provide information for AOD (Atmospheric Optical Depth) that is a measure of the aerosol loading in the entire atmospheric column. They could provide information for sources, typically not included in the models, such as dust storms and wild fires. Together with aerosol parameters AOD, AAI (Atmospheric Aerosol Index) etc., some columnar gaseous parameters are also retrieved (e.g. columnar values of NO2 and SO2 ). The purpose of this work is to establish a methodology for satellite data assimilation in the current Bulgarian Chemical Weather Forecast System and to discuss some preliminary results from simulations for two periods of one-month. Up to now, satellite data are not used in relation to air quality in Bulgaria. This work is the first attempt to benefit from available satellite data for better understanding and simulating air quality parameters in the country.

2 The Modelling System The Bulgarian Chemical Weather Forecast System [6, 10, 11] is based on the state of the art WRF/CMAQ modelling chain. WRF (Weather Research Forecast) v3.6 [9] is used as meteorological pre-processor to the CMAQ (Community Multi-scale Air Quality) model, v4.6 [3]. The nesting capabilities of both models are used; they are run over five nested domains—Europe (EU, 81 km horizontal resolution), Balkan Peninsula (BP, 27 km), Bulgaria (BG, 9 km), Sofia district (SD, 3 km) and Sofia city (SC, 1 km). WRF is fed by the forecast/reanalysis production of the US NCEP (National Center for Environmental Prediction). As far as this data is global, it provides the initial and boundary meteorological conditions for WRF as well. The chemical boundary conditions over the mother domain (Europe) are set according to the climatic profiles, embedded in the CMAQ software. The presumption is that the possible errors decrease inside the domain because of the continuous action of pollution sources. The boundary conditions for each of the other domains are determined from the senior one. The emissions are based on the inventory, provided by TNO (The Netherlands Organization for Applied Scientific Research) for 2010 for Europe [7]. For Bulgaria, national emission inventories for 2015 are used. Special emission interface processors

Effects of Satellite Data Assimilation …

5

are created as to impose speciation and time variation profiles upon the inventory values. The physical parameterizations selected for WRF are the well-known and widely used Kain-Fritsch scheme for the cumulus parameterization, YSU (Yonsei University) scheme for PBL (Planetary Boundary Layer), the WRF single-moment 6-class microphysics scheme (WSM6), RRTM (Rapid Radiative Transfer Model) for longwave radiation, Noah scheme as land-surface model, RADM scheme for wet deposition of gases and particulates. The exploited version of CMAQ uses CBIV (Carbone bound v.4) as gas-phase chemical mechanism and ISORROPIA 1.7 for inorganic aerosol thermodynamics/partitioning. The aerosol module of CMAQ represents particulate matter (PM) using three modes. The Aitken and accumulation mode represent PM2.5 (aerosols with diameter equal or less than 2.5 µm), called also fine particles. A coarse mode represents PM with diameter greater than 2.5 µ and equal to or less than 10 µ. Thus, PM10 is modelled as the sum of the fine and the coarse-mode PMs. The chemical species treated are sulfate, nitrates, ammonium, elemental carbon, organic carbon, sea salt, crustal materials, and secondary aerosols of biogenic and anthropogenic origin.

3 Satellite Data Assimilation Description As far as there is no feedback between WRF and CMAQ, the meteorological model is run once in the beginning of the assimilation. During the assimilation CMAQ is run twice—normal 24-h run and part-day integration from Hs (satellite passing hour) to the end of the day. The aim of the assimilation is the creation of new concentration fields for Hs (new initial condition—IC).

3.1 AOD Calculations AOD is not routinely calculated in BgCWFS as well as in the most of the CTMs. Different algorithms are described in the literature. Results from testing some of them for calculating AOD at 550 nm are reported in [12]. Finally, the FlexAOD software [4, 5] is chosen for post-processing of CMAQ estimated profiles of aerosols species. The algorithm is based on the Mie theory and the tool has a version consistent with the CMAQ output. It is well known that models usually underestimate the particulate matter concentrations. The assimilation of satellite retrieved AOD is supposed to improve this shortcoming.

6

D. Syrakov et al.

3.2 Assimilation Technology For assimilation we consider aerosol and gas data provided by GOME-2 instrument (Global Ozone Monitoring Experiment-2) on board of MetOpA, MetOpB and MetOpC satellites. An objective analysis scheme (called further ANALIZ) is used for changing model estimated fields as to assimilate the satellite retrieved data: AOD and columnar densities of NO2 (NO2 _C) and SO2 (SO2 _C). The objective analysis scheme was specifically created based on calculation of autocorrelation functions and application of spline interpolation. The autocorrelation functions are determined for each day and each pollutant from the satellite data. The analyzed data is used to calculate the correction factors between model estimated and satellite assimilated parameters—different factors for AOD, NO2 _C and SO2 _C. The gridded values of the correction factors (CFs) are then used to change the concentrations of different pollutants (Fig. 1). The correction factor of AOD is applied to all particle variables and is one and the same for all levels. The correction factors of NO2 and SO2 are applied to the respective variable profiles (again one and the same values of CFs for all layers). In such a way, new fields for the satellite passage hour are produced serving as initial condition for further CMAQ integration. It is obvious that the surface values of PMs, NO2 and SO2 are respectively changed. BgCWFS is modified in a way to organize pseudo-operational (near real time) satellite data assimilation. The main data stream for one day is as follows: 1.

2.

Performs full-day CMAQ integration using as initial conditions the last hour concentrations from the previous day. Concentrations for the whole day are obtained—CONC(00 ÷ 24). Prepares new initial condition (IC) for satellite passage hour Hs performing: a.

Prepares the satellite data as input to ANALIZ (for Hs).

Fig. 1 Scheme for the use of satellite retrieved AOD in BgCWFS

Effects of Satellite Data Assimilation …

b. c. d. e. f. g. h. i. j. 3. 4.

7

Extracts and saves CONC(00 ÷ Hs − 1) from CONC(00 ÷ 24). Extracts and saves CONC(Hs). Prepares input for FlexAOD (for the whole day). Runs FlexAOD. Prepares AOD data as input to ANALIZ (for Hs). Prepares NO2 and SO2 columnar data as input to ANALIZ (for Hs). Runs ANALIZ (for AOD, NO2 _C and SO2 _C for Hs). Calculates Correction Factor (CF) for AOD, NO2 and SO2 (for Hs). Modifies PMs, NO2 and SO2 fields by respective CF—new IC for Hs.

Performs a part-day CMAQ integration for h = Hs ÷ 24 − CONC(Hs ÷ 24). Concatenates CONC(00 ÷ Hs − 1) with CONC(Hs ÷ 24:00) − new CONC(00 ÷ 24).

The assimilation is taking place in the first three model domains of BgCWFS (EU, BP and BG). The remaining two domains (SD and SC) obtain satellite influence via their boundary conditions, calculated from the senior domains. This is made because of the specific characteristic resolution of GOME-2 data (about 40 km). Such assimilation, to our knowledge, is made for the first time worldwide. According to the numerous publications, satellite data assimilation has been applied for one domain, only. Another novelty is the simultaneous assimilation of AOD, NO2 _C and SO2 _C.

4 Results As far as only limited amount of pollutants are monitored and the measurements are made at the ground, mainly, a post-processing of obtained data is performed consisting in extracting the surface values of several key pollutants as well as producing compositions as PM10, PM2.5, AOD and columnar values of NO2 (NO2 _C) and SO2 (SO2 _C). This is made for both sat- and mod-runs, the results are saved as so called ARCH-files (ARCHs and ARCHm, respectively).

4.1 Effects of Satellite Data Assimilation on the Daily Evolution of Surface Concentrations We discuss here effects of the assimilation exploiting the difference “sat-mod” of several variables, for different days and domains. The main behavior of this difference is as follows: • In the first hours of the day, the differences remaining from the previous day are relatively small, with positive and negative values.

8

D. Syrakov et al.

• At hour Hs (satellite overpass fixed to 09:00 UTC) a disturbance appears like an instantaneous area source of pollution (similar to fire or other accidental release of pollutants). • The shape and the structure of this disturbance are quite irregular and it appears approximately over the territory covered by satellite measurements (often not over the whole model area). • Similarly to pollution distribution, the disturbances evolve with time and move as if due to atmospheric transport. • Usually, its extreme values decrease and for some variables become small or disappear by the end of the day. • Often, the disturbances remain to the next day, especially in the bigger modelling domains. An example of this behavior is shown in Fig. 2 (satellite overpass hour Hs = 9:00 UTC). Of course, the details of this common behavior of the “sat–mod” difference vary very much with the different days, different variables, different domains, and for different seasons, as well.

Fig. 2 Evolution of the disturbance in the AOD “sat-mod” difference caused by satellite data assimilation over domain BP (Balkan Peninsula), 27 km resolution, at h = 0, 8, 9 (Hs), 10, 14, 23, on August 22, 2017

Effects of Satellite Data Assimilation …

9

Fig. 3 Maps of dAOD fields on 01.08.2017 at time step h = Hs for the first 3 BgCWFS domains

The AOD-differences (dAOD) by the two versions of BgCWFS at the moment Hs follow, more or less the satellite data coverage. This can be seen in Fig. 3, where dAOD is shown in the first three BgCWFS domains for 01.08.2017. One can notice that the three fields are quite consistent. More or less similar is the spatial distribution of dPM10. Note that PM10 is directly influenced by the ratio AODsat/AODmod (not the differences but the ratio) at the same hour. On Fig. 4 the evolution of dPM10 for a particular day on the biggest BgCWFS domain is shown.

Fig. 4 Maps of dPM10 on 09.02.2019 at time steps h = 0, 8, 10 (Hs + 1), 16, domain EU

10

D. Syrakov et al.

The satellite data for NO2 _C cover bigger areas than for the AOD ones. The coverage by SO2 _C data is smaller, but it is still bigger than for AOD. The AOD coverage is smaller in the winter month than in the summer one. The dNO2 behavior is slightly different from dAOD. Up to the hour Hs of a certain day, the differences “satmod” are around zero. At Hs, fragmentary spots appear with positive and negative values, predominantly in the first two modelling domains. At h = Hs + 1, the positive spots practically disappear and the negative ones decrease their extreme values. Possible reason for this behavior is the high reactivity of NO2 . This tendency is expressed better at summer because of the higher temperatures. Figure 5 shows the time evolution of the dNO2 spatial distribution in domain Bulgaria. Figure 6 shows the time evolution of dSO2 fields for domain Bulgaria on 07.02.2019. The decrease of the differences with the time is smaller than for August 2017; small positive differences remains for the next day. Interesting fact is the appearance of high values of dSO2 around the upper border of the domain. This is example of the influence of the senior domain—satellite data assimilation from domain BP is transferred to domain BG via the boundary conditions. Such an effect is observed for other days and pollutants. The final example for the effect of satellite data assimilation is the evolution of ozone surface concentrations, Fig. 7. Like the other parameters directly influenced by the satellite data assimilation (PMs, NO2 , SO2 ) the dO3 behavior is also characterized by “explosion” (higher values with respect to previous hour), but at time Hs + 1,

Fig. 5 Maps of dNO2 on 03.08.2017 at different time steps, h = 8, 9 (Hs), 10, 11, domain BG

Effects of Satellite Data Assimilation …

11

Fig. 6 Maps of dSO2 for domain BG on 07.02.2019 at different time steps, h = 0, 9 (Hs), 10, 23

i.e. one hour later than satellite passage. This delay is due to the fact that ozone is a secondary pollutant and some time is necessary for reactions (mainly with already changed NO2 ) leading to this deviation from the usual daily evolution of the difference “sat-mod”. The spots with dO3 are quite irregular on the maps, mainly positive, and their decrease lasts much longer, often continuing during the next day. An increase of the maxima during the following one or two hours after Hs is also possible. Figure 7 shows the dO3 time evolution: the remains from the previous day at h < Hs, the appearance of a new disturbance at h = Hs + 1 due to assimilation of satellite NO2 _C data, the increase of its intensity and further decrease several hours later.

4.2 Monthly Means and Domain Means for Domains BP and BG In this section the comparison between the two versions of BgCWFS (mod and sat) is presented as maps of monthly mean concentrations for the different pollutants, and also as Tables with domain averaged values in domains Balkan Peninsula and Bulgaria.

12

D. Syrakov et al.

Fig. 7 Maps of dO3 for 3.08.2017 at different time steps: h = 0, 9 (Hs), 10, 11, 18, 23

August 2017 Domain Balkan Peninsula. The spatial distribution of mean monthly surface concentrations is shown in Figs. 8 and 9 for PM10 and PM2.5, respectively. They show results without satellite data assimilation, “mod”, and with satellite data retrievals,

Fig. 8 Mean monthly concentrations of PM10 for August 2017; domain BP

Effects of Satellite Data Assimilation …

13

Fig. 9 Mean monthly concentrations of PM2.5 for August 2017; domain BP

“sat”. The PMs increase by the “sat” version, especially in the southern and eastern part of the domain, as expected effect by Saharan intrusions occurring in this month. Analogous pictures (not shown here) have been analyzed for all interesting parameters as gases NO2 , SO2 , and O3 as well as for the columnar variables AOD, NO2 _C and SO2 _C. The surface NO2 concentrations (not shown here) show very small changes between the two versions, well evident are spots with higher values near big cities or industrial areas (i.e. Istanbul, Bucharest, TPP Maritsa East Complex in southern Bulgaria, etc.). The SO2 concentrations seem to have been increased by the “sat” version, big emission sources are well identified by both BgCWFS versions. The domain mean concentrations by the two versions and the changes (absolute values and in %) are presented in Table 1. Domain Bulgaria. The spatial distribution of mean monthly surface concentrations of PM10 and PM2.5, obtained by the two versions of BgCWFS, is shown in Figs.10 and 11. The PMs are increased by the “sat” version, especially in the eastern part of the domain. The increasing is by more than 100%, for NO2 there are almost no changes in the domain mean values. Table 1 Domain mean concentrations for August 2017, domain BP Units

Mod

Sat

Sat-mod

%

NO2

µgm−3

3.15

O3

µgm−3

101.41

SO2

µgm−3

2.61

PM10

µgm−3

14.09

PM2.5

µgm−3

11.43

AOD

–

0.17

0.37

NO2 _C

µgcm−2

0.2

0.21

0.01

5.00

SO2 _C

µgcm−2

0.36

1.22

0.86

238.89

3.15

0

0.00

6.99

6.89

2.68

102.68

28.9

14.81

105.11

26.18

14.75

129.05

0.2

117.65

108.4 5.29

14

D. Syrakov et al.

Fig. 10 Mean monthly concentrations of PM10 for August 2017; domain BG

Fig. 11 Mean monthly concentrations of PM2.5 for August 2017; domain BG

The respective domain mean concentrations of all interesting parameters and the comparison between both versions of BgCWFS for domain Bulgaria and for August 2017 are shown on Table 2. February 2019 Domain Balkan Peninsula. The effect of satellite data assimilation is graphically demonstrated in Figs. 12 and 13 for PM10 and for PM2.5, respectively. The PMs are increased by the “sat” version, especially in the southern and eastern part of the domain, but also in the north-western part. The NO2 concentrations show small changes between the two versions mainly for the northern part of the domain, well evident also in this test case are spots with higher values near big cities or Table 2 Domain mean concentrations for August 2017, domain BG Units

Mod

NO2

µgm−3

2.72

O3

µgm−3

95.43

SO2

µgm−3

PM10

µgm−3

PM2.5

µgm−3

AOD

Sat

Sat-mod

%

2.74

0.02

0.74

102.26

6.83

7.16

2.47

5.23

2.76

111.74

8.68

18.88

10.2

7.83

18

10.17

129.89

–

0.15

0.34

0.19

126.67

NO2 _C

µgcm−2

0.18

0.19

0.01

5.56

SO2 _C

µgcm−2

0.3

1.14

0.84

280.00

117.51

Effects of Satellite Data Assimilation …

15

Fig. 12 Mean monthly concentrations of PM10 for February 2019; domain BP

Fig. 13 Mean monthly concentrations of PM2.5 for February 2019; domain BP

industrial areas. The SO2 concentrations are increased by the “sat” version through the whole domain. The domain mean concentrations of all concerned parameters for February 2019 and for domain Balkan Peninsula are presented in Table 3. The assimilation of satellite Table 3 Domain mean concentrations for February 2019, domain BP Units

Mod

Sat

Sat-mod

%

−1.51

−26.77

83.58

1.77

2.16

3.92

10.49

6.57

167.60

23.38

33.7

10.32

44.14

20.68

30.91

10.23

49.47

–

0.12

0.23

0.11

91.67

µgcm−2

0.5

0.31

−0.19

−38.00

µgcm−2

0.39

1.7

1.31

335.90

NO2

µgm−3

5.64

4.13

O3

µgm−3

81.81

SO2

µgm−3

PM10

µgm−3

PM2.5

µgm−3

AOD NO2 _C SO2 _C

16

D. Syrakov et al.

data leads to about 45% increase in PMs, but more than 150% increase in SO2 , and small decrease (4%) in NO2 surface concentrations. Compared to values for the summer month, all concentrations are higher. The domain mean values are increased by BgCWFS_sat, for SO2 _C more than 3 times, for AOD—by nearly 100%, while for NO2 _C the mean domain values are slightly decreased. Domain Bulgaria. The monthly mean fields of PM10 and PM2.5 for February 2019 are presented in Figs. 14 and 15, respectively. The PMs are increased by the “sat” version, especially in the northern and eastern parts of the domain. The domain mean values (Table 4) show an increase for PMs by the sat version by about 40%, and this increase is to a smaller extent compared to the summer month. Most probably this is influenced by the lower availability of satellite data in the winter

Fig. 14 Mean monthly concentrations of PM10 for February 2019; domain BG

Fig. 15 Mean monthly concentrations of PM2.5 for February 2019; domain BG

Table 4 Domain mean concentrations for February 2019, domain Bulgaria Units

Mod

Sat

Sat-mod

%

−1.43

−26.00

81.22

1.96

2.47

11.38

6.62

139.08

18.61

26.51

7.9

42.45

17.65

25.52

7.87

44.59

0.12

0.22

0.1

83.33

µgcm−2

0.51

0.3

−0.21

−41.18

µgcm−2

0.46

1.82

1.36

295.65

NO2

µgm−3

5.5

4.07

O3

µgm−3

79.26

SO2

µgm−3

4.76

PM10

µgm−3

PM2.5

µgm−3

AOD

–

NO2 _C SO2 _C

Effects of Satellite Data Assimilation …

17

month (more clouds, precipitations etc.). The domain mean SO2 value increases by more than 130%, while the one for NO2 has a decrease by about 25% (Table 4).

5 Conclusions We presented results obtained by BgCWFS model system running in pseudooperational mode for assimilation of satellite data produced by GOME-2 instrument on board of MetOP A, B and C satellites. Two months—a summer one (August 2017) and a winter one (February 2019)—have been simulated using all five nested domains of the system. The analysis and discussions were along two lines. The first one focused on the time evolution of hourly changes in key pollutant concentrations for selected days, using the difference in “sat-mod” values (i.e. values obtained with assimilation and without assimilation of satellite data). The second type of analysis was based on monthly mean concentrations of PM10, PM2.5, NO2 , SO2 , O3 and columnar parameters (AOD, NO2 _C and SO2 _C) in the outputs “mod” and “sat” for the domains Balkan Peninsula and Bulgaria. The time evolution of the “sat-mod” difference revealed some common elements. At the time of satellite overpass hour Hs (fixed to 09:00 UTC) a disturbance in the spatial distribution of the “sat-mod” difference appears, that resamples a new source of pollution. The shape and the structure of this disturbance are quite irregular and evolve with time (usually decrease) and change location due to atmospheric transport. The extreme values of the difference “sat-mod” decrease and for some parameters become usually small or disappear towards the end of the day. The differences for the bigger domains (Europe or Balkan Peninsula) were present also at the beginning of the next day. The domain mean surface concentrations in both domains were affected by the assimilation of satellite data resulting in more than 110% increase in PMs in summer and about 50% in winter. The respective increases for SO2 are more than 100% in August and about 150% in the winter month. The increases in columnar values of SO2 (SO2 _C) are much bigger—around 250% in the summer and 300% in the winter. The ozone shows small increase—7% in August and 2–3% in February. The only parameters that show negative difference “sat-mod” are NO2 and its columnar value NO2 _C in the winter. In fact, their domain mean monthly differences are 0% for NO2 in the summer and +5% for NO2 _C. During the winter month these decreases are about −25% for NO2 and about −40% for NO2 _C. Further investigations are required to understand the advantages and the weaknesses of the BgCWFS modelling system with satellite data assimilation, using both comparisons with observations and model inter-comparisons to data from air quality databases. Acknowledgements This study was carried out in the framework of the project SIDUAQ, funded by the European Space Agency (ESA) through contract No. 4000124150/18/NL/SC.

18

D. Syrakov et al.

Deep gratitude is due to ESA services providing data from GOME-2 instrument on board of MetOp-A, B and C satellites as well to USA EPA and USA NCEP for providing free software and data. Many thanks to TNO for European inventory data.

References 1. Benedetti, A., Morcrette, J.-J., Boucher, O., Dethof, A., Engelen, R.J., Fisher, M., Flentjes, H., Huneeus, N., Jones, L., Kaiser, J.W., Kinne, S., Mangold, A., Razinger, M., Simmons, A.J., Suttie, M., GEMS-AER Team: Aerosol analysis and forecast in the ECMWF integrated forecast system. Part II: data assimilation. J. Geophys. Res. 114, D13205 (2009) 2. Boucher, O.: Atmospheric Aerosols: Properties and Climate Impacts. Springer, Dordrecht. ISBN 9789401796484 (2015) 3. Byun, D., Schere, K.L.: Review of the governing equations, computational algorithms and other components of the models-3 community multiscale air quality (CMAQ) modeling system. Appl. Mech. Rev. 59, 51–77 (2006) 4. Curci, G.: FlexAOD: a chemistry-transport model post-processing tool for a flexible calculation of aerosol optical properties. In: Proceedings of the 9-th International Symposium on Tropospheric Profiling, L’Aquila, Italy, September 2012, ISBN: 978-90-815839-4-7 (2012) 5. Curci, G., Hogrefe, C., Bianconi, R., Im, U., Balzarini, A., Baró, R., Brunner, D., Forkel, R., Giordano, L., Hirtl, M., Honzak, L., Jiménez-Guerrero, P., Knote, C., Langer, M., Makar, P. A.,Pirovano, G., Pérez, J. L., San José, R., Syrakov, D., Tuccella, P., Werhahn, J., Wolke, R., Žabkar, Zhang, R. J., Galmarini, S.: Uncertainties of simulated aerosol optical properties induced by assumptions on aerosol physical and chemical properties: an AQMEII-2 perspective. Atmos. Environ. 115, 541–552 (2015) 6. Gadzhev, G., Ganev, K., Miloshev, N.: Numerical study of the atmospheric composition climate of Bulgaria—validation of the computer simulation results. Int. J. Environ. Pollut. 57(3–4), 189–201 (2015) 7. Kuenen, J.J.P., Visschedijk, A.J.H., Jozwicka, M.,Denier van der Gon, H.A.C.: TNO-MACC_II emission inventory; a multi-year (2003–2009) consistent high-resolution European emission inventory for air quality modelling. Atmos. Chem. Phys. 14(20), 10963–10976 (2014) 8. Park, R.S., Song, C.H., Han, K.M., Park, M.E., Lee, S.-S., Kim, S.-B., Shimizu, A.: A study on the aerosol optical properties over East Asia using a combination of CMAQ-simulated aerosol optical properties and remote-sensing data via a data assimilation technique. Atmos. Chem. Phys. 11(23), 12275–12296 (2011) 9. Skamarock, W.C., Klemp, J.B.: A time-split non-hydrostatic atmospheric model. J. Comput. Phys. 227(7), 3465–3485 (2008) 10. Syrakov, D., Prodanova, M., Slavov, K., Etropolska, I., Ganev, K., Miloshev, N., Ljubenov, T.: Bulgarian system for air pollution forecast. J. Int. Sci. Publ. Ecol. Safety 7(1), 325–334 (2013) 11. Syrakov, D., Prodanova, M., Etropolska, I., Slavov, K., Ganev, K., Miloshev, N., Ljubenov, T.: A multi-domain operational chemical weather forecast system. In: Lirkov I. et al. (eds.) LargeScale Scientific Computing, LNCS 8353, pp. 413–420. Springer-Verlag Berlin Heidelberg (2014) 12. Syrakov, D., Prodanova, M., Georgieva, E., Dimitrova, M., Spassova, T., Atanassov, D., Veleva, B., Nedkov, R.: Aerosol optical depth calculations using the Bulgarian chemical weather forecast system. Bulg. J. Meteorol. Hydrol. 23(2), 31–36 (2019)

Degree-Days and Agro-meteorological Indices in CMIP5 RCP8.5 Future Climate—Results for Central and Southeast Europe Hristo Chervenkov , Georgi Gadzhev , Vladimir Ivanov, and Kostadin Ganev Abstract The present paper is continuation of our recent study and analyzes the potential changes of residential heating and cooling degree-days as well as three stakeholder-relevant indices of agro-meteorological change (growing season length, sum of the active and sum of the effective temperatures) for Central and Southeast Europe over near past (1975–2004), near (2021–2050) and far (2070–2099) future periods. All indicators were calculated from the output data of our simulations with the regional climate model RegCM driven by the ERA-Interim reanalysis for the near past and by the global circulation model HadGEM2-ES under RCP8.5 CMIP5 radiative forcing scenario for the future periods. The validation of the model-based indices against their counterparts, computed from the observational dataset E-OBS, shows that the model reproduces their spatial variability and magnitude generally well. A linear bias correction of the considered indices is also demonstrated. Consistent with the general trend of the mean and extreme temperatures over the region, the study reveals a decrease of the heating degree days and considerable increase of the cooling degree days and the agro-meteorological indices practically over the whole domain in the future. The detected changes are fairly not symmetrical - the relative increase of the cooling degree days is significantly bigger than the decrease of the heating degree-days. H. Chervenkov (B) National Institute of Meteorology and Hydrology, Tsarigradsko Shose blvd 66, 1784 Sofia, Bulgaria e-mail: [email protected] URL: http://www.meteo.bg G. Gadzhev · V. Ivanov · K. Ganev National Institute of Geophysics, Geodesy and Geography—Bulgarian Academy of Sciences, acad. Georgi Bonchev str., Bl. 3, 1113 Sofia, Bulgaria e-mail: [email protected] URL: http://www.geophys.bas.bg V. Ivanov e-mail: [email protected] K. Ganev e-mail: [email protected] © The Author(s), under exclusive license to Springer Nature Switzerland AG 2021 N. Dobrinkova and G. Gadzhev (eds.), Environmental Protection and Disaster Risks, Studies in Systems, Decision and Control 361, https://doi.org/10.1007/978-3-030-70190-1_2

19

20

H. Chervenkov et al.

Keywords Heating and cooling degree-days · Agro-meteorological indices · CMIP5 RCP8.5 · Bias correction · Regional climate simulation

1 Introduction Nowadays there is a strong degree of agreement that the climate change is the defining challenge of our time. It will exert influence on the ecosystems, on all sectors of the international economy, and on the human health and quality of life [25, 26, 39]. The global warming tendencies and the linked regional climatic changes over Central and Southeast (CSE) Europe have been widely studied in the last decades based on in situ measurements [1, 2, 16], assimilated surface observations [4, 12, 36], reanalysis [6] global models [5, 38] and regional climate models [9, 29, 34]. Most of these studies are focused on the second half of the twentieth and the first decade of the twenty-first century, clearly evidencing that, similarly to the global and continental trends, the regional climate got warmer during the period. There is also an overall consensus that the projected changes in the mean and extreme (i.e. minimum and maximum) temperatures stand out in the region indicating a considerable intensification of heat stress in the future [10, 11, 35, 39]. Beside the mentioned effects, the ongoing and projected future climate changes have direct and indirect impact on managed systems like heating, ventilating and air-conditioning industry [3, 28, 41] as well as the agriculture [23, 24, 30]. Space heating and cooling is responsible for a large fraction of European energy use [17]. Agriculture is probably the sector most dependent on climate. Agricultural production is highly dependent on weather conditions and extreme weather events can have a dramatic impact on the crop yield [23, 24, 37]. The increase of temperature in the region, together with more frequent severe winters and summer heat waves may lead to a change in energy consumption and agricultural production [28, 41]. The linkage of the ambient daily mean (tg), minimum (tn) and maximum (tx) temperatures and the energy needs for air-conditioning or heating buildings as well as the crop productivity can be quantified by means of numerical indicators, calculated from these input parameters. They are rough surrogates for how climate change is likely to affect both sectors [22, 41]. The present study analyses the potential changes of residential heating and cooling degree-days (HDD and CDD) as well as three stakeholder-relevant indices of agrometeorological (AM) change (growing season length, sum of the active and sum of the effective temperatures) for CSE Europe over near past (1975–2004), near (2021– 2050) and far (2070–2099) future periods. All indicators were calculated from the output data of our simulations in 20 km grid spacing with the regional climate model (RCM) RegCM driven by the global circulation Model (GCM) HadGM under the CMIP5 radiative forcing scenario RCP8.5. The present paper is natural continuation of our recent work, documented in [27] and fits in the same conceptual framework.

Degree-Days and Agro-meteorological Indices in CMIP5 RCP8.5 …

21

The RCP8.5, being the scenario with the strongest radiative forcing among all others, deserves special attention due to the most expressive manifestation of the projected climate changes [39, 42]. The paper is structured as follows: Concise description of the RCP8.5 scenario and the used model set-up is in Sect. 2. The theoretical background of the applied indicators is described in Sect. 3. The core of the article is Sect. 4, titled ‘Calculations and Results’. The concise conclusion remarks are in Sect. 5.

2 RCP8.5 Scenario and Model Set-Up In climate change research, scenarios describe plausible trajectories of climate conditions and other aspects of the future [42, 43]. Along with information on other related conditions such as land use and land cover, emissions scenarios provide inputs to climate models. The Coupled Model Intercomparison Project (CMIP) is a standard experimental protocol for studying the output of coupled atmosphere-ocean general circulation models (CAOGCMs) which provides a community-based infrastructure in support of climate model diagnosis, validation, intercomparison, documentation and data access [42]. The fifth phase of CMIP, CMIP5 experiment uses new emission scenarios called representative concentration pathways (RCP) [33, 43] to assess the interactions between the human activities on the one hand and the environment on the other hand, and their evolution. They are named according to the radiative forcing level at 2100 [33], with the numbers representing the 2100 radiative forcing increase relative to pre-industrial levels in W m−2 . Among other RCPs, the RCP8.5 is the scenario with the highest concentration of greenhouse gases (GHG): it predicts a continuous rise of GHG emissions until 2100, causing a CO2 equivalent larger than 1370 ppm and a temperature increase close to 4 ◦ C. Note that RCP8.5 assumes radiative forcing levels continue rising after the end of the twenty-first century. As in [27], the simulation of the future climate was carried out using RegCM4.4, a limited-area, hydrostatic, compressible, sigma-p vertical coordinate model maintained at the International Centre for Theoretical Physics (ICTP) in Trieste, Italy [21]. The model is flexible, portable and easy to use, combining efficiency and high performance skill. It can be applied to any region of the world, with grid spacing of down to about 10 km (hydrostatic limit), for a wide range of studies. Subsequently, it is widely exploited and there are a number of previous studies that evaluated the model performance around the world (see [19, 21, 34] and references therein). The entire experiment covered the near past (1975–2004), near (2021–2050) and far (2070–2099) future periods, with initial and boundary conditions taken from the Hadley Centre Global Environment Model version 2, Earth-System configuration (HadGEM2-ES) [14] CAOGCM. In the comprehensive study [35], which uses also HadGEM2-ES as driving model, is stated that this CAOGCM is characterized by a relatively good level of performance among CMIP5 models for most regions (including the Mediterranean region).

22

H. Chervenkov et al.

Our modelling group has a previous, partially project-driven, experience in the exploration of the model RegCM for simulation of the near past, present and projected future climate [18, 20] as well as in assessments, based on RegCM-derived indices [26].

3 Heating, Cooling Degree-Days and AM Indices The HDD and CDD are, similarly to the climate indices [1, 44], an attempt to objectively extract information from daily weather data (observations or model output as in case) that answers questions concerning energy demand and/or consumption in the business and residential heating and cooling sector [40]. Thus, they are likely to display the same types of variability as the temperature data on which they are based which makes them a common climatological indicators. The units of measurement of the HDD and CDD are degree-days which, according the proper proposal in [28], will be noted henceforth as °D. The heating and cooling requirements for a given structure at a specific location are considered, in some degree and beside the influence of the other factors, proportional to the number of HDDs and CDDs at that location. The method assumes that the energy needs for a building are proportional to the difference between the daily mean and extreme temperatures and a base temperature (tb). The base temperature is the outdoor temperature below or above which heating or cooling is needed. In terms of degree-days, the annual energy consumption, Q year (W day), can be calculated according [3] as: Q year =

K tot D D, η

(1)

where K tot is the total heat-transfer coefficient of the building in W °C−1 , η is the dimensionless efficiency of the heating or cooling system and D D is the value of degree-days for heating or cooling. In contrast of some collections of climate indices [44], the theoretical formulation of the CDD and HDD is not standardized; their computation can be performed in different ways, depending on the nature and scope of the study as well as availability of input data. Computation methods range from simple approaches, based on monthly or annual temperature, to more sophisticated models [40, 41]. As in [27], in the present study we use the developed in the United Kingdom Meteorological Office (UKMO) [13] and successfully applied in [40, 41] method. According it, daily HDD and CDD are calculated based on a comparison of tn, tg and tx with the selected in advance tb, taking account of fluctuations of daily air temperature around the base temperature, as well as the asymmetry between daily mean temperature and diurnal temperature variations, as shown on Table 1. Hence the study is on annual basis, we summed the daily values. As in the original proposal [13], tb is set on 15.5 °C for the HDD- and on 22.0 °C for the CDDcomputation. Other definitions of these indicators, based solely on tg, have a jump

Degree-Days and Agro-meteorological Indices in CMIP5 RCP8.5 …

23

Table 1 UKMO methodology for computing daily HDD and CDD Condition HDD = CDD = tx ≤ tb (uniformly cold day) tg ≤ tb < tx (mostly cold day) tn < tb < tg (mostly warm day) tn ≥ tb (uniformly warm day)

tb − tg (tb − tn)/2 − (tx − tb)/4 (tb − tn)/4 0 (no heating is required)

0 (no cooling is required) (tx − tb)/4 (tx − tb)/2 − (tb − tn)/4 tg − tb

discontinuity when daily mean temperature falls below the base temperature. The methodology of UKMO does not exhibit such a discontinuity [17]. It is worth to emphasize also, that in present work the daily mean temperature is independent input parameter, rather than estimated as arithmetic average between tn and tx as in [40, 41]. In the last decades has been done significant research in many countries on temperatures critical to plants, and this, along with the aggregate evaluation of thermal resources, has made possible a substantially more accurate determination of climatic heat provision to crops [31]. The effects of climate change clearly appear in agriculture and forestry in the considered region. Production of these sectors is strongly influenced by the climate-related measures as the growing season length (GSL), accumulated active and effective temperatures (AAT and AET). These quantities are valuable AM indicators relevant for cultivated plants phenology and active growth of crops [4, 24, 30, 37]. According the common definition of the European Climate Assessment & Dataset (ECA&D) project [15] and the Expert Team on Climate Change Detection and Indices (ETCCDI, [44]), the GSL is the annual count of days between first span of at least 6 days with tg > tb and first span after July 1 (in Northern Hemisphere) of at least 6 days with tg < tb. In this definition tb = 5 °C which is threshold temperature for the cold-tolerant species. The threshold temperature for the thermophile species is 10 °C and in the present study, due to the geographical location of the region, we apply this value. The units of measurement of the GSL are, obviously, days. As the GSL, the AAT and AET are calculated also on annual basis and are defined as: i=io i=io   tg(i), AE T = max (tg(i) − tb, 0) , (2) A AT = i=iu

i=iu

where tg(i) is the mean daily temperature in the day of year (DOY) i, iu is the DOY of the start, and io the DOY of the end (cessation) of the GSL. The units of measurement of the AAT are degree-days, noted as the units of the HDD and CDD °D. The degree day method, which expresses numerically the relationship of plant development and growth to atmospheric temperature, was developed in the United States in the first half of the twentieth century. Total active and effective temperatures, subsequently the values of AAT and AET as well as ranges for GSL, have been

24

H. Chervenkov et al.

established for many crops. These methods of expressing crop heat requirements are widely used for agricultural climate evaluation in the former Soviet Union, Bulgaria, Poland, Romania, and a number of other countries [37] which motivates their selection in the present study.

4 Calculations and Results First, the RegCM-output parameters tn, tg and tx are mapped onto regular 0.25° × 0.25° lat–lon grid. The considered indices are computed according the definitions in Sect. 3, by purposely-build by the authors procedures. The analysis in the present study is focused on the multiyear (i.e. over the whole 30-year long periods) means (MM). The credibility of the model set-up to reproduce the considered indices should be examined. The validation of the results is performed in the traditional way, comparing of the model-based indices against their counterparts, computed from the observational dataset E-OBS v19.0 [15], accepted as reference. The relative bias (RB), i.e. the metrics (3) R B = (I M − I R )/I R , where I M and I R are the model and reference values of any of the considered indices respectively, is presented on Fig. 1. Similarly to some ETCCDI-indices, for example, the ‘tropical nights’, the CDD is practically meaningful only in low elevation areas (i.e. below 1000 m) which are particularly exposed to persistent and intense warm spells in summer [4, 12, 38].

Fig. 1 From left to right: MM for the 1975–2004 of HDD, CDD, GSL, AAT, AET from E-OBS (first row) and RegCM-output (second row). The units of HDD, CDD, AAT and AET are 100 °D and of GSL—days. The RB (in %) is shown on the third row

Degree-Days and Agro-meteorological Indices in CMIP5 RCP8.5 …

25

Figure 1 shows that the model reproduces the spatial variability and magnitude of the all indices, except CDD, relatively well—the spatially prevailing RB is in the range −15 to 15%. The RegCM simulations, as these from any other regional climate model, often show some, in certain cases considerable, deviations from observations [9, 21, 34]. This common methodological problem is partly inherited from the driving global model and has led to the development of a number of correction approaches, known with the common name bias correction (BC). BC first of all aims to adjust selected statistics of a climate model simulation to better match observed statistics over the present-day reference period. The basic assumption is that bias changes are negligible compared to climate change or, equivalently, that the bias itself is time-invariant [32]. The general view in the expert community is that the bias-corrected climate change signal is more reliable compared with the uncorrected one. Subsequently, BC model output is more suitable for impact assessments and thus we will apply it on the considered indices. Following the proposed in [32] and adopted in [7, 8] notation, the simulated present-day model time series of length N of chosen variable will be denoted as p p xi , the corresponding reference time series as yi . The mean of the uncorrected p model over the considered near past period (i.e. 1975–2004) μraw can be estimated p p as μˆ raw = xi (the hat denotes traditionally the estimator and the bar—averaging in p p p time), the corresponding real mean μr eal as μˆ r eal = yi . The most simple approach used for BC is the so-called delta change approach. It is widely used in climate impact research [32]. In its most basic application, performed in the present work also, a time series of future climate is generated as: f

p

f

p

xi,corr = yi + (xi − xi ). f

(4)

p

The quantity xi − xi is actually the model derived climate change signal. Thus, Eq. 4 could be treated as observed time series corrected with the simulated climate change. The method based on Eq. 4 is called correspondingly additive delta change method. The delta change approach and its modifications are investigated in [7, 8] considering ETCCDI climate indices. One of the newly proposed idea was to apply Eq. 4 directly on the indices, rather on the parameters (i.e. tn, tg, tx and the daily precipitation sum), used for their calculation. One of the basic findings in [7, 8] is that this approach is reasonable in the case of the thermal indices and thus we will implement it here, as shown on Figs. 2 and 3. The relative climate changes (RCC), which is defined similarly to the RB in Eq. 3: RCC = (I F − I P )/I P ,

(5)

where I F and I P are the model simulation output for the future (near or far) and the near past of the considered indices respectively, are also presented on these figures.

26

H. Chervenkov et al.

Fig. 2 From left to right: MM for the NF of HDD, CDD, GSL, AAT and AET from the raw RegCM output (first row) and after the BC (second row). The units are as on Fig. 1. The RCC (in %) are shown on the third row

Fig. 3 Same as Fig. 2 but for the FF

Most apparently, the spatial patterns of the uncorrected and corrected versions of all indices, shown on Figs. 2 and 3 are practically identical. This result is direct consequence of the generally good agreement between the reference and the model output for the near past, shown on Fig. 1 and commented above. The RCC demonstrates the definite evolution of the considered indicators—distinct and spatially dominating increase of the ‘warm’ indicators (CDD, GSL, AAT and AET) and decrease of the ‘cold’ one (i.e. the HDD). It is worth emphasizing, that for the both future periods the relative increase of the CDD is significantly bigger than the relative decrease of the HDD—the decrease of the HDD in the FF, which is almost homogeneous distributed, is −50 to −30% and increase of the CDD is practically everywhere above 90%. The pan-European and comprehensive study [41] reveals similar tendencies. Comparing the differences in degree-days between the periods 2041–2070 and 1981–

Degree-Days and Agro-meteorological Indices in CMIP5 RCP8.5 …

27

2010, the authors discover overall increase of the CDD in the projected future climate which peaks over the Mediterranean region and the Balkans. Conversely, the HDD is expected to fall: the decrease in HDD over the same region is (on average) about −200 °D under RCP4.5. However, in relative terms, the decrease in HDD is largest in Southern Europe, where values are projected to be reduced (on average) by 30% under RCP4.5. The local study [28], outlines coherent outcomes despite the different modelling and scenario set-up. Although not directly comparable, our results are in principal agreement with the outcomes of these and some older studies.

5 Conclusion It should be noticed that this research, which is the continuation of our previous [27], does not quantify future changes in heating or cooling residential energy demand and the general agro-meteorological conditions, but gives an overview of projected changes in the considered indicators which point to the sign and trend of changes in these sectors. RegCM has proven to reproduce the magnitude and spatial variability of HDD, as well as GSL, AAT and AET very well and thus the applied bias correction do not alter significantly the result. Our main findings show unambiguously a projected general decrease in HDD and all AM indicators over CSE Europe, which quantitatively is stronger expressed as in the scenarios with weaker radiative forcing (i.e. RCP2.6 and RCP4.5) in the far future. Conversely, the CDD is expected to increase. The detected changes, which agrees with most recent studies [28, 41], are direct consequence of the expected general temperature tendencies in the region and are natural continuation of the tendencies revealed from the analysis of historical records of the near past [2, 5, 12, 16, 36, 38]. As emphasized in [17], a decrease in the demand for space heating can significantly decrease overall energy use in Europe, but this gain can be offset in part or completely by an increase in cooling demand. Furthermore, heating is delivered to end users in different ways (individual boilers powered by oil, gas and coal, and electricity and district heating), whereas cooling is supplied currently almost exclusively through electricity. As a result, a given change in cooling demand is generally associated with larger costs, a larger change in primary energy needs and larger impacts on the peak capacity of supply networks than the same change in heating demand. The estimated changes of the AM indices in projected future climate could be also prerequisite for deep ecological and economical consequences. Longer growing seasons, as well as bigger AAT and AET, may allow for a greater diversity of crops (including those with long maturation periods), and the potential for multiple harvests on the same land. Conversely, both irrigation needs and the risk from invasive species, pests and pathogens may increase [22]. The results of the present study could be useful as scientific basis for the long-range policy of energy management and of the agricultural sector within CSE Europe.

28

H. Chervenkov et al.

Acknowledgements The authors would express their gratitude of the institutions which provides free of charge software and data (ICTP, UKMO, ECA&D, MPI-M). This work has been carried out in the framework of the National Science Program “Environmental Protection and Reduction of Risks of Adverse Events and Natural Disasters”, approved by the Resolution of the Council of Ministers № 577/17.08.2018 and supported by the Ministry of Education and Science (MES) of Bulgaria (Agreement № D01-322/18.12.2019) and by the Bulgarian National Science Fund (grant DN-14/3/13.12.2017). This work has been accomplished with the financial support by the Grant № BG05M2OP001-1.001-0003, financed by the Science and Education for Smart Growth Operational Program (2014–2020) and co-financed by the European Union through the European structural and Investment funds.

References 1. Alexander, L.V., et al.: Global observed changes in daily climate extremes of temperature and precipitation. J. Geophys. Res. 111(D5) (2006). https://doi.org/10.1029/2005jd006290 2. Bartholy, J., Pongrácz, R.: Comparing tendencies of some temperature related extreme indices ˝ on global and regional scales. IDOJÁRÁS 110, 35–48 (2006) 3. Buyukalaca, O., Bulut, H., Yılmaz, T.: Analysis of variable-base heating and cooling degreedays for Turkey. Appl. Energy 69, 269–283 (2001) 4. Chervenkov, H., Slavov, K.: STARDEX and ETCCDI climate indices based on E-OBS and CARPATCLIM; Part two: ClimData in use. In: Nikolov, G., et al. (eds.) NMA 2018. LNCS, vol. 11189, pp. 368–374 (2019). https://doi.org/10.1007/978-3-030-10692-841 5. Chervenkov, H., Slavov, K.: Historical climate assessment of temperature-based ETCCDI Climate indices derived from CMIP5 simulations. C. R. Acad. Bulg. Sci. 73(6), 784–790 (2020). https://doi.org/10.7546/CRABS.2020.06.05 6. Chervenkov H., Slavov K.: ETCCDI climate indices for assessment of the recent climate over southeast Europe. In: Dimov, I., Fidanova, S. (eds.) Advances in High Performance Computing. HPC 2019. Studies in Computational Intelligence, vol. 902. Springer, Cham (2021). https:// doi.org/10.1007/978-3-030-55347-0_34 7. Chervenkov, H. Spiridonov, V.: Bias correcting of selected ETCCDI climate indices for projected future climate. In: Lirkov, I., Margenov, S. (eds.) Large-Scale Scientific Computing. LSSC 2019. Lecture Notes in Computer Science, vol. 11958, pp. 292–299 (2020). https://doi. org/10.1007/978-3-030-41032-2_33 8. Chervenkov, H., Spiridonov, V.: Sensitivity of selected ETCCDI climate indices from the calculation method for projected future climate. In: Dimov, I., Fidanova, S. (eds.) Advances in High Performance Computing. HPC 2019. Studies in Computational Intelligence, vol. 902, pp. 413–427. Springer, Cham (2021). https://doi.org/10.1007/978-3-030-55347-0_35 9. Chervenkov, H., Ivanov, V., Gadzev G., Ganev K.: Sensitivity study of different RegCM4.4 model set-ups—recent results from the TVRegCM experiment. Cybern. Inf. Technol. 5(17), 17–26 (2017) 10. Chervenkov H., Ivanov V., Gadzhev G., Ganev K.: (2020) Assessment of the future climate over Southeast Europe based on CMIP5 ensemble of climate indices - Part one: Concept and methods. In: Gadzhev G., Dobrinkova, N. (eds.) Proceeding of 1st International Conference on Environmental Protection and disaster RISKs - Part One, ISBN978-619-7065-38-1 144-156 (2020). https://doi.org/10.48365/envr-2020.1.13 11. Chervenkov H., Ivanov V., Gadzhev G., Ganev K.: Assessment of the future climate over Southeast Europe based on CMIP5 ensemble of climate indices - Part two: Results and discussion. In: Gadzhev G., Dobrinkova, N. (eds.) Proceeding of 1st International Conference on Environmental Protection and disaster RISKs - Part One, ISBN978-619-7065-38-1 157-169 (2020). https://doi.org/10.48365/envr-2020.1.14

Degree-Days and Agro-meteorological Indices in CMIP5 RCP8.5 …

29

12. Cheval, S., Birsan, M.-V., Dumitrescu, A.: Climate variability in the Carpathian Mountains Region over 1961–2010. Global Planet. Change 118, 85–96 (2014). https://doi.org/10.1016/j. gloplacha.2014.04.005 13. CIBSE: Degree-days: theory and application. Technical Manual 41. Chartered Institution of Building Services Engineers, London, UK (2006). ISBN-10: 1-903287-76-6. http://www. degreedaysforfree.co.uk/pdf/tm41.pdf 14. Collins, W.J., Bellouin, N., Doutriaux-Boucher, M., Gedney, N., Halloran, P., Hinton, T., Hughes, J., Jones, C.D., Joshi, M., Liddicoat, S., Martin, G., O’Connor, F., Rae, J., Senior, C., Sitch, S., Totterdell, I., Wiltshire, A., Woodward, S.: Development and evaluation of an Earth-system model–HadGEM2. Geosci. Model Dev. Discuss. 4, 997–1062 (2011). https:// doi.org/10.5194/gmdd-4-997-2011 15. Cornes, R., van der Schrier, G., van den Besselaar, E.J.M., Jones, P.D.: An ensemble version of the E-OBS temperature and precipitation datasets. J. Geophys. Res. Atmos. 123 (2018). https://doi.org/10.1029/2017JD028200 16. Croitoru, A.-E., Holobaca, I.-H., Lazar, C., Moldovan, F., Imbroane, A.: Air temperature trend and the impact on winter wheat phenology in Romania. Clim. Change 111, 393–410 (2012). https://doi.org/10.1007/s10584-011-0133-6 17. European Environment Agency: Heating and Cooling Degree Days. https://www.eea. europa.eu/data-and-maps/indicators/heating-degree-days-2/assessment (2019). Accessed 4 Aug 2020 18. Gadzhev, G., Georgieva, I., Ganev, K., Ivanov, V., Miloshev, N., Chervenkov, H., Syrakov, D.: Climate applications in a virtual research environment platform. Scalable Comput.: Pract. Exp. 19(2), 107–118 (2018). https://doi.org//10.12694/scpe.v19i2.134 19. Gadzhev, G., Ivanov, V., Ganev, K., Chervenkov, H.: TVRegCM Numerical Simulations— Preliminary Results. In: Lirkov, I., Margenov, S. (eds.) Large-Scale Scientific Computing. LSSC 2017. Lecture Notes in Computer Science, vol. 10665. Springer, Cham (2018). https:// doi.org/10.1007/978-3-319-73441-5_28 20. Gadzhev, G., Ivanov, V., Valcheva, R., Ganev, K., Chervenkov, H.: HPC simulations of the present and projected future climate of the Balkan region. In: Dimov, I., Fidanova, S. (eds.) Advances in High Performance Computing. HPC 2019. Studies in Computational Intelligence, vol. 902. Springer, Cham (2021). https://doi.org/10.1007/978-3-030-55347-0_20 21. Giorgi, F., Coppola, E., Solomon, F., Mariotti, L., Sylla, M.B., Bi, X., Elguindi, N., Diro, G.T., Nair, V., Giuliani, G., Turuncoglu, U.U., Cozzini, S., Guttler, I., O’Brien, T.A., Tawfic, A.B., Shalaby, A., Zakey, A.S., Steiner, A.L., Stordal, F., Sloan, L.C., Brankovic, C.: RegCM4: model description and preliminary tests over multiple CORDEX domains. Clim. Res. 52, 7–29 (2012). https://doi.org/10.3354/cr01018 22. Harding, A.E., Rivington, M., Mineter, M.J., Tett, S.F.B.: Agro-meteorological indices and climate model uncertainty over the UK. Clim. Change 128, 113–126 (2015). https://doi.org/ 10.1007/s10584-014-1296-8 23. Harkness, C., Semenov, M.A., Areal, F., Senapati, N., Trnka, M., Balek, J., Bishop, J.: Adverse weather conditions for UK wheat production under climate change. Agric. Forest Meteorol. 282–283, 107862 (2020). https://doi.org/10.1016/j.agrformet.2019.107862 24. Hatfield, J.L., Prueger, J.H.: Agroecology: implications for plant response to climate change. In: Yadav, S.S., Redden, R.J., Hatfield, J.L., Lotze-Campen, H., Hall, A.E. (eds.) Crop Adaptation to Climate Change, pp. 27–43. Wiley, West Sussex, UK (2011) 25. IPCC: Climate change: synthesis report. In: Core Writing Team, Pachauri, R.K., Meyer, L.A. (eds.) Contribution of Working Groups I, II and III to the Fifth Assessment Report of the Intergovernmental Panel on Climate Change, 151 pp. IPCC, Geneva, Switzerland (2014) 26. Ivanov, V., Gadzhev, G., Ganev, K., Chervenkov, H.: Sensitivity of the simulated heat risk in southeastern Europe to the RegCM model configuration preliminary results climate. In: Lirkov, I., Margenov, S. (eds.) Large-Scale Scientific Computing. LSSC 2019. Lecture Notes in Computer Science, vol. 11958, pp. 340–347 (2020). https://doi.org/10.1007/978-3-03041032-2_39

30

H. Chervenkov et al.

27. Ivanov, V., Chervenkov, H., Gadzhev, G., Ganev, K.: Degree-days and agro-meteorological indices in projected future climate over southeast Europe. In: Proceedings of the 20th International Multidisciplinary Scientific GeoConference SGEM 2020, Albena, Bulgaria, 16–25 Aug (2021, in press) 28. Jankovi´c, A., Podrašˇcanin, Z., Djurdjevic, V.: Future climate change impacts on residential ˝ heating and cooling degree days in Serbia. IDOJÁRÁS Q. J. Hung. Meteorol. Serv. 123(3), 351–370 (2019) 29. Kotlarski, S., et al.: Regional climate modeling on European scales: a joint standard evaluation of the EURO-CORDEX RCM ensemble. Geosci. Model Dev. 7, 1297–1333 (2014). https:// doi.org/10.5194/gmd-7-1297-2014 30. Linderholm, H.W.: Growing season changes in the last century. Agric. Forest Meteorol. 137, 1–14 (2006). https://doi.org/10.1016/j.agrformet.2006.03.006 31. Luo, Q.: Temperature thresholds and crop production: a review. Clim. Change 109, 583–598 (2011). https://doi.org/10.1007/s10584-011-0028-6 32. Maraun, D.: Bias correcting climate change simulations—a critical review. Curr. Clim. Change Rep. 2, 211–220 (2016). https://doi.org/10.1007/s40641-016-0050-x 33. Moss, R.H., et al.: The next generation of scenarios for climate change research and assessment. Nature 463(7282), 747–756 (2010). https://doi.org/10.1038/nature08823 34. Pieczka, I., Pongrácz, R., André, K.S., Kelemen, F.D., Bartholy, J.: Sensitivity analysis of different parameterization schemes using RegCM4.3 for the Carpathian region. Theor. Appl. Climatol. 1–14 (2016). https://doi.org/10.1007/s00704-016-1941-4 35. Pieczka, I., Bartholy, J., Pongrácz, R., André, K.S.: Validation of RegCM regional and HadGEM ˝ global climate models using mean and extreme climatic variables. IDOJÁRÁS 123(4), 409–433 (2019) 36. Pongrácz, R., Bartholy, J., Szabo, P., Gelybó, G.: A comparison of the observed trends and simulated changes in extreme climate indices in the Carpathian Basin by the end of this century. Int. J. Global Warm. 1(1/2/3), 336–355 (2009). https://doi.org/10.1504/IJGW.2009.027097 37. Seemann, J., Chirkov, Y.I., Lomas, J., Primault, B.: Agrometeorology. Springer, New York (1979). https://doi.org/10.1007/978-3-642-67288-0 38. Sillmann, J., Röckner, E.: Clim. Change 86, 83 (2008). https://doi.org/10.1007/s10584-0079308-6 39. Sillmann, J., Kharin, V.V., Zwiers, F.W., Zhang, X., Bronaugh, D.: Climate extremes indices in the CMIP5 multimodel ensemble: Part 2. Future climate projections. J. Geophys. Res. Atmos. 118, 2473–2493 (2013). https://doi.org/10.1002/jgrd.50188 40. Spinoni, J., Vogt, J., Barbosa, P.: European degree-day climatologies and trends for the period 1951–2011. Int. J. Climatol. 35, 25–36 (2015). https://doi.org/10.1002/joc.3959 41. Spinoni, J., Vogt, J.V., Barbosa, P., Dosio, A., McCormick, N., Bigano, A., Füssel, H.M.: Changes of heating and cooling degree-days in Europe from 1981 to 2100. Int. J. Climatol. 38, e191–e208 (2018). https://doi.org/10.1002/joc.5362 42. Taylor, K.E., Stouffer, R.J., Meehl, G.A.: An overview of CMIP5 and the experiment design. Bull. Amer. Meteor. Soc. 93, 485–498 (2012). https://doi.org/10.1175/BAMS-D-11-00094.1 43. van Vuuren, D.P., et al.: The representative concentration pathways: an overview. Clim. Change 109, 5–31 (2011). https://doi.org/10.1007/s10584-011-0148-z 44. Zhang, X., Alexander, L., Hegerl, G.C., Jones, P., Tank, A.K., et al.: Indices for monitoring changes in extremes based on daily temperature and precipitation data. WIREs Clim. Change 2, 851–870 (2011). https://doi.org/10.1002/wcc.147

Assessment of the Joint Quantiles of Temperature and Precipitation in CMIP5 Future Climate Projections over Europe Hristo Chervenkov , Georgi Gadzhev , Vladimir Ivanov, and Kostadin Ganev Abstract The present study assesses the changes in the exceedances of the joint extremes of temperature and precipitation quantiles as well as the trend magnitude and statistical significance of these changes. Following the Beniston’s idea, the combination of cool/dry, cool/wet, warm/dry and warm/wet modes in projected future climate over Europe up to the end of the twenty first century is investigated in consistent manner. These modes are defined as excedances of fixed quantile thresholds, the lower and the upper quartile respectively. The use of joint quantiles allows an exploration of climate statistics that in many instances would be overlooked by simply analyzing single thresholds of temperature or precipitation. The used for the computation of the quantiles data for the mean 10-day temperature and 10-day precipitation sum are obtained as ensemble multi-model median from the bias-corrected output of 5 CMIP5 global models, forced with all 4 RCP emission scenarios. The model output is accessed from the section of the Inter Sectoral Impact Model Intercomparison Project in the Copernicus Data Store. Generally, the obtained results are coherent with the consolidated outcomes of the most recent studies, considering the projected future changes of the mean temperature and the precipitation across Europe. Key finding is, however, the revealed steady and statistically significant increase of the number of the extreme warm and dry events over the whole Mediterranean basin. H. Chervenkov (B) National Institute of Meteorology and Hydrology, Tsarigradsko Shose blvd 66, 1784 Sofia, Bulgaria e-mail: [email protected] URL: http://www.meteo.bg G. Gadzhev · V. Ivanov · K. Ganev National Institute of Geophysics, Geodesy and Geography—Bulgarian Academy of Sciences, acad. Georgi Bonchev str., Bl. 3, 1113 Sofia, Bulgaria e-mail: [email protected] URL: http://www.geophys.bas.bg V. Ivanov e-mail: [email protected] K. Ganev e-mail: [email protected] © The Author(s), under exclusive license to Springer Nature Switzerland AG 2021 N. Dobrinkova and G. Gadzhev (eds.), Environmental Protection and Disaster Risks, Studies in Systems, Decision and Control 361, https://doi.org/10.1007/978-3-030-70190-1_3

31

32

H. Chervenkov et al.

The consequences of such tendency could be manifold, including several adverse effects on the ecosystems as well as on managed systems (e.g., agriculture and water supply sector). Keywords Joint exceedances · Temperature and precipitation · Quantile thresholds · CMIP5 ensemble · RCP scenario

1 Introduction A changing climate leads to changes in the frequency, intensity, spatial extent, duration, and timing of weather and climate extremes, and can result in unprecedented extremes [19]. The extreme weather phenomena are discussed in all reports of the Intergovernmental Panel on Climate Change (see, for example, [12]). The rare occurrence of extremes makes it necessary to investigate long data records to determine significant changes in the frequency and intensity of extreme events. There are various methods to investigate extreme events, but statistically consistent approach is to focus on the tails of the probability density function (PDF) of the considered atmospheric variable [1, 2]. Generally speaking, quantities that characterize aspects of the far tails of the distribution tend to be more relevant to society and natural systems than measures that characterize aspects of the distribution that occur more frequently. This is because the more extreme an event, the more likely it is to cause societal or environmental damage [27]. Daily mean temperature and daily precipitation sum, noted for sake of brevity tg and rr henceforth, are core climatic parameters particularly involved in determining climate change impacts on society and ecosystems. Following the leading idea of Beniston [1–3], the present article exploits the 25 and 75% quantile thresholds of the tg and rr, noted tg X 25, tg × 75 rr × 25 and rr × 75 correspondingly, in order to define particular modes of heat and moisture. Four modes are investigated, defined by joint exceedances above or below these thresholds that serve to define “cool/dry” (CD), “cool/wet” (CW), “warm/dry” (WD) and “warm/wet” (WW) regimes. The use of joint PDFs, in this instance, those of temperature and precipitation, provides insight into the behavior of particular modes of heat and moisture that the analysis of the statistics of each variable taken individually does not [1]. To this end, coupled atmosphere-ocean general circulation models (CAOGCMs) are appropriate tools to simulate past, present, and future climate states. Thus, CAOGCMs are able to generate long time series that can be used for model evaluation and also for analyses of possible future changes in extreme events [20]. In the current study, data for the mean daily temperature and daily precipitation sum are obtained as ensemble multi-model median, noted as MMX50 henceforth, from the bias-corrected output of 5 CMIP5 CAOGCMs, forced with all 4 RCP emission scenarios. The quantiles are subsequently calculated and used in the present assessment.

Assessment of the Joint Quantiles of Temperature and Precipitation in CMIP5 …

33

The paper is structured as follows: Concise description of the CMIP5 emission scenarios, the primary source of the input data as well as the used models are in Sect. 2. The core of the article is Sect. 3, where the performed calculations and the obtained results are explained and discussed. Short conclusion remarks as well as outlook for further continuation is placed in the last Sect. 4.

2 CMIP5 Scenarios, Models and Input Data Coupled Model Intercomparison Project (CMIP) is a standard experimental protocol for studying the output of CAOGCMs. CMIP provides a community-based infrastructure in support of climate model diagnosis, validation, intercomparison, documentation and data access [23]. CMIP5 utilizes a new, fourth generation, set of emission scenarios referred to as Representative Concentration Pathways (RCPs) [16, 25]. They are based on radiative forcing trajectories, rather than socioeconomic “storylines” as in the previous generations and are named according to the radiative forcing level at 2100. There are four RCPs: 2.6, 4.5, 6.0, and 8.5, with the numbers representing the 2100 radiative forcing increase relative to pre-industrial levels in W m-2 [22]. The used in this study tg and rr are part of the database of the Inter Sectoral Impact Model Intercomparison Project (ISIMIP 1, https://www.isimip.org/protocol/), Fast Track simulation round, available on the Copernicus Data Store (CDS). This database was recently exploited by the authors for regional climate study [8, 9]. It contains collection of 26 climate variables produced from the bias-corrected output of 5 CMIP5 CAOGCMs according Table 1. The database spans over the period 1950–2099 (historical run up to 2005 and CMIP5 simulations for 2011–2099), downscaled to a 0.5°× 0.5° lat-lon grid and covers the global land area. The present study is based entirely on the ensemble multi-model statistics, namely MMX50, rather than the simulation output of the individual models. Although an

Table 1 Considered in ISIMIP fast track CMIP5 models Model acronym Institution, Country GFDL-ESM2M HadGEM2-ES IPSL-CM5A-LR MIROC-ESM-CHEM NorESM1-M

Geophysical Fluid Dynamics Laboratory, USA Met Office Hadley Centre, UK Institut Pierre-Simon Laplace, France AORI, NIES, JAMSTEC, Japan Norwegian Climate Centre, Norway

Spat. resolution (Lon × Lat ∼ Levels) 144 × 90L24 192 × 145L40 96 × 96L39 128 × 64L80 144 × 96L26

34

H. Chervenkov et al.

equal weighting does not incorporate known differences among models in their fidelity in reproducing various climatic conditions, a number of research studies [11, 14] have found that the multi-model ensemble statistics with equal weighting like the mean and median are superior in any single model in reproducing the presentday climate [22]. Subsequently, this is a modern common approach, adopted in many recent studies [5, 6, 17, 21]. Note that the models used in this study differ from the models in [5–7]. ISIMIP utilizes significantly fewer models, than the applied in [17, 21] but these studies considers only CMIP5 RCP2.6, RCP4.5 and RCP8.5 (i.e. not RCP6.0) and CMIP3 SRES A2 scenario, correspondingly. The main goal of ISIMIP is to offer reliable global climatological data for agroclimatic impact assessments but most of the included variables have universal applicability. They have been calculated for the complete matrix of 5 CAOGCMs×4 RCPs combinations. In addition, as a proxy for historical observations, the “Watch Forcing Data methodology applied to ERA-Interim (WFDEI)” [26] were used to generate historical indicators. This subset is available in the CDS at the same spatial resolution of ISIMIP climate datasets, covers the time range of 1979 to 2013 and its 30 year long part 1981–2010 (baseline period) is used in the present study as reference for the current climate.

3 Performed Calculations and Obtained Results After the download from the CDS the ISIMIP-datasets are significantly postprocessed. The most essential steps of this first stage of the performed calculations are: • The datasets for each model and RCP which are downloadable in 30-years time slices are merged in common data streams for 2011–2099. • The datasets for the tg and rr are joined in common netCDF4 files. • Due to storage constrains only a spatial subset over Europe is preserved. All netCDF manipulations are performed with the powerful tool Climate Data Operators (https://code.zmaw.de/projects/cdo). Due to agrometeorological reasons, the temporal resolution of the tg and rr is 10 days. The original study [1] is based on daily data, but decadal data is still reasonable, and coarser resolution in time could lead to over-smoothing. All CAOGCMs listed in Table 1 use 360-day year and thus one year contains 36 data records (i.e. decades).

3.1 Ensemble Spatial Patterns and Temporal Evolution The first step of the second stage of of the performed calculations is the computation of the lower and upper quantile thresholds of tg and rr namely tg × 25, tg × 75, rr × 25 and rr × 75. While there is a consensus view that the 10 and 90% quantiles

Assessment of the Joint Quantiles of Temperature and Precipitation in CMIP5 …

35

Fig. 1 From left to right: tg × 25, tg × 75 (units: ◦ C), rr × 25 and rr × 75 (units: mm)

define an extreme in the PDF of the considered variable (e.g., [12]), the values are set here, as in [1], at 25 and 75% in order to capture a larger number of events. In [1] is argued that these thresholds enable to focus on particular modes of variability that can have perceptible impacts on environmental (e.g., hydrology) and managed systems (e.g., agriculture) that the use of the 50% percentile (the median) would not. The use of joint quantiles allows an exploration of climate statistics that in many instances would be overlooked by simply analyzing single quantile thresholds of temperature or precipitation. Hence the reference period 1981–2010 is 30-years long the threshold quantiles are computed for any 30 × 36 = 1080 long time series for each grid cell individually. Figure 1 shows the spatial distributions of the quantile thresholds. The spatial patterns of the thermal quantiles is dominated by continental gradient from Northeast to Southwest for tg × 25 and from North to South for tg × 75. The spatial pattern of the precipitation quantiles is more complex: The southern part of the domain is, as a whole, dryer. There is also broad maximum of the rr × 75 over Central and West Europe and along the west coasts to the Atlantic ocean and to the Mediterranean sea. Next, the joint exceedance of the PDFs of mean temperature and precipitation is obtained by counting the frequency of exceedance, for each year in the period 1981–2099, below or above the four combinations of heat and moisture quantiles, i.e., tg × 25/rr X 25, tg × 25/rr × 75, tg × 75/rr × 25, and tg × 75/rr × 75, that define respectively the CD, CW, WD, and WW modes. Our analysis starts with a comparison of the spatial patterns of the multiyear means of the CMIP5 projections of the four modes for 2070–2099 with their counterparts for the reference period. In order to intercompare them, the multi-model ensemble median (MM X 50) for all 4 scenarios is superimposed to the median for the reference period (1981–2010), as shown on Figs. 2 and 3. The multiyear means are presented as relative share in respect to the total annual number of records (i.e. 36). The most apparent result from the analysis of Figs. 2 and 3 is the steady decrease of the cold-related modes and especially CD. The projected changes of the both modes demonstrates gradual reduction from RCP2.6 to RCP8.5, i.e. proportional to the radiative forcing. The long-term changes of the other modes are more complex according their spatial heterogeneity and non-negligible differences from scenario to scenario. Due to the warming tendency over the continent, it is reasonable to expect generally

36

H. Chervenkov et al.

Fig. 2 Multiyear means of the medians of the modes CD, WW, CW and WD (in %) in the first, second, third and fourth row correspondingly for the reference period in the first column and multiyear means for 2070–2099 for RCP2.6, RCP4.5, RCP6.0 and RCP8.5 in the second, third, fourth and fifth column respectively

positive (i.e. increase of the number of events) changes of the WW. In fact, the WW increases over the bigger part of the domain in the scenarios RCP2.6–RCP6.0. In contrast, the change in RCP8.5 is prevailing negative except the NE domain corner and the Scandinavian Peninsula. The possible reason could be rooted in the evidenced in many studies [10, 17, 21] overall drying tendency over the pointed region which, in the case, overwhelms the warming. The spatial patterns of the WD demonstrates the strongest areal contrast - the change over the bigger part of Europe in RCP2.6–RCP6.0 is weak negative (i.e. decrease of the WD-cases in the future); over the southern edge, including North Africa, parts of the Balkan Peninsula, Asia Minor and the Near East - strong positive. Transition zone is practically absent. The reduction of the WD over Central and North Europe is, evidently, caused mainly by the increase in RCP2.6– RCP6.0 of the precipitation totals there [21]. The almost opposite tendency in RCP8.5 could be linked, as stated above, to the coherent warming, which is significant and projected precipitation decrease. The temporal evolution over the whole period of interest 1981–2099 of the areaweighted averages (AA) over land of the considered modes, are shown on Fig. 4. The step-wise discontinuity between the reference and the projected future is notable in the temporal evolution of all modes especially of CD. This fact could be linked to the well-known and documented in many articles (see, for example, [20, 21] and references therein) model deficiencies in the simulation of weather extremes, especially precipitation extremes, which even the bias-correction can not

Assessment of the Joint Quantiles of Temperature and Precipitation in CMIP5 …

37

Fig. 3 From left to right: absolute changes of the RCP2.6, RCP4.5, RCP6.0 and RCP8.5 relative to the reference period in the of the modes CD, WW, CW and WD (in %) in the first, second, third and fourth row correspondingly

Fig. 4 Temporal evolution of the AAs of the CD, WW, CW and WD (in %) in the first, second, third and fourth row correspondingly for the reference (black line) and ensemble MMX50 for the RCP2.6 (blue), RCP4.5 (green), RCP6.0 (yellow) and RCP8.5 (red). Solid lines indicate the running 3-year means

38

H. Chervenkov et al.

(at least not completely) dampen. The evolution lines of the CD, WW and partially CW are practically overlapping for all RCPs. The WW demonstrates the biggest abrupt changes from year to year. Most steady appears the increase of the WD, with clear difference from scenario to scenario.

3.2 Trend Analysis The importance of assessing trends of key variables is often emphasized in the modern climatology and, subsequently, it is essential part of many recent studies (see, for example, [1, 5–7, 17, 21] and references therein). The magnitude of the trend is estimated with the Theil-Sen Estimator (TSE) [18, 24], which is preferably used as a superior alternative of the ordinary least squares [4]. The statistical significance is analyzed with the non-parametric Mann-Kendall (MK) test [13, 15]. As the TSE, The MK test is a rank-based procedure, especially suitable for non-normally distributed data, data containing outliers and nonlinear trends. Both methods are practically standard tools for trend analysis in the climatology. In the present study, they are applied for every grid point time series and each scenario individually. The results from the trend analysis of the considered temperature/precipitation modes are shown in Fig. 5. The absence of statistically significant trend over substantial part of the domain for all modes and scenarios is the most notable result of the analysis of Fig. 5. In contrast to the expectations for the CD mode, practically there is no trend under any scenario, except relatively small regions in South and North. The reason is rooted in the relatively small value of this mode even in the reference period as showed in Fig. 2. Thus, in the conditions of generally warmer climate, CD drops close to zero (i.e. constant), causing absence of trend. The WW is the single one mode which demonstrates (statistically significant) trend with opposite signs. Over part of CSE Europe in RCP8.5 the trend is negative (i.e. tend to lower number of events) and in the far North it is positive. Over the residual, much bigger part of the domain there is not significant trend. It is worth to emphasize that the ISIMIP-collection contains the index warm and wet days. Our preliminary evaluations reveals that it demonstrates steady increase in all scenarios proportionally to the radiative forcing. Unlike the WW mode considered here, however, it is calculated on daily, rather than decadal basis. It is well known that the daily excedances, both of temperature and precipitation sums, are more frequent, hence the longer summarization period smooths the extremes. This example demonstrates the importance of standardized definitions of the climate indicators being in use worldwide [27]. The CW mode shows prevailing negative trend but this trend is statistically significant over continental Europe over small isolated regions. The picture for the WD is most homogeneous according the spatial structure and most consistent across all scenarios. The positive trend magnitude is apparently proportional to the radiative forcing. It is also statistically significant over the substantial part of the domain in RCP8.5. The rate of the of the WD-cases increase

Assessment of the Joint Quantiles of Temperature and Precipitation in CMIP5 …

39

Fig. 5 Trend magnitude (units: decades/10 yr) of the MMX50s of the CD, WW, CW and WD in the first, second, third and fourth row correspondingly for RCP2.6, RCP4.5, RCP6.0 and RCP8.5 in the first, second, third and fourth column respectively. Stippling indicates grid points with changes that are not significant at the 5% significance level

is essential over the whole Mediterranean basin and vast surrounding territories. This fact could play major role in the future due to the evidenced in many studies [6, 17, 19, 21] drought vulnerability of the region.

4 Conclusion Based on the availability of new sources of information representing the state of the art global climate change simulations in the frame of the CMIP5 project which are free accessible from the Copernicus Data Store, we present a assessment of the joint quantiles of temperature and precipitation for the projected future climate over Europe. The study confirms the suitability of the database created from ISIMIP 1 products for the performed assessment in computationally feasible way as has been already shown in [8, 9]. Beside its methodological transparency, the adopted approach has at least two other merits. First, as stated in [27] and demonstrated here, the quantile thresholds allow for spatial comparisons over large regions with complex topography because

40

H. Chervenkov et al.

they sample the same part of the probability distribution of temperature and precipitation at each location. They also can account for isolated missing values in a relatively straightforward fashion. The percentiles itself are robust estimators of the sample—they are much less sensitive from, say, the arithmetic mean to outliers [4]. Such values, however, could be find even in time series of data from model output. Second, which is probably more important, this approach is non-parametric. This is essential because some of the atmospheric variables are typical examples for non-normally distributed data. As a whole, the obtained results are in principal agreement with the documented in several publications [7–9, 17, 19, 21] projected pan-European changes of the temperature and precipitation reflecting at the same time the complexity of the synergetic tendencies of the both atmospheric variables. Key finding is, however, the projected steady increase of the extreme warm and dry events over the Mediterranean basin. Keeping in mind the high drought vulnerability of the region, this could lead to several adverse effects on the ecosystems as well as on managed systems (e.g., agriculture and water supply sector). The study could be continued in many aspects. It is reasonable, for example, to enrich the general picture with regional details, obtained from exploitation of the output from regional climate models in similar manner. More in-depth investigations on seasonal basis could be also performed, always keeping in mind the high ecological and social importance of the climatological research. Acknowledgements Hence this study is entirely based on free available data and software, the authors would like to express their deep gratitude to the primary CMIP5 model output vendors as well as all other organizations and institutions (MPI-M, UNI-DATA, CDS), which provides free of charge software and data. This work has been carried out in the framework of the National Science Program “Environmental Protection and Reduction of Risks of Adverse Events and Natural Disasters”, approved by the Resolution of the Council of Ministers № 577/17.08.2018 and supported by the Ministry of Education and Science (MES) of Bulgaria (Agreement № D01-322/18.12.2019) and by the Bulgarian National Science Fund (grant DN-14/3/13.12.2017). This work has been accomplished with the financial support by the Grant № BG05M2OP001-1.001-0003, financed by the Science and Education for Smart Growth Operational Program (2014–2020) and co-financed by the European Union through the European structural and Investment funds.

References 1. Beniston, M.: Trends in joint quantiles of temperature and precipitation in Europe since 1901 and projected for 2100. Geophys. Res. Lett. 36, L07707 (2009). https://doi.org/10.1029/ 2008GL037119 2. Beniston, M.: Decadal-scale changes in the tails of probability distribution functions of climate variables in Switzerland. Int. J. Climatol. 29, 1362–1368 (2009). https://doi.org/10.1002/joc. 1793 3. Beniston, M., Stephenson, D., Christensen, O., et al.: Future extreme events in European climate: an exploration of regional climate model projections. Climatic Change 81, 71–95 (2007). https://doi.org/10.1007/s10584-006-9226-z

Assessment of the Joint Quantiles of Temperature and Precipitation in CMIP5 …

41

4. Chervenkov, H., Slavov, K.: Theil-Sen estimator vs. ordinary least squares – trend analysis for selected ETCCDI climate indices. Compt. Rend. Acad. Bulg. Sci. 72(1), 47–54 (2019). https:// doi.org/10.7546/CRABS.2019.01.06 5. Chervenkov, H., Slavov, K.: Historical climate assessment of temperature-based ETCCDI climate indices derived from CMIP5 simulations. C. R. Acad. Bulg. Sci. 73(6), 784–790 (2020). https://doi.org/10.7546/CRABS.2020.06.05 6. Chervenkov, H., Slavov, K.: Historical climate assessment of precipitation-based ETCCDI climate indices derived from CMIP5 simulations. C. R. Acad. Bulg. Sci. 73(7), 942–948 (2020). https://doi.org/10.7546/CRABS.2020.07.06 7. Chervenkov H., Slavov K.: ETCCDI thermal climate indices in the CMIP5 future climate projections over Southeast Europe. In: Proceedings of the 14th Annual Meeting of the Bulgarian Section of SIAM, Studies in Computational Intelligence (2020) (in press) 8. Chervenkov H., Ivanov V., Gadzhev G., Ganev K.: Assessment of the future climate over Southeast Europe based on CMIP5 ensemble of climate indices - Part one: concept and methods. In: Gadzhev G., Dobrinkova, N. (eds) Proceeding of 1st International Conference on Environmental Protection and disaster RISKs - Part one, ISBN978-619-7065-38-1 144–156 (2020). https://doi.org/10.48365/envr-2020.1.13 9. Chervenkov H., Ivanov V., Gadzhev G., Ganev K., (2020) Assessment of the future climate over Southeast Europe based on CMIP5 ensemble of climate indices - Part two: results and discussion. In: Gadzhev G., Dobrinkova, N. (eds) Proceeding of 1st International Conference on Environmental Protection and disaster RISKs - Part one. ISBN978-619-7065-38-1 157–169. https://doi.org/10.48365/envr-2020.1.14 10. Dai, A.: Drought under global warming: a review. WIREs Clim. Change 2, 45–65 (2011). https://doi.org/10.1002/wcc.81 11. Herger, N., Abramowitz, G., Knutti, R., Angélil, O., Lehmann, K., et al.: Selecting a climate model subset to optimise key ensemble properties. Earth Syst. Dyn. 9, 135–151 (2018). https:// doi.org/10.5194/esd-9-135-2018 12. Solomon, S. et al.: The Physical Science Basis. Contribution of working group I to the fourth assessment report of the intergovernmental panel on climate change. Intergovernmental Panel on Climate Change (IPCC): Climate Change 2007. Cambridge University Press, Cambridge, U. K. (2007) 13. Kendall, M.G.: A new measure of rank correlation. Biometrika 30, 81–93 (1938) 14. Knutti, R.: The end of model democracy? Climatic Change 102, 395–404 (2010). https://doi. org/10.1007/s10584-010-9800-2 15. Mann, H.B.: Nonparametric tests against trend. Econometrica 13, 245–259 (1945) 16. Moss, R.H., et al.: The next generation of scenarios for climate change research and assessment. Nature 463, 747–756 (2010). https://doi.org/10.1038/nature08823 17. Orlowsky, B., Seneviratne, S.I.: Elusive drought: uncertainty in observed trends and short- and long-term CMIP5 projections. Hydrol. Earth Syst. Sci. 17(5), 1765–1781 (2012). https://doi. org/10.3929/ethz-b-000073994 18. Sen, P.K.: Estimates of the regression coefficient based on Kendall’s tau. J. Am. Stat. Assoc. 63, 1379–1389 (1968) 19. Seneviratne, S., Nicholls, N., Easterling, D., Goodess, C., Kanae, S., Kossin, J., Luo, Y., Marengo, J., McInnes, K., Rahimi, M., Reichstein, M., Sorteberg, A., Vera, C., Zhang, X., et al.: Changes in climate extremes and their impacts on the natural physical environment. In: Field, C.B., Barros, V., pp. 109–230 (2012). https://doi.org/10.1017/CBO9781139177245.006 20. Sillmann, J., Röckner, E.: Indices for extreme events in projections of anthropogenic climate change. Climatic Change 86, 83–104 (2008). https://doi.org/10.1007/s10584-007-9308-6 21. Sillmann, J., Kharin, V.V., Zhang, X., Zwiers, F.W., Bronaugh, D.: Climate extremes indices in the CMIP5 multimodel ensemble: Part 2. Future climate projections. J. Geophys. Res. Atmos. 118, 2473–2493 (2013). https://doi.org/10.1002/jgrd.50188 22. Sun, L., Kunkel, K.E., Stevens, L.E., Buddenberg, A., Dobson, J.G., Easterling, D.R.: Regional surface climate conditions in CMIP3 and CMIP5 for the United States: differences, similarities, and implications for the U.S. national climate assessment. NOAA Technical Report NESDIS, vol. 144, 111 pp. (2015). https://doi.org/10.7289/V5RB72KG

42

H. Chervenkov et al.

23. Taylor, K.E., Stouffer, R.J., Meehl, G.A.: An overview of CMIP5 and the experiment design. Bull. Amer. Meteor. Soc. 93, 485–498 (2012). https://doi.org/10.1175/BAMS-D-11-00094.1 24. Theil, H.: A rank-invariant method of linear and polynomial regression analysis. I, II, III, Nederl. Akad. Wetensch., Proc. 53, 386–392, 521–525, 1397–1412 (1950) 25. van Vuuren, D.P., et al.: The representative concentration pathways: an overview. Clim. Chang. 109, 5–31 (2011). https://doi.org/10.1007/s10584-011-0148-z 26. Weedon, G.P., Balsamo, G., Bellouin, N., Gomes, S., Best, M.J., Viterbo, P.: The WFDEI meteorological forcing data set: WATCH forcing data methodology applied to ERA-reanalysis data. Water Resour. Res. 50(9), 7505–7514 (2014) 27. Zhang, X., Alexander, L., Hegerl, G.C., Jones, P., Tank, A.K., et al.: Indices for monitoring changes in extremes based on daily temperature and precipitation data. WIREs Clim. Change 2, 851–870 (2011). https://doi.org/10.1002/wcc.147

Coastal Boundary-Layer Characteristic During Night Time Using a Long-Term Acoustic Remote Sensing Data Damyan Barantiev

and Ekaterina Batchvarova

Abstract The study of the Planetary Boundary Layer vertical structure in coastal areas is of particular importance due to the fact that a large number of urban areas and their industrial activities are located on the shores of the seas, oceans or large lakes. Based on long-term (August 2008–October 2016) sodar measurements at a Bulgarian Black Sea coastal site, the mean characteristics of the two main types of nocturnal air flows (marine and land air masses) are obtained. Typical parameters for the investigated region, such as the heights of the marine, the internal and planetary boundary layers, as well as wind and turbulence vertical structure details are revealed exploring this high spatial (10 m) and temporal (10 min) resolution data. The observation site is near the town of Ahtopol in Southeast Bulgaria. The analyses are based on averaging of the measured profiles of 12 output sodar parameters and calculated Buoyancy Production mean profiles. The seasonal variability of all characteristics is explored. The nocturnal land air masses are found to be with neutral and slightly stable stratification, Planetary Boundary-Layer height of 410–430 m and corresponding Surface-Layer height of 50–80 m. The nocturnal marine air masses are found to be with neutral and slightly unstable stratification, Internal BoundaryLayer height of about 40–50 m and a nocturnal marine Planetary Boundary-Layer height of about 300 m. The study contributes to disclosure and understanding the coastal nocturnal wind and turbulence regime in a region with modest observation networks. The obtained results can be also used for evaluation of the various theoretical, mesoscale and air quality models performance. Keywords Sodar · Remote sensing data · Wind and turbulent profiles · Black sea · Coastal area · PBL · Climatological studies

D. Barantiev (B) · E. Batchvarova Climate, Atmosphere and Water Research Institute—Bulgarian Academy of Sciences, CAWRI—BAS, 66, Tsarigradsko Shose blvd, 1784 Sofia, Bulgaria e-mail: [email protected] © The Author(s), under exclusive license to Springer Nature Switzerland AG 2021 N. Dobrinkova and G. Gadzhev (eds.), Environmental Protection and Disaster Risks, Studies in Systems, Decision and Control 361, https://doi.org/10.1007/978-3-030-70190-1_4

43

44

D. Barantiev and E. Batchvarova

1 Introduction The significant technological development of ground-based instruments for remote measurements of atmospheric characteristics during the last years places them among the reliable and indispensable instruments in a number of innovative scientific methods for the study of the main meteorological parameters and turbulence in the Planetary Boundary Layer (PBL) [1–6]. Substantial interest in studying the coastal PBL is the complexity of the processes observed in air masses transformation due to the sharp change in the physical characteristics of the surface. The challenges in describing coastal processes are related to Internal Boundary Layer (IBL) formation and its evolution as a sublayer of PBL when marine airflow comes over land and define its spatial scales with a height dependent on the distance from the shore [7, 8]. Due to the formation of such sublayers with different thermodynamical states and the presence of local circulation in the coastal zones (breeze), the atmosphere is complexly stratified, resulting in intricate processes of air pollution dispersion compared to regions with homogeneous surface [9]. A number of scientific experiments with doppler lidars, sodars, high meteorological masts, surface and aerological measurements aim to provide data to evaluate the mesometeorological models performance in coastal areas and to ensure further development of parameterizations. Wilczak et al. [10] present the results of a complex experiment to study varying scales of airflows in California’s coastal zone, with a center in Santa Barbara. The experiment was conducted on September 20, 1985. Doppler wind lidars, sodars, multiple stations with surface observations, and radio sounding were used. Part of the sodars measurements were also carried out at the platforms in the ocean. These data were used to verify the accuracy of mesometeorological models’ simulations. An IBL study in a complex coastal area was based on airborne lidar measurements, mesometeorological modeling with CSU-RAMS and the use of analytical models for the Pasific 93 experiment, Vancouver, Canada. Good agreement between of the modelled and measured height of the IBL was obtained despite its complex structure in space and time [11]. Batchvarova and Gryning [12] describe the development of an IBL in the Athens area at the time of the international experiment MEDCAPHOT-TRACE 1994 to study the processes of distribution, photochemical transformation and transmission of pollutants with the ultimate goal of forming a strategy for better ambient air quality for the 2004 Olympic Games. Through a numerous of surface observations and vertical sounding systems with tethered balloons, IBL height of 400 m is found at 4 km from the coast and of 700 m at 13 km from the coast. De Leo et al. [13] analyze sodar data and mesometeorological modeling of the breeze circulation in the area of Lamezia Terme, in the Tyrrhenian Sea in Calabria in the summer of 2007. The sodar data included in the work shows a clear alternation of positive vertical wind speeds during the day with airflow from the sea to the land (from the west) and negative at night from land to sea (east) airflow.

Coastal Boundary-Layer Characteristic During …

45

Prabha et al. [14] describe observations with sodar and the formation of a Thermal Internal Boundary Layer (TIBL) in sea breeze in India for 10 days in February 1998 under conditions of switching from winter to summer circulation. The sodar is located on land at a distance of 5 km from the shore, and the difference in land and water temperatures is from 0.5 to 3.5 K. In this situation, the measured TIBL height through the peak in the vertical dispersion profile is about 150–200 m. With higher temperature amplitude on the two main surfaces, the TIBL’s height is expected to increase much faster with the sea breeze spread overland. Petenko et al. [15] present a study of the turbulent characteristics and the height of the PBL at Concordia Station (Dome C) in Antarctica. The main measuring instrument is a sodar with an exceptional resolution of 2 m height and a range of about 200 m. This work discusses in detail the use of data as a source for the study of turbulence in PBL. It can be noted that in the literature there is research dedicated to the coastal boundary layer using sodar measurements in many places around the world, but mainly for short periods or even days. These studies prove the great capabilities of acoustic remount sounding to study the wind and turbulent structure of the PBL, particularly in coastal areas. The analysis of long-term data proposed in this paper discloses the coastal nocturnal wind regime and turbulent structure in a region with modest observation networks in Bulgaria, but are of importance for coastal climatology in general. Except for theoretical research, the created database can be used for regime studies and evaluation of model performance in coastal areas.

2 Measuring Site and Equipment Meteorological Observatory (MO) Ahtopol is located in the south-eastern part of the Bulgarian Black Sea coast (Fig. 1) at about 2 km southeast of the town of Ahtopol. The observatory falls into the Black Sea coastal Strandzha climate region, which is a part of the Black Sea climatic sub-region of Continental-Mediterranean climatic zone in Bulgaria [16]. Well expressed breeze circulation in the warm half of the year is typical for the climate in the region. Local circulation in the study area is observed throughout the year, whereas during the cold season a lower frequency and smaller time and spatial scales are registered [17]. MO Ahtopol is situated primarily on a flat grassland at about 400 m inland and 30 m height above sea level. The coast line is stretching out from NNW to SSE with a steep about 10 m high cost (Fig. 1). The observations are performed with an acoustic mono-static Doppler remote sensing system—SCINTEC Flat Array Sodar MFAS with a frequency range of 1650–2750 Hz, 9 emission/reception angles (0°, ±9.3°, ±15.6°, ±22.1°, ±29°), a vertical range from 150 to 1000 m and a vertical resolution of 10 m. The accuracy for wind speed is 0.1–0.3 ms−1 and for wind direction is 2–3° [18]. The sodar system is mounted on the roof of MO Ahtopol at about 4.5 m above

46

D. Barantiev and E. Batchvarova

Fig. 1 Location of MO Ahtopol in southeastern Bulgaria on Google Earth (42° 5 3.37 N, 27° 57 4.49 E)

the ground. The data are recorded every 10 min and the average period is 20 min. The first measurement level is 30 m and the maximum vertical range for these study reaches 700 m.

3 Data and Analysis Overview 3.1 Data Availability The exploration of coastal PBL characteristics in this work covers a 3014-day period from 1 August 2008 to 31 October 2016. The continuity of operation of the sodar was disturbed by frequent accidents of the main power supply and practically hampered remote technical support of the installed equipment by the lack of Internet access in observatory until 2011. In the summer months of 2008 and 2009, the sodar was stopped during the night hours. From 1st August to 31st October 2008 sodar measurements were made from 7:00 a.m. to 6:00 p.m. From 8th December 2008 to 27th July 2009 the sodar functioned continuously and after that until December 2009 only during the day, from 7:30 a.m. to 9:40 a.m. After December 2009 the sodar was worked continuously. The data availability of the atmospheric acoustic sounding is presented in Table 1. The low data availability (yellow markings) during the summers of 2008 and 2009 is mainly due to the restrictions placed on the night mode of operation while in the other months it is due to the occurrence of interruption of electricity at MO Ahtopol. Data with availability above 70% is given in green, between 40 and 70% in yellow, and below 40% in red. During the years of operating mode, the manufacturer provided periodic updates to the sodar software, resulting in a gradual increase in the optimal height of the output profiles while preserving the strict data quality control

Coastal Boundary-Layer Characteristic During …

47

Table 1 Monthly data availability and maximum effective height reached during the sodar operating mode for the period 01.08.2008–10.31.2016 I

II

III

IV

V

VI

2008 max range [m]

-

-

-

-

-

-

VII

VIII

IX

X

XI

XII

2009 max range [m]

99.9% 520

99.6% 520

99.2% 680

96.7% 680

98.3% 680

99.3% 680

2010 max range [m]

97.5% 680

98.3% 680

89.0% 680

68.5% 680

96.8% 680

86.4% 680

2011 max range [m]

96.8% 510

94.8% 460

99.8% 510

96.7% 510

96.8% 510

96.2% 510

2012 max range [m]

75.3% 620

95.8% 620

99.7% 620

96.7% 620

99.3% 620

93.3% 620

99.1% 700

31.9% 620

96.6% 720

30.6% 670

23.3% 640

100.0% 720

2013 max range [m]

96.6% 720

96.4% 720

50.6% 720

100.0% 620

98.3% 680

54.7% 670

95.3% 680

96.3% 680

96.7% 720

58.0% 720

74.7% 720

61.7% 720

2014 max range [m]

100.0% 720

99.9% 720

99.7% 720

100.0% 720

97.1% 720

99.8% 720

90.0% 720

59.4% 720

51.7% 750

98.5% 750

97.5% 750

98.0% 730

2015 max range [m]

99.9% 750

96.3% 750

94.9% 750

97.8% 750

69.5% 750

73.3% 750

57.4% 590

97.9% 750

99.5% 750

99.9% 750

65.9% 750

95.1% 730

2016 max range [m]

80.1% 750

70.0% 750

76.1% 750

49.9% 750

35.6% 750

93.2% 1000

99.9% 1000

93.0% 1000

75.0% 1000

45.4% 750

45.2% 520

40.3% 520

57.7% 520

59.5% 520

88.4% 520

94.6% 680

57.5% 680

59.9% 680

58.4% 680

96.7% 680

96.0% 680

99.9% 560

98.3% 510

99.9% 510

92.6% 510

99.6% 510

99.8% 510

92.2% 510

99.9% 510

38.8% 560

78.8% 620

99.2% 620

81.3% 620

-

-

build into the sodar software and the spatial and temporal resolution of the data selected at the beginning of the atmospheric sounding at MO Ahtopol (July 2008). The maximum height specified in the measurement settings is not guaranteed. The actual height (effective) of the output profiles is determined by the availability of turbulent temperature inhomogeneous in the atmosphere above the sodar, which can return the signals to the antenna receiver. The actual sodar range is also presented graphically in Table 1 with one bar filled at a height of 510 m, 2 bars over 610 m, 3 bars over 710 m and 4 bars over 810 m. In this paper all profiles are studied up to 700 m due to low availability of data above that height.

3.2 Analysis In this climatological study of coastal PBL vertical structure with high spatial and temporal resolution the direction from 0 to 120° is determined for marine air masses, while for the air masses from the land the direction is in the range between 170 and 290°. The entire 3014-day study period contains total of 2708 days with measurements, representing nearly 90% availability from all days from the total period. A summary of the analysis of the nocturnal marine air masses and those from the land is presented in Table 2. Conducted analyses of all land (brown color in Table 2) and marine (blue color in Table 2) air masses are deepened with study by season (cold and warm part of the years) for exploring seasonal variability of coastal PBL characteristics. The cold part of the years is defined from November to March and the warm period from May to September (Table 2—first column). Due to the presence of breeze circulation in the studied area, the number of profiles involved in the averaged characteristics of land air masses (10.5% from the total period) are significantly higher than marine air masses profiles (2.5% from the total period). The conditions for filtering the data for various analyses (fourth column at Table 2) are as follows:

48

D. Barantiev and E. Batchvarova

Table 2 Types of long-term analyses of night coastal air masses by wind direction and different conditions Period of analysis from 08/2008 to 10/2016 (total - 3014 days). Days with observaƟons - 2708 days Analysis of land air masses /170 ÷ 290 Deg/ Nocturnal air masses /21 ÷ 05h/

Height [m]

Available profiles

CondiƟons

700/ 630/ 550*/ 450*/ 45 549/ 8 671/ 39*/ 244*/ 3/ 2/ 1*/ 1*/

(45 549 profiles) 10.5 %

350*/ 320*/ 300*/

1 388*/ 2 027*/ 2 571*/

1*/ 1*/ 1*/

from the total period

270*/ 250*/ 150*

3 523*/ 4 278* / 8022*

1*/ 1*/ 1*

Nocturnal /November÷ March/

700/ 630/ 550*/ 450*/ 18 616/ 4 458/ 21*/ 178*/ 3/ 2/ 1*/ 1*/

(18 616 profiles) 4.3 %

350*/ 320*/ 300*/

799*/ 1 224*/ 1 596*/

from the total period

270*/ 250*/ 150*

2 143*/ 2540* / 4262*

1*/ 1*/ 1*

730/ 610/ 550*/ 450*/

18 769/ 2 603/ 18*/ 37*/

3/ 2/ 1*/ 1*/

350*/ 320*/ 300*/

379*/ 504*/ 602*/

1*/ 1*/ 1*/

270*/ 250*/ 150*

811*/ 1 021* / 2 311*

1*/ 1*/ 1*

Height [m]

Available profiles

CondiƟons 3/ 2/ 1*/ 1*

Nocturnal /May ÷ September/ (18 769 profiles) 4.3 % from the total period Analysis of marine air masses /0 ÷ 120 Deg/ Nocturnal air masses /21 ÷ 05h/

1*/ 1*/ 1*/

700/ 590/ 550*/ 450*/

10 919/ 3 731/ 3*/ 9*

(10 919 profiles) 2.5 %

350*/ 320*/ 300*/

25*/ 55*/ 89*/

1*/ 1*/ 1*

from the total period

270*/ 250*/ 150*

196*/ 400*/ 3 120*

1*/ 1*/ 1*

700/ 340/

3 939/ 1 141/

3/ 2/

320*/ 300*/

22*/ 36*/

1*/ 1*/ 1*

Nocturnal /November÷ March/ (3 937 profiles) 0.9 % from the total period Nocturnal /May ÷ September/

270*/ 250*/ 150*

91*/ 159* / 854*

1*/ 1*/ 1*

700/ 590/ 550*/ 450*/

4 666/ 1 863/ 3*/ 9*

3/ 2/ 1*/ 1*

(4 666 profiles) 1.1 %

350*/ 320*/ 300*/

21*/ 26*/ 35*/

1*/ 1*/ 1*

from the total period

270*/ 250*/ 150*

54*/ 151* / 1 654*

1*/ 1*/ 1*

• 1*—continuous profiles to fixed heights with simultaneous availability of 12 sodar output parameters (wind direction/WD/, wind speed and its dispersion/WS, sigWS/, vertical wind speed and its dispersion/W, sigW /, horizontal wind speed components and their dispersions /U, sigU, V, sigV /, eddy dissipation rate /EDR/, turbulent intensity /TI/ and turbulent kinetic energy /TKE/); • 2—continuous profiles with a minimum height of 110 m and simultaneous availability of 12 sodar output parameters; • 3—profiles consisting of a minimum of 3 points in height satisfying the wind direction condition and permitting an interruption only for lack of data. The complexity of different fulfilled conditions in the extracted profiles (1*, 2 or 3) determine the number of profiles involved in the various analyses (Table 2—third column) and the maximum height to which the average profiles reach (Table 2— second column). The lowest availability of profiles involved in averaging of coastal PBL characteristics is under condition 1*, due to its strength as filter—continuous profiles of 12 different simultaneous available parameters to fixed heights (150, 250, 270, 300, 320, 350, 450 and 550 m). The complexity of fulfilling this condition lies in the fact that measurements with remote sensing instruments may fail at some levels and lead to missing data at given altitudes or the algorithms for signal quality of the instrument may reject data at some levels.

Coastal Boundary-Layer Characteristic During …

49

As there is a need for aerosols in the atmosphere for the doppler lidars for uninterrupted data availability in profiles, there is a need for temperature inhomogeneities for the sodar. Thus, in areas with very low aerosol content, the lidars work is hampered. Similarly, the sodar work is hampered in areas with homogeneous temperature conditions. These features lead to different data availability in the averaged profiles at different heights and so impose the use of different analysis—all profiles with available data at least three points in height defined by condition 3 (though intermittently) or filtered data—samples that meet different conditions as 1* and 2. Using conditions 2 and 3 results in averaged profiles with not fixed-size and different data availability involved in calculating of certain mean value at a corresponding height while the condition 1* results in averaging of the same numbers of values across the entire height of the fixed-size averaged profile.

4 Averaged Vertical Profiles of Nocturnal Coastal Boundary Layer Estimating the PBL height through the turbulent profiles characteristics (sigW, TKE, BP, etc.) has been suggested in recent years in studies based on data from remote sensing measurements [5, 19]. In the absence of temperature and humidity profiles in the atmosphere, such analyses make it possible to retrieve more information from wind data. While with the lidars the calculations of sigW are based on a backscatter signals from aerosols, the measurement of turbulent characteristics with the sodars is direct. In this paper, all profiles of the turbulent parameters are presented and explored first as information from all long-term measurements with given characteristics and then divided by cold and warm part of the year. The nocturnal PBL falls better in the range of the sodar which allows to assess its characteristics, including its height.

4.1 Nocturnal Land Air Masses All nocturnal land air masses and condition 3 Characteristics of all nocturnal land air masses are shown in Fig. 2 (averaged 45,549 individual profiles representing 10.5% of the total period—Table 2) through 12 averaged profiles (red lines with color dots) and their dispersions (green area). The color dots have indicated the availability of the individual profiles involved in the averaging at given altitudes. The lowest profiles availability was observed in 4 parameters (sigWS, sigU, sigV and TI) due to the fact that they are calculated from the sodar software as a second statistical moment of the corresponding sodar parameters, requiring higher quality control of the individual measurements compared to the first statistical moments, such as the wind components. An approximate linear increase in the wind

50

D. Barantiev and E. Batchvarova

Fig. 2 Averaged nocturnal land air masses characteristics with condition 3. Mean profiles and their dispersions from left to right and from top to bottom: WD, WS, sigWS, W, sigW, U, sigU, V, sigV, EDR, TI, TKE

speed profile is observed up to a height of 540 m above the surface (1318 individual profiles available at that height) resulting in a more than quadruple value compared to the lowest point of the averaged profile (30 m above the surface). Positive values after 340 m are seen in the vertical wind speed profile (12,227 individual profiles) and increase faster above 500 m. Slightly expressed peaks in the shape of the sigW, EDR and TKE profiles are observed at 430 m with respectively 3625, 1756 and 3011 individual profiles involved in the averaged outputs at this altitude. The surface layer (SL) height is defined between 50 and 100 m with sigW, EDR and TKE characteristics change. Nocturnal land air masses and condition 2 during the warm part of the year Averaged nocturnal land air masses during the period from May to September are presented in Fig. 3. The individual profiles used for the calculations of the mean values are close to 30% of all registered nocturnal land air masses with the same performed condition 2 (Table 2). Well expressed changes in the shape of averaged profiles after 400 m are observed at almost all graphs. The peaks in the sigW, EDR and TKE mean profiles are at the same height of 410 m above the ground (98 individual profiles involved at this height). Almost a linear decrease in the averaged TI profile is observed after the main peak at 140 m (2412 individual profiles) and it is interrupted by a second smaller peak at 430 m (59 individual profiles). This type of profiles,

Coastal Boundary-Layer Characteristic During …

51

Fig. 3 Averaged nocturnal land air masses characteristics during warm part of the year with imposed condition 2. Details as in Fig. 2

characterizing nocturnal land air masses during the warmer part of the year, is the reason for the less pronounced peak observed in the mean sigW, EDR and TKE profiles in Fig. 2 at an altitude of 430 m. Close to the ground at a height between 50 and 80 m the SL height is defined with changes in sigW, EDR, TI and TKE profiles.

4.2 Nocturnal Marine Air Masses Nocturnal marine air masses during the cold part of the year and condition 3 The mean values of up to 3937 individual profiles are used for the averaged profiles and their dispersions output at the 12 graphs in Fig. 4. The characteristics of nocturnal marine air masses meeting conditions 3 from November to March are presented. Noticeable peaks are seen at the averaged sigW and TKE profiles (respectively 406 and 289 individual profiles) with a change in the character of the profiles at a height of 300 m above the ground. Such a pronounced peak is also observed with the average EDR profile, but at height of 340 m (103 individual profiles). The nocturnal marine air masses during the cold part of the year are defined by the winter profiles and partly by these in the transition seasons at which the synoptic conditions with eastern and northeastern wind components were dominant. Therefore, in this case, we associate

52

D. Barantiev and E. Batchvarova

Fig. 4 Averaged nocturnal marine air masses characteristics during cold part of the year with performed condition 3. Details as in Fig. 2

the observed sigW and TKE peaks at a height of 300 m, with a slightly unstable or neutral marine PBL height, due to the relatively warm sea surface during the cold part of the year. A weakly pronounced positive peak at 40–50 m from the ground at TI profile, supported by the almost constant values of sigW up to 50 m and EDR up to 40 m, as well as the observed weak peaks of WD, WS and W at 40 m can be associated with height of the IBL formed by a dominant factor the surface roughness change. Using the dispersion of the vertical wind speed sigW (σw ) measured from the sodar at different altitudes (z), the Buoyancy Production (BP) profiles can be derived from (1): β=

σw 3 z

(1)

Averaged BP profiles characterizing the nocturnal marine air masses during the cold part of the year with performed condition 1* and 3 (left and right graphic respectively) are derived from Eq. (1) and are presented in Fig. 5, confirming the comments made so far about observed peaks and altitudes in the averaged profiles and their dispersions (Fig. 4). The flow of Archimedes force (buoyancy flow) or the assessment of turbulence generation due to convection is expected and has high values near the ground and decreases with height to reach the value distinctive for

Coastal Boundary-Layer Characteristic During …

53

Fig. 5 Averaged buoyancy production profiles from nocturnal marine air masses with performed conditions 1* (a) and condition 3 (b) from November to March

the higher layer in the atmosphere. High BP values can also be expected in the interaction or entrainment zone over a convective boundary layer where transport of warmer air masses from the stable layer aloft takes place. The averaged nocturnal BP profile that is calculated from 159 individual profiles at which the condition 1* was fulfilled (Fig. 5a) decreases its values to 40 m, and thus with Condition 3 (Fig. 5b) to 50 m. These negative peaks are an indicator of convective or neutral IBL and non-disturbed marine air masses over it during the night. Under condition 3 (Fig. 5b) there is a pronounced peak at 300 m above ground, thus confirming that at night in the cold season the height of the marine PBL (slightly unstable or neutral) is about 300 m.

4.3 Thermodynamic State of Nocturnal Air Masses The seasonal diagrams of stability classes probabilities at different heights are shown in Fig. 6. The atmospheric stability classes according to the Pasquill–Gifford classification using the σφ method [20] are defined by two main types of nocturnal individual profiles with applied condition 2. The sum of all probabilities at a given altitude is considered as 100% based on all available time series with disposable data for the respective altitude. According to the classification used, the probabilities range mainly from 38 to 78% with mean value of 59% for the slightly stable stratification (E) of the nocturnal land air masses during the warm part of the year (Fig. 6a). Neutral stratification (D) can be indicated as the second dominant thermodynamical state of the atmosphere for this type of air masses with probability distribution values in the rage mainly from 17 to 29% and maximum of about 62% at 570 m and 22% mean value. The dominant class of atmospheric stratification for nocturnal land air masses during

54

D. Barantiev and E. Batchvarova

Fig. 6 Diagrams of stability classes probabilities at different heights (warm season—a, b; cold season—c, d) of nocturnal land (a, c) and marine (b, d) air masses with performed condition 2

the cold part of the year (Fig. 6c) can be indicated as the slightly stable (E) with more than 50% probability after 70 m and more than 70% after 200 m AGL with 63% mean value throughout the acoustic sounding layer. In this type of air masses, the results of stability classes probabilities with height have determined the neutral stratification again as the second dominant class with an average value of 20%. The stable stratification (F) is presented in both types of nocturnal land air masses as the third dominant class of atmospheric stratification with mean probability about 16% for the warm period and 11% for the cold one. The mean probability values of extremely unstable (A) classes have been respectively about 2 and 3% while for moderately unstable classes (B) have been below 0.7 and 0.5%.

Coastal Boundary-Layer Characteristic During …

55

The changes in the stratification of nocturnal marine air masses during the warm and cold part of the year are shown on the right side of Fig. 6. Compared to the graphs of land profiles, the marine air masses have revealed more unstable thermodynamical state of the atmosphere. Neutrally stratified atmosphere (D) is observed as dominant stratification in nocturnal marine air masses with minimum probability values of 33% and mean 49% for the warm part of the year and minimum 46% and mean 54% for the cold pat of the year. The largest instability and the variations of thermodynamical state of all cases considered in Fig. 6 are observed in the averaged profiles of nocturnal marine air masses during the cold season (Fig. 6d).

5 Conclusions The presented results for all nocturnal land air masses with applied condition 3 show a super position of all seasonal profiles with imposed minimum restriction in their selection. Reducing the number of selected profiles only during the warm part of the year with applied condition 2, the achieved results are confirmed. The analysis is based on 82% of the data on nocturnal land air masses. The nocturnal air masses from the land are characterized by prevailing conditions between neutral and slightly stable stratification. Indications of a nocturnal stable PBL height at 410–430 m and corresponding SL height of 50–80 m are found. Prevailing conditions of neutral and slightly unstable stratification are revealed in the results for nocturnal marine air masses. Indications of nocturnal IBL height of about 40–50 m and a nocturnal marine PBL height of about 300 m are found. This study is based on reliable long term acoustic remote sensing data and it is of great importance for diagnostic or prognostic air pollution modelling and could support related adequate decision-making actions of governments and business for better air quality, citizens health protection and tourism development in a coastal zone. Such analyses allow not only climatological studies for a number of PBL parameters, but also to assess the fraction of time when theoretical profiles can be used in coastal areas. The performance of weather and climate models in coastal areas can be improved if assessed with such types of long-term wind and turbulence vertical profiles data. Acknowledgements This work was partially supported by the Bulgarian Ministry of Education and Science under the National Research Programme “Young scientists and postdoctoral students” approved by DCM #577/17.08.2018 and by the National Science Fund of Bulgaria, Contract KP06-N34/1 “Natural and anthropogenic factors of climate change—analyses of global and local periodical components and long-term forecasts”.

56

D. Barantiev and E. Batchvarova

References 1. Cimini, D., Marzano, F.S., Visconti, G.: Integrated Ground-Based Observing Systems. Springer-Verlag, Berlin Heidelberg (2011) 2. Coulter, R.L., Kallistratova, M.A.: Two decades of progress in SODAR techniques: a review of 11 ISARS proceedings. Meteorol. Atmos. Phys. 85, 3–19 (2004). https://doi.org/10.1007/ s00703-003-0030-2 3. Engelbart, D., Monna, W., Nash, J., Mätzler, C.: Integrated ground-based remote-sensing stations for atmospheric profiling: COST action 720: final report. In: Engelbart, D., Monna, W., Nash, J., Mätzler, C. (eds.) p. 398. Publications Office of the European Union—COST Office, Luxembourg (2009) 4. Emeis, S.: Surface-based remote sensing of the atmospheric boundary layer, vol. 40, 1st edn. Atmospheric and Oceanographic Sciences Library. Springer Netherlands (2011) 5. Illingworth, A., Ruffieux, D., Cimini, D., Lohnert, U., Haeffelin, M., Lehmann, V.: COST Action ES0702 Final Report: European Ground-Based Observations of Essential Variables for Climate and Operational Meteorology. In: COST Action ES0702 EG-CLIMET, p. 141. COST Office, PUB1062 (2013) 6. Peña, A., Floors, R.R., Sathe, A., Gryning, S.-E., Wagner, R., Courtney, M., Larsén, X.G., Hahmann, A.N., Hasager, C.B.: Ten years of boundary-layer and wind-power meteorology at Høvsøre, Denmark. Boundary-Layer Meteorol. 158(1), 1–26 (2016). https://doi.org/10.1007/ s10546-015-0079-8 7. Batchvarova, E.: Theoretical and Experimental Studies of the Atmospheric Boundary Layerover Different Surface Types. National Institute of Meteorology and Hydrology (NIMH) and the Bulgarian Academy of Sciences (BAS) (2006) 8. Hsu, S.A.: A note on estimating the height of the convective internal boundary layer near shore. Bound.-Layer Meteorol. 35(4), 311–316 (1986). https://doi.org/10.1007/BF00118561 9. Simpson, J.E.: Sea Breeze and Local Winds. Cambridge University Press, Cambridge, UK (1994) 10. Wilczak, J.M., Dabberdt, W.F., Kropfli, R.A.: Observations and numerical model simulations of the atmospheric boundary layer in the Santa Barbara coastal region. J. Appl. Meteorol. 30(5), 652–673 (1991). https://doi.org/10.1175/1520-0450(1991)030%3c0652:OANMSO% 3e2.0.CO;2 11. Batchvarova, E., Cai, X., Gryning, S.-E., Steyn, D.: Modelling internal boundary-layer development in a region with a complex coastline. Bound.-Layer Meteorol. 90(1), 1–20 (1999). https://doi.org/10.1023/A:1001751219627 12. Batchvarova, E., Gryning, S.-E.: Wind climatology, atmospheric turbulence and internal boundary-layer development in Athens during the MEDCAPHOT-TRACE experiment. Atmos. Environ. 32(12), 2055–2069 (1998). https://doi.org/10.1016/S1352-2310(97)00422-6 13. De Leo, L., Federico, S., Sempreviva, A.M., Pasqualoni, L., Avolio, E., Bellecci, C.: Study of the development of the sea breeze and its micro-scale structure at a coastal site using a multitone sodar system. In: 14th International Symposium for the Advancement of Boundary Layer Remote Sensing, 23–25 June 2008, Copenhagen, Denmark 2008. IOP Conference Series: Earth and Environmental Science, p. 9. 2008 IOP Publishing Ltd. 14. Prabha, T.V., Venkatesan, R., Mursch-Radlgruber, E., Rengarajan, G., Jayanthi, N.: Thermal internal boundary layer characteristics at a tropical coastal site as observed by a mini-SODAR under varying synoptic conditions. J. Earth Syst. Sci. 111(1), 63–77 (2002). https://doi.org/10. 1007/BF02702223 15. Petenko, I., Argentini, S., Casasanta, G., Kallistratova, M., Sozzi, R., Viola, A.: Wavelike structures in the turbulent layer during the morning development of convection at Dome C, Antarctica. Bound.-Layer Meteorol. 1–19 (2016). https://doi.org/10.1007/s10546-016-0173-6 16. Sabev, L., Stanev, S.: Climate Regions of Bulgaria and Their Climate, vol. V. State Publishing House “Science and Art”, Sofia, Bulgaria (1959)

Coastal Boundary-Layer Characteristic During …

57

17. Barantiev, D., Batchvarova, E., Novitsky, M.: Breeze circulation classification in the coastal zone of the town of Ahtopol based on data from ground based acoustic sounding and ultrasonic anemometer. Bulgarian J. Meteorol. Hydrol. (BJMH) 22(5) (2017) 18. ScintecAG: Scintec Flat Array Sodars—Hardware Manual (SFAS, MFAS, XFAS) including RASS RAE1 and windRASS, Version 1.03 ed. Scintec AG, Germany (2011) 19. Illingworth, A.J., Cimini, D., Gaffard, C., Haeffelin, M., Lehmann, V., Löhnert, U., O’Connor, E.J., Ruffieux, D.: Exploiting existing ground-based remote sensing networks to improve highresolution weather forecasts. Bull. Am. Meteor. Soc. 96(12), 2107–2125 (2015). https://doi. org/10.1175/BAMS-D-13-00283.1 20. Bailey, D.T.: Meteorological monitoring guidance for regulatory modeling applications. In: Standards, O.O.A.Q.P.A. (ed.) p. 171. United States Environmental Protection Agency (EPA), Research Triangle Park, NC 27711 (2000)

Climatological Study of Extreme Wind Events in a Coastal Area Damyan Barantiev , Ekaterina Batchvarova , Hristina Kirova, and Orlin Gueorguiev

Abstract Long-term sodar measurements (Aug 2008–Oct 2016) of wind and turbulence profiles with high spatial (10 m) and temporal (10 min) resolution were performed at the southern Bulgarian Black Sea coast. This data has provided an opportunity to define “rare” values of meteorological parameters within their statistical distributions and to identify them as extreme events according to the Intergovernmental Panel on Climate Change. The statistical analysis of wind speed profiles has been performed for the eight-year period using the two parameter Weibull distribution. The determination of the ninety-percentile of this statistical distribution (at every sodar measurement level from 30 up to 600 m) has given values (“rare” events) that have defined the theoretical extreme wind speed profile (reference profile). On this basis, the extreme profiles during the reviewed period have been determined. Analysis of the distribution of the situations with extreme weather events by months and hours for the entire period has been performed. The multiple time series with the registered extreme profiles have been used to derive averaged parameters defining the vertical structure of the coastal boundary layer during extreme events. The thermodynamic state of the coastal boundary layer according to the Pasquill-Gifford classification has been revealed. Keywords Sodar · Wind profiles · Turbulent characteristics · Extreme events · Black sea · Coastal PBL · Weibull distribution · Climatological study

1 Introduction The society requires increased accuracy in time and space of forecasts; of observation data; of early warnings for dangerous and extreme weather events; and of climate D. Barantiev (B) · E. Batchvarova Climate, Atmosphere and Water Research Institute, Bulgarian Academy of Sciences (CAWRI—BAS), 66, Tsarigradsko Shose blvd, Sofia 1784, Bulgaria e-mail: [email protected] H. Kirova · O. Gueorguiev National Institute of Meteorology and Hydrology (NIMH), 66, Tsarigradsko Shose blvd, Sofia 1784, Bulgaria © The Author(s), under exclusive license to Springer Nature Switzerland AG 2021 N. Dobrinkova and G. Gadzhev (eds.), Environmental Protection and Disaster Risks, Studies in Systems, Decision and Control 361, https://doi.org/10.1007/978-3-030-70190-1_5

59

60

D. Barantiev et al.

models’ predictions in order to plan the future of the Planet. During the last years, many governments invest in the development of effective systems for observations and forecasting of hazardous meteorological events for prevention of socio-economic lost through adequate management and reduction of risks. The extreme phenomena and the specific thresholds for the extreme values of a corresponding climatic variable (e.g. wind speed), by definition, vary from place to place for natural reasons (different climates), because the extreme value of given meteorological parameter at one place, can be within the normal range at another place. A reason to apply different values is their application to socio-economic activities and needs, which also differ with climate. Despite their rarity, the extreme weather events are of a dangerous nature and could be harmful to human health, infrastructure, economy and even cause loss of human life [1, 2]. The advancement in ground-based remote sensing (GBRS) instruments proved them to be a tool to achieve more accurate spatial, qualitative and quantitative assessments of the processes within the planetary boundary layer (PBL). Information about an object or phenomenon is acquired without making physical contact and allows data to be collected for hazardous or inaccessible areas by remote sensing [3]. The capabilities of GBRS instruments to detect extreme wind events have been explored within the frame of SafeWind project at the Danish National Test Center for Large Wind Turbines at Høvsøre, Denmark [4]. Measurements in flat coastal terrain from two different types of lidars (continuous wave and a pulsed lidar) and a reference 116.5 m tall meteorological mast with cup anemometers were used for analysis of cup-lidar data comparisons in the experimental campaign. Wind data taken at 40, 60, 80 and 100 m above the ground showed that both lidars are capable for the maximum wind speed (WS) value determination within a 10-min averaging period up to an underestimation of about 10% with respect to the cup anemometers. Moreover, the probability density function (PDF) and the cumulative distribution function (CDF) of the time difference of the maximum WS between different instruments have been studied and comparisons of the gust factor have been discussed too. The experimental data showed better results for the pulsed lidar measurements with the same maximum at about 50% of the time and a comparable gust factor to that of the cup anemometers. In Gottschall et al. [5], measurements results from a 100 m tall reference meteorological mast and two pulsed Doppler wind lidar units (one mounted on a floating table simulating motions of different possible offshore platforms and the other used as fixed reference instrument) were studied. In addition, based on the undisturbed wind data motion compensation algorithms were developed allowing correction of the affected measurement. Thus, the possibilities of the wind lidar mounted on a floating platform to determine properly turbulence as well extreme wind events were discussed. The results from the experimental campaign at the test site Tauche near Berlin showed that the 10-min mean values of horizontal WS agree quite well between both lidars but the turbulence intensity, as well as the recorded extreme wind values and the gust factors within a 10-min period values of the tilted lidar, was increased with respect to the fixed reference one. The developed compensation algorithm allowed to assess and control the impact of the motions of the floating platform.

Climatological Study of Extreme Wind Events in a Coastal Area

61

A month of remote sensing data (April 2001) with high spatial and temporal resolution from Doppler sodar (DSDPA.90-24 METEK make) was used to describe the WS values in an extensive plateau (horizontal homogeneity area) in the North of Spain at the Low Atmosphere Research Centre (CIBA) [6]. This work included analysis of hourly means, daily WS evolution, WS power law, WS distribution by 10-min means of four WS classes, wind roses characteristics at 100 m for light and moderate winds, and four methods of fitting the two-parameter Weibull distribution function— linear regression by cumulative frequency, moments, maximum likelihood and quartiles. The results showed a sharp contrast between day and night cycle of WS due to strong convection during the day and the stratification stability during the night. Wind distribution revealed two prevailing directions at the frequent moderate winds considering synoptic forces affecting the Iberian Peninsula at the time of the measurements campaign. The close values of calculated Weibull parameters by the four different methods approved their usefulness for wind data analysis and practical purposes. A Radio Acoustic Sounding System (RASS) extension of the sodar provided temperature profiles. All sodar and RASS data were compared with meteorological mast measurements (10 min averages of temperature, WS, and wind direction at 100 m height and only temperature at 51 m) in Pérez et al. [7]. The results showed good comparability of sodar WS and RASS temperature measurements with those from the mast with satisfactory linear regression. For the wind direction measurements, statistical treatment of circular data was used and satisfactory correlation calculated by means of a nonparametric statistical test was obtained. Such type of equipment is reliable and irreplaceable for a number of studies and innovative scientific methods of research on the mean meteorological parameters and turbulence in the PBL [8–14]. Over the past decade an integration of a number of GBRS devices within a uniform European network for observations was initialized through research and collaboration within several COST Actions: COST Action (EG-CLIMET) ES0702, COST Action 720, COST Action (TOPROF) ES1303 and COST Action (PROBE) CA18235 [12, 15–18]. Through the development of the necessary calibration standards for GBRS instruments, the establishment of procedures for maintenance and automatic control of data quality, common data formats, and protocols for exchange of data, these COST actions aim at standardization of the use of GBRS data for better weather forecast and thus to increase of the quality of all meteorological products in service of society. Unfortunately, the modern GBRS methods are not developed in operational mode in Bulgaria, though such data are needed for fundamental theoretical research, climatological studies, evaluation of meteorological models, economic and air quality activities.

62

D. Barantiev et al.

2 Study Area, Experiment Equipment, and Data Overview 2.1 Measuring Side Data collection and analysis in the coastal PBL using acoustic sounding starts in Bulgaria from the summer of 2008 at the Meteorological Observatory (MO) Ahtopol, located on the coast of south-eastern Bulgaria (Fig. 1—yellow pin). These high special and temporal resolution measurements allow to start climatological studies of the coastal PBL in Bulgaria [19–22]. The climatic zoning of Bulgaria presented in Sabev, Stanev [23] attributed the studied area into the Black Sea coastal Strandzha climate region, which is under the influence of the Black Sea climatic sub region of the Continental-Mediterranean climate zone in Bulgaria. Typically, in this climate region, well expressed breeze circulation in the warm part of the year is observed, whereas during the cold part of the year a lower frequency and smaller time and spatial scales of the coast circulation is registered [24]. MO Ahtopol is located on a flat grassy terrain at 30 m height above sea level and at about 400 meters inland. The shore near the observatory is steep, about 10 meters high cliffs and the coastline stretches from north-northwest to south-southeast (Fig. 1).

Fig. 1 Location of MO Ahtopol in Bulgaria on Google Earth (42° 5 3.37 N, 27° 57 4.49 E) with views of the terrain (right) and sodar system on the roof of the administrative building of MO Ahtopol (left)

Climatological Study of Extreme Wind Events in a Coastal Area

63

2.2 Measuring Instrument Used To study the wind and turbulence profiles in coastal PBL, a multibeam acoustic monostatic Doppler system for sounding the atmosphere is used—SCINTEC Flat Array Sodar MFAS with frequency range 1650–2750 Hz, 9 angles of emission/reception (0°, ±9.3°, ±15.6°, ±22.1°, ±29°), vertical range from 150 to 1000 m, with 10 m resolution, first level of measurement 30 m. The accuracy for WS is 0.1–0.3 ms−1 , and for the wind direction is 2°–3° [25]. The sodar system is installed on the roof of the administrative building of the MO Ahtopol (Fig. 1) at an approximate height of 4.5 m. During the study period its settings undergo a number of changes, mainly related to the update of the operation software. Thanks to these changes, more reliable operation is achieved during atmosphere sounding and its vertical range is increased while maintaining the data resolution. The basic settings do not undergo significant changes and the quality control of the data is maintained, as throughout the period the sodar records every 10 min with averaging period of 20 min (running averages) with a vertical resolution of 10 m. Due to the operating software updates over the years, the vertical range of the sodar in 2008 reached 520 m in height, while in 2016 it was 1000 m. In summary, the effective sodar range (the height to which wind profiles are measured) depend mainly on the turbulent inhomogeneities in the atmosphere, on the spatial and temporal resolution setup and operational software updates.

2.3 Data Availability In this work we explore 3014-day period (1 August 2008–31 October 2016) of acoustic soundings in the coastal PBL with 341,971 profiles corresponding to 78.8% time coverage. The monthly data availability and maximum effective height reached by the sodar are presented in Table 1. Data with availability below 40% is given in red, between 40 and 70% in yellow, and above 70% in green. The maximum vertical range for this study is set to 600 m due to the lower data availability above this height. The actual sodar range is also presented graphically in Table 1 with filled bars at a height—over 810 m (4 bars), over 710 m (3 bars), over 610 m (2 bars). The continuity of the operation of the sodar during the study period was disturbed by frequent accidents of the main power supply on the territory of the MO Ahtopol until 2011 (Table 1—yellow and red colors). During the second half of 2008 and 2009, the sodar was stopped during the night hours. From 1st August to 31st October 2008 the measurements were made in the daytime, from 7:00 a.m. to 6:00 p.m. From 8th December 2008 to 27th July 2009 the sodar worked continuously and after that until December 2009 only during the day, from 7:30 a.m. to 9:40 p.m. After December 2009 the sodar functioned with no restrictions. The operational data records from the sodar until 26 September 2014 were made in local time (i.e. during the cold half of the year/UTC + 2/, and during the warm half/UTC + 3/). After this period, the

64

D. Barantiev et al.

Table 1 Data availability from the acoustic sounding of the atmosphere and maximum effective height reached during the study period on the territory of the MO Ahtopol I

II

III

IV

V

VI

VII

2008 max range [m]

-

-

-

-

-

-

-

2009 max range [m]

99.9% 520

99.6% 520

99.2% 680

96.7% 680

98.3% 680

99.3% 680

2010 max range [m]

97.5% 680

98.3% 680

89.0% 680

68.5% 680

96.8% 680

2011 max range [m]

96.8% 510

94.8% 460

99.8% 510

96.7% 510

96.8% 510

2012 max range [m]

75.3% 620

95.8% 620

99.7% 620

96.7% 620

2013 max range [m]

96.6% 720

96.4% 720

50.6% 720

100.0% 620

2014 max range [m]

100.0% 720

99.9% 720

99.7% 720

2015 max range [m]

99.9% 750

96.3% 750

94.9% 750

2016 max range [m]

80.1% 750

70.0% 750

76.1% 750

VIII

IX

X

XI

XII

45.2% 520

40.3% 520

57.7% 520

59.5% 520

88.4% 520

94.6% 680

57.5% 680

59.9% 680

58.4% 680

96.7% 680

96.0% 680

86.4% 680

99.9% 560

98.3% 510

99.9% 510

92.6% 510

99.6% 510

99.8% 510

96.2% 510

92.2% 510

99.9% 510

38.8% 560

78.8% 620

99.2% 620

81.3% 620

99.3% 620

93.3% 620

99.1% 700

31.9% 620

96.6% 720

30.6% 670

23.3% 640

100.0% 720

98.3% 680

54.7% 670

95.3% 680

96.3% 680

96.7% 720

58.0% 720

74.7% 720

61.7% 720

100.0% 720

97.1% 720

99.8% 720

90.0% 720

59.4% 720

51.7% 750

98.5% 750

97.5% 750

98.0% 730

97.8% 750

69.5% 750

73.3% 750

57.4% 590

97.9% 750

99.5% 750

99.9% 750

65.9% 750

95.1% 730

49.9% 750

35.6% 750

93.2% 1000

99.9% 1000

93.0% 1000

75.0% 1000

45.4% 750

-

-

sodar data were recorded only in winter time, i.e. UTC + 2. Most of the analyses performed in this work are independent of the time, but for the needs of the extreme weather events distribution analysis by month and hour for the entire period, the entire time series is converted to UTC + 2.

3 Methodology The analysis for extreme winds is related to the derivation of a theoretical extreme profile (reference profile) of the WS on the basis of which the extreme speed profiles are determined. For this purpose, an analysis of the WS distribution in height was performed by processing the values of WS for each individual profile both in height (every 10 m) and in WS intervals from 0 to 40 ms−1 through 1 ms−1 as follows: • • • • •

calm—at values of WS > or = 0 ms−1 , but < 0.5 ms−1 interval 1 ms−1 —at values of WS > or = 0.5 ms−1 , but < 1.5 ms−1 interval 2 ms−1 —at values of WS > or = 1.5 ms−1 , but < 2.5 ms−1 … interval 40 ms−1 —at values of WS > or = 39.5 ms−1 ;

The WS probability distribution for all WS profiles measured at MO Ahtopol is obtained by histograms at each level with applied PDF also known as the two parameter Weibull distribution [26] described in Eq. (1) based of the method of maximal likelihood [27]. The two distribution parameters are given in Eqs.: (2) and (3) [28], respectively, for a shape parameter k,which has a non-dimensional value defining the shape of the probability density distribution curve and the scale parameter c assuming the dimension of the variable and representing the 63.2th percentile of the distribution [29]. 



Climatological Study of Extreme Wind Events in a Coastal Area

e−(u/c) , u ≥ 0 0, u < 0

(1)

 N u ki ln u i − u kN ln u N 1  ln u i −  N  k k N i=1 i=1 u i − u N

(2)

f (u; c, k) = f (x) = 

k

−1

   k u k−1

65

c

c

k

N  =

i=1

N  1  k c = u − u kN N i=1 i



k

(3)

The two-parameter Weibull distribution is one of the most commonly used and preferred distributions in wind potential estimates and has been studied in a number of scientific papers [30–34]. In this study we use it to employ the definition of the Intergovernmental Panel on Climate Change [1, 35] for extreme events (exceedance over a relatively low threshold) through determination of the 90th percentile of the performed statistical distribution for all heights. Thus, the “reference” values, at/above which the WS is considered as extreme, are defined. In this way, “reference” wind profile is derived and extreme wind events are determined by comparing it with the actual profiles during the study period. Only profiles with at least ten points with values equal to or greater than those of the reference profile are used. Finally, an “extreme wind “data set based on sodar data is created containing 10,854 extreme wind profiles representing about 3.2% of the soundings performed.

4 Results WS histograms with applied Weibull distributions, certain 90th percentile and statistical data for different measurement levels for the study period in Ahtopol are shown in Fig. 2. The pink bar, on the side of each histogram, is indicator for data availability at each of the displayed levels. The number of measurements involved in the derivation of the statistical graphs in Fig. 2 decreases sharply in height due to the lower number of profiles reaching the corresponding level of measurement. At 50 m (Fig. 2—top left), a 99.2% data availability is achieved, while at 350 m (Fig. 2—bottom left) this value is already 35%, and at 550 m (Fig. 2—bottom right) only 4.4%, which is equal to just over 15,000 profiles reaching this height. A characteristic change with height of the WS histograms is observed, which is expressed by the typical shift of the maximum to higher WS values. The highest probability (nearly 27%) is observed at winds falling in the interval defined as 1 ms−1 at 50 m above the ground level (AGL), while at 550 m there are two pronounced maxima determined by the intervals 4 and 11 ms−1 (about 8% of cases for each interval). Due to the larger variability of WS with height, a “widening” of the graphs in the Weibull distributions applied to the respective histograms (Fig. 2—red curves) and in the histograms themselves is

66

D. Barantiev et al.

Fig. 2 Histograms of WS (green bars), two-parameter Weibull distributions (red curves), certain values equal to or exceeding the 90th percentile of the probability density of Weibull distributions (blue squares), availability of data (pink bar) and derived statistical data. Top left—level 50 m, top right—200 m, bottom left—350 m, bottom right—550 m

observed (Fig. 2—green bars). The extent to which the Weibull distribution graph is “shrunk” is determined by the shape parameter k (2)—the larger value corresponds to more “wider” distribution graph. For each Weibull distribution, actual measured values equal to or greater than its 90th percentile (blue squares on the red curve) are presented, which according to the IPCC definition are defined as extreme values of WS [1, 35]. In addition, to each of the graphs in Fig. 2 statistical values for the respective heights are derived, such as minimum, maximum and average values of WS, extreme value determined by the 90th percentile, as well as values of the two parameters of the respective Weibull distribution. By summarizing the information presented in Fig. 2 from 30 to 600 m AGL with 10 m resolution the changes in the probability distributions of WS in height with the attached “reference” profile of extreme values of WS (minimum extreme values— determined by the 90th percentile of statistical distributions in height) for the whole considered period are obtained and presented at Fig. 3. The color bar shows the probability distribution values changes in height, as its color range is limited to 10% in order to achieve better visualization of the results (values of probability distributions above 10% are colored as 10%, but values under 0.2% are still not viewable). The area with maximum values of the probability distribution of WS in height is clearly observed at the presented graph, as up to 170 m the probability is highest for WS up to 3 ms−1 . At 300 m, high probability is observed in 

Climatological Study of Extreme Wind Events in a Coastal Area

67

Fig. 3 Diagram of WS probability distributions at different heights (color section of altitude and speed intervals) with available data (green bars) and applied reference profile for extreme values of WS (black profile), determined by the 90th percentile of the statistical distributions in height

winds of 3 to 6 ms−1 . At 600 m the probability is highest for winds speeds 4–5 ms−1 and 10–12 ms−1 . From the green bars showing the change in the number of profiles reaching a certain height, it can be seen that their availability begins to decrease sharply after 180 m AGL, with a maximum of data observed in the layer from 90 to 140 m. Almost 5% lower data availability is observed at the first level of the sodar measurements (30 m AGL) than the layer at which the maximum data is. The values of the reference extreme WS profile (REWSP) (Fig. 3—black profile) increase with height, as the extreme values at 30 m AGL are about 4 ms−1 , and at 600 m—about 20 ms−1 . Close to the ground, where the availability of data is significant, two small peaks are observed in the reference profile (at heights of 50 and 80 m), differing against the background of gradually increasing extreme values in height. These relatively sharp changes in the reference values determined by the 90th percentile of the respective statistical distributions can be associated with corresponding changes of the probability density curve shape of the Weibull distribution graphs at these heights, expressed respectively by relevant changes in the shape parameter. REWSP is located primarily in the area with a probability distribution of WS between 1 and 4% up to 200 m and between 1 and 2% higher up. This may be associated with the typically larger and more varied presence of higher WS in height, leading to lower values and shifting the maximum of the probability distribution to higher speed ranges. At least 50% probability is available in total for the first two speed intervals (calm and 1 ms−1 ) up to 40 m AGL due to which relatively low values of REWSP close to the surface are observed.

68

D. Barantiev et al.

Fig. 4 Number of extreme wind speed profiles reported during the different years of the study at MO Ahtopol by month (left) and hour of the day (right)

The number of extreme wind speed profiles during the different years of the study period by month and hour of the day is shown in Fig. 4. The number of extreme wind speed profile (EWSP) for the period is 10,854 or 3.2% of all profiles included in the statistics. It can be noted that 2012 (Fig. 4—red) was the windiest year (excluding 2008 and 2016 for which the data do not cover a calendar year). In 2010 only 13 EWSP were registered in December around 8 a.m. The intensity of the atmospheric circulation dynamics, over the studied area, is evident from the variability of the number of profiles during the different months and hours of the day. EWSP are observed in all months and hours, and the maxima can be localized in the cold half of the year and during the day when the processes in the atmosphere are more intense. The biggest number of extreme profiles (over 1500) is observed in January, and the calmest months are June and July, when the number of extreme profiles is just over 200. Calm and windy periods at Ahtopol can be also identified, namely July 2009–August 2011 was a calm period, while September 2011–June 2014 and April 2015–October 2016 were windy periods. Calmer periods are revealed at 6-7 a.m. and 5-6 p.m., which are related to the start and end of the day during cold seasons (when a bigger number of EWSP are reported) and calm periods in the beginning and end of sea breeze during warm seasons. Most windy are the periods 10 p.m.–1 a.m. and 11 a.m.–1 p.m., which can be related to the maximal development of the local circulation in both directions. The vertical structure of the coastal PBL in the study area during extreme winds phenomena is presented in Fig. 5 by averaged profiles and their dispersions of 12 output parameters from the sodar measurements—from left to right and from top to bottom: wind direction (WD), extreme wind speed profile (WS), extreme speed profile dispersion (sigWS), vertical wind speed (W), vertical wind speed dispersion (sigW ), horizontal (western) component of the extreme profile (U), dispersion of the western component (sigU), horizontal (southern) component of the extreme profile (V ), dispersion of southern component (sigV ), eddy dissipation rate (EDR), turbulent

Climatological Study of Extreme Wind Events in a Coastal Area

69

Fig. 5 Averaged profiles and their dispersions from a sample with a maximum of 10 854 selected extreme wind speed profiles

intensity (TI) and turbulent kinetic energy (TKE). Different number of values are used for the derivation of the average values at different measurements levels. The color bar (on the right side of the graphs) shows the number of profiles involved in calculations of the averaged values and their dispersions at a certain level, as it is on a logarithmic scale for better visualization of the results due to the lower availability of profiles in height. The WD profile graph (Fig. 5) shows north-northwest direction in the registered extreme winds up to 130 m (where most of the profiles are concentrated), which is related to wind blowing parallel to the coastline. Gradual change to west direction is observed between 130 and 200 m. Higher up, the WD is southwesterly. The greatest dispersion of this profile is observed in the first 130 m, after which it decreases gradually. The averaged extreme wind speed profile (EWSP) is characterized by a relatively constant dispersion in height, and an almost linear increase in values. Velocities from 8 to 12 ms−1 up to 130 m, and their rapid increase in height reaching values close to 24 ms−1 up to 600 m are observed. In the profile of W, positive values up to 160 m altitude are observed, after which negative values are recorded decreasing to −1.3 ms−1 at 600 m. In almost all presented averaged profiles, changes in the profiles shape in the layer 40–60 m are observed. In the EDR profile a sharp decrease of the values up to 90 m is observed. Also, of interest is the sigW profile with slight peak at a height of 150–160 m (where a sharper change of WD and faster decrease of the W values are observed), followed by almost constant values up to 300 m, and an increase up to the second main peak between 440 and 490 m. At this

70

D. Barantiev et al.

Fig. 6 Averaged Buoyancy Production profile (left) and diagram of stability classes probabilities at different heights (right) during extreme wind events from Aug 2008 to Oct 2016

height sigW reaches maximum values. These peaks are also reflected in the shape of the other averaged turbulent profiles. Following Illingworth et al. [12] the PBL height at Ahtopol during extreme wind situations is between 440 and 490 m. The height with peculiarities in the profiles in Fig. 5 of 40–60, 90 and 150 m can be related to Internal Boundary Layer (IBL) and surface layer (SL) heights, or very low nocturnal PBL height in cold seasons. Two more parameters characterizing the structure of the coastal boundary layer are shown in Fig. 6. In the left graph the averaged profile of Buoyancy Production (BP) is presented considering its active role in the TKE production and momentum transfer. The involved BP profiles in the averaging are obtained using the dispersion of the vertical wind speed sigW (σ w ) profiles derived from sodar at different altitudes (z) with Eq. (4) [8]: β=

σw3 z

(4)

In the BP profile a sharp decline in the values in the first 90 m followed by almost constant values up to 130 m and two peaks (slight at 150 m and main between 440 and 490 m) coinciding in height with the peaks of sigW and TKE are observed. In general, the average Archimedes force decreases in height, with the maximum values observed near the ground (where the turbulent heat flow is maximum) and the minimum in the extreme high parts of the average profile. A diagram of stability classes probabilities at different heights is shown in Fig. 6 (right). The profiles of atmospheric stability classes according to the Pasquill-Gifford classification using the σ φ method [36] are used as an additional output from the sodar measurements to retrieve the thermodynamic state of the atmosphere during extreme wind events. Only time series with available stability classes data at respective altitude are used for the consideration of the 100% probability. The dominant class of atmospheric stratification can be indicated as the slightly stable (E) with more than 50% probability almost throughout the acoustic sounding layer and with 100% probability after 500 m. The next dominant class according to the

Climatological Study of Extreme Wind Events in a Coastal Area

71

Pasquill-Gifford classification used is neutral (D) with probability distribution values in the range between 9 and 58%. About 20% of averaged probability distribution is registered as D stability class for the whole measurement layer while for the E stability class 66% can be pointed as averaged probability distribution with minimum value of about 42% at 470 m. The stable stratification (F) is presented with maximum value of about 18% close to the surface and with averaged probability distribution of about 13% for the entire layer. The probability distributions of the other three classes of atmospheric stability are present with very low values close to the surface, which is the reason why only weakly unstable class (C) is visualized in the right graph of Fig. 6 with maximum value about 2% and averaged about 0.2%. The maximum values of moderately (B) and extremely unstable (A) classes have been below 0.2%.

5 Conclusions The derived averaged characteristics of the coastal PBL in extreme winds conditions are pioneering result based on unique set of GBRS observation data at a Bulgarian Black Sea coastal site. In this paper a super position of different types of air masses is considered which allows to assess the vertical structure and height of the PBL (determined between 440 and 490 m by the main peaks at graphs of sigW, EDR and TKE in Fig. 5 and graph of PB in Fig. 6), the IBL height of 40–60 m at high winds and SL height or very low nocturnal PBL height in cold seasons of 90 or 150 m. The dominant thermodynamic state of the coastal boundary layer according to the Pasquill-Gifford classification can be characterized as slightly stable (E) with averaged probability distribution of 66% for the whole measurement layer followed by neutral (D) stratification with 20% and stable stratification (F) with 13%. This climatic study reveals that quiet and windy periods can be of different length, which is likely related to different types of prevailing weather structures. The observed winter highs and summer lows of extreme winds are likely due to the presence of local coastal circulation. The distinct morning and evening lows in the average distribution by hour are related to transitions between night and day and start or end of sea breeze. The distribution of EWSP by year, month and hour of the day is important climatological feature, which is of great use for economic activities (such as construction of buildings, wind energy potential, air quality, climate comfort for citizens and tourists, etc.) and municipal emergency plans and actions in cases of meteorological extreme phenomena. The methodology used in this study is suitable for variety of remote sensing instruments with long-term data sets, while allowing both the evaluation of diagnostic or prognostic numerical models under extreme events conditions and integration into synergy systems for monitoring, control and robust decision making in the challenge of increasing climate change rates in recent years. Acknowledgements The work is within the frame of research projects DM 14/1 26-05-2020 (REPLICA—extReme Events and wind ProfiLe In a Coastal Area) project, funded by National

72

D. Barantiev et al.

Science Fund of Bulgaria and it was partially supported by the Bulgarian Ministry of Education and Science under the National Research Programme “Young scientists and postdoctoral students” approved by DCM # 577 /17.08.2018. The contribution of E. Batchvarova is supported by the National Science Fund of Bulgaria, Contract KP-06-N34/1 30-09-2020 “Natural and anthropogenic factors of climate change—analyses of global and local periodical components and long-term forecasts”.

References 1. IPCC: Climate Change 2001: The Scientific Basis. Contribution of Working Group I to the Third Assessment Report of the Intergovernmental Panel on Climate Change. In: Houghton, J.T., Ding, Y., Griggs, D.J., Noguer, M., van der Linden, M., P.J., Dai, X., Maskell, K., Johnson, C.A. (eds.). p. 881. Cambridge University Press, United Kingdom and New York, NY, USA (2001) 2. IPCC: Managing the Risks of Extreme Events and Disasters to Advance Climate Change Adaptation. A Special Report of Working Groups I and II of the Intergovernmental Panel on Climate Change. In: Field, C.B., Barros, V., Stocker, T.F., Dahe, Q., Dokken, D.J., Ebi, K.L., Mastrandrea, M.D., Mach, K.J., Plattner, G.-K., Allen, S.K., Tignor, M., Midgley, P.M. (eds.). Cambridge University Press, Cambridge, UK, and New York, NY, USA (2012) 3. Prasad, S.T.: Remotely Sensed Data Characterization, Classification, and Accuracies Remote Sensing Handbook, vol. 1. CRC Press, (2015) 4. Sathe, A., Courtney, M., Mann, J., Wagner, R.: How good are remote sensors at measuring extreme winds?. Paper presented at the EWEA Conference, Brussels, Belgium, 14–17 March 2011 5. Gottschall, J., Wolken-Möhlmann, G., Lange, B.: About offshore resource assessment with floating lidars with special respect to turbulence and extreme events. J. Phys. Conf. Ser. 555(1), 012043 (2014). https://doi.org/10.1088/1742-6596/555/1/012043 6. Pérez, I.A., Garc´ıa, M.A., Sánchez, M.L., de Torre, B.: Analysis of height variations of sodarderived wind speeds in Northern Spain. J. Wind Eng. Ind. Aerodyn. 92(10), 875–894 (2004). https://doi.org/10.1016/j.jweia.2004.05.002 7. Pérez, I.A., Sánchez, M.L., Garc´ıa, M.A., de Torre, B.: Comparison between measurements obtained with a meteorological mast and a RASS sodar. In: 3rd International Conferences on Experiences with Automatic Weather Stations, Torremolinos, Malaga, Spain, 19–21 Feb 2003 8. Engelbart, D., Monna, W., Nash, J., Mätzler, C.: COST 720 Final Report: Integrated groundbased remote-sensing stations for atmospheric profiling. Office for Official Publication of the European Communities, Luxembourg (2009) 9. Cimini, D., Marzano, F.S., Visconti, G.: Integrated Ground-Based Observing Systems. Springer, Berlin Heidelberg (2011) 10. Emeis, S.: Surface-based remote sensing of the atmospheric boundary layer, 1st ed. Atmospheric and Oceanographic Sciences Library. Springer Berlin Heidelberg, New York (2010) 11. Peña, A., Floors, R.R., Sathe, A., Gryning, S.-E., Wagner, R., Courtney, M., Larsén, X.G., Hahmann, A.N., Hasager, C.B.: Ten years of boundary-layer and wind-power meteorology at Høvsøre, Denmark. Bound.-Layer Meteorol. 158(1), 1–26 (2016). https://doi.org/10.1007/s10 546-015-0079-8 12. Illingworth, A., Ruffieux, D., Cimini, D., Lohnert, U., Haeffelin, M., Lehmann, V.: COST action ES0702 final report: European ground-based observations of essential variables for climate and operational meteorology. In: COST Action ES0702 EG-CLIMET, p. 141. COST Office, PUB1062 (2013) 13. Coulter, R.L., Kallistratova, M.A.: Two decades of progress in SODAR techniques: a review of 11 ISARS proceedings. Meteorol. Atmos. Phys. 85, 3–19 (2004). https://doi.org/10.1007/ s00703-003-0030-2

Climatological Study of Extreme Wind Events in a Coastal Area

73

14. Bradley, S., Antoniou, I., Hünerbein, S.v., Kindler, D., Noord, M.d., Jørgensen, H.: SODAR calibration procedure (final reporting on WP3, EU WISE project NNE5-2001-297) In: Stuart Bradley, p. 69. The University of Salford, Salford, Greater Manchester, UK, (2005) 15. Engelbart, D., Monna, W., Nash, J., Mätzler, C.: Integrated ground-based remote-sensing stations for atmospheric profiling. In: Engelbart, D., Monna, W., Nash, J., Mätzler, C. (eds.), COST action 720: Final Report, p. 398. Publications Office of the European Union - COST Office, Luxembourg (2009) 16. Illingworth, A., Ruffieux, D., Haeffelin, M., O’Connor, E., Cimini, D., Potthast, R.: COST Action Final Achievement Report ES1303: Towards operational ground based profiling with ceilometers, doppler lidars and microwave radiometers for improving weather forecasts (TOPROF). In: Illingworth, A., Ruffieux, D., Haeffelin, M., O’Connor, E., CIMINI, D., Potthast, R. (eds.). COST Association AISBL, Brussels, Belgium, (2017) 17. Illingworth, A., Cimini, D., Gaffard, C., Haeffelin, M., Lehmann, V., Löhnert, U., O’Connor, E.J., Ruffieux, D.: Exploiting existing ground-based remote sensing networks to improve highresolution weather forecasts. Bull. Am. Meteor. Soc. 96(12), 2107–2125 (2015). https://doi. org/10.1175/BAMS-D-13-00283.1 18. Cimini, D., Haeffelin, M., Kotthaus, S., Löhnert, U., Martinet, P., O’Connor, E., Walden, C., Coen, M.C., Preissler, J.: Towards the profiling of the atmospheric boundary layer at European scale—introducing the COST Action PROBE. Bull. Atmos. Sci. Technol. 1(1), 23–42 (2020). https://doi.org/10.1007/s42865-020-00003-8 19. Barantiev, D., Novitsky, M., Batchvarova, E.: Meteorological observations of the coastal boundary layer structure at the Bulgarian Black Sea coast. Adv. Sci. Res. (ASR) (6), 251–259 (2011). https://doi.org/10.5194/asr-6-251-2011 20. Batchvarova, E., Barantiev, D., Novitsky, M.: Costal boundary layer wind profile based on SODAR data—Bulgarian contribution to COST Acton ES0702. Paper presented at the The 16th International Symposium for the Advancement of Boundary-Layer Remote Sensing—ISARS Boulder, Colorado, USA, 5–8 June 2012 21. Novitsky, M., Kulizhnikova, L., Kalinicheva, O., Gaitandjiev, D., Batchvarova, E., Barantiev, D., Krasteva, K.: Characteristics of speed and wind direction in atmospheric boundary layer at southern coast of Bulgaria. Russ. Meteorol. Hydrol. 37(3), 159–164 (2012). https://doi.org/ 10.3103/S1068373912030028 22. Barantiev, D., Batchvarova, E., Novitsky, M.: Exploration of the coastal boundary layer in ahtopol through remote acoustic sounding of the atmosphere. Paper presented at the 2nd National Congress on Physical Sciences and 41st National Conference on Physics Education Matters, Sofia, Bulgaria, 25–29 Sept 2013 23. Sabev, L., Stanev, S.: Climate regions of Bulgaria and their climate, vol. V. State Publishing House “Science and Art”, Sofia, Bulgaria (1959) 24. Barantiev, D., Batchvarova, E., Novitsky, M.: Breeze circulation classification in the coastal zone of the town of Ahtopol based on data from ground based acoustic sounding and ultrasonic anemometer. Bulgarian J. Meteorol. Hydrol. (BJMH) 22(5), 24 (2017) 25. ScintecAG: Scintec Flat Array Sodars - Hardware Manual (SFAS, MFAS, XFAS) including RASS RAE1 and windRASS, Version 1.03 ed. Scintec AG, Germany (2011) 26. Papoulis, A., Pillai, S.U.: Probability, Random Variables, and Stochastic Processes, 4th edn. McGraw-Hill Europe (2002) 27. Indhumathy, D., Seshaiah, C.V., Sukkiramathi, K.: Estimation of Weibull parameters for wind speed calculation at Kanyakumari in India. J. Innov. Res. Sci. Eng. Technol. 3(1) (2014) 28. Sornette, D.: Critical Phenomena in Natural Sciences: Chaos, Fractals Selforganization and Disorder: Concepts and Tools. Springer, Berlin (2004) 29. Papanchev, T.: A modified approach for parameters estimation of the Weibull distribution for interval data and zero or few failures. J. Notices Union Sci. Varna, Technical Sciences Series (2013) 30. Gryning, S.-E., Batchvarova, E., Floors, R.R., Peña, A., Brümmer, B., Hahmann, A.N., Mikkelsen, T.: Long-term profiles of wind and Weibull distribution parameters up to 600 m in a rural coastal and an inland suburban area. Bound.-Layer Meteorol. 150(2), 167–184 (2014). https://doi.org/10.1007/s10546-013-9857-3

74

D. Barantiev et al.

31. Lun, I.Y.F., Lam, J.C.: A study of Weibull parameters using long-term wind observations. Renew. Energy 20(2), 145–153 (2000). https://doi.org/10.1016/s0960-1481(99)00103-2 32. He, Y., Monahan, A.H., Jones, C.G., Dai, A., Biner, S., Caya, D., Winger, K.: Probability distributions of land surface wind speeds over North America. J. Geophys. Res. Atmos. 115(D4) (2010). https://doi.org/10.1029/2008JD010708 33. Wijnant, I.L., van den Brink, H.W., Stepek, A.: North Sea Wind Climatology Part 1: A Review of Existing Wind Atlases, vol. TR-342, p. 66. Royal Netherlands Meteorological Institute Ministry of Infrastructure and the Environment, De Bilt, Netherlands, (2014) 34. Stevens, M.J.M., Smulders, P.T.: The estimation of the parameters of the Weibull wind speed distribution for wind energy utilization purposes. Wind Eng. 3(2), 132–145 (1979) 35. IPCC: Climate Change 2007: Synthesis Report. Contribution of Working Groups I, II and III to the Fourth Assessment Report of the Intergovernmental Panel on Climate Change. In: Team, C.W., Pachauri, R.K., Reisinger, A. (eds.), p. 104. IPCC, Geneva, Switzerland, Printed in Sweden (2007) 36. Bailey, D.T.: Meteorological monitoring guidance for regulatory modeling applications. In: Standards, O.o.A.Q.P.a. (ed.), p. 171. United States Environmental Protection Agency (EPA), Research Triangle Park, NC 27711 (2000)

Anthropogenic and Solar Influence on Temperature over Bulgaria Yavor Chapanov

Abstract The Sun is the main source of energy for all Earth’s geosystems, including climate, weather, mean sea level, winds, precipitation, and etc., mainly through Total Solar Irradiance (TSI), whose variations during solar activity cause various changes on the Earth surface. Climate processes, interactions between atmosphere-and ocean system, various local, regional and global hydrological cycles are the main mediator between solar activity and a number of geophysical processes on the Earth surface. The temperature at the Earth surface is widely used climate index, whose variations consist of significant seasonal oscillations, trend and long-term cycles. The global warming due to greenhouse gases grout produces significant temperature rise in the last decades, while the solar activity cycles drive periodic oscillations of the temperature. The variations of temperature over Bulgaria, due to anthropogenic and solar influences, is investigated by means of several long time series of meteorological observations. The changes of seasonal components of temperature and long-term oscillations are analyzed in narrow frequency bands by means of the Partial Fourier Approximation (PFA). These temperature variations are compared with the corresponding cycles of solar activity. The determined linear trends of temperature rise in the last decades are associated with the anthropogenic factors of the global warming. Keywords Solar activity · Temperature · Climate variations

1 Introduction Climate change (recently commonly referred to global or anthropogenic warming), changes in Mean Sea Level (MSL) and ice melt have been extremely important areas of research in recent years. Measurements show that the Earth’s mean temperature and MSL are rising, mainly due to increasing anthropogenic greenhouse gas emissions. Rising ocean heat content and corresponding ocean thermal expansion is an

Y. Chapanov (B) Climate, Atmosphere and Water Research Institute, Bulgarian Academy of Sciences (CAWRI-BAS), 66, Blvd Tzarigradsko Chaussee, Sofia 1784, Bulgaria © The Author(s), under exclusive license to Springer Nature Switzerland AG 2021 N. Dobrinkova and G. Gadzhev (eds.), Environmental Protection and Disaster Risks, Studies in Systems, Decision and Control 361, https://doi.org/10.1007/978-3-030-70190-1_6

75

76

Y. Chapanov

important element of climate change and sea-level rise. The significant contributions to climate variations are both anthropogenic and solar activity influences on temperature, ice melting, MSL, precipitation, winds and etc. The solar activity affects terrestrial systems by means of direct radiation over Earth surface, solar wind, and the solar magnetic field. The TSI cycles are the main source of climate indices variations. The solar wind directly affects Earth magnetic field, ionosphere and atmosphere. The variations of solar magnetic field modulate solar wind and Cosmic Rays (CR) in the frame of the heliosphere. The cosmic rays near Earth are modulated by Earth magnetic field variations, too. Recently a new mechanism of climate modulation, based on CR variations, has been proposed [1–4]. This mechanism is based on chain processes near tropopause by ozone production, temperature variations, followed by vertical winds and water content change. The last step of this chain affects surface temperature, because the atmospheric water is one of the most powerful greenhouse gas. This model provides an explanation for the cascade processes in which CR, whose total energy is relatively small, cause climatic effects with much more energy. The solar activity cycles modulate CR directly by the heliosphere and indirectly by the geomagnetic field changes, whose effect is visible mainly at high latitudes. A significant part of cosmic rays consists of charged solar particles, whose interaction with the Earth atmosphere is preceded by concentrating effect of geomagnetic field over polar regions, while the most energetic galactic cosmic rays affect all Earth regions. The cosmic ray intensity is controversial to the TSI variations. The TSI is strong during solar activity maximum, when the cosmic ray intensity has minimum and vice versa. So, during TSI maxima, the warming processes occur on Earth surface, and during TSI minima the thermal cycle amplitudes are amplified, due to cooling effects of cosmic rays. The variations of Earth temperature are affected directly by TSI cycles and indirectly by solar wind and solar magnetic field. These solar cycles are presented by the indices of TSI, sunspot numbers (also known as Wolf’s number) and North-South (N-S) solar asymmetry. The knowledge of temperature variations and cycles is important in various fields of human activity. An aspect of this knowledge is assessment of agrometeorological conditions in national agricultural lands. The determination of statistically significant tendency of average annual air temperature is important for spring crop growing [5, 6]. The climate change and coming with them abiotic stressors, a consequence of extreme weather conditions affects agriculture plant productivity [7, 8]. The goal of this study is to determine the anthropogenic trends and changes of seasonal and long-term components of temperature at several Bulgarian meteorological stations, their connection with the cycles and harmonics of solar activity in order to improve long term forecasts. The temperature variations tor the period 1900–2005 from 5 Bulgarian stations are analyzed in [9] and compared with some solar harmonics and cycles. The linear trends, seasonal components and long-term oscillations of the temperature are determined. Almost all of periodic components of observed temperature are highly correlated, including the seasonal variations and long-terms. The global warming signature is discovered in variations of seasonal amplitudes after 1975, where the amplitude rates are between 0.10 and 0.15 °C/year. The global warming signals were not

Anthropogenic and Solar Influence on Temperature over Bulgaria

77

detected in [9] from long-term variations of temperature, because the anthropogenic rise of temperature during the last decades stay hidden in long-term oscillations determined by the PFA. In the new research will be involved extended time series up to epoch 2020 from stations Sofia, Plovdiv, Sliven, Burgas and Varna. The time series from Sofia station is significant larger than the used one in [9]. It covers time interval 133 year for the period 1887–2020 and will help in precise determination of solar-climate influences. The mean temperature will be determined by averaging in 2-year sliding window and applied to all available data from this century may help in better detection of the anthropogenic signals in modern temperature measurements over Bulgaria.

2 Data 2.1 Meteorological Data The used data consist of centennial time series of monthly temperature variations from 5 Bulgarian stations in Sofia for the period 1887.0–2020.4, in Plovdiv, Sliven, Varna, and Burgas for the period 1899.0–2020.7 (Fig. 1). All data are available in [10].

2.2 Solar Data The solar data are presented by the Total Solar Irradiance (TSI) variations and N-S solar asymmetry (Fig. 2a). The daily reconstruction of TSI since 1850 is a composite of SATIRE-T reconstruction from [11] for 1850 to 22 August 1974; and SATIRE-S reconstruction from [12, 13] for 23 August 1974 onwards. The 0.1-year values of TSI (Fig. 2b) are calculated by means of robust Danish method [14–16]. This method allows to detect and isolate outliers and to obtain accurate and reliable solution for the mean values. The index S a of N-S solar asymmetry is calculated by formula Sa =

(A N − A S ) , (A N + A S )

(1)

where AN is the total area of the sunspots over Northern solar hemisphere and AS — the total area of the sunspots over Southern solar hemisphere. The data of sunspot area since 1874 are observed by the Royal Greenwich Observatory and merged after 1976 with the US Air Force (USAF) and the US National Oceanic and Atmospheric Administration (NOAA) data by D. Hathaway.

78

Y. Chapanov

Fig. 1 Mean monthly temperature at stations Sofia, Plovdiv, Sliven, Varna and Burgas

3 Methods The time series spectra are determined by the well-known Fast Fourier Transform (FFT). The periodical variations are derived from the data by means of partial Fourier approximation based on the Least-Squares (LS) estimation of Fourier coefficients. The Partial Fourier approximation F(t) of time series is given by F(t) = f 0 + f 1 (t − t0 ) +

n  k=1

ak sin k

2π 2π (t − t0 ) + bk cos k (t − t0 ), P0 P0

(2)

Anthropogenic and Solar Influence on Temperature over Bulgaria

79

Fig. 2 Reconstructed daily values of TSI variations (a); normal points of TSI variations at 0.1-year intervals (b); N-S solar asymmetry index (c)

where P0 is the period of the first harmonic, t 0 —the mean epoch of observations, f 0 , f 1 , ak and bk are unknown coefficients and n is the number of harmonics of the partial sum, which covers all oscillations with periods between P0 /n and P0 . The application of the LS estimation of Fourier coefficients needs at least 2n + 2 observations, so the number of harmonics n is chosen significantly smaller than the number N of sampled data f i . The small number of harmonics n yields to LS estimation of the coefficient errors. This estimation is the first essential difference with the classical Fourier approximation. The second difference is the arbitrary choice of the period of first harmonic P0 , instead of the observational time span, so the estimated frequencies may cover the desired set of real oscillations. This method allows a flexible and easy separation of harmonic oscillations into different frequency bands by the formula B(t) =

m2  k=m 1

ak sin k

2π 2π (t − t0 ) + bk cos k (t − t0 ), P0 P0

(3)

where the desired frequencies ωk are limited by the bandwidth 2π m 1 2π m 2 ≤ ωk ≤ , P0 P0

(4)

After estimating the Fourier coefficients, it is possible to identify a narrow frequency zone presenting significant amplitude, and defining a given cycle. Then

80

Y. Chapanov

this cycle can be reconstructed in time domain as the partial sum limited to the corresponding frequency bandwidth. Doing this for terrestrial and solar time series, we shall identify their respective cycles, isolate and compare the common ones. The PFA Method is applied to analyze common cycles of TSI, N-S solar asymmetry and Sofia station time series. These time series cover 133.2-year time interval for the period 1887.1–2020.3 for TSI temperature variations. The N-S solar asymmetry time series cover 133.2-year time interval for the period 1875.5–2008.7. The analyzed time series consist of 1586 monthly data points, reduced to 1333 values after linear interpolation at 0.1-year intervals. The PFA performs estimation of 150 harmonics with the accuracy better than 0.09 °C for temperature; 0.02 for N-S solar asymmetry and 5 mW/m2 for TSI. The seasonal amplitude of temperature variations AT for each year is calculated by A T = Tmax − Tmin ,

(5)

where the T max is the maximal summer temperature of a given station, and T min —the minimal winter temperature. The epoch of seasonal values is chosen to be the middle month of T max . The time series of seasonal amplitude of temperature at 0.1-year time interval are calculated by means of cubic spline interpolation. The obtain curve is rather smooth, so the accuracy of estimated Fourier harmonics is better than 0.003 °C.

4 Results The main results of this study are described by analyses of FFT spectra; seasonal and long-term variations of the temperature; linear trends due to anthropogenic global warming; influence of TSI and N-S solar asymmetry harmonics on temperature oscillations and variations of seasonal amplitude.

4.1 Time Series Spectra The time series spectra are calculated by the Fast Fourier Transform (FFT) and they are shown in Fig. 3. It is remarkable that the temperature spectra are almost coherent for interannual oscillations. All time series have significant seasonal components and oscillations with periods 2.2, 2.5, 2.8, 3.4, 4.0, 5.2, 7.0 and 10.5 years, so a high correlation between the long-term variations of temperature is expected.

Anthropogenic and Solar Influence on Temperature over Bulgaria

81

Fig. 3 FFT spectra of mean monthly temperature at stations Sofia, Plovdiv, Sliven, Varna and Burgas

4.2 Anthropogenic Effects Anthropogenic influence on mean temperature. The mean temperature is determined by data averaging in 2-year sliding window (Fig. 4). The average time series

Fig. 4 Mean temperature at stations Sofia (black), Plovdiv (red), Sliven (blue), Varna (green) and Burgas (yellow), determined in 2-year sliding window

82

Y. Chapanov

are highly correlated with some common decadal cycles. The time series variations of mean temperature at stations Plovdiv, Sliven, Varna and Burgas are almost identically, while the mean temperature at station Sofia is approximately 2 °C low. The mean temperature variations consist of significant decadal cycles, parabolic trend during the period 1900–1970, almost horizontal part in 1970–1980 and visible rise after 1980 (Fig. 4). The anthropogenic influence on mean temperature is determined by the linear trends for the whole data time span and for the periods 1980–2020. The linear trend for the period 1900–1980 is a horizontal line, so the temperature for this period is not sensitive to the global warming and corresponding anthropogenic rise is practically zero. Significant anthropogenic effect on temperature rise exists after 1980, when the linear rates of temperature over Bulgarian stations are between 0.02 and 0.06 °C/year. The results are shown in Table 1. Anthropogenic influence on seasonal amplitudes. The time series of seasonal temperature variations are determined by formula (5) and cubic spline interpolation at 0.1-year time intervals (Fig. 5). The seasonal components from all stations have similar behavior of periodic variations. They have common 5-year and 10-year cycles, modulated by long-term decadal oscillations. The amplitude of seasonal oscillations of the temperature varies mostly between 20 and 30 °C for the period 1900–1970, when the mean annual rate is about 0.01 °C/year (Table 2). This value is rather small in order to associate it with the anthropogenic effects of global warming. The signature of anthropogenic influence on temperature over Bulgaria is detectable after 1970, when the seasonal amplitudes rise more than 5 °C with mean rates between 0.05 and 0.09 °C/year (Fig. 5). Table 1 Linear trends in time series of average temperature, in [°C/year]

Station/Period 1900–2020 1900–1980 1980–2020 Sofia

0.007

−0.005

0.042

Plovdiv

0.003

−0.007

0.024; (1980–2015)

Sliven

0.007

0.0

0.053

Burgas

0.006

0.0

0.054

Varna

0.009

0.003

0.057

Fig. 5 Seasonal variations of temperature at stations Sofia (black), Plovdiv (red), Sliven (blue), Varna (green) and Burgas (yellow)

Anthropogenic and Solar Influence on Temperature over Bulgaria Table 2 Linear trends in time series of seasonal amplitudes, in [°C/year]

83

Station/Period 1900–2020 1900–1970 1970–2020 −0.005

0.012

0.086

0.002

0.012

0.084; (1970–2015)

Sliven

0.004

0.013

0.054

Burgas

−0.001

0.008

0.046

Varna

−0.004

0.005

0.049

Sofia Plovdiv

4.3 Solar Effects The influence of solar harmonics on temperature variations is often hidden, because the major part of solar signals is absorbed by cooling process of evaporation, especially in the case of decadal cycles. Nevertheless, some interannual and decadal cycles of the temperature are sensitive to the solar harmonics. The solar-temperature influence will be demonstrated on data from a single station, because the oscillations are highly correlated with a lot of identical cycles between different stations. The station Sofia is chosen, because its measurements cover the largest time interval. Solar influence on temperature variations. Common TSI and temperature cycles. TSI harmonics have excellent agreement with the corresponding oscillation of temperature from Sofia in 5 frequency bands and partial correlation from a decadal band (Fig. 6). The TSI harmonics and oscillation of temperature at Sofia station have equivalent cycles with periods from bands 5.1–5.3; 7.0–7.4; 14.8–16.6; and 16.6–19.0 years. The oscillations with periods 10.2–11.1 years have phase reverse during 1895–1920. The TSI influence on temperature oscillations with periods 22.2–26.6 years is broken for the period 1935–1965. The correlation coefficients of these cycles are frequency dependent, so it is not possible to detect them directly from the original time series and to create a common model of solar-climate influences. Common cycles of N-S solar asymmetry and temperature. The common cycles between N-S solar asymmetry and temperature at station Sofia are shown in Fig. 7. these cycles have good agreement for the oscillations with periods from bands 7.8– 8.3 years with phase deviation for 1885–1920; 8.3–8.9; 9.2–10.2 years with phase reverse during 1970–1990; 11.1–12.1 years with phase reverse during 1935–1965; 12.1–13.3 years; and 19.0–22.2 years. Solar influence on seasonal variations. The solar influence on seasonal variations of the temperature over Bulgaria is studied by means of solar indices and the variations of seasonal amplitude of temperature at station Sofia for the period 1887–2020. The TSI and seasonal amplitude have excellent agreement in 7 interannual and decadal frequency bands, while the interaction between N-S solar asymmetry and seasonal amplitude is discovered in 2 frequency bands.

84

Fig. 6 Common cycles of TSI (in W/m2 ) and temperature at station Sofia

Y. Chapanov

Anthropogenic and Solar Influence on Temperature over Bulgaria

Fig. 7 Common cycles of N-S solar asymmetry and temperature at station Sofia

85

86

Y. Chapanov

Fig. 8 Common interannual cycles of TSI (in W/m2 ) and seasonal amplitude of temperature at station Sofia

Common TSI and seasonal cycles. The common interannual TSI and seasonal cycles are presented in Fig. 8. They have good excellent agreement for the oscillations with periods 4.75–4.92; 7.0–7.4 and 8.9–9.5 years. The common decadal cycles of TSI and seasonal amplitude of temperature at station Sofia are presented in Fig. 9. These cycles cover oscillations with periods 12.1–13.3; 13.3–14.8; 14.8–16.6; and 19.0–22.2 years. All common TSI and seasonal cycles have frequency dependent coefficients of linear regression, so it is not possible to create a uniform linear model of TSI-seasonal influences. Common cycles of N-S solar asymmetry and seasonal amplitude. The common cycles of N-S solar asymmetry and seasonal amplitude are detected in 2 frequency bands (Fig. 10). The interannual cycles of N-S solar asymmetry and seasonal amplitude have periodicity from band 4.1–4.3 years, the decadal cycles are from band 26.6–33.3 years.

5 Conclusions The temperature variations for the period 1887–2020 from meteorological station Sofia are analyzed and compared with some solar harmonics and cycles. The linear trends, seasonal components and long-term oscillations of the temperature from 4 Bulgarian stations: Plovdiv, Sliven, Burgas and Varna are determined for the period

Anthropogenic and Solar Influence on Temperature over Bulgaria

87

Fig. 9 Common decadal cycles of TSI (in W/m2 ) and seasonal amplitude of temperature at station Sofia

Fig. 10 Common cycles of N-S solar asymmetry and seasonal amplitude of temperature at station Sofia

88

Y. Chapanov

1899–2020. Almost all of periodic components of observed temperature are highly correlated, including the seasonal variations and long-terms. The linear trends of the centennial time series of temperature match the tendency of temperature rise due to the global warming after 1970 for the seasonal components and after 1980 for the average temperature. The anthropogenic influences on temperature variations over Bulgaria have annual rates of average temperature between 0.02 and 0.06 °C/year and cumulative effects up to 2 °C. The seasonal amplitude rates due to anthropogenic influences are between 0.05 and 0.09 °C/year and cumulative effects about 5 °C. More precise determination of anthropogenic effects on temperature is expected after remove the decadal and interannual oscillations from observational series. The next step of this research should include estimation of parabolic terms in temperature trends. The TSI and N-S solar asymmetry harmonics have good agreement with the variations of temperature and seasonal amplitudes in various narrow frequency bands, whose periods are between 4.1 and 33.3 years. These results may improve climatic models and some long-term forecasts. Acknowledgements The study is supported by the National Science Fund of Bulgaria, Contract KP-06-N34/1 “Natural and anthropogenic factors of climate change—analyzes of global and local periodical components and long-term forecasts”.

References 1. Kilifarska, N.A., Haight, J.D.: The impact of solar variability on the middle atmosphere in present day and pre-industrial atmospheres. J. Atmos. Solar Terr. Phys. 67(3), 241–249 (2005) 2. Kilifarska, N.A., Tassev, Y.K., Tomova, D.Y.: Cosmic ray showers and their relation to the stratospheric sudden warmings. Sun Geosphere 3(1), 10–17 (2008) 3. Kilifarska, N.A.: Long–term variations in the stratospheric winter time ozone variability— 22 year cycle. C. R. l’Académie Bulgare Sci. 64(6), 867–874 (2011) 4. Velinov, P.I.Y., Mateev, L., Kilifarska, N.A.: 3-D model for cosmic ray planetary ionisation in the middle atmosphere. Ann. Geophys. 23(9), 3043–3046 (2005) 5. Georgieva, V., Shopova, N., Kazandjiev, V.: Assessment of conditions in South Bulgaria for spring crop growing using agrometeorological indices. In: AIP Conference 2075 (1), id.120014 (2019) 6. Kazandjiev, V., Shopova, N., Georgieva, V.: Hydrothermal conditions during vegetation season and spring crop growing in Plovdiv region. J. Balkan Ecol. 21(1), 23–38 (2018) 7. Shopova, N., Alexandrov, V., Todorova, G.: Cluster analysis of the highest daily temperature amplitudes in some agricultural regions of Southeastern Bulgaria. In: Proceedings of the First Science Conference “Climate, Atmosphere and Water Resources in the Face of Climate Change, Sofia, pp. 64–72 (2019) 8. Slavcheva-Sirakova, D., Shopova, N., Kostasdinov, K., Filipov, S., Velichkova, K.: Climate analysis and effects of abiotic stress on salad grounded in underground greenhouse and outdoor and effects of organic fertilizers in the fight with stress factors. In: Proceedings of the Conference “Agriculture for Life, Life for Agriculture”, Bucharest (2020) 9. Chapanov, Y.: Variations of temperature over Bulgaria and their connection with solar cycles. In: Proceedings of the 1st International Conference on Environmental protection and Disaster Risks, Sofia, Bulgaria (2020)

Anthropogenic and Solar Influence on Temperature over Bulgaria

89

10. https://www.stringmeteo.com 11. Krivova, N.A., Vieira, L.E.A., Solanki, S.K.: Reconstruction of solar spectral irradiance since the Maunder minimum. J. Geophys. Res. 115(A12112), 1–11 (2010) 12. Yeo, K.L., Krivova, N.A., Solanki, S.K., Glassmeier, K.H.: Reconstruction of total and spectral solar irradiance from 1974 to 2013 based on KPVT, SoHO/MDI and SDO/HMI observations. Astron. Astrophys. 570(A85), 1–18 (2014) 13. Yeo, K.L., Krivova, N.A., Solanki, S.K.: Solar cycle variation in solar irradiance. Space Sci. Rev. 186, 137–167 (2014) 14. Juhl, J.: The “Danish Method” of weight reduction for gross error detection. In: XV ISP Congress Proceedings, Commission III, Rio de Janeiro (1984) 15. Kegel, J.: Zur Lokalizierung grober Datenfehler mit Hilfe robuster Ausgleichungsfervahren. Vermessungstechnik 35, 348–350 (1987) 16. Kubik, K.: An error theory for the Danish method. In: ISP Symposium, Community III, Helsinki (1982)

Chemical Characteristics of Precipitation and Cloud Water at High Elevation Site in Bulgaria Elena Hristova , Blagorodka Veleva , Krum Velchev, and Emilia Georgieva

Abstract The aim of this work is to present and discuss newly obtained data for the chemical composition of precipitation (RW) and cloud water (CW) at a high-elevation site in Bulgaria. Sampling of RW and CW was organized in 2017 and 2018 during field experiments at Cherni Vrah, the highest peak in Vitosha Mountain. Passive collectors designed and constructed at NIMN were used. All collected samples (118) were analyzed for acidity (pH), conductivity (EC), main anions—SO4 2− , NO3 − , Cl− , ammonium ions (NH4 + ), macro and micro elements (Na, K, Mg, Ca, Fe, Si, Zn, Cu). The average pH values for both types of samples were in the acidity range ( NO3 − > Cl− for both type of samples and Ca > K > NH4 + > Na > Si > Mg > Zn > Fe > Cu (RW) and Ca > NH4 + > K > Mg > Na > Si > Zn > Fe > Cu (CW). The VWM total ionic content for cloud water samples (23.1 mg L−1 ) was around three times higher than this for precipitation samples (7.23 mg L−1 ). The major anion for both types of samples was sulphate with contribution to the total ionic concentration (TIC) above 30%. The second element is Ca for precipitation samples (18%) and NO3 − for cloud water samples (25%). The VWM total ionic content for cloud water samples (23.1 mg L−1 ) was around three times higher than this for precipitation samples (7.23 mg L−1 ). The major anion for both types of samples was sulphate with contribution to the total ionic concentration (TIC) above 30%. The second element is Ca for precipitation samples (18%) and NO3 − for cloud water samples (25%). The total ionic content (TIC) of precipitation and cloud water samples as frequency distribution is presented on Fig. 5.

98

E. Hristova et al.

Table 1 Volume weighted mean (VWM), average, minimum, and maximum concentrations and number of the samples (N) for precipitation samples mg L−1

Precipitation VWM

Average

Min

Max

N

0.89

1.36

0.05

26.94

63

1.21

2.12

0.18

15.83

67

2−

2.26

3.35

0.32

20.28

67

nss_SO4 2−

2.23

3.30

0.32

19.70

67

Ca

1.28

1.85

0.08

15.83

67

K

0.73

0.92

0.10

8.87

62

Mg

0.09

0.14

0.01

1.40

67

Na

0.30

1.27

0.10

15.64

23

Cu

0.01

0.01

0.01

0.05

28

Fe

0.01

0.01

0.01

0.06

33

Si

0.06

0.25

0.06

1.33

20

Zn

0.06

0.08

0.01

0.81

67

NH4 +

0.42

0.59

0.03

3.90

64

TIC

7.27

10.72

1.13

68.29

67

Cl− NO3 − SO4

Table 2 Volume weighted mean (VWM), average, minimum, and maximum concentrations and number of the samples (N) for cloud water samples mg L−1

Cloud water VWM

Average

Min

Cl−

1.17

1.17

0.10

8.05

40

NO3 −

5.69

5.69

0.15

23.13

40

9.13

9.13

0.59

39.52

40

SO4 2− nss_SO4

2−

Max

N

9.10

9.10

0.59

39.33

40

Ca

3.55

3.55

0.32

24.95

40

K

0.82

0.91

0.11

8.68

36

Mg

0.29

0.29

0.04

1.37

40

Na

0.29

0.50

0.06

3.20

23

Cu

0.01

0.01

0.01

0.03

15

Fe

0.02

0.03

0.01

0.14

21

Si

0.16

0.24

0.06

1.45

27

Zn

0.15

0.15

0.01

1.77

39

NH4 +

1.86

1.91

0.11

8.28

39

23.10

23.10

3.93

90.14

40

TIC

Chemical Characteristics of Precipitation and Cloud Water …

99

Fig. 5 Frequency distribution of total ionic content in cloud and rain water samples

The total ionic content in CW and RW was in the ranged from 1.1 to 68 mg L−1 and from 4 to 90 mg L−1 , respectively. The median of TIC in CW samples was 16.5 mg L−1 and 6.2 mg L−1 for RW samples. As shown by the frequency distribution of TIC, a fraction of cloud and precipitation samples have concentrations between 1 and 20 mg L−1 (89% of the RW and 62% of the CW). The percentage of samples in the concentration range 20–40 mg L−1 was higher for the CW than for the RW (25% and 6%, respectively). Only 2.6% of the CW samples had TIC in the range 80–100 mg L−1 . The variation in the concentrations of all studied elements is shown as Box Plot presented in Fig. 6. The ion composition of RW and CW was dominated by NH4 + , Ca, nss_SO4 2− and NO3 − , which made up more than 63% and 75% of the total ionic content. As expected, concentrations of analyzed elements were higher in cloud water than in precipitation samples.

Fig. 6 Concentrations of the studied elements in RW (left) and CW (right) samples

100

E. Hristova et al.

The concentrations of the main acidifying ions—nssSO4 2− for the study period ranged from 0.6 to 39 mg L−1 for CW and from 0.3 to 20.3 mg L−1 for RW. Concentrations of NO3 − vary from 0.15 to 23.1 mg L−1 for CW and from 0.18 to 15.8 mg L−1 for RW. NH4 + ion concentrations for CW and RW samples are ranged from 0.01 to 8.3 mg L−1 and from 0.03 to 3.9 mg L−1 . The lowest variations in concentrations were observed for Fe, Cu and Zn. Their concentration is ranged from 0.005 to 1.8 mg L−1 . High variation in Cl and Na concentrations are observed in RW samples. The obtained Cl concentration ranged from 0.05 to 27 mg L−1 and for Na from 01 to 15.6 mg L−1 .

3.3 Acidification and Neutralization Potentials The acidification potential of precipitation is usually due to the presence of H2 SO4 , HNO3 and organic acids [1, 9, 21], and the neutralization of these species occurs in the presence of NH3 and CaCO3 . The analysis of linear regression applied to the set of variables with acidification (SO4 2− and NO3 − ) and neutralization potential (Ca2+ and NH4 + ) is presented in Fig. 7. The correlation coefficient for precipitation samples (0.74) was higher than one found for cloud water samples (0.66), indicating that the contribution of other ionic species in the neutralization and acidification processes in the samples of atmospheric precipitation was not significant. The mean values of fractional acidity (FA) and the neutralization factor, calculated by Eqs. 4 and 5, for the precipitation and cloud water samples are summarized in Table 3. The fractional acidity (FA) ratio indicates whether the acidity generated by strong acids (H2 SO4 and HNO3 ) is neutralized or not, with the value of 1 for non-neutralized precipitation. Deviations from unity quantify the percentage of neutralization. The FA for the precipitation samples is 0.77 and for cloud water is 0.36, which is indicates that

Fig. 7 Linear regression of sum (SO4 2− , NO3 − ) and sum (Ca2+ , NH4 + ) in mgL−1 for RW (left) and CW (right) samples

Chemical Characteristics of Precipitation and Cloud Water …

101

Table 3 Mean fractional acidity (FA) and neutralization factor (NF) for precipitation (RW) and cloud water samples (CW) FA

NFCa

NFNH+

NFMg

RW

0.77

0.35

0.18

0.03

CW

0.36

0.52

0.16

0.03

4

there is not full neutralization of acidifying substance by alkaline constituents. This explained lower value of pH in this study. The neutralization factors (NF) of Ca, NH4 + , and Mg for RW samples were 0.35, 0.18, and 0.03 and for CW samples 0.52, 0.16, and 0.03, respectively. From these data it is obvious that the major neutralizing element is Ca followed by NH4 + .

3.4 Long Range Transport Effects for Some Selected Periods The origin of the air masses was examined by using back-trajectories from the model HYSPLIT for three periods: 19–20 March 2018, 30 June–4 July 2018 and 27-th of August 2018. The synoptic situation during the period 19–20 March 2018 is characterised by Saharan outbreak towards the Balkans, associated with coloured rain and orange snow in many parts of Eastern Europe. At Cherni Vrah the arriving air masses were from south (S)-southwest (SW) (Fig. 8) on 19.03 (precipitation sample) and from West (W) on 20.03 (cloud water sample) (Fig. 8). From all analysed samples collected in 2018, the highest concentrations of Cl and Na were obtained in the precipitation samples from 19 March. The TIC of the precipitation sample was 64.1 mg L−1 with

Fig. 8 Back-trajectories and contribution of different elements in precipitation and cloud waters samples for the period 19–20 March 2018

102

E. Hristova et al.

42% contribution of Cl, 24% of Na, 3% ss_SO4 2− (sea salt SO4 2− ) and only 6% of the sulphates from anthropogenic source (nss_SO4 2− ) (Fig. 8). A significant difference in the SO4 2− concentrations between RW and CW was observed. The TIC of the cloud water sample on 20 March 2018 was 3.9 mg L−1 containing 28% nss_SO4 2− , following by 28.3% Ca, 15.9% Cl and 11.1% K. These results indicated aged air masses with sea salt aerosols (Cl and Na) and mineral dust (Ca and Si) associated with Saharan origin. The coefficient of neutralization suggested Ca (0.36) as major neutralizing element in CW samples while the NH4 + ion (1.77) is for the precipitation sample. The synoptic situation for the second period, 30 June–4 July 2018, was characterized by the influence of the slowly moving Mediterranean cyclone “Nefeli” crossing the country from south to northeast. The atmospheric conditions in the first part of the period were highly unstable with heavy rains and thunderstorms in many places in Bulgaria, while the end of the period was marked by increased surface pressure and occasional convective precipitations [22]. The TIC in precipitation sample collected on 30 June (48.5 mg L−1 ) is higher than this for cloud water sample collected on 3 July (36.7 mg L−1 ). The trajectory analysis shows that on 30 of June the transport of air masses to the Cherni Vrah is from north while they were from W, NW on 3 of July (Fig. 9). Generally, nss_SO4 2− was found to be the dominant ion in both samples: RW (36%) and CW (43%). The contribution of NO3 − and NH4 + ions in the CW sample (23 and 12%) were higher than in the RW sample (16 and 5%). The contribution of Ca is three times higher in the RW sample than observed in the CW sample. For this selected period the TIC is consisted mainly of nss_SO4 2− , NO3 − , NH4 + and Ca (RW-83% and CW-88%). The analysis of neutralization factors for precipitation samples showed that Ca is the most dominant neutralized element (0.63). This analysis for cloud water samples presented lower neutralization from NH4 + (0.18) followed by Ca (0.16). The back-trajectory analysis for the 3rd case (27 of August 2018) shows that the transport of air masses to the sampling site were mainly from E, SE presenting influence from local or regional pollution sources. The synoptic situation for the country was characterized by a low-pressure system with very unstable atmospheric conditions. Frequent precipitations, somewhere also intense, were observed over the whole country. The TIC in this case is 5.5 and 63.3 mg L−1 for precipitation and cloud water samples, respectively. The most dominant ion in both types of samples was nss_SO4 2− with very high contribution to the TIC (49.8% for CW and 36.2% for the RW). This high contribution of the sulphates is related probably to the main point emission sources of SO2 located in south-east of Bulgaria. The second ion with high contribution to the TIC in RW was NO3 − (26%) followed by Ca (15%). The contribution of NO3 − to the TIC of CW is almost the same (27%). The contribution of NH4 + to the TIC of CW (2.7%) was lower than one for RW (9.9%). The analysis of neutralization factors in this case showed that Ca is the dominant neutralized element for RW and CW (Fig. 10).

Chemical Characteristics of Precipitation and Cloud Water …

103

Fig. 9 Back-trajectories and contribution of different elements in precipitation (30 Jun 2018) and cloud water (3 Jul 2018) samples

4 Conclusions New results for the chemical composition of precipitation (RW) and cloud water (CW) at the high-elevation site Cherni Vruh were presented. The results were based on 40 cloud and 78 rainwater samples collected and analysed in the period June 2017–November 2018. The comparison of the chemical content of the precipitation

104

E. Hristova et al.

Fig. 10 Back-trajectories and contribution of different elements in precipitation and cloud water samples on 27 August 2018

and cloud water presented systematic differences concerning the pH, the electric conductivity, and concentrations of most elements. The frequency analysis showed that 100% of the cloud and 98% of the precipitation samples have pH value in the acidity range (150 µg m−3 ) not only in the direct proximity of the main streets but also in many adjacent streets and finally we can also find areas with contaminant concentrations below 100 µg m−3 . In the southern area of the domain, the concentration of NO2 increase in some buildings that are to the right of the green open area. This result also shows that the dispersion of NO2 (and other pollutants) is strongly associated with the wind direction in and presence of the buildings, which is consistent with the results of this research [32]. This work shows that the concentration of pollutants in urban areas depends largely on traffic emissions and air flow patterns, which, in turn, is strongly affected by the heterogeneity of the urban layout, as described in this paper [33]. In the results we can see that some eddies are observed in the corners of the buildings where shear stresses are common and responsible for some turbulences in the flow, this phenomenon was described by [33].In general, PALM4U is very sensitive to emissions and their dispersion achieving an adequate spatial variability of NO2 concentrations for the study period. The reproduction of spatial gradients near heavy traffic is crucial in cities with high vehicle density and high NO2 concentration levels such as Madrid. After the model performance to reproduce wind flow and pollutant dispersion was previously evaluated, the next objective was to demonstrate that the modeling approach used in the present study was appropriate to evaluate multiple air pollution mitigation strategies because the modelling tool is so sensitive to any change. Urban vegetation impacts on air quality through influencing pollutant deposition and dispersion. Choice of urban vegetation is crucial when using vegetation as an ecosystem

The Use of LES CFD Urban Models and Mesoscale …

195

service for air quality improvements. Trees have cooling capacity and can change the energy balance in the atmosphere through latent heat flux. The micro-meteorology will affect to the pollutant dispersion. In this experiment we will compare the effects of broad leaf and needle leaf trees in NO2 concentrations during an air pollution episode using the PALM4U model capabilities. In the base simulation, named BROAD because it has been considered that all the trees in the park area are broad-leaf type we have changed the type of trees to needleleaf and it has run again the PALM4U model, this new simulation has been named NEEDLE. In order to know the impact or effect that the new trees would produce, we have calculated the differences between both simulations (BROAD-NEEDLE) and the result is shown in Fig. 4. The spatial distribution of the differences (%) in the NO2 concentrations are shown for the day 27/12/2016 at 20:00, when the high peak of NO2 concentration in the episode occurred. It can be seen how in the area of the park where the trees are located, if the trees were of the needle-leaf type, the concentrations of NO2 could increase up to 2%, but in the surrounding streets and other more distant areas where the concentrations of NO2

Fig. 4 Spatial distribution of the differences between BROAD and NEEDLE PALM4U simulation (%) in the NO2 concentration for the day 27/12/2016 at 20:00

196

R. San Jose et al.

are very high due to road traffic, this type of tree could reduce the contamination up to 15%. As can be seen, the impacts (negative and positive) are very diverse, distributed heterogeneously throughout the area of 1 km by 1 km, which shows that any action to reduce pollution can have contrary effects to those desired at some points and that the change even if it is very local can affect large distances. All this makes it necessary to use this type of tools to have an initial assessment of the mitigation measures to be adopted. The change of the type of trees, affects two fundamental processes in the atmospheric pollution, in the energy balance of the flows and in the velocity of deposition of the pollutants. In the model of land surface at the time of calculating the latent flow, in the case of the trees broad-leaf the minimum canopy resistance is of 175 s/m and for the needle-leaf it is of 500 s/m. Then the reduction of latent heat when using needle-leaf trees, will produce an increase of the sensible flow and an increase of the temperature in the zone. This causes the energy balance of the entire zone to change and effects can be produced in any part of the domain, as shown in the figure. In addition, needle trees facilitate deposition, since they need an optimal temperature (12º) lower than broad-leaf trees (20º) to open their stomata and facilitate deposition thanks to greater stomach conductivity. It is clear that the inclusion of new trees will produce changes in the urban pollution but it is impossible to know a prior how will be the impacts. The results are agreeing with other studies where deposition velocities were higher on needles than on broadleaves [34] and that trees considerably affected the accumulation of transported pollutants [35].

5 Conclusions An integrated urban air quality modelling system has been implemented. The tool includes a traffic emission model (which implement the EMEP-Tier 3 methodology) based on the SUMO model for traffic flows, a pollutant transport and chemistry model (WRF/Chem) and the CFD PALM4U model (LES mode). The modelling system has been used to simulate an episode of high NO2 concentrations in the city of Madrid during December 2016 with high spatial resolutions (WRF/Chem 25-5-1 km and PALM4U 5 m). The WRF/Chem 1 km evaluation has been satisfactory, with good values in the correlation coefficients, although at some local points the system has not been able to reach the maximum peaks of NO2 concentrations measured by monitoring stations. We added a CFD model (5 m resolution)—PALM4U model— into the system to reproduce more accurately the hot-spots. The PALM4U 5 m simulation reproduces pollutant dispersion in the presence of boundary and top conditions supplied by the WRF/Chem 1 km simulation, real building morphology, vegetation and hourly emissions. The results obtained allow us to confirm that the modeling tool (WRF/Chem-PALM4U) presented is capable to reproduce successfully the distribution of the concentration of pollutants over urban areas. The comparison with the observational data at E. Aguirre monitoring station shows that the model improves the peak values substantially. An experiment has

The Use of LES CFD Urban Models and Mesoscale …

197

been run to analyse the effects of two type of trees: broad and needle leafs (natural based solutions). Results show that the effects of urban vegetation on local air quality can be very complex and have a substantial impact in areas apart more than 500 m from the changed trees. The integrated modelling system is suitable for testing and evaluating NBS mitigation strategies on a scale of meters and obtaining information on their effectiveness. Acknowledgements The UPM authors thankfully acknowledge the computer resources, technical expertise and assistance provided by the Centro de Supercomputación y Visualización de Madrid (CESVIMA). The UPM authors thankfully acknowledge the computer resources, technical expertise and assistance provided by the Red Española de Supercomputación.

References 1. Pope, C., Dockery, D.: Health effects of fine particulate air pollution: lines that connect. J. Air Waste Manag. Assoc. 56, 709–742 (2006) 2. Caplin, A., Ghandehari, M., Lim, C., Glimcher, P., Thurston, G.: Advancing environmental exposure assessment science to benefit society. Nat. Commun. 10 (2019) 3. Liu, S., McKeen, S., Hsie, E., Lin, X., Kelly, K., Bradshaw, J., Sandholm, S., Browell, E., Gregory, G., Sachse, G., Bandy, A., Thornton, D., Blake, D., Rowland, F., Newell, R., Heikes, B., Singh, H., Talbot, R.: Model study of tropospheric trace species distributions during PEMWest a. J. Geophys. Res. Atmos. 101, 2073–2085 (1996) 4. Beevers, S., Kitwiroon, N., Williams, M., Carslaw, D.: One way coupling of CMAQ and a road source dispersion model for fine scale air pollution predictions. Atmos. Environ. 59, 47–58 (2012) 5. Choudhary, H., Tarlo, S.: Airway effects of traffic-related air pollution on outdoor workers. Curr. Opin. Allergy Clin. Immunol. 14, 106–112 (2014) 6. Mukul, S., Alamgir, M., Sohel, M., Pert, P., Herbohn, J., Turton, S., Khan, M., Munim, S., Reza, A., Laurance, W.: Combined effects of climate change and sea-level rise project dramatic habitat loss of the globally endangered Bengal tiger in the Bangladesh Sundarbans. Sci. Total Environ. 663, 830–840 (2019) 7. Resler, J., Krˇc, P., Belda, M., Juruš, P., Benešová, N., Lopata, J., Vlˇcek, O., Damašková, D., Eben, K., Derbek, P., Maronga, B., Kanani-Sühring, F.: PALM-USM v1.0: A new urban surface model integrated into the PALM large-eddy simulation model. Geoscientific Model Dev. 10, 3635–3659 (2017) 8. Buccolieri, R., Jeanjean, A., Gatto, E., Leigh, R.: The impact of trees on street ventilation, NOx and PM2.5 concentrations across heights in Marylebone Rd street canyon, central London. Sustain. Cities Soc. 41, 227–241 (2018) 9. Baik, J., Kim, J.: On the escape of pollutants from urban street canyons. Atmos. Environ. 36, 527–536 (2002) 10. Berardi, U., GhaffarianHoseini, A., GhaffarianHoseini, A.: State-of-the-art analysis of the environmental benefits of green roofs. Appl. Energy 115, 411–428 (2014) 11. Neft, I., Scungio, M., Culver, N., Singh, S.: Simulations of aerosol filtration by vegetation: Validation of existing models with available lab data and application to near-roadway scenario. Aerosol Sci. Technol. 50, 937–946 (2016) 12. Pataki, D.E., Carreiro, M.M., Cherrier, J., Grulke, N.E., Jennings, V., Pincetl, S., Pouyat, R.V., Whitlow, T.H., Zipperer, W.C.: Coupling biogeochemical cycles in urban environments: ecosystem services, green solutions, and misconceptions. Front. Ecol. Environ. 9(1), 27e36 (2011)

198

R. San Jose et al.

13. Grell, G., Peckham, S., Schmitz, R., McKeen, S., Frost, G., Skamarock, W., Eder, B.: Fully coupled “online” chemistry within the WRF model. Atmos. Environ. 39, 6957–6975 (2005) 14. Fast, J., Gustafson, W., Easter, R., Zaveri, R., Barnard, J., Chapman, E., Grell, G., Peckham, S.: Evolution of ozone, particulates, and aerosol direct radiative forcing in the vicinity of Houston using a fully coupled meteorology-chemistry-aerosol model. J. Geophys. Res. 111, (2006) 15. San José, R., Pérez, J., Balzarini, A., Baró, R., Curci, G., Forkel, R., Galmarini, S., Grell, G., Hirtl, M., Honzak, L., Im, U., Jiménez-Guerrero, P., Langer, M., Pirovano, G., Tuccella, P., Werhahn, J., Žabkar, R.: Sensitivity of feedback effects in CBMZ/MOSAIC chemical mechanism. Atmos. Environ. 115, 646–656 (2015) 16. Zaveri, R., Peters, L.: A new lumped structure photochemical mechanism for large-scale applications. J. Geophys. Res. Atmos. 104, 30387–30415 (1999) 17. Zaveri, R., Easter, R., Fast, J., Peters, L.: Model for simulating aerosol interactions and chemistry (MOSAIC). J. Geophys. Res. 113, (2008) 18. Binkowski, F., Shankar, U.: The regional particulate matter model: 1. Model description and preliminary results. J. Geophys. Res. 100, 26191 (1995) 19. Easter, R.: MIRAGE: Model description and evaluation of aerosols and trace gases. J. Geophys. Res. 109, (2004) 20. Williams, J., Landgraf, J., Bregman, A., Walter, H.: A modified band approach for the accurate calculation of online photolysis rates in stratospheric-tropospheric chemical transport models. Atmos. Chem. Phys. 6, 4137–4161 (2006) 21. Mlawer, E., Taubman, S., Brown, P., Iacono, M., Clough, S.: Radiative transfer for inhomogeneous atmospheres: RRTM, a validated correlated-k model for the longwave. J. Geophys. Res. Atmos. 102, 16663–16682 (1997) 22. Lin, Y., Farley, R., Orville, H.: Bulk parameterization of the snow field in a cloud model. J. Climate Appl. Meteorol. 22, 1065–1092 (1983) 23. Grell, G., Dévényi, D.: A generalized approach to parameterizing convection combining ensemble and data assimilation techniques. Geophys. Res. Lett. 29, 38-1–38-4 (2002) 24. National Centers for Environmental Prediction, 2015: NCEP GFS 0.25 degree global forecast grids historical archive. National Center for Atmospheric Research, Computational and Information Systems Laboratory, accessed 16 August 2019, https://doi.org/10.5065/D65 D8PWK 25. Maronga, B., Gryschka, M., Heinze, R., Hoffmann, F., Kanani-Sühring, F., Keck, M., Ketelsen, K., Letzel, M., Sühring, M., Raasch, S.: The parallelized large-eddy simulation model (PALM) version 4.0 for atmospheric and oceanic flows: model formulation, recent developments, and future perspectives. Geoscientific Model Dev. 8, 2515–2551 (2015) 26. Kanda, M., Inagaki, A., Miyamoto, T., Gryschka, M., Raasch, S.: A new aerodynamic parametrization for real urban surfaces. Bound.-Layer Meteorol. 148, 357–377 (2013) 27. Letzel, M., Krane, M., Raasch, S.: High resolution urban large-eddy simulation studies from street canyon to neighbourhood scale. Atmos. Environ. 42, 8770–8784 (2008) 28. Park, S., Baik, J., Lee, S.: Impacts of mesoscale wind on turbulent flow and ventilation in a densely built-up urban area. J. Appl. Meteorol. Climatol. 54, 811–824 (2015) 29. Abd Razak, A., Hagishima, A., Ikegaya, N., Tanimoto, J.: Analysis of airflow over building arrays for assessment of urban wind environment. Build. Environ. 59, 56–65 (2013) 30. Krajzewicz, D., Erdmann, J., Behrisch, M., Bieker, L.: Recent development and applications of SUMO—simulation of urban mobility. Int. J. Adv. Syst. Measure. 5(3&4), 128–138 (2012) 31. Hood, C., MacKenzie, I., Stocker, J., Johnson, K., Carruthers, D., Vieno, M., Doherty, R.: Air quality simulations for London using a coupled regional-to-local modelling system. Atmos. Chem. Phys. 18, 11221–11245 (2018) 32. Wang, Y., Zhou, Y., Zuo, J., Rameezdeen, R.: A computational fluid dynamic (CFD) simulation of PM10 dispersion caused by rail transit construction activity: a real urban street canyon model. Int. J. Environ. Res. Public Health. 15, 482 (2018) 33. Gonzalez Olivardia, F., Zhang, Q., Matsuo, T., Shimadera, H., Kondo, A.: Analysis of Pollutant Dispersion in a Realistic Urban Street Canyon Using Coupled CFD and Chemical Reaction Modeling. Atmosphere. 10, 479 (2019)

The Use of LES CFD Urban Models and Mesoscale …

199

34. Hwang, H., Yook, S., Ahn, K.: Experimental investigation of submicron and ultrafine soot particle removal by tree leaves. Atmos. Environ. 45, 6987–6994 (2011) 35. Salmond, J., Williams, D., Laing, G., Kingham, S., Dirks, K., Longley, I., Henshaw, G.: The influence of vegetation on the horizontal and vertical distribution of pollutants in a street canyon. Sci. Total Environ. 443, 287–298 (2013)

Urban Heat Island and Future Projections: A Study in Thessaloniki, Greece Stavros Keppas, Daphne Parliari, Serafeim Kontos, Anastasia Poupkou, Sofia Papadogiannaki, Paraskevi Tzoumaka, Apostolos Kelessis, and Melas Dimitrios

Abstract Future climate simulations have been produced for three 5-year periods until the end of twenty-first century using the WRF-ARW numerical weather prediction model for the greater area of Thessaloniki in the framework of the forecasting System for urban heaT Island effect (LIFE-ASTI) programme. In the present study, we analyse the characteristics of heat wave days in present and future at a central urban region of Thessaloniki and a rural region around the city in order to investigate the urban heat island effect under extreme heat. The number of heat wave days until 2100 is expected to increase by >12 times more than in the present. It is notable that more than 60% of the heat wave days within the urban area will be characterized by minimum temperatures ≥30 °C, while this percentage will be ~12% for the rural area. Finally, while in the present the urban heat island intensity during heat wave days presents mostly values 1–3 °C, in the future the intensity will be larger, in a few cases exceeding even 6 °C. Keywords Climate change · Heat waves · Urban heat island · WRF model

1 Introduction In recent decades, environment and humanity are experiencing tangible alterations in climate. Each of the last three decades has been gradually warmer than any preceding decade since 1850; climate change, mainly due to human influence, causes adverse impacts on human and natural systems [1]. Based on scientific evidence, it is believed that most of the warming observed over the last 50 years is attributable to human activities [2]. According to [3] extreme events such as heat waves, cold waves, floods, heavy precipitations, drought, tornadoes, and tropical cyclones are becoming more S. Keppas (B) · D. Parliari · S. Kontos · A. Poupkou · S. Papadogiannaki · M. Dimitrios Laboratory of Atmospheric Physics, School of Physics, Aristotle University of Thessaloniki, Thessaloniki 541124, Greece e-mail: [email protected] P. Tzoumaka · A. Kelessis Department of Environment, Municipality of Thessaloniki, Kleanthous 18, Thessaloniki 54642, Greece © The Author(s), under exclusive license to Springer Nature Switzerland AG 2021 N. Dobrinkova and G. Gadzhev (eds.), Environmental Protection and Disaster Risks, Studies in Systems, Decision and Control 361, https://doi.org/10.1007/978-3-030-70190-1_14

201

202

S. Keppas et al.

frequent as years go by, and although an extreme weather episode would normally be as rare as the 10th or 90th percentile of a probability density function estimated from observations [4], a significant risk for human and natural systems is posed. Global mean temperature is increasing and will continue to increase over the twenty-first century, while heat waves (HWs) will be more frequent, more severe and longer lasting in many regions [5]. More specifically, countries that belong to the temperate zones of Europe are very likely to experience several heat wave events [1]; in the meantime cold waves and frost days will subsequently become fewer [2, 6]. Under the trend of increasing temperature, extreme events will increase in frequency and heat waves, which are considered as extreme today, will be the normality in the future. Numerical modeling techniques are a valuable tool when attempting to assess the future projection of interdependent processes; on the other hand, simulating extreme weather events and projecting them in future periods can be a rather challenging task. The WRF model is suitable for multi-physics approach [7] and in many cases it is used to assess heat wave events, as in Stegehuis et al. [8], Wang et al. [9], Tewari et al. [10]. Modeling capabilities can also be extended to include thermal comfort simulations during high-temperature weather events and the synergy between heat waves and Urban Heat Island. Numerous cities around the world suffer from the adverse consequences of raised temperatures, UHI and significant thermal discomfort, as summarized in [11]. Several researches have focused on Mediterranean area: for instance Athens, Greece was the central point in Giannaros [12], in Giannaros et al. [13] and Founda and Santamouris [14] and Thessaloniki case, Greece, was depicted in Giannaros and Melas [15]. Heat waves are related to serious environmental and medical problems generating also serious economic consequences. McGregor et al. [16] claims that HWs present variable impacts on society e.g., rise in mortality, increased strain on infrastructure (power, water and transport), possible rise in social disturbance and wider impacts such as effects on the retail industry, levels of street crime and tourism. Nevertheless, the most prominent and well documented effect of extremely high temperatures is a subsequent increase in human morbidity and mortality (indicatively: [17–21]). One of the most prominent examples of a severe heat wave with clearly documented impact on human mortality was the heat wave that stroke Europe in August 2003. In France, the mean maximum temperature from August 1st to 20th 2003 exceeded the seasonal normal values by 11–12 °C on nine consecutive days, and 15,000 excess deaths were attributed to extremely high temperatures [22]. These detrimental effects can be further aggravated by rapid urbanization and deterioration of air quality due to human activities, e.g. [23]. The adverse impacts of HWs are more pronounced in urban areas due to higher population density and Urban Heat Island (UHI) phenomenon, which refers to the higher temperatures observed in urban environments compared to their rural surroundings [24]. As UHIs are characterized by increased temperatures, they can potentially interact with HWs intensifying the extreme heat in cities. Founda and Santamouris [14] found a positive feedback between UHIs and HWs when they investigated five HW episodes during summer 2012 in Athens, Greece. Such conditions have been linked to increased mortality

Urban Heat Island and Future Projections …

203

rates, due to an increased exposure to high temperatures which tend to exacerbate the threats to human health posed by thermal stress. The combined effects of climate change, rapid urban sprawl and UHIs introduce an emerging health concern for the urban residents.

2 Data and Methodology 2.1 Regional Climate Model This study is based on simulations conducted through the WRF-ARW numerical weather prediction model. The physics options used here, are shown in Table 1. The selection of schemes follows previous regional climate simulations, which were focused on the investigation of the Urban Heat Island (UHI) effect in Greece [24]. The WRF regional climate model was run in the framework of the forecasting System for urban heaT Island effect (LIFE-ASTI) project (https://app.lifeasti.eu/) using four two-way nested domains with horizontal grid resolutions of 50 km (d01), 10 km (d02) and 2 km (d03, d04). For the purposes of this study, only outputs referred to the d03 are analysed (Fig. 1). In addition, the single-layer urban canopy model (SLUCM) used in order to simulate the urban areas in a finer detail [25, 26]. In total, three datasets were produced each of them referring to one of the periods 2006–2010 (used as reference period), 2046–2050 and 2096–2100. Table 1 Physics parametrizations used in the present study Physics

Parametrization

References

Microphysics (clouds)

WRF single-moment 6-class (WSM6)

Hong and Lim [27]

Cumulus (convection)a

Kain-Fritsch (KF)

Kain [28]

Planetary boundary layer

YSU scheme

Hong, Noh and Dudhia (2006, MWR) [48]

Surface layer

Monin-Obukhov (Janjic Eta) Sceme

Monin and Obukhov [31]; Janjic [32]

Land surface

Noah model

Tewari et al. [33]

Short-wave radiation

RRTMG

Iacono et al. [34], JGR

a Cumulus

parametrization only used in d01 and d02

204

S. Keppas et al.

Fig. 1 The four two-way nested domains over Europe

2.2 Model Inputs For the simulations of the d01, the land use input comes from the Global Land Cover by National Mapping Organizations version 1 (GLCNMO v1). The GLCNMO v1 was developed on the basis of MODIS-TERRA 2003 data with a resolution of 30 arcseconds [35]. For all the other three domains, the Corine Land Cover (CLC) version 2012 was used. The CLC12 dataset is a Copernicus land cover product acquired between 2011 and 2012 with a spatial resolution of 100–250 m classifying the land use into 44 different classes. In order to simulate the present and future climate, boundary conditions from the National Centre for Atmospheric Research (NCAR) and the community Earth System Model (CESM), which was participated in phase 5 of the Coupled Model Intercomparison Experiment (CMIP5). The data were used in six hourly intervals with a spatial resolution of 1°. It should be noted that the future climate simulations were based in the worst case or ‘business as usual’ high emissions scenario, which is the Representative Concentration Pathway 8.5 (RCP8.5; [36]).

2.3 The Heat Wave Day Criteria, the UHI and the Discomfort Index (DI) This study is focused on the identification and comparison of the heat wave days in the wider region of Thessaloniki in northern Greece, which were occurred within

Urban Heat Island and Future Projections …

205

the three 5-year periods simulated by the WRF model. In order to compare all these three periods on the same basis, we used a simple methodology instead of a percentile method (e.g [37–39]). The criteria for the distinguishing of the heat wave days were adopted from [40] and they are: Tmax ≥ 37 ◦ C Tavg ≥ 31 ◦ C where Tmax is the daily maximum temperature (in °C) and Tavg is the daily average temperature (in °C). These criteria were applied in two grid points of the outputs of the simulations (Fig. 2). The one is selected to represent the urban area of Thessaloniki centre, while the other is a representative of a rural area (reference point) around the city. It should be highlighted that the selected points are considered as points of similar altitude (33 m), located close to the sea but in different land use areas (continuous urban fabric and agricultural area respectively). It should be noted that UHI intensity was calculated via the subtraction between the reference point temperature and the urban point temperature. In order to estimate and compare the heat sensation in the present and future, we used the Thom’s Discomfort Index [41]: DI = T2m − 0.55 · (1 − 0.01 · RH2m ) · (T2m − 14.5)

Fig. 2 The two points, which were selected for the comparisons of the study

206

S. Keppas et al.

where T2m is the temperature at 2 m (in °C) and RH2m is the relative humidity at 2 m (in %).

3 Results The number of the identified heat wave days, at the urban and the reference grid point respectively, is shown in Fig. 3, as simulated for the three periods of study. The urban area exhibited 3.25 more heat wave days comparing to the reference area of Michaniona during the reference period (13 and 4 heat wave days respectively). This ratio remains almost identical (3.2) for the middle of the current century (2046–2050), but decreases at the end of the century to 2.4 (161 and 66 heat wave days for urban and reference area). The fact that the ratio decreases means that the characteristics of the reference site will change, becoming somewhat similar to those of the urban area. In general, it seems that heat wave days will almost double in number by the middle of the century. However, there will be a dramatic and significant increase by the end of the twenty-first century with 12.3 and 16.5 times more heat wave days in the urban and reference area respectively. Significant average temperature increase, regardless of the land use, is indisputable under the RCP8.5 scenario in a global scale until the end of the twenty-first century (e.g. [42, 43]). In Fig. 4a–b, it seems that the maximum temperature during a heat wave day presents a similar trend for both urban and reference point. In particular, at

Fig. 3 Heat wave days as simulated at the urban (Thessaloniki centre; red bars) and the reference (Michaniona; green bars) point for the three simulation periods of 2006–2010, 2046–2050 and 2096–2100

Urban Heat Island and Future Projections …

207

Fig. 4 Graphs showing the maximum temperature at the urban a and reference b point. The graphs c and d display the minimum temperature respectively

the end of the century the maximum temperature will exceed 40 °C in the 20% of the heat wave days in and out of the urban area. However, there is a significant difference between urban and reference point regarding the minimum temperature. According to Fig. 4c–d, the minimum temperature, which during the reference period demonstrates values 6 °C (Fig. 6b). The maximum UHI intensity in the reference and the 2046–2050 period mostly fluctuates between 2 and 4 °C, while within 2096–2100 presents a wider range with most of heat wave days (~80%) exhibiting an intensity of 2–5 °C (Fig. 6c). Moreover, while maximum UHI intensity during heat wave days occurs in the morning (00-03UTC) and late in the afternoon (15UTC) in the reference period, the time range broadens till 2100 with many cases of maximum UHI intensity within 00-03UTC (~33%) and 12-18UTC (~60%) (Fig. 6d).

Urban Heat Island and Future Projections …

209

Fig. 6 UHI intensity relative frequency at a 00UTC and b 12UTC during common heat wave days between urban and reference points. Figure c displays the daily maximum UHI intensity and figure d shows the time of the day that the maximum intensity occurs. The different red-ish shades refer to different simulation period

4 Conclusions For the present study, we used the outputs of WRF simulations for three different periods, a reference period (2006–2010), and two periods in the future one in the middle of the twenty-first century (2046–2050) and another at the end of it (2096– 2100). The WRF model ran under the worst-case scenario RCP8.5. This study aims to a brief analysis and comparison of the traits of the heat wave days in the present and future, and the investigation of the UHI effect in the greater area of Thessaloniki. Here we sum up the most interesting results of this work: • Regarding the number of heat wave days, it is expected that they will be almost doubled by 2050 and be 12.3–16.5 times more by 2100. • Both Tmax and Tmin of Heat Wave Days in and out of the city of Thessaloniki will increase in the future. However, Tmin within the urban area in 2100 will be, in more than half of the cases, >30 °C. • UHI intensity in the future remains similar to now (1–3 °C). However, there is a notable number of heat wave days (~22%) with intensities of 3–>6 °C by 2100, especially in the afternoon. • The maximum UHI intensity is usually 2–4 °C, but demonstrating a wider range in the future. The daily period of the occurrence of the maximum UHI intensity seems to broaden in the future with many occurrences in the afternoon (60% until the end of the century).

210

S. Keppas et al.

As the Mediterranean Sea is expected to be fairly vulnerable in the climate change (e.g. [45–47]), future works, as part of the project LIFE-ASTI, may include more comprehensive analysis of the future climate in the two Mediterranean cities Thessaloniki and Rome. Acknowledgements This work was funded by the LIFE Programme of the European Union in the framework of the project “Implementation of a forecasting system for urban heat island effect for the development of adaptation strategies—LIFE ASTI”. Results presented in this work have been produced using the Aristotle University of Thessaloniki (AUTh) High Performance Computing Infrastructure and Resources.

References 1. Field, C.B., Barros, V.R., Mastrandrea, M.D., Mach, K.J., Abdrabo, M.K., Adger, N., Anokhin, Y.A., Anisimov, O.A., Arent, D.J., Barnett, J., Burkett, V.R.: Summary for policymakers. In: Climate Change 2014: Impacts, Adaptation, and Vulnerability. Part A: Global and Sectoral Aspects. Contribution of Working Group II to the Fifth Assessment Report of the Intergovernmental Panel on Climate Change, pp. 1–32. Cambridge University Press (2014) 2. Houghton, J.T., Ding, Y., Griggs, D.J., Noguer, M., Van der Linden, P.J., Xiaosu, D.: Contribution of working group 1 to the third assessment report of the intergovernmental panel on climate change (IPCC) (2001) 3. TT-DEWCE, W.W.O.: Guidelines on the definition and monitoring of extreme weather and climate events: draft version. First Review by TT-Dewce (2016) 4. Cubasch, U., Wuebbles, D., Chen, D., Facchini, M. C., Frame, D., Mahowald, N.W.J.G., Winther, J.G.: Introduction. In: Plattner, K., Tignor, M., Allen, S.K., Boschung, J., Nauels, A., Xia, Y., Bex, V., Midgley, P.M. (eds.) Climate Change 2013: The Physical Science Basis. Contribution of Working Group I to the Fifth Assessment Report of the Intergovernmental Panel on Climate Change. Cambridge University Press, Cambridge, UK, and New York (2013). http:// www.climatechange2013.org/images/report/WG1AR5_Chapter01_FINAL.pdf 5. Change, I.C.: Synthesis report. Contribution of working groups I. II and III to the Fifth Assessment Report of the Intergovernmental Panel on Climate Change, 151(10.1017) (2014) 6. Feyen, L., Russo, S., Naumann, G., Formetta, G., Forzieri, G., Girardello, M.: Global warming and human impacts of heat and cold extremes in the EU (2020) 7. Mooney, P.A., Mulligan., F.J., Fealy, R.: Evaluation of the sensitivity of the weather research and forecasting model to parameterization schemes for regional climates of Europe over the period 1990–95. J. Clim. 26(3), 1002–1017 (2013). https://doi.org/10.1175/JCLI-D-11-006 76.1 8. Stegehuis, A.I., Vautard, R., Ciais, P., Teuling, A.J., Miralles, D.G., Wild, M.: An observationconstrained multi-physics WRF ensemble for simulating European mega heat waves. Geoscientific Model Development 8(7), 2285–2298 (2015). https://doi.org/10.5194/gmd-8-22852015 9. Wang, P., Hui, P., Xue, D., Tang, J.: Future projection of heat waves over China under global warming within the CORDEX-EA-II project. Clim. Dyn. 53(1–2), 957–973 (2019). https:// doi.org/10.1007/s00382-019-04621-7 10. Tewari, M., Chen, F., Kusaka, H., Miao, S.: Coupled WRF/Unified Noah/Urban-Canopy Modeling System. NCAR WRF Documentation. Retrieved from http://www.ral.ucar.edu/res earch/land/technology/urban/WRF-LSM-Urban.pdf (2007) 11. Santamouris, M.: Heat island research in Europe: The state of the art. Adv. Build. Energy Res. 1(1), 123–150 (2007). https://doi.org/10.1080/17512549.2007.9687272

Urban Heat Island and Future Projections …

211

12. Giannaros, C.: Sensitivity analysis and optimization of a mesoscale atmospheric model. Aristotle University of Thessaloniki (2018) 13. Giannaros, T.M., Melas, D., Daglis, I.A., Keramitsoglou, I., Kourtidis, K.: Numerical study of the urban heat island over Athens (Greece) with the WRF model. Atmos. Environ. 73, 103–111 (2013). https://doi.org/10.1016/j.atmosenv.2013.02.055 14. Founda, D., Santamouris, M.: Synergies between Urban Heat Island and Heat Waves in Athens (Greece), during an extremely hot summer (2012). Sci. Rep. 7(1), 1–11 (2017) 15. Giannaros, T.M., Melas, D.: Study of the urban heat island in a coastal Mediterranean City: the case study of Thessaloniki, Greece. Atmos. Res. 118, 103–120 (2012). https://doi.org/10. 1016/j.atmosres.2012.06.006 16. McGregor, G.R., Felling, M., Wolf, T., Gosling, S.: The social impacts of heat waves. (2007) 17. Analitis, A., Michelozzi, P., D’Ippoliti, D., De’Donato, F., Menne, B., Matthies, F., Atkinson, R.W., Iñiguez, C., Basagaña, X., Schneider, A., Lefranc, A.: Effects of heat waves on mortality: effect modification and confounding by air pollutants. Epidemiology 15–22 (2014) 18. Cusack, L., de Crespigny, C., Athanasos, P.: Heatwaves and their impact on people with alcohol, drug and mental health conditions: a discussion paper on clinical practice considerations. J. Adv. Nurs. 67(4), 915–922 (2011) 19. D’Ippoliti, D., Michelozzi, P., Marino, C., De’Donato, F., Menne, B., Katsouyanni, K., Kirchmayer, U., Analitis, A., Medina-Ramón, M., Paldy, A., Atkinson, R.: The impact of heat waves on mortality in 9 European cities: results from the EuroHEAT project. Environ Health 9(1), 37 (2010) 20. Semenza, J.C., McCullough, J.E., Flanders, W.D., McGeehin, M.A., Lumpkin, J.R.: Excess hospital admissions during the July 1995 heat wave in Chicago. Am. J. Prev. Med. 16(4), 269–277 (1999) 21. Theoharatos, G., Pantavou, K., Mavrakis, A., Spanou, A., Katavoutas, G., Efstathiou, P., Mpekas, P., Asimakopoulos, D.: Heat waves observed in 2007 in Athens, Greece: synoptic conditions, bioclimatological assessment, air quality levels and health effects. Environ. Res. 110(2), 152–161 (2010) 22. Fouillet, A., Rey, G., Laurent, F., Pavillon, G., Bellec, S., Guihenneuc-Jouyaux, C., Clavel, J., Jougla, E., Hémon, D.: Excess mortality related to the August 2003 heat wave in France. Int. Arch. Occup. Environ. Health 80(1), 16–24 (2006) 23. Katsouyanni, K., Pantazopoulou, A., Touloumi, G., Tselepidaki, I., Moustris, K., Asimakopoulos, D., Poulopoulou, G., Trichopoulos, D.: Evidence for interaction between air pollution and high temperature in the causation of excess mortality. Arch. Environ. Health An Int. J. 48(4), 235–242 (1993) 24. Giannaros, C., Nenes, A., Giannaros, T.M., Kourtidis, K., Melas, D.: A comprehensive approach for the simulation of the Urban Heat Island effect with the WRF/SLUCM modeling system: the case of Athens (Greece). Atmos. Res. 201, 86–101 (2018) 25. Kusaka, H., Kondo, H., Kikegawa, Y., Kimura, F.: A simple single-layer urban canopy model for atmospheric models: comparison with multi-layer and slab models. Bound.-Layer Meteorol. 101(3), 329–358 (2001) 26. Kusaka, H., Kimura, F.: Coupling a single-layer urban canopy model with a simple atmospheric model: impact on urban heat island simulation for an idealized case. J. Meteorol. Soc. Jpn. Ser. II, 82(1), 67–80 (2004) 27. Hong, S.Y., Lim, J.O.J.: The WRF single-moment 6-class microphysics scheme (WSM6). AsiaPacific J Atmos Sci 42(2), 129–151 (2006) 28. Kain, J.S.: The Kain-Fritsch convective parameterization: an update. J. Appl. Meteorol. 43(1), 170–181 (2004) 29. Pleim, J.E.: A combined local and nonlocal closure model for the atmospheric boundary layer. Part I: Model description and testing. J. Appl. Meteorol. Climatol. 46(9), 1383–1395 (2007a) 30. Pleim, J.E.: A combined local and nonlocal closure model for the atmospheric boundary layer. Part II: Application and evaluation in a mesoscale meteorological model. J. Appl. Meteorol. Climatol 46(9), 1396–1409 (2007b)

212

S. Keppas et al.

31. Monin, A.S., Obukhov, A.M.: Basic laws of turbulent mixing in the surface layer of the atmosphere. Contrib. Geophys. Inst. Acad. Sci. USSR 151(163), (1954) 32. Janji´c, Z.I.: The surface layer in the NCEP Eta model. In: Eleventh Conference on Numerical Weather Prediction, pp. 19–23. Amer. Meteor. Soc. Norfolk, VA (1996) 33. Tewari, M., Chen, F., Wang, W., Dudhia, J., LeMone, M. A., Mitchell, K., Ek, M., Gayno, G., Wegiel, J., Cuenca, R.H.: Implementation and verification of the unified NOAH land surface model in the WRF model. In 20th Conference on Weather Analysis and Forecasting/16th Conference on Numerical Weather Prediction (Vol. 1115, No. 6). Seattle, WA: American Meteorological Society (2004) 34. Iacono, M.J., Delamere, J.S., Mlawer, E.J., Shephard, M.W., Clough, S.A., Collins, W.D.: Radiative forcing by long-lived greenhouse gases: calculations with the AER radiative transfer models. J. Geophys. Res. Atmos. 113(D13) (2008) 35. Tateishi, R., Uriyangqai, B., Al-Bilbisi, H., Ghar, M.A., Tsend-Ayush, J., Kobayashi, T., Kasimu, A., Hoan, N.T., Shalaby, A., Alsaaideh, B., Enkhzaya, T.: Production of global land cover data–GLCNMO. Int. J. Digit. Earth 4(1), 22–49 (2011) 36. Meinshausen, M., Smith, S.J., Calvin, K., Daniel, J.S., Kainuma, M.L.T., Lamarque, J.F., Matsumoto, K., Montzka, S.A., Raper, S.C., Riahi, K., Thomson, A.G.J.M.V.: The RCP greenhouse gas concentrations and their extensions from 1765 to 2300. Clim. Change 109(1–2), 213 (2011) 37. Hartmann, D.L., Tank, A.M.G.K., Rusticucci, M.: IPCC fifth assessment report, climate change 2013: the physical science basis. Ipcc Ar5, 5, 31–39 (2013) 38. Robinson, P.J.: On the definition of a heat wave. J. Appl. Meteorol. 40(4), 762–775 (2001) 39. Tolika, K.: Assessing heat waves over Greece using the excess heat factor (EHF). Climate 7(1), 9 (2019) 40. Metaxas, D.A., DA, M.: Heat waves from a synoptic point of view (1980) 41. Thom, E.C.: The discomfort index. Weatherwise 12(2), 57–61 (1959) 42. Diffenbaugh, N.S., Giorgi, F.: Climate change hotspots in the CMIP5 global climate model ensemble. Clim. Change 114(3–4), 813–822 (2012) 43. Stocker, T.F., Qin, D., Plattner, G.K., Tignor, M., Allen, S.K., Boschung, J., Nauels, A., Xia, Y., Bex, V. and Midgley, P.M.: Climate change 2013: the physical science basis. Contribution of working group I to the fifth assessment report of the intergovernmental panel on climate change, 1535 (2013) 44. Medina-Ramon, M., Schwartz, J.: Temperature, temperature extremes, and mortality: a study of acclimatisation and effect modification in 50 US cities. Occup. Environ. Med. 64(12), 827–833 (2007) 45. Adloff, F., Somot, S., Sevault, F., Jordà, G., Aznar, R., Déqué, M., Padorno, C., AlvarezFanjul, E.: Mediterranean Sea response to climate change in an ensemble of twenty first century scenarios. Clim. Dyn. 45(9–10), 2775–2802 (2015) 46. Lejeusne, C., Chevaldonné, P., Pergent-Martini, C., Boudouresque, C.F., Pérez, T.: Climate change effects on a miniature ocean: the highly diverse, highly impacted Mediterranean Sea. Trends Ecol. Evol. 25(4), 250–260 (2010) 47. Raitsos, D.E., Beaugrand, G., Georgopoulos, D., Zenetos, A., Pancucci-Papadopoulou, A.M., Theocharis, A., Papathanassiou, E.: Global climate change amplifies the entry of tropical species into the eastern mediterranean sea. Limnol. Oceanogr. 55(4), 1478–1484 (2010) 48. Hong, S.Y., Noh, Y., Dudhia, J.: A new vertical diffusion package with an explicit treatment of entrainment processes. Mon. Weather Rev. 134(9), 2318–2341 (2006)

Chemical Characteristics of Flue Gas Particulates: An Experimental Investigation Tsvetelina Petrova, Iliyana Naydenova, Ricardo Ferreira, Yordanka Karakirova, and Mário Costa

Abstract The present work aims at characterizing particulate matter (PM) of different size, emitted during biomass gasification in a drop tube furnace (DTF) at 1000 °C. The elemental composition was determined using X-ray fluorescence (XRF) and Electron Paramagnetic Resonance (EPR) analyses. Overall 19 elements were determined and the relative mass concentration of their oxides was identified as macro- (above 3%) and micro-concentration (below 3%). The elements Fe, Mn and S were found in each type of particulates, regardless of the used biomass and gasifying agent. The dominant macro component of char (cyclone particles >10 µm) was Ca (50.56–100% of the total CaO), followed by K, Fe, S, Mn and Cr. Only colza char contained significant portion of P and much lower Fe. The primary macro constituents of PM10–2.5 were Fe, Mn and S. The volatile ash compounds K and Cl are typical constituents of the submicron sized ultrafine particles (UFP), when biomass from agricultural residue was gasified. This confirms the hypothesis that elements, having low boiling point significantly influence UFP formation through the nucleation. Two EPR spectra were obtained for the char samples: a broad signal with g ≈ 2.1–2.6, and a narrow sharp signal with g ≈ 2.002–2.003. The broad EPR signal was attributed to the paramagnetic metal ions Fe3+ and Mn2+ , which was in agreement with the XRF analysis. The narrow signal was attributed to the appearance of soot particles. Keywords Particulate matter · Biomass gasification · XRF analysis · EPR analysis

T. Petrova (B) · I. Naydenova College of Energy and Electronics, Technical University of Sofia, 8 Kliment Ohridski Blvd, bl. 2, 2307-B, 1000 Sofia, Bulgaria e-mail: [email protected] R. Ferreira · M. Costa Mechanical Engineering Department, Universidade de Lisboa, Instituto Superior Técnico, IDMEC, Avenida Rovisco Pais 1, 1049-001 Lisbon, Portugal Y. Karakirova Bulgarian Academy of Sciences, Institute of Catalysis, Acad. G. Bonchev Str., bl. 11, 1113 Sofia, Bulgaria © The Author(s), under exclusive license to Springer Nature Switzerland AG 2021 N. Dobrinkova and G. Gadzhev (eds.), Environmental Protection and Disaster Risks, Studies in Systems, Decision and Control 361, https://doi.org/10.1007/978-3-030-70190-1_15

213

214

T. Petrova et al.

1 Introduction Good understanding of the nature and source of particulate matter (PM) is of crucial importance for developing effective measures for their reduction in the European cities, continuously exceeding the limit values [1]. Currently in Bulgaria, certain attention is drawn on the chemical composition and source apportionment of PM10 and PM2.5 , sampled in the atmospheric air. Chuturkova [2] assessed the annual, monthly and maximum concentrations of PM10 and PM2.5 in atmospheric air, within 2007–2014. This investigation involved urbanized and industrial city areas with overall three monitoring stations: city background, transportoriented, and industrial-oriented. Vlaknenski et al. [3] traced the background air pollution with PM10 in three medium-sized urban areas of Central North Bulgaria during 2007–2010. Veleva et al. [4, 5] measured the daily concentration of PM10 in a period of four seasons in Bulgaria, as well as the chemical composition of the collected samples through Energy Dispersive X-Ray Fluorescent (EDXRF) analysis. The authors identify more than 23 chemical elements in the structure of the collected PM10 . Later on, Veleva et al. [6] report the PM10 mass concentration and chemical composition, where more than 20 elements (P, S, Cl, K, Ca, Ti, V, Cr, Mn, Fe, Ni, Cu, Zn Br, Rb, Sr, Y, Zr, Cd, Sn, Sb, I, Ba, Pb) are detected. The PM is sampled during six experimental campaigns in Sofia, within the winter and the summer periods of 2012, 2013 and 2014. The exceeded PM10 and PM2.5 concentrations in the urban atmosphere is still a challenge for many European Member States, struggling to find efficient solution for various insufficiently controlled processes, such as residential heating and transport [7–9]. The role of the local, national and regional PM sources and their specific properties must also be considered. Pateraki et al. [10] chemically characterized PM2.5 and PM1 , sampled from Greater Athens Area. The authors extracted 20 different PM-bound polycyclic aromatic hydrocarbons (PAH). Zalakeviciute et al. [11] identified 28 different elements in the structure of PM10 , as well as the existence of ions, such as SO4 2− , NO3 − and NH4 + . Gvero et al. [12] measured the daily average PM10 concentrations (µg m−3 ) in Banja Luka city area, with dominant private households, and determined the PM chemical composition. Juda-Rezler et al. [13] studied the PM2.5 for one calendar year and gave the seasonal concentration of PM2.5 , 19 trace elements in PM2.5 , as well as PM2.5 sources apportionment confirmed the primar role of the following emission sources: residential combustion, exhaust traffic emissions and non-exhaust traffic emissions. Chernishev et al. [14] investigated the particle size distribution and the chemical composition, e.g. the fraction of particles with structured carbon (crystalline phase state) in the sampled PM10 , which originated from the exhausts of two-wheeled vehicles that are typically used in the territory of Vladivostok, Russia. According to [14] environmentally persistent free radicals (EPFRs) exist in significant concentration in the atmospheric PM. These EPFRs are primarily emitted from combustion and thermal processing of organic materials, in which the organic combustion by-products interact with transition metal-containing particles to form a free radical-particle pollutant. Vejerano et al. [15] investigated EPFRs in

Chemical Characteristics of Flue Gas Particulates …

215

atmospheric PM2.5 . The experimentally measured hydroxyl radicals (•OH) were generated from aqueous suspension of ambient PM2.5 , using EPR spectrometry and spin trapping. Thus, significant levels of hydroxyl radicals (•OH) were obtained, without the addition of H2 O2 . The relation between outdoor and indoor air pollution is investigated in a pilot study of Antova et al. [16]. The authors experimentally measure the concentration of the main pollutants, identified in indoor air, such as: PM10 and PM2.5 , ozone, carbon monoxide, carbon dioxide, formaldehyde, nitrogen dioxide, volatile organic compounds, moisture and mold presence in school classrooms. The study confirmed that a large percentage of the schoolchildren in the early school age occupy classrooms where both pollutants—carbon dioxide (90% of the classrooms) and fine particulate matter (50% of the classrooms) significantly exceed the recommended concentrations. The pollutants’ concentration can be directly related to ventilation intensity, classroom occupancy, student activity, heating fuel type and building location. The typical urban air pollutants affect the respiratory and cardiovascular systems. The release of high levels of PM to the atmosphere greatly concerns the human health and the environment [17]. Therefore, the European legislation [1] sets limit values for particularly harmful substances, such as PM10 and PM2.5 and many others. However, recent surveys show that PM of submicron size is even more dangerous for humans than the monitored PM10 and PM2.5 . This type of particulates is often defined as UFP and is related to the anthropogenic emission sources [18]. The toxicity of UFP in air pollution mostly comes from inhalation exposure through the respiratory system, causing significant toxic and health effects. The nose and bronchioles cannot filter out UFP efficiently, thus they have high pulmonary deposition efficiency and could reach the deep inside of the lungs [19]. According to Chen et al. [18] and Cheng [20] the human nose filters 80% of 1 nm particles during resting breathing. The UFP (PM0.1 ) generally enter the body through the lungs but translocate to essentially all organs. Compared to fine particles (PM2.5 ), they cause more pulmonary inflammation and are retained longer in the lungs. Their toxicity is increased with smaller size, larger surface area, adsorbed surface material, and the related physical and chemical characteristics of the particles. Exposure to PM0.1 induces cough and worsens asthma [21, 22], and according to Bhardawaj et al. [23] particles greater than 1 µm aerodynamic diameters generally linger on epithelial. Biomass is one of the renewable energy sources that can be utilized in different ways, thus serving as valuable feedstock for the chemical industry. It is highly diverse, with different chemical composition, depending on its origin. One way for utilizing biomass is gasification. In order to obtain valuable products, such as chemicals, syngas, etc., biomass is processed in atmosphere with low presence of oxygen [24, 25]. The gasification can be carried out, using different gasifying agents—O2 , oxygen-depleted air, CO2 , steam or mixtures of these compounds. Usually, moderate to high temperature and pressure, and/or existence of catalyst maximize the yield of the desired products, like CO and H2 [26]. Of scientific, environmental, health and social concern is to study the flue gas, emitted while utilizing concrete biomass types that are annually available in large

216

T. Petrova et al.

quantity. Along with the desired valuable products, resulting from the applied utilization process, harmful by-products are emitted. The flue gases of such processes are typical source of pollutants, including PM of different size and chemical composition. Currently, biomass is gaining significant interest on the Bulgarian energy market, but still detailed research is needed about the emission modes and factors of harmful pollutants, like PM with a particular accent on their physical and chemical characteristics. The present work aims to characterize particulates obtained during biomass gasification in a DTF reactor. In focus were the cyclone particulates with size >10 µm (for simplicity, further in the text they are named char particles) as well as the sampled fine PM and UFP. For that purpose, three types of biomass were gasified, using two different gasifying agents. A particular interest was drawn on the effects of biomass on the PM elemental composition. The elemental composition of the examined particulates was obtained through XRF analysis. In addition, EPR technique was used to study the existence of paramagnetic constituents and the presence of carbon–centered particles in the char. Because this work is still in progress, herein only some preliminary results are presented.

2 Materials and Methods The selected biomass consists of a woody biomass (softwood and bark) and agriculture residue (colza and sunflower husks (SFH)) that are used for production of solid biofuels in the Member State. The investigated biomass is characterized through proximate, ultimate, ash and calorimetric analyses. The results are presented elsewhere [27]. The gasification installation consists of a biomass feeding system, gas supply system (O2 , N2 , CO2 ), vertical DTF reactor, particulate matter sampling system and system for sampling and analyzing the gaseous products. The gasification was carried out at 1000 °C and two different agents—(1) 1% O2 and 99% N2 and (2) 1% O2 , 94% N2 and 5% CO2 . The biomass feed rate (15 g/h) was constant throughout the entire experiment. The PM sampling system consists of two cyclones and a Dekati DLPI 13-stage cascade low pressure impactor. The first cyclone was an in-house made in Instituto Superior Técnico, Lisboa, Portugal [25]. The second one was a commercial cyclone, S110 by Dekati® . The role of both cyclones was to capture the particles with diameter larger than 10 µm. The 13-stage cascade (impactor) collects the particulate matter with sizes between 30 nm and 10 µm, with a 50% efficiency cut-off at 13 aerodynamic diameters of PM. The present study examines the char particles and the PM collected on five filters (with D50% = 0.265 µm (PM0.265 ), D50% = 0.65 µm (PM0.65 ), D50% = 1 µm (PM1 ), D50% = 2.5 µm (PM2.5 ) and D50% = 10 µm (PM10 )). For simplicity, further in the text, the above described particulates are entitled as UFP (PM0.265 , PM0.65 , PM1 ) and PM (PM2.5 and PM10 ).

Chemical Characteristics of Flue Gas Particulates …

217

Energy dispersive X-ray fluorescence spectrometer EDX-720 was used to determine the elemental composition of the sampled particulates. The XRF analysis was carried out at room temperature with regard to chemical elements with atomic numbers from 11 (sodium) to 92 (uranium). The EPR spectra were recorded at room temperature on a JEOL JES-FA 100 EPR spectrometer operating at a frequency of 9.5 GHz (X–band), equipped with a standard TE011 cylindrical resonator. The magnetic field was modulated with a frequency of 100 kHz and amplitude of 0.1 mT. The microwave power was 1 mW. The char samples were placed into a quartz tube (3 mm inner diameter) and were fixed in the cavity center.

3 Results and Discussion 3.1 XRF Analysis The results from the XRF analyses of the flue gas particulates, sampled during biomass gasification, are qualitatively presented in Table 1. It is an overview of the distribution of chemical elements, that were typically identified in the investigated particulates (char, PM and UFP). In general, Fe, Mn and S were present in all studied particulates, whereas P and Ca were found only in the larger particles (char, PM10 and PM2.5 ). Figures 1 and 2 present the relative mass concentration (%) of the obtained oxides, which were conditionally separated in two groups: (a) oxides in macro concentration (above 3%—Fig. 1); (b) oxides in micro concentration (below 3%—Fig. 2). Among the investigated particulate matter, the oxides of Fe, K, Mn, Ca, Cl, S, P were in macro concentration, whereas the oxides of Ti, Co, Ni, Pb, Cu, Zn, Se, Hg were in micro concentration. Similar observations are reported by Veleva et al. [6] in atmospheric PM10 , using EDXRF technique. The authors observed the following macro elements: P, S, Cl, K, Ca, Ti, V, Fe (ng.m−3 ), while the micro elements are: Cr, Mn, Ni, Cu, Zn, Br, Rb, Sr, Zr, Ag, Cd, Sn, Sb, I, Ba, Pb (ng.m−3 ). According to Obernberger et al. [28] the major ash forming elements (Al, Ca, Fe, K, Mg, Na, P, Si, Ti), typically measured in biomass combustion, are responsible for the ash melting, deposit formation and corrosion. The Cl and S are present in high plant concentrations. Both elements are forming submicron sized PM—the UFP, and are well known for causing deposit formation and corrosion. In addition, Cl is known for causing HCl and polychlorinated dibenzo-p-dioxins and/or dibenzofurans, while S—the SOX emissions [28]. In the present work, the effect of biomass elemental and ash composition on the flue gas particulates (of different size) was considered|. According to the ash analysis [27], Ca and K were the dominant ash constituents in all biomass types. Their relative mass distribution in the biomass ash is as follows: CaO is 33.29 wt % (softwood), 20.44 wt % (SFH) and 29.07 wt % (colza); K2 O: 15.76 wt % (softwood), 28.78 wt

O2 /N2

Colza

O2 /N2 /CO2

Colza

Sunflower husks

Softwood

Biomass

Gasifying agent

x

PM2.5

x

PM0.265 x

x

PM0.65 x

x

PM1

PM10

x

PM2.5

Char

x

x

PM0.265 x

x

PM0.65

PM10

x

PM1

Char

x

x

PM0.265

PM2.5

x

PM0.65

x

x

PM1

x

x

PM2.5

PM10

x

PM10

Char

x

Fe

Char

Particulates

x

x

x

x

x

x

x

x

x

x

x

x

x

x

x

x

x

x

K

Table 1 Elemental composition of the examined particulates

x

x

x

x

x

x

x

x

Ca

x

x

x

x

x

x

x

x

x

x

x

x

x

x

x

x

x

x

x

x

x

Mn

x

x

x

x

x

Cl

x

x

x

x

x

x

x

x

x

x

x

x

x

x

x

x

x

x

x

x

x

S

x

x

x

P

x

x

x

x

x

x

x

x

x

x

x

x

x

x

x

x

x

x

x

x

Cr

x

x

x

x

x

x

x

x

x

x

x

x

x

x

x

x

x

x

x

Ti

x

x

x

x

x

x

x

x

x

x

x

x

x

x

x

x

x

x

x

Zn

x

x

x

x

x

x

x

x

x

x

x

x

x

x

x

x

Cu

x

x

x

x

x

x

x

x

x

x

x

x

Ni

x

x

x

x

Co

x

x

x

x

x

Pb

x

x

x

x

x

x

Se

x

Hg

x

Mo

Si

(continued)

x

Pt

218 T. Petrova et al.

Gasifying agent

Sunflower husks

Softwood

Biomass

Table 1 (continued)

x x x x

PM2.5

PM1

PM0.65

PM0.265

x

PM0.265 x

x

PM0.65 x

x

PM1

PM10

x

PM2.5

Char

x

x

PM0.265 x

x

PM0.65

PM10

x

PM1

Char

Fe

Particulates

x

x

x

x

x

x

x

x

x

x

x

x

x

x

K

x

x

x

x

x

x

Ca

x

x

x

x

x

x

x

x

x

x

x

x

x

x

x

Mn

x

x

x

x

x

Cl

x

x

x

x

x

x

x

x

x

x

x

x

x

x

x

S

x

x

x

P

x

x

x

x

x

x

x

x

x

x

x

x

x

x

Cr

x

x

x

x

x

x

x

x

x

x

x

x

x

x

Ti

x

x

x

x

x

x

x

x

x

x

x

x

x

x

x

Zn

x

x

x

x

x

x

x

x

x

x

x

x

x

Cu

x

x

x

x

x

x

x

x

x

Ni

x

x

x

x

Co

x

x

x

x

x

x

x

Pb

x

x

x

x

x

Se

x

Hg

x

Mo

Pt

x

Si

Chemical Characteristics of Flue Gas Particulates … 219

220

T. Petrova et al.

Fig. 1 Content of oxides in macro concentration (XRF-analysis)

% (SFH) and 17.80 wt % (colza). Expectedly, the major char constituent was CaO (50.56–100% of the total CaO), regardless of the biomass and the gasifying agent, followed by K, Fe, S, Mn and Cr. According to Gao et al. [29], the majority (83.9– 97.5%) of Mg and Ca retains in the chars during biomass pyrolysis/combustion. Potassium (K) is present in significant composition in agriculture residue, because it is among the typical plant’s nutrients. It plays significant role for combating plant diseases, as it increases the plants’ resistance to some pathogens; it also increases cell wall thickness; provides greater tissue stiffness and promotes rapid recovery after injury [30]. The overall sunflower and colza growth and yield decrease in case of potassium deficiency in the soil [31, 32]. The sunflower crops are very sensitive to soil K deficiency and the lack of K often results in both low seed yields and oil concentrations. Sunflower plants have higher resistance to drought and salinity stresses when supplied with adequate quantity of K [33]. The soil fertilization also improves the oil quality, in terms of total unsaturated fatty acid and protein in achene of edible sunflower [34].

Chemical Characteristics of Flue Gas Particulates …

221

Fig. 2 Content of oxides in micro concentration (XRF-analysis)

The ashes of the studied biomass types contained considerable amount of P and Mg. They were found to increase in the following order: P2 O5 : 5.78 wt % (softwood) 4 is present in Fig. 5b.

4 Results Figure 6 shows the graph of the time change of the b-value in the studied spatial window. The following parameters are set for the construction in the Zmap program:

Spatial Variation of Precursory Seismic Quiescence Observed …

237

Fig. 5 a Graph of the magnitude-frequency distribution Mc >4 for the selected landfill; b plots of cumulative number of events versus time for test polygon

Fig. 6 Time change of the b—value of earthquakes with M l >3.5 within the studied area; - moment of occurrence of the investigated event

method—Max Curvature; sample window size = 500 number of events; min of events = 50; window overlap (%) = 4; bootstraps = 200; smot plot = 5. The graph of the value of b (Fig. 6) shows a maximum (b = 1.95) around 1998, which suggests increased heterogeneity and reduced voltages [25]. Since 2000 the b-value decreases continuously and in 2003 reaches its minimum b = 1.35. The studied earthquake occurs in a period of decreasing value of b. A significant decrease in the value of b may be associated with an increasing effective level of stress before major earthquakes [10, 13]. In addition to the change in time, the spatial changes in a spatial circle of the parameter b are analyzed. The results are presented in Fig. 7, (a) maximum likelihood estimations and (b) least squares estimations. The spatial fluctuations of b for the studied area were estimated for the period from 1964 to 30.03.2011. Spatial differences in the value of b illustrate variability in plan, and relatively low values can determine the places where an earthquake would most likely occur [26, 27]. The zones with relatively low values of b (0.9–1.2) are clearly delineated and the epicenter of the earthquake falls into them. According to [13] and [28] low values of b indicate that fault stresses accumulate in these zones until the main event is activated. Note that the least squares

238

E. Oynakov et al.

Fig. 7 Spatial distribution of the b-parameter, calculated by the methods: a maximum likelihood; b least squares; —epicentre of the earthquake from 01.04.2011

method gives better results. In order to confirm the result and get a better idea of the area of seismic calm, the radius f the studied area was increased to r = 200 km. A period of reduction from 01.01.2000 to 28.02.2011 of the value of b (Fig. 6) was selected, after which the spatial distribution of b (Fig. 8a) was built again. This result confirms the decrease of the value of b around the epicenter of the studied earthquake for the determined period. The spatial distribution of the a-parameter characterizing the seismic activity before the earthquake showed (estimated for the period from 2000 to 28.02.2011) that the epicenter falls in an area with relatively low seismic activity (a = 7–7,5), but which is close to the zone with relatively high activity (west of the epicenter, a = 8–9) (Fig. 8b).

Fig. 8 a Spatial distribution of the b-parameter, calculated by the methods maximum likelihood for the selected polygon with radius R = 200 km; b Spatial distribution of the a-parameter; — epicenter of the earthquake from 01.04.2011

Spatial Variation of Precursory Seismic Quiescence Observed …

239

Fig. 9 Z-statistics for the studied area in a circle with radius R =100 km, around the epicenter of the earthquake from 01.04.2011; —epicenter of the earthquake

The spatial distribution of the Z-parameter before the earthquake of 01.04.2011 (Fig. 9) is calculated for the same area (R = 100 km), comparing two time periods: 1st period from 30.03.1997 to 30.03.2005 and 2nd period from 30.03.2005 until 30.03.2011. The high (positive) Z-values of the maps can be interpreted as a decrease in the flow rate of seismic events (seismic lull) compared to the first period, and the low (negative) Z-values represent an increase in velocity. Earthquake density and distribution is a critical factor in interpreting Z-value variations. Large areas of constant value could show the same density of earthquakes for different periods of time, may show a homogeneous degree of seismicity in this area. In the Fig. 9 the epicenter falls in an area with relatively high values of (Z = 4–5), which means that the selected period (30.03.2005 to 30.03.2011) before the earthquake is a period of relative seismic lull. These relatively high values of Z = 4–5 show 99% reliability of the result. High values of Z = 5–6 are also observed in the northwestern part of the landfill, which may be due to the low density of earthquakes in this part of the area. To check, the Z-value was calculated in a polygon excluding the northwestern part. In this case, the epicenter falls exactly in the zone of relative seismic lull (Fig. 9, Z ≈ 4). For the verification of the result, the Z-value was calculated in a circle whit radius R = 200 km for first period from 01.012005 to 01.01.2008 and second period from 01.01.2008 to 01.01.2011. In this case, the epicenter falls exactly in the zone of relative seismic lull (Fig. 10a; Z ≈ 4). The percentage comparison simply calculates the changes between the rates of the second and first periods. The range of this function is from −100% (no earthquakes in second period) to infinity (no earthquakes in first period). The percentage change in the area around the epicenter of the event is −51~−52%, which clearly shows that around the epicenter of the earthquake during the second period a zone of relative seismic calm is formed. As long as this percentage change is greater than 30%, a

240

E. Oynakov et al.

Fig. 10 a z-statistics for the studied area in a circle with radius r = 200 km, around the epicentre of the earthquake from 01.04.2011; b percent change of second to first period; c number of events in first period; d number of events in second period; —epicentre of the earthquake

significant reduction in the number of events is in force, which makes it statistically significant. This can also be seen when comparing the spatial distribution of the number of seismic events for the first (Fig. 10c) and the second (Fig. 10d) period. For the first period (01.01.2005–01.01.2008) the epicenter is in an area with 100–102 earthquakes and in the second period (01.01.2008–01.01.2011) the epicenter is in an area with 48–50 earthquakes.

5 Conclusions The temporary change in the b-value for the period 1964–2020 shows a minimum in the value of b (b = 1.35), preceding the earthquake of April 1, 2011 (Ml = 6.2) by about 7 years.

Spatial Variation of Precursory Seismic Quiescence Observed …

241

Large decreases in the value of b possibly associated with increasing effective levels of stress before major earthquakes. These significant decreases in the value of b can lead to an increase in effective stress before major events. An increase in the b-value after these earthquakes may mean an increase in the heterogeneity of the earth’s crust and a decrease in shear stress. The change in the spatial distribution of the b value before the earthquake shows that the area with an abnormally low value of b covers the epicenter of the studied earthquake. These low values of b can be interpreted as a potentially locked or high-stress zone before major earthquakes. The epicenters of the earthquakes are located in areas of relatively high value of the parameter Z ≈ 4.2, which indicates a statistically reliable determination of an area with relatively seismic “calm” before the earthquake. Moreover, detecting these two precursory anomalies in relatively large regions might be in relevant to the preparation zone of the 2011 Crete event. Therefore, a decrease in the value of b and seismic attenuation anomalies can be an indicator of strong stress release and these changes can be interpreted as predictors of strong seismic events in studied region. Acknowledgements This work has been carried out in the framework of the National Science Program “Environmental Protection and Reduction of Risks of Adverse Events and Natural Disasters”, approved by the Resolution of the Council of Ministers No 577/17.08.2018 and supported by the Ministry of Education and Science (MES) of Bulgaria (Agreement No D01-322/18.12.2019).

References 1. Wyss, M., Habermann, R.E.: Precursory seismic quiescence. Pure Appl. Geophys. 126, 319– 332 (1988) 2. Wyss, M., Martirosyan, A.: Seismic Quiescence Before the M 7, 1988. Spitak earthquake, Armenia (1998) 3. Console, R., Montuori, C., Murru, M.: Statistical assessment of seismicity patterns in Italy: are they precursors of subsequent events? J. Seismol. 4, 435–449 (2000) 4. Wiemer, S., Wyss, M.: Seismic quiescence before the landers (M = 7.5) and big bear (M = 6.5), 1992 earthquakes. Bull. Seismol. Soc. Am 84(3), 900–916 (1994) 5. Tsukakoshi, Y., Shimazaki, K.: Decreased b-value prior to the M 6.2 Northern Miyagi, Japan, earthquake of 26 July 2003. Earth Planet. Space 60, 915–924 (2008) 6. Bridges, D.L., Gao, S.S.: Spatial variation of seismic b-values beneath Makushin Volcano, Unalaska Island. Alaska. Earth Planet Sci Lett 245, 408–415 (2006) 7. Nuannin, P., Kulha´nek, O., Persson, L.: Spatial and temporal b-value anomalies preceding the devastating off coast of NW Sumatra earthquake of December 26, 2004. Geophys. Res. Lett. 32, L11307 (2005). https://doi.org/10.1029/2005gl022679 8. Wiemer, S., Wyss, M.: Mapping the frequency-magnitude distribution in asperities: an improved technique to calculate recurrence times? J. Geophys. Res. 102(15), 115–128 (1997) 9. Wyss, M., Stefansson, R.: Nucleation points of recent main shocks in southern Iceland mapped by b-values. Bull. Seismol. Soc. Am. 96, 599–608 (2006). https://doi.org/10.1785/0120040056 10. Wu, Y.M., Chang, C.H., Zhao, L., Teng, T.L., Nakamura, M.: A comprehensive relocation of earthquakes in Taiwan from 1991 to 2005. Bull. Seismol. Soc. Am. 98(3), 1471–1481 (2008)

242

E. Oynakov et al.

11. Gutenberg, B., Richter, C.F.: Frequency of earthquakes in California. Bull. Seismol. Soc. Am. 34, 185–188 (1944) 12. Utsu, T.: A method for determining the value of b in the formula log n a bM showing the magnitude-frequency relation for earthquakes. Geophys. Bull. Hokkaido Univ. 13, 99–103 (1965). (in Japanese with English summary) 13. Schorlemmer, D., Wiemer, S., Wyss, M.: Earthquake statistics at Parkfield, Stationarity of b values. J. Geophys. Res. 109: B12307 (2004). https://doi.org/10.1029/2004jb003234 14. Wiemer, S.: A program to analyse seismicity: ZMAP. Geophys. Res. Lett. 72, 373–382 (2001) 15. Aki, K.: Maximum likelihood estimate of b in the formula log N = a-bM and its confidence limits. Bull. Earthq. Res. Inst. Tokyo Univ. 43, 237–239 (1965) 16. Wiemer, S., Wyss, M.: Minimum magnitude of completeness in earthquake catalogs: examples from Alaska, the western United States, and Japan. Bull. Seismol. Soc. Am. 90(4), 859–869 (2000) 17. Wiemer, S., Wyss, M.: Mapping spatial variability of the frequency–magnitude distribution of earthquakes. Adv. Geophys. 45, 259–302; Wu YM, Chiao LY (2006) Seismic quiescence before the 1999 (2002) 18. Habermann, R.E.: Man-made changes of seismicity rates. Bull. Seism. Soc. Am. 77, 141–159 (1987) 19. Maeda, K., Wiemer, S.: Significance test for seismicity rate changes before the 1987 Chibatoho-oki earthquake (M6.7). Japan. Ann. Geofis 42(5), 833–850 (1999) 20. Damanik, R., Andriansyah Putra, H.E., Zen, M.T.: Variations of b-values in the Indian Ocean— Australian plate subduction in south Java Sea. In: Proceedings of the Bali 2010 International Geosciences Conference and Exposition, Bali, Indonesia, pp. 19–22 (2010) 21. University of Athens- http://dggsl.geol.uoa.gr/en_index.html 22. Taylor, S.R., Denny, M.D.: An analysis of spectral differences between Nevada Test Site and Shagan River nuclear explosions. J. Geophys. Res. Solid Earth 96(B4), 6237–6245 (1991) 23. Woessner, J., Wiemer, S.: Assessing the quality of earthquake catalogues: estimating the magnitude of completeness and its uncertaint. Bull. Seismol. Soc. Am. 95(2), 684–698 (2005). https:// doi.org/10.1785/0120040007 24. Schorlemmer, D., Woessner, J.: Probability of detecting an earthquake. Bull. Seismol. Soc. Am. 98(5), 2103–2117 (2008). https://doi.org/10.1785/0120070105 25. Görgün, E., Zang, A., Bohnhoff, M., Milkereit, C., Dresen, G.: Analysis of Izmit aftershocks 25 days before the November 12th 1999 Düzce earthquake. Turkey. Tectonophysics 474(3–4), 507–515 (2009) 26. Schorlemmer, D., Neri, G., Wiemer, S., Mostaccio, A.: Stability and significance tests for bvalue anomalies: example from the Tyrrhenian Sea. Geophys. Res. Lett. 30(16), 1835 (2003). https://doi.org/10.1029/2003GL017335 27. Westerhaus, M., Wyss, M., Yilmaz, R., Zschau, J.: Correlating variations of b values and crustal deformation during the 1990s may have pinpointed the rupture initiation of the Mw = 7.4 Izmit earthquake of 1999 August 17. Geophys. J. Int. 148, 139–152 (2002) 28. Motaghi, K., Hessami, K., Tatar, M.: Pattern recognition of major asperities using local recurrence time in Alborz Mountains, Northern Iran. J. Seismol. 14, 787–802 (2010). https://doi. org/10.1007/s10950-0109201-z 29. Wyss M., Martirosyan A.H.: Seismic quiescence before the M 7, 1988, Spitak earthquake, Armenia Geophys. J. Int 134(2), 329–340 (1998). https://doi.org/10.1046/j.1365-246x.1998. 00543.x

Earthquake Ground Motion Scenarios for the City of Ruse Dimcho Solakov , Stela Simeonova, Plamena Raykova, Boyko Rangelov, and Constantin Ionescu

Abstract Global seismic risk and vulnerability to earthquakes are increasing steadily as urbanization and development occupy more areas that are prone to effects of strong earthquakes. The assessment of seismic hazard and generation of earthquake scenarios is the first link in the prevention chain and the first step in the evaluation of the seismic risk. In the present study both deterministic and probabilistic earthquake scenarios for the city of Ruse are generated. The study is guided by the perception that usable and realistic, based on both local seismic history and tectonic conditions, ground motion maps have to be produced for urban areas. The consideration of the earthquake scenarios into the policies for seismic risk reduction will allow focusing on the prevention of earthquake effects rather than on the activities following the disasters. Keywords Seismic hazard assessment · Deterministic and probabilistic earthquake scenarios · Intermediate depth earthquakes · GMPE’s · Vrancea seismic source

1 Introduction Earthquakes are the deadliest of the natural disasters affecting the human environment; indeed, catastrophic earthquakes have marked the whole human history. The uncontrolled growth of mega cities in highly seismic areas around the world is often D. Solakov (B) · S. Simeonova · P. Raykova National Institute of Geophysics, Geodesy and Geography, Bulgarian Academy of Sciences (NIGGG-BAS), Sofia, Bulgaria e-mail: [email protected] S. Simeonova e-mail: [email protected] B. Rangelov University of Mining and Geology, Sofia, Bulgaria C. Ionescu National Institute of Research and Development for Earth Physics, Calugareni street, 12, 077125 Magurele, Romania © The Author(s), under exclusive license to Springer Nature Switzerland AG 2021 N. Dobrinkova and G. Gadzhev (eds.), Environmental Protection and Disaster Risks, Studies in Systems, Decision and Control 361, https://doi.org/10.1007/978-3-030-70190-1_17

243

244

D. Solakov et al.

associated with the construction of seismically unsafe buildings and infrastructures and undertaken with an insufficient knowledge of the regional seismicity peculiarities and seismic hazard. The first link in the prevention chain and the first step in the evaluation of the seismic risk are seismic hazard assessment and generation of earthquake scenarios for urban areas. The territory of Bulgaria (situated in the eastern part of the Balkan Peninsula) represents a typical example of high seismic risk area. Over the centuries, Bulgaria has experienced strong earthquakes. Moreover, some of the Europe s strongest twentieth century earthquakes occurred in Bulgaria. Impressive seismic activity developed in the SW Bulgaria in the time period 1904–1906. The seismic sequence started on the 4th of April 1904 with two catastrophic earthquakes within 23 min (the first quake at 10 h 05 min with MW = 7.1 considered as a foreshock and the second one at 10 h 26 min with MW = 7.6 the main shock). Along the Maritsa valley (central part of Southern Bulgaria), in 1928 a sequence of three destructive earthquakes occurred. However, no such large earthquakes occurred in Bulgaria since 1928, which may induce non-professionals to underestimate the earthquake risk. Moreover, the seismicity of the neighboring countries, like Greece, Turkey, former Yugoslavia and Romania (especially Vrancea-Romania intermediate earthquakes), influences the seismic hazard in Bulgaria. In the present study are described the basic tools for estimating the earthquake ground-shaking hazard in urban areas and for producing suitable map representations. The work is a comprehensive earthquake damage scenario study for the city of Ruse. The city is situated in the northeastern part of the country, on the right bank of the Danube in the mouth of Rusenski Lom River. It is the most significant Bulgarian river port. Ruse is an administrative, transport and tourism center of Bulgaria. In the study, earthquake ground motion scenarios for the city of Ruse—one of the largest earthquake-prone cities in Bulgaria are generated using both deterministic and probabilistic approach.

2 Earthquake Scenarios The work on scenarios is guided by the perception that usable and realistic (also in the sense of being compatible with seismic histories of cities that are several centuries long) ground motion maps had to be produced for urban areas. The approaches adopted for developing of ground motion hazard maps (earthquake scenarios)—both deterministic and probabilistic-should include the following stages: 1. 2.

Compilation of regional seismotectonic data base (region within a radius of 150 km surrounding the considered urban area); Analysis of regional seismotectonics (identification of the tectonic feature(s) capable of generating the scenario earthquake with the associated magnitude level, and representation of seismic sources for the different methods of analysis.

Earthquake Ground Motion Scenarios for the City of Ruse

3.

4. 5. 6.

245

Seismo-tectonic zonation, the aim of which is to highlight the relationships between seismicity and active geological structures, forms an indispensable part of seismic-hazard assessment; Geotechnical characterization of the studied area; Seismic hazard assessment; Cross-validation of the local ground-shaking representations obtained through the use of different methods.

A flowchart showing the main ingredients of the approaches for developing earthquake scenarios and their mutual relationships is illustrated in Fig. 1. By deterministic scenario it is meant a representation of the severity of ground shaking over an urban area, using one or more hazard descriptors. Such representation can be obtained: either from the assumption of a “reference earthquake” specified by a magnitude or an epicentral intensity, associated to a particular earthquake source— or, directly, showing values of local macroseimic intensity generated by a damaging, real earthquakes of the past. In the first case, the local ground shaking levels typically need to be evaluated through attenuation relations for the selected parameters. In the second case, the most common situation is that of a single or few intensity values that concisely describes the severity of the effects on the built environment caused by the historical earthquake selected as representative. Perhaps the single most important feature of the approach is that the deterministic scenario must be consistent with the seismic history of the city and with known active faults close to the city. The improvement of the knowledge on the past seismicity is essential to lower the uncertainties in hazard assessments.

Fig. 1 Flowchart showing the main ingredients of the approach for earthquake scenario generation

246

D. Solakov et al.

The probabilistic seismic hazard is the probability that various levels of strong ground motion will be exceeded during a specified time period at a site. The ground motion levels may be expressed in terms of peak ground acceleration (velocity, displacement) and/or peak response spectral amplitudes for a range of frequencies. The analysis methodology is based on the conception that the seismic hazard at a site is a function of three main components: the space geometry of seismic sources, the characteristics and statistics of their seismicity and the characteristics of seismic wave propagation in the region. The resultant hazard at a specified site is obtained by integrating the effects of ground motion from earthquakes of different size occurring at different locations within different seismic source regions and with different frequencies of occurrence. Experience of some countries in damage and loss estimation studies suggests the use of a deterministic approach as a valid support to the decision makers because it improves the ability to predict the amount and location of damage. Furthermore, the deterministic seismic hazard evaluation is preferred in some studies to a probabilistic one because it is deemed to better accounting for the complex geology of the area and associated local site amplification, and for the first-order effects on ground motion expected in the near field of a large seismic source. When both deterministic and probabilistic results are obtained, deterministic assessments can be used as a check against probabilistic assessments in terms of the reasonableness of the results, particularly when small annual frequencies of exceedance are considered. The probabilistic results allow deterministic values to be evaluated within a probabilistic framework.

3 Regional Seismo-Tectonic Setting The tectonic situation in the Eastern Mediterranean is dominated by the collision of the Arabian and African plates with the Eurasian [1–3]. The models of movement of the continental plates indicate that the Arabian plate is moving in North-Northwest direction relative to Eurasia with velocity of about 18–25 mm/y. The African continental plate is moving in north direction relative to Eurasia with velocity of about 20–25 mm/y (among others in [4]). The leading edge of the African plate is subducted on the Greek arc under the Eurasian plate. The recent tectonics of Bulgaria is determined by the geotectonic of the region in which dominate the processes of extension with general direction north–south. Analysis of the obtained horizontal velocities of the permanent GNSS stations and geodynamic networks shows that the horizontal velocities increase from 1 to 2 mm/y in the region of Northern and Eastern Bulgaria till 10 mm/y in the region of the Chalkidiki Peninsula (illustrated in Fig. 2). On the territory of Northern Bulgaria, the horizontal velocities are practically negligible which confirms that this region belongs to the Eurasian plate while Central Western and Southwestern Bulgaria belong to the so-called South Balkan Extensional

Earthquake Ground Motion Scenarios for the City of Ruse

247

Fig. 2 Horizontal velocity field relative to stable Eurasia in the region of the Eastern Mediterranean (modified from [4])

Region—a transition zone between Eurasian continental plate and Aegean microplate to the south and the Anatolian in the southeast (among others in [4]). The seismic hazard for the city of Ruse is controlled by the Vrancea intermediatedepth source in Romania that is located at about 150 km to the N-NW from the city (among others in [5, 6]). The Vrancea seismogenic zone of Romania is a very peculiar seismic source, often described as unique in the world, and it represents a major concern for most of the northern part of Bulgaria. The events generated in this seismogenic zone are characterized by relatively deep hypocentres and wide area of macroseismic impact. In this area, strong intermediate focused earthquakes are being realized with depth 90–230 km. The strongest known events, occurred in the Vrancea seismogenic zone are the following earthquakes: the 1802 quake with magnitude MW = 7.9 [7], the 1940 MW = 7.7 (ROMPLUS catalogue available at https://web.infp.ro/#/romplus), and the 1977 quake, MW = 7.5 (according GCMT catalogue—[8, 9]). Situated at distances larger than 200 km from the Vrancea zone, several cities in the northern Bulgaria suffered many damages due to high energy Vrancea intermediate-depth earthquakes (as illustrated in Fig. 3).

248

D. Solakov et al.

Fig. 3 Damages in Bulgaria caused by the 1977 (MW = 7.5) Vrancea earthquake

4 Soil Properties The city of Ruse is situated in a flat area. The area is characterized by a high level of groundwater. Under dynamic loads of different intensity, the soil base can be subjected to liquefaction, sand volcanism and local landslides. The representation

Fig. 4 Average shear-wave velocity in the upper 30 m of the soil/rock profile for the city of Ruse (from [10])

Earthquake Ground Motion Scenarios for the City of Ruse

249

Fig. 5 Global Slope-Based Vs30 model for the city of Ruse, according to USGS

of the soil properties of the city (Fig. 4) was defined by the engineering parameter Vs30 —average shear-wave velocity in the upper 30 m of the soil/rock profile. The Vs30 values for the city of Ruse, presented in Fig. 4 are based on the results derived in the frame of the project “Environmental Protection and Reduction of Risks of Adverse Events and Natural Disasters". Four blocks with different properties are outlined throughout the city. Figure 4 shows that the values of Vs30 slightly vary (from 190 to 550 m/s) throughout the territory of the city of Ruse [10]. Additionally, the soil properties of the city of Ruse were defined by the engineering parameter Vs30 that is obtained from the USGS slope-based global map for the city (https://earthquake.usgs.gov/data/vs30). The territory of the city is represented by several points with varying values for Vs30 between 184 and 582 m/s, presented in Fig. 5.

5 Selection of the Ground Motion Attenuations (GMPE’ S) An essential element in seismic hazard analyses is the ability to estimate strong ground motion from a specified set of seismological parameters. This estimation is carried out using a Ground Motion Prediction Equation (GMPE), or what is also referred to as an attenuation relation. A GMPE is a mathematical equation that relates a given strong-motion parameter to one or more parameters of the earthquake rupture, wave propagation path and local site conditions, referred to as seismological parameters. GMPE’s are essential in seismic hazard studies, which are a key step for the evaluation of seismic risk and loss estimation for a region.

250

D. Solakov et al.

The selection process of adequate GMPE’s started from a comprehensive list of available ground motion models to which a set of exclusion criteria was applied according to the similarity of the tectonic environment, clarity of the underlying datasets, frequency ranges of the predictive equations, the regression model, the definition of the prediction variables, international recognition, etc. Six attenuation relationships for shallow earthquakes (CB14-[11]; AS14-[12]; BA14-[13]; CY14-[14]; AB14-[15]; CF15-[16]) and 3 for intermediate depth quakes (BC Hydro-[17]; Y97-[18]; VAC-[19]) were selected on the base of general criteria [20]. The suitability of selected GMPE’s for shallow and intermediate depth earthquakes is tested on the base of a large dataset of ground motion data from Balkan countries and Italian strong motion networks, and System of Accelerographs for Seismic Monitoring of Equipment and Structures deployed in Kozloduy NPP. The dataset is composed of 424 three-component accelerograms from 54 in magnitude range 3.7 ≤ MW ≤ 6.9 shallow earthquakes occurred in Balkan region and Italy, and 8 moderate-to-large intermediate depth earthquakes (as presented in [20]). The method introduced by [21] is chosen for testing the models against the observational data. Scherbaum provides a ranking criterion based on information theory. This technique is based on the probability for an observed ground motion to be realized under the hypothesis that a model is true. It provides one value, the negative average log-likelihood LLH that reflects the fit between data and model: LLH = −

N 1  log2 (g(xi ), N i=1

(1)

where N is the number of observations xi , and g is the probability density function predicted by the GMPE’s (normal distribution). The selected models provide low LLH values, less than or equal to 2.0 implying a good fit with the observational data (as seen from Tables 1 and 2). In the last column of the tables are given the weights of the GMPE, that are used in probabilistic hazard assessment. Table 1 The LLH values for the selected six attenuation models for shallow earthquakes

All—251 observations

Mw > 4.5—137 observations

Weight

GMPE’s

LLH

LLH

CB14

1.734

1.714

0.18

AS14

1.835

1.888

0.167

BA14

1.846

1.879

0.166

CY14

1.685

1.636

0.186

AB14

1.863

1.957

0.154

CF15

1.898

2.028

0.147

Earthquake Ground Motion Scenarios for the City of Ruse Table 2 The LLH values corresponding to the selected three attenuation models for intermediate depth earthquakes

251

All—173 observations

Mw > 6—97 observations

Weight

GMPE’s

LLH

LLH

BCHydro

1.307

1.272

0

Y97

1.423

1.252

0

VAC

1.343

1.073

1

The lowest LLH values are estimated for GMPE’s CY14 for shallow and BCHydro for intermediate depth earthquakes that are 1.685 and 1.307, respectively. The GMPE VAC [19] is the best fit to observations for strong Vrancea earthquakes-the lowest LLH value of 1.073 [20]. The GMPE VAC relationship is used in the calculations.

6 Deterministic Earthquake Scenario for the City Ruse A deterministic earthquake scenario in macroseismic intensity is generated for the city of Ruse (presented in [10]). The scenario is based on both seismic histories of the city and tectonic environment. Seismic history of Ruse shows that the hazard for the city is mainly influenced by the intermediate depth quakes occurred in the region of Vrancea (Romania). The strongest documented seismic impacts on the city of Ruse until nowadays are from the 1940 and the 1977 quakes, that are generated in Vrancea seismogenic zone in Romania. Distribution of macroseismic effects along the city is estimated separately for each quake on the base of the available observations and documented damages generated by the two Vrancea earthquakes. The intensity map for 1977 quake is presented in Fig. 6. The observed distribution of intensity function along the city of Ruse in both cases is identical (though a scant information about the earthquake of 1940). Impacts of the both earthquakes are with the highest intensity in coastal parts of the city and strong effects are observed in the western and central parts of the city of Ruse. The soil properties of the Ruse municipal area were incorporated in the hazard evaluation by using the map of Vs30 values for the city of Ruse that is presented in Fig. 4. The ground motions in PGA are calculated using the GMPE VAC [19]. The estimated peak ground accelerations are converted into intensity applying the following relations [22]: I = 3.146 log(PGAmax ) + 0.375

(2)

I = 3.058 log(PGAgm ) + 0.731

(3)

252

D. Solakov et al.

Fig. 6 Observed macroseismic impacts on the city of Ruse of the 1977 Vrancea earthquake (among others in [5])

where PGAmax is the maximum of two horizontal components, PGAgm is the geometric mean of two horizontal components in cm/s. Figure 6 illustrates observed macroseismic intensities (EMS) for the city of Ruse for Vrancea earthquake with MW = 7.5 at depth 94 km and the hypocentral distance 240 km. The generated scenario earthquake (presented in Fig. 7) is consistent with the conclusion reached after the 1977 Vrancea earthquake—the consequences decrease with distance from the Danube River [6, 23]. Vrancea earthquakes that affected the city of Ruse with an intensity of 5 and higher are presented in Table 3. In the Table 3 are presented the following parameters: the predicted PGAgm , the predicted and observed intensity I (in MSK), the available observed horizontal accelerations at a Ruse site and difference between observed and predicted parameter values. The data presented in Table 3 show a good agreement between observed and predicted intensities. The predicted intensities that are based on the observed peak ground acceleration for 30.5.1990 earthquake—6.9 and 6.8, respectively (presented in columns 5 and 6) overestimate the observed intensity—6 (column 3).

Earthquake Ground Motion Scenarios for the City of Ruse

253

Fig. 7 Deterministic earthquake scenario for the city of Ruse in intensity

Table 3 Observed and predicted ground motions from intermediate depth Vrancea earthquakes Date

MW

Iobs*

13.9.1903

6.3

5

6.10.1908

7.1

5

22.10.1940

6.5

5

10.11.1940

7.7

4.3.1977

7.5

30.8.1986 30.5.1990 * From

PGAgm pr

PGAmax obs** /Ipr

PGAgm obs** /Ipr

Ipr

Iobs-Ipr

29.4

5.2

−0.2

72.5

6.4

−1.4

29.7

5.2

−0.2

7

94.9

6.8

0.2

7

103.2

6.9

0.1

7.23

6

79.9

75.5/6.3

6.5

−0.5

6.95

6

53.3

114.5/6.9

6.0

0.0

100.1/6.8

[5] and [24], ** From [25]

7 Probabilistic Seismic Hazard Evaluation for the City of Ruse PSHA was developed in the late 1960s and early 1970s at the Universidad National Autonoma de Mexico (UNAM) and the Massachusetts Institute of Technology (MIT) and PSHA has now become the most widely used approach for estimating seismicdesign loads. Probabilistic approaches were developed to provide a more systematic method to deal with the “uncertainty in the number, sizes, and locations of future earthquakes” [26] than the arbitrary selections used previously [27]. Probabilistic estimates of

254

D. Solakov et al.

design ground motion levels are derived directly from the hazard analysis. In probabilistic terms seismic hazard is defined as the likelihood that various levels of ground motion will be exceeded at a site during a specified time period. The constituent models of the Probabilistic Seismic Hazard Methodology are the following models: (1) seismic source model; (2) earthquake recurrence frequency; (3) ground motion attenuation models; and (4) ground motion occurrence probability at a site [28]. In the present study the PSHA was performed by using a version of machine code EQRISK [29] that is applied in seismic hazard assessment for Bulgaria (among others in [4]). The main difference from the original code consists in using calculation procedures for coordinate transformation and distance integration presented in [30, 31].

7.1 Seismic Source Model A key component of seismic hazard assessment is the creation of seismic source model, which demands translating seismotectonic information into a spatial approximation of earthquake location and recurrence. The seismicity in 150 km region surrounding the city of Ruse is associated within 12 seismic sources. The regional seismic source model used in seismic hazard analysis for the city of Ruse is illustrated in Fig. 8 and specified in Table 4. In this particular

Fig. 8 Seismic source model (modified from [4])

Earthquake Ground Motion Scenarios for the City of Ruse

255

Table 4 Seismic source parameters used in the hazard analysis Source

Gr*

Mmax

Mmax

a

b

Observed

Estimated

1

1.83

0.8

5.2

7.0

2

2.52

0.8

6.0

6.9

3

1.84

0.8

4.4

7.0

4

2.18

0.8

6.8

7.1

5

2.25

0.8

5.0

7.1

6

1.84

0.8

5.7

6.2

7

2.03

0.8

6.3

7.2

8

2.02

0.8

4.9

6.2

9

3.66

1

6.4

6.9

10

2.9

0.94

5.4

6.5

11

3.81

0.9

6.8

7.5

12

4.5

0.85

7.9

8

Gr*—estimated parameters of the magnitude frequency relationship (the Gutenberg-Richter relationship) Mmax —estimated maximum magnitude

study, a seismic source model that is derived from the seismic zoning of Bulgaria [4] is used. The model includes all seismic sources that substantially influence the seismic hazard of the city of Ruse. The compiled model consists of areal and mixed type sources (including area and fault sources with different minimum and maximum magnitudes). For each seismic source are defined: geometry, earthquake distribution in the source, earthquake recurrence frequency and the maximum potential earthquake magnitude. All model parameters for PSHA are summarized in Table 4. The areal sources define regions that are assumed to have uniform seismicity characteristics distinct from neighboring zones and are exclusive of active faults that are identified [28]. Those source type displays lows to moderate disperse seismicity. A mix-source type is a seismic source in which one or more active faults are defined within a boundary of area source [28]. For each of them two separate sets of minimum and maximum magnitudes (one for the area source and the other for fault source) are determined. The area source is defined as a region of the background seismicity in which earthquakes up to maximum magnitude of 6.0 are modeled as uniformly distributed random events, with recurrence frequency statistically determined on the base of the catalogue of earthquakes in the area. The largest earthquakes (M > 6.0) are modeled as occurring on the faults (narrow area) defined inside the area source with recurrence frequency determined for the area source. The maximum magnitudes for fault sources are presented in Table 4.

256

D. Solakov et al.

7.2 Seismicity Parameters The seismicity statistics are defined by specifying the magnitude-frequency relation with the annual seismic rate parameter, λ, (number of earthquakes with magnitudes between Mmin and Mmax ), the b-value of the Gutenberg-Richter exponential relation, the minimum magnitude, Mmin and the maximum possible magnitude, Mmax . Parameters of cumulative frequency-magnitude relation, Mmin (minimum magnitude used for determination of earthquake frequency occurrence) and Mmax (maximum observed magnitude) for each seismic source in the region (within a radius of 150 km) surrounding the city of Ruse that were used in the seismic hazard analysis are presented in Table 4 (details are presented in [4]). Earthquake frequency throughout the 150 km region surrounding the city of Ruse is determined using the cumulative form of Gutenberg-Richter relationship of occurrence frequencies: Log N(M) = a − b M

(4)

where: N(M)-number of earthquakes with magnitude equal to or greater than M occurring in a seismic source per unit time, the parameter a (10a-the total number of earthquakes with M > 0) and b (rate of seismicity) are constant that are determine on the base of the compiled catalogue. The b-value of the Gutenberg-Richter relationship is estimated using the following two approaches: standard least square method (LSQ) applied to the cumulative number of earthquakes [32]; and the maximum likelihood estimate (MLH) of bvalue for discrete data with different periods of completeness of the catalogue for different magnitudes [33]. The following relation is used for parameter a (of Gutenberg-Richter relationship) estimation:   N (M1 ≤ M ≤ M2 ) = 10(a−bM1 ) 1 − 10(−b(M2 −M1 ))

(5)

It is assumed that the catalogue used in the computation is complete for magnitude larger than or equal to M1 (all earthquakes with M ≥ M1 are reported). The a value of Gutenberg-Richter relationship is estimated for MW (moment magnitude). The limiting size earthquake that can occur in each seismic source is a very important parameter in seismic hazard analysis, especially at low probability levels. Maximum earthquake potential (Mmax ) can be evaluated on the base of the following data: dimensions of the structure and slip rate, maximum historical earthquake, paleoseismological data and earthquake frequency. In practice professional judgment of experts for evaluation of the maximum earthquake potential is applied. The expert judgments are based on available geological, geophysical and seismological information for the region. Seismicity parameters a, b and Mmax have been considered in the PSHA Logic tree.

Earthquake Ground Motion Scenarios for the City of Ruse

257

7.3 Treatment of Uncertainties The epistemic variability (uncertainties) comes from statistical or modeling variations. The large uncertainties in seismic hazard result from lack of knowledge about earthquake cause, characteristics, ground motions, i.e. from uncertainties in seismic input. Uncertainties are expressed as a confidence level for the hazard results. Modern methods of seismic hazard analysis incorporate uncertainties into the analysis to assess their impact on the estimate of the expected level of seismic hazard as well as the uncertainty in that estimate. The commonly accepted procedure of treating uncertainty is to postulate a set of hypotheses. A probability value (weight) is assigned to each hypothesis, based on analyst’s degree of belief and expert opinion. A seismic hazard curve representing the annual frequency of exceeding a specified ground motion parameter is generated for each hypothesis. Thus, the result of uncertainty in seismic input is a family of hazard curves representing a range of discrete alternatives. Logic trees approach [34, 35] is used to treat uncertainties in input assumptions. The logic tree formulation for seismic hazard analysis involves specifying discrete alternatives for states of nature or parameter values and specifying the relative likelihood that each discrete alternative is correct value of the input parameter. In the present study for accounting the epistemic uncertainties in the seismic input a logic tree approach is used in the seismic hazard assessment for the city of Ruse. For the PSHA analysis of the city of Ruse, the following 4 logic tree levels have been used (details are presented in [4]): Level 1—Seismic source model In the present study seismic source model presented in Fig. 8 is considered; Level 2—Seismic sources A logical tree is generated for each seismic source. Four sublevels: Parameters of the magnitude frequency relationship; Maximum magnitude; Fault plane solution; Dip angle. Level 3—Ground Motion Prediction Equations (GMPEs) Two sublevels: 1. 2.

GMPEs for shallow earthquake with weights that are presented in Table 2; The GMPE VAC for intermediate depth earthquakes with the lowest LLH value for strong Vrancea earthquakes—weight 1;

258

D. Solakov et al.

Level 4—Soil properties of the city of Ruse (defined by the engineering parameter V s30 ) Two sublevels—the VS30 maps for the city of Ruse that are presented in Figs. 4 and 5.

7.4 Probabilistic Earthquake Scenarios for the City of Ruse Probabilistic seismic analysis (PSHA) for the city of Ruse was performed using the model of seismic sources presented in Fig. 9 and specified in Table 4. Values (in [g]) of the ground motion parameters were calculated using six GMPE’s for shallow earthquakes that are presented in Table 1 and VAC relationship for intermediate depth earthquakes (Table 2). For propagating the epistemic uncertainties through the probabilistic seismic hazard analysis 4 level logic tree for each seismic source is developed. The probabilistic scenario maps have been generated in terms of Peak Ground Velocity (PGV), Peak Ground Acceleration (PGA) and response spectral accelerations, SA(T) for two values of the vibration period T-0.3 and 1.0 s. for 475 and 95 years return periods. Figures 9 and 10 display the seismic hazard for a 475 years return period expressed in PGA and PGV, respectively.

Fig. 9 Probabilistic earthquake scenario for the city of Ruse in PGA

Earthquake Ground Motion Scenarios for the City of Ruse

259

Fig. 10 Probabilistic earthquake scenario for the city of Ruse in PGV

The estimated PGA and PGV for the city of Ruse are varying between 0.175– 0.25 g and 28.5–30.5, respectively (as seen in Figs. 9 and 10). The generated probabilistic earthquake scenarios support the conclusion that the consequences decrease with distance from the Danube River [6, 23]. To compare the results of the deterministic and probabilistic analysis, the parameters evaluated by the two methods (presented in Table 5) are considered. As can be seen from the table, the deterministic values are within the range of probabilistic ones for 95 years return period. Moreover, deterministic estimates indicate that the probabilistic results are reasonable. Table 5 Deterministic and probabilistic estimates for the city of Ruse Deterministic PGA (g)

0.11

PGV (cm/s)

Probabilistic for return period 475 years

Vrancea*

0.10–0.14

0.17–0.25

77–86

12–14

28–31

94–99

SA0.3 (g)

0.25

0.21–0.34

0.40–0.64

85–92

SA1.0 (g)

0.09

0.05–0.11

0.13–0.22

78–99

* Vrancea

12.2

Probabilistic for return period 95 years

intermediate earthquakes contribution to the 475 years hazard in %

260

D. Solakov et al.

8 Conclusion The main conclusions from this study can be summarized as follows: Estimation of earthquake scenarios for a city is one of the most important challenges in the field of seismology. Moreover, it is a key step for the evaluation of seismic risk and loss estimation for a city. Both deterministic (in macroseismic intensity) and probabilistic in (PGA, PGV and of response spectral accelerations) earthquake scenarios are developed for the city of Ruse. The probabilistic scenarios are generated for 95 and 475 years return periods. Vrancea intermediate source dominates significantly the 475 years probabilistic hazard. The deterministic estimates coincide with the probabilistic ones for 95 years return period; The most endangered areas in the city of Ruse are identified in the present study. Important management decisions regarding planning and need for research related to destruction and casualty prevention in the event of a strong earthquake can be made, considering the results of the study. Acknowledgements The present study has been carried out in the framework of the National Science Program “Environmental Protection and Reduction of Risks of Adverse Events and Natural Disasters”, approved by the Resolution of the Council of Ministers № 577/17.08.2018 supported by the Ministry of Education and Science of Bulgaria (Agreement № DO-230/06-12-2018).

References 1. McKenzie, D.P.: Plate tectonics of the mediterranean region. Nature 226, 239–243 (1970) 2. Jackson, J.: McKenzie, D: The relationship between plate motions and seismic moment tensors, and the rates of active deformation in the Mediterranean and Middle East. Geophys. J. 93, 45–73 (1988) 3. Jackson, J.: Partitioning of strike-slip and convergent motion between Eurasia and Arabia in eastern Turkey. J. Geophys. Res. 97, 12471–12479 (1992) 4. Solakov D., Simeonova S., Trifonova P., Georgiev I., Raykova P., Metodiev M., Aleksandrova I.: Building Seismic Risk Management, Part 2: Regional Seismotectonic Model and Model of Seismic Sources. BAS Publ. House, S., 21–45 (2019). (In Bulgarian). 5. Simeonova, S., Aleksandrova, I., Solakov, D., Popova, I., Georgieva, G.: Observed macroseismic effects from intermediate Vrancea, Romania earthquakes (1940, 1977) on the territory of the town of Rousse. Proc. Geosci. 323–326 (2006). (Sofia) 6. Solakov, D., Simeonova, S., Aleksandrova, I., Popova, I., Georgieva, G.: Earthquake Scenarios: cases study for the cities of Ruse and Vratsa. In: Proceedings of 5th Congress of Balkan Geophysical Society—Belgrade, Serbia 10–16 May 2009, cp-126-00083 (2009).https://doi. org/10.3997/2214-4609-pdb.126.6497 7. Oncescu, M.C., Mârza, V.I., Rizescu, M., Popa, M.: The Romanian earthquake catalogue between 6 984–1996. In: Wenzel, F., Lungu, D., Novak, O. (eds.) Vrancea earthquakes: tectonics, hazard and 7 risk mitigation, pp. 43–49. Kluwer Academic Publishers, Dordrecht (1999) 8. Dziewonski, A., Chou, T., Woodhouse, J.: Determination of earthquake source parameters from waveform data for studies of global and regional seismicity. J. Geophys. Res. 86, 2825–2852 (1981). https://doi.org/10.1029/JB086iB04p02825

Earthquake Ground Motion Scenarios for the City of Ruse

261

9. Ekström, G., Nettles, M., Dziewonski, A.M.: The global CMT project 2004–2010: Centroidmoment tensors for 13,017 earthquakes. Phys. Earth Planet. Inter. 200–201, 1–9 (2012). https:// doi.org/10.1016/j.pepi.2012.04.002 10. Dimcho, S., Simeonova, S., Raykova, P., Rangelov, B., Ionescu, C.: Earthquake scenarios for the city of Ruse. In: Proceeding of 1st International conference on Environmental protection and disaster Risks, vol. 2, pp. 371–380 (2020). ISBN 978-619-7065-38-1. https://doi.org/10. 48365/envr-2020.1.34 11. Campbell K. W. and Y. Bozorgnia: NGA-West2 ground motion model for the average horizontal components of PGA, PGV, and 5%-damped linear acceleration response spectra. Earthq. Spectra 30(3), 1087–1115 (2014). https://doi.org/10.1193/062913EQS175M 12. Abrahamson, N., W. Silva, R. Kamai: Summary of the ASK14 ground motion relation for active crustal regions. Earthquake Spectra, 30(3):1025–1055 (2014). https://doi.org/10.1193/ 070913EQS198M 13. Boore, D.M., Stewart, J.P., Seyhan, E., Atkinson, G.M.: NGA-West 2 equations for predicting PGA, PGV, and 5%-damped PSA for shallow crustal earthquakes. Earthq. Spectra 30(3), 1057– 1085 (2014). https://doi.org/10.1193/070113EQS184M. 14. Chiou, B.S.J., Youngs, R.R.: Update of the Chiou and Youngs NGA model for the average horizontal component of peak ground motion and response spectra. Earthq. Spectra 30(3), 1117–1153 (2014). https://doi.org/10.1193/072813EQS219M. 15. Akkar, S., Sandikkaya, M.A., Bommer, J.J.: Erratum to: empirical ground-motion models for point and extended-source crustal earthquake scenarios in Europe and the Middle East. Bull. Earthq. Eng. 12(1), 389–390 (2014). https://doi.org/10.1007/s10518-013-9508-6. 16. Cauzzi, C., Faccioli, E., Vanini, M., Bianchini, A.: Updated predictive equations for broadband (0:01–10 s) horizontal response spectra and peak ground motions, based on a global dataset of digital acceleration records. Bull Earthq Eng 13(6):1587–1612 (2015b). https://doi.org/10. 1007/s10518-014-9685-y. 17. Abrahamson, N., Gregor N., Addo, K.: BC Hydro ground motion prediction equations for subduction earthquakes. Earth. Spectra 32(1), 23–44 (201, 2016, February 6) 18. Youngs, R.R., Chiou, S.J., Silva, W.J., Humphrey, J.R.: Strong ground motion attenuation relationships for subduction zone earthquakes. Seism. Res. Lett. 68 (1), 58–73 (1997) 19. Vacareanu, R., Radulian, M., Iancovici, M., Pavel, F., Neagu, C.: Fore-arc and back-arc ground motion prediction model for Vrancea intermediate depth seismic source. J. Earthq. Eng. 19(3), 535–562 (2015) 20. Solakov, D., Simeonova, S., Raykova, P., Oynakov, E., Aleksandrova, I.: GMPEs used in seismic hazard assessment for Bulgaria-selection and testing in Bulgaria. In: Conference. In: Proceedings of 10th Congress of the Balkan Geophysical Society, pp. 1–5 (2019, September).https:// doi.org/10.3997/2214-4609.201902658 21. Scherbaum, F., Delavaud, E., Riggelsen, C.: Model selection in seismic hazard analysis: an information-theoretic perspective. Bull. Seism. Soc. Am. 99(6), 3234–3247 (2009) 22. Ardeleanu, L., Neagoe C., Ionescu C.: Empirical relationships between macroseimic intensity and instrumental ground motion parameters for the intermediate-depth earthquakes of Vrancea region. Romania Nat. Hazards (2020). https://doi.org/10.1007/s11069-020-04070-0 23. Brankov, G., Bonchev, E., Ignatiev, N., Boncheva, H., Christoskov, L., Paskaleva, I.: Vrancea earthquake in 1977. Its after-effects in the People’s Republic of Bulgaria. Publishing House of the Bulgarian Academy of Sciences, Sofia, 428 p. (1983). (In Bulgarian) 24. Alexandrova, I.: Macroseismic field modelling for the territory of Bulgaria. Ph.D. thesis, 77p. (2015). (In Bulgarian) 25. Report GPhI: Seismic zoning of the Republic of Bulgaria, in accordance with the requirements of Eurocode 8 (interim report), pp 207 (2007) (in Bulgarian) 26. Cornell, C.: Engineering seismic risk analysis. BSSA 5, 1583 (1968) 27. Bommer, J., Abrahamson, N.: why do modern probabilistic seismic-hazard analyses often lead to increased hazard estimates? Bull. Seismol. Soc. Am. 96(6), 1967–1977 (2006) 28. Thenhaus P., Campbell, K.: Seismic hazard analysis. In: Chen, W., Scawthorn, C. (eds.) Earthquake Engineering Handbook, pp. 8-1–8-50. CRC Press, Boca Raton, Florida (2003)

262

D. Solakov et al.

29. McGuire, R.: FORTRAN Computer Program for Seismic Risk Analysis, U.S. Geol. Surv. Open-File Rep. 76–67, U.S., Rep. 1–90 (1976) 30. Bender, B., Perkins, D.: SEISRISK II: A computer program for seismic hazard estimation. Open file report, USGS, pp. 82–293 (1982) 31. Bender, B., Perkins, D.: SEISRISK III: A computer program for seismic hazard estimation. USGS Bull. 1772 (1986) 32. Gutenberg, B., Richter, C.F.: Frequency of earthquakes in California. Bull. Seism. Soc. Am. 34, 185–188 (1944) 33. Weihert, D.: Estimation of earthquake recurrence parameters for unequal observation periods for different magnitudes. Bull. Seism. Soc. Am. 70, 1337–1346 (1980) 34. Kulkarni R., Youngs, R., Coppersmith, K.: Assessment of confidence intervals for results of seismic hazard analysis. In: Proceedings of the 8 World Conference on Earthquake Engineering, San Francisco, California, vol. 1, pp. 263 (1984) 35. Coppersmith K., Youngs, R.: Capturing uncertainty in probabilistic seismic hazard assessments within intraplate environments. In: Proceedings of the 3rd National Conference on Earthquake Engineering, Charleston, August 24–28, vol. I, pp. 301 (1986)

Precipitation Chemistry in Bulgaria During Saharan Dust Outbreaks Emilia Georgieva , Elena Hristova , and Blagorodka Veleva

Abstract The objective of this work is to investigate the influence of Saharan dust events on the chemical composition of rain samples collected at three sites in Bulgaria during 2017–2018. Saharan dust intrusions were identified through a combination of satellite retrieved aerosol data and results from dust forecasting models and from backward trajectory model. The chemical composition of the samples (acidity pH, conductivity EC, main ions and elements) is analysed in view of the direction of the approaching air masses—“direct” influence (south-west), and “indirect” influence from other directions and regions, already impacted by Saharan dust. The samples were characterised by pH from 4.1 to 7.4, elevated values for EC (max 202 µS cm−1 ) and for Si, Ca, Fe, Mg concentrations. For cases with direct influence Si and Ca values were up to 1.5 and 25 mg l−1 . In most of the indirect cases increased concentrations of sulphate, nitrate and ammonium were observed (up to 39.5, 23.1 and 8.3 mg l−1 ). Keywords Precipitation chemistry data · Saharan dust · Satellite AOD

1 Introduction Sand and Dust Storms are recognized as hazardous meteorological events that impact the society in many ways—soil and agriculture, ecosystems, air quality and human health, aviation, visibility, solar power production and other socio-economical activities. These events present a unique form of natural hazard in that the source and the impact regions can be separated by great distances [1]. Mineral dust particles can be lifted by strong winds from bare dry soils into the atmosphere and being transported E. Georgieva (B) · E. Hristova · B. Veleva National Institute of Meteorology and Hydrology, 66, Tsarigradsko Shose Blvd, 1784 Sofia, Bulgaria e-mail: [email protected] E. Hristova e-mail: [email protected] B. Veleva e-mail: [email protected] © The Author(s), under exclusive license to Springer Nature Switzerland AG 2021 N. Dobrinkova and G. Gadzhev (eds.), Environmental Protection and Disaster Risks, Studies in Systems, Decision and Control 361, https://doi.org/10.1007/978-3-030-70190-1_18

263

264

E. Georgieva et al.

downwind affecting regions hundreds to thousand kilometers away. It is estimated that between 1000 and 3000 Tg of mineral dust is uplifted into the atmosphere annually, with Saharan desert being the largest global contributor [2]. The atmosphere of the Mediterranean Basin is highly influenced by the Sharan Dust intrusions, as two of the main transport paths of the emitted dust particles is northward to Europe, and eastward to Middle East [3]. The airborne dust particles are removed from the atmosphere by settlement (dry deposition) or are washed out by rains (wet deposition). The last mechanism is prevailing for fine grained particles and distances far-away from the source regions [4]. There are various effects of the deposited dust particles: they are capable of modifying the soil properties, can act as fertilizes in marine ecosystems [5] and can neutralize atmospheric acidity and reduce acid rains, [6]. Although Saharan dust events are detected with higher frequency in the Mediterranean countries and southern Europe, other northern and central parts of the continent are also influenced [7, 8]. The Balkans are not regarded as a dusty region, but the location in the so called D1B zone of the Saharan dust-fall map, [4], implies that Saharan dust can be incorporated in the soil system and change its structure. The frequency of Saharan Dust outbreaks in Bulgaria is about 20% over annual days, as estimated by 10 years data for particulate matter concentrations with diameter less than 10 µm (PM10) at the regional background station Rozhen in the southern part of the country, [9]. The maximum of Saharan outbreaks towards Bulgaria is in spring and autumn, as found in a recent study based on satellites data for the period 2005–2018, [10]. This study indicates that on average the days with Saharan outbreaks are 10–13 for the months March, April, May, and can be 20 and more for the same months in specific years. As in Bulgaria during spring also the precipitations are frequent, it could be expected that their chemical composition is influenced by Saharan dust. At the National Institute of Meteorology and Hydrology (NIMH) a monitoring network for acidity of precipitations has been established, [11]. In the last years the chemical analysis of the rain water samples was extended with analysis of main ions, macro and microelements, providing, thus, a possibility to investigate the characteristics of precipitation chemistry during Saharan dust outbreaks in the country. The purpose of this study is to analyse the influence of Saharan dust outbreaks on the chemical composition of rain samples collected at three sites in Bulgaria during field campaigns in 2017–2018. Another objective is to discuss precipitation chemistry in view of typical pathways of the dust loaded air masses.

2 Methodology The procedures used for the collection of precipitation samples and their chemical analysis, as well as the methods applied for identification of Saharan dust intrusions are briefly outlined.

Precipitation Chemistry in Bulgaria During Saharan Dust Outbreaks

265

Fig. 1 Geographical map of Bulgaria with sampling sites (orange square—Sofia, blue triangle— Cherni vruh, red circle—Ahtopol)

2.1 Precipitation Samples and Their Chemical Analysis The collection of precipitation samples was organised during field campaigns in 2017–2018 at three meteorological stations located in different environment (Fig. 1). Two of the stations are in the western part of the country: urban one—SofiaNIMH (42.655 N, 23.384 E, 586 m a.s.l.), and a mountain one—peak Cherni Vruh (42.6167 N, 23.2667 E, 2286 m a.s.l.). The third station is rural one in southeast Bulgaria near the Black Sea coast—Ahtopol (42.084 N, 27.952 E, 26 m a.s.l.). Daily precipitation samples in Sofia and Ahtopol were collected with an automatic wet only device (WADOS), at Cherni Vruh a passive bulk sampler was operated, made of polyethylene terephthalate funnel that was washed every day with deionized water to avoid dry deposition. The collected samples were further analysed for acidity-pH, conductivity-EC, main anions Cl− , SO4 2− , NO3 − , cation NH4 + and elements Na, K, Mg, Ca, Fe, Si, Zn, Cu. More details and detection limits for the analysed elements are given in [12].

2.2 Identification of Saharan Dust Intrusions A combination of modelling results and observational data was applied in order to identify the days characterised by Saharan dust outbreaks in Bulgaria in 2017 and 2018. For all the dates with available precipitation chemistry data at the three stations, an analysis was carried out, involving the following information:

266

E. Georgieva et al.

• Results for dust optical depth, dust surface concentrations, dust dry and wet depositions forecasted by the models at the Barcelona Supercomputing Centre (BSC)— BSC-DREAM8b [13], NMMB/BSC-Dust [14], the horizontal resolution is 0.3° × 0.3°; • Results for aerosol optical depth (AOD) and dust surface concentrations, forecasted by the ensemble model at the World Meteorological Organization Sand and Dust Storm Warning Advisory and Assessment System (WMO SDS-WAS) Regional Center for Northern Africa, Middle East and Europe on a grid with resolution 0.5° × 0.5° [15]; • Results for the Dust AOD and total AOD at 550 nm, and PM10 concentrations forecasted by the global CAMS-ECMWF model [16] over Europe on a grid resolution 0.125° × 0.125°, available through the Copernicus Atmosphere Monitoring Service (CAMS) and the European Centre for Medium Weather Forecast (ECMWF) [17]; • Results for dust and PM10 over Europe obtained by the CAMS regional air quality ensemble model, with horizontal resolution 0.1° × 0.1°, [18]; • Maps based on multi-model results at global scale with resolution 0.1° × 0.1° at the Marine Meteorology Division of the Naval Research Laboratory USA (NRL), [19, 20]; • HYSPLIT air mass backward-trajectories [21, 22] calculated at three arrival heights (500, 1500 and 3000 m a.g.l.) for 96 h, using NCEP GDAS meteorological input with resolution 0.5° × 0.5°, and reanalysis data; • Satellite data for AOD (level 3 MODIS Terra and Aqua globally on a grid 0.1° × 0.1°), for Aerosol absorbing index and Dust optical depth from MetOP satellites; • Observed PM10 concentrations at two background stations in mountain areas in Bulgaria—Kopitoto (BG0070A, 1321 m a.s.l.) and Rozhen (BG0053R, 1720 m a.s.l.).

3 Results and Discussion 3.1 Selected Dates and Samples The analysis of the origin of the dust loaded air masses indicated that the daily precipitation samples can be grouped into two main categories—with “direct” influence, i.e. approaching flow from southern directions, mainly from south-west, and with “indirect” influence, associated with other directions and respective regions, already impacted by Saharan dust. Table 1 presents details for the samples used in this study and the type of influence attributed to them.

Precipitation Chemistry in Bulgaria During Saharan Dust Outbreaks

267

Table 1 Date, location and type of influence for the precipitation samples used in this study, A—direct influence, B—indirect influence No

Date

Sofia

ChVruh

Ahtopol

Type

1.

04.06.2017

2.

06.06.2017

+

+

A A

3.

03.07.2017

+

B

4.

04.07.2017

5.

20.09.2017

6.

07.02.2018

7.

08.02.2018

8.

02.03.2018

9.

05.03.2018

10.

06.03.2018

11.

19.03.2018

+

A

12.

20.03.2018

+

A

13.

20.03.2018

14.

21.03.2018

+

A

15.

22.03.2018

+

A

16.

23.03.2018

+

17.

23.03.2018

+

B

18.

28.03.2018

+

A

19.

06.04.2018

+

A

20.

16.04.2018

+

A

21.

24.04.2018

+

22.

24.04.2018

23.

05.05.2018

+

24.

10.06.2018

+

25.

15.06.2018

26.

30.06. 2018

+ +

B A

+

A +

+

A A

+ +

A A

+

A

B

B +

B B B +

+

B B

3.2 Examples for Type A and Type B of Saharan Dust Outbreaks We show here results as maps and precipitation chemistry data for two typical cases/days influenced by Saharan outbreaks of type A and type B. The first case (16.04.2018) was characterized by dust loaded air masses approaching Bulgaria from south-west (A). The forecast by the SDS-WAS ensemble model (Fig. 2a) showed that the Dust AOD extends over the whole Balkan Peninsula. The accumulated dust wet depositions, estimated by the NMMB/BSC-Dust model

268

E. Georgieva et al.

Fig. 2 Case 16.04.2018: a median dust AOD from SDS-WAS ensemble model; b dust wet deposition from NMMB/BSC-dust model

(Fig. 2b) was concentrated in the western part of the Balkan Peninsula, indicting, thus, that the sampling sites Sofia and Cherni Vruh might be influenced. The origin of the air masses arriving in Sofia at altitudes 1500 and 3000 m a.g.l, was in Northern Africa, as suggested by the HYSPLIT back-trajectories (Fig. 3a). The Dust AOD, retrieved by the IASI instrument on MetOpA satellite, shows higher values south of Bulgaria and in the eastern parts of the country (Fig. 3b). Satellite data were missing over most of the Balkans (white areas) due to the presence of clouds and rain. The chemical composition of the precipitation sample from Sofia is shown in Fig. 4. For this case the total ionic concentration (TIC) in the precipitation sample was 54.1 mg.l−1 , and consisted from 79% of the following elements: nssSO4 2− , NO3 − and Ca. The pH and EC values of this sample were very high (7.4 and 132.7 µS cm−1 ). The most abundant element was Ca followed by nssSO4 2− and NO3 − .

Fig. 3 Case 16.04.2018: a HYSPLIT back trajectories, and b dust AOD a from IASI instrument on MetOpA

Precipitation Chemistry in Bulgaria During Saharan Dust Outbreaks

269

Fig. 4 Concentrations of elements in precipitation sample from Sofia on 16 April 2018

The concentrations of Si, K and Mg were also high, with contribution to the TIC of 3%, 2% and 2%, respectively. The contribution of NH4 + is 3%. The second case (30.06.2018) was characterised by an intrusion from north, northwest towards Bulgaria, but the regions in Central and Western Europe were impacted by Saharan dust few days before. Thus, indirect influence of dust loaded air masses was expected (B). The models at BSC and the system CAMS-ECMWF suggested transport of dust loaded air masses from Northern Africa towards western and central Europe (Fig. 5a, b). Satellite data were limited in Eastern Europe due to cloudiness (Fig. 6b), however the dust transport towards the western part of the continent was well evident. The HYSPLIT back trajectories for the sampling site Cherni Vruh, presented as frequencies for the period 22.06–30.06.2018 in Fig. 6a, indicate that the air masses were from north and north-westerly directions, but also from southern direction. The chemical composition of the precipitation sample from Cherni Vruh for this second case (Fig. 7) is characterised by high concentrations of nssSO4 2− , Ca, K and NH4 + . The measured pH and EC are 4.9 and 50 µS cm−1 . The TIC is 48.5 mg l−1

Fig. 5 Case 30.06.2018 (type B): a dust AOD and wind vectors at 700 hPa from NMMB/BSC-dust model; b dust AOD from CAMS-ECMWF model on 29.06.18 at 18:00 UTC

270

E. Georgieva et al.

Fig. 6 Case 30.06.2018 (type B): a HYSPLIT frequency of back trajectories (22.06–30.06.18), and b dust AOD from IASI instrument on MetOpA

Fig. 7 Concentrations of elements in precipitation sample from Cherni Vrah on 30 June 2018

composed from 84% of the following elements: nssSO4 2− , NO3 − and Ca. The contribution of NH4 + is 5% and of Si is only 1%. Comparing the chemical analysis data for these two samples belonging to the groups A and B, one can notice that for the first one (A-type, or direct influence) the values of Cl− , Mg, Na and Si were higher, for the second case (B-type, indirect influence) NH4 + values were higher. Common for both samples are the high values of nssSO4 2− and Ca. Before discussing the average precipitation chemistry data for the two groups A and B, synoptic parameters and aerosol loads for the two groups were investigated through composite maps, i.e. maps representing all days in the group.

Precipitation Chemistry in Bulgaria During Saharan Dust Outbreaks

271

3.3 Composite Maps for the Groups A and B The synoptic situation for the two groups was analysed using the National Centers for Environmental Prediction/National Center for Atmospheric Research (NCEP/NCAR) Reanalysis—2 global dataset, [23]. Composite daily maps for Europe were constructed to highlight the main features typical for the two groups. The geopotential height at 700 hPa, the wind vectors and speed at 700 hPa, and the columnar precipitable water amount are shown in Fig. 8. The level 700 hPa (around 3000 m a.s.l.) was selected as the dust transport occurs mainly aloft, [24]. The geopotential height at 700 hPa for group A indicated low pressure system over Central Europe and a center with higher pressure over the southern Mediterranean and North Africa, that led to south-westerly flows towards Bulgaria (Fig. 8a, c). For the group B a high pressure ridge extended from North Africa towards the western and central part of Europe, while there was a low pressure trough over south-eastern Europe that led to flow from northern origin towards Bulgaria (Fig. 8b, d). The winds at 700 hPa over the Balkans were stronger for the A group. The columnar precipitable water amount for group A had higher values in a region south of Bulgaria, while for group B the precipitation region was more wide spread, extending to the west, north and east of the country. The horizontal distribution of the dust aerosols for the two groups was analysed using the CAMS-ECMWF global model [15, 16] forecasted values for the respective days. The maps for the average Dust AOD for the two groups (Fig. 9 a, b) showed the dust load south of the country for A, and enhanced dust load north-west of the country for B. The maps for the sulphate AOD (Fig. 9 c, d) indicated higher load over Bulgaria and south of the country for group A, while for the group B the sulphate rich areas were extended to the north-west and north of the country.

3.4 Precipitation Chemistry Analysis The pH frequency distribution (Fig. 10) shows that about 70% of all samples (both groups) were in the neutral and acidity range. The pH values ranged from 4.1 to 7.4.40% of the precipitation samples in A were in the acidity range (pH < 5.0), 6.7%—in the slightly acidic range (5.0–5.5) and 6.7%—in the alkaline range (>7.0). The pH frequency analysis for B samples showed that in 30% pH was in the range 4.5–5.0, 10% were in the slightly acidic range, and 20% were in slightly alkaline range. pH value higher than 6.2 were not observed for B samples. pH values in the neutral range are observed in 40% for B samples, and in 33% for A samples. The mean pH value for both groups of samples from Sofia is 6.05 and is higher than the average multiyear pH (5.12) estimated from the precipitation chemistry network

272

E. Georgieva et al.

Fig. 8 Synoptic situation–composite maps for group A (left) and group B (right): a, b geopotential height (m) at 700 hPa; c, d wind vectors (arrows) and wind speed, (ms−1 ) at 700 hPa; and e, f columnar precipitable water (kgm−2 )

for the period 2002–2019. The same is observed for the samples from Ahtopol—mean for both groups pH is 5.34, while the multiyear one is 5.16. Some statistical parameters (mean, standard deviation, minimum and maximum) were estimated for the pH and EC, the precipitation chemistry elements and the TIC, for the samples defined as A and B (Table 2). The contribution of different elements in both A and B samples to the total ionic concentration is presented in Fig. 11. The most abundant ionic species for samples

Precipitation Chemistry in Bulgaria During Saharan Dust Outbreaks

273

Fig. 9 CAMS-ECMWF averaged maps for A (left) and B (right): a, b dust AOD, c, d sulphate AOD, (Generated using Copernicus Atmosphere Monitoring Service Information (2020)

Fig. 10 Frequency distribution of pH for samples in A (red) and B (blue)

in both categories was nssSO4 2− , followed by Cl− and Ca for A category, and Ca and NO3 − for B category. The total ionic concentration in both A and B samples consisted mainly of nssSO4 2− , NO3 − and NH4 + (A—44% and B—58%). For the group A (flow mainly from south-west) the air masses passing over the Mediterranean Sea were enriched with sea salt aerosol. The evident high correlation between the elements Cl and Na (0.99) conferred this. 31% of the TIC consisted of Cl and Na. The percentage of terrigenous elements (Ca, K, Mg and Si) of the TIC in the A samples is 23.3%. High correlation was obtained also between Ca and Si (0.90), Ca and Fe (0.84), and between nssSO4 2− and Ca (0.73). The higher contribution of sulphates, nitrates and ammonium ions in B samples can be explained by the enrichment of air masses with substances of anthropogenic origin. The correlations between nssSO4 2− and NH4 + (0.7), and between NO3 − and

274

E. Georgieva et al.

Table 2 Statistics (mean, standard deviation, minimum and maximum) for the samples in A (N = 16) and B (N = 10), concentrations in mg.l−1 Element mg.l−1 pH (−) EC (µS/cm)

Mean A

SD A

Min A

Max A

Mean B

SD B

Min B

Max B

5.5

0.9

4.1

7.4

5.5

0.5

4.7

6.2

38.7

34.1

16.4

132.7

29.8

22.5

7.4

76.0

Cl

4.55

7.04

0.44

26.94

1.67

1.93

0.11

6.41

NO3

3.09

3.96

0.17

15.83

4.52

4.84

0.88

15.75

SO4

5.91

5.54

1.67

20.28

5.93

5.45

0.41

17.55

nssSO4

5.64

5.40

1.59

19.70

5.88

5.44

0.41

17.52

Ca

3.89

4.66

1.07

17.66

4.59

5.03

0.75

15.83

K

1.20

2.37

0.08

8.87

0.79

0.74

0.23

2.35

Mg

0.46

0.42

0.14

1.40

0.34

0.24

0.06

0.86

Na

2.40

4.27

0.11

15.64

0.48

0.86

0.10

2.60

Cu

0.02

0.02

0.01

0.05

0.01

0.00

0.01

0.01

Fe

0.06

0.09

0.01

0.30

0.02

0.02

0.00

0.05

Si

0.28

0.36

0.06

1.50

0.24

0.16

0.07

0.55

Zn

0.11

0.24

0.01

0.81

0.05

0.04

0.01

0.11

NH4

0.90

1.00

0.13

3.90

0.92

0.80

0.06

2.33

TIC

22.28

22.28

6.58

68.87

18.72

16.95

2.53

48.50

Fig. 11 Contribution (in %) of different elements in precipitation samples for A and B groups

NH4 + (0.9) indicate contribution from secondary formed aerosols ((NH4 )2 SO4 and NH4 NO3 ).Very high correlations were found between nssSO4 2− and Ca (0.91), and between Ca and K (0.95), indicating that the main source of those ions are from terrigenous origin (e.g. gypsum—CaSO4 ), [25]. 30.4% of the TIC for the samples of B group consisted of Ca, K, Mg, and Si.

Precipitation Chemistry in Bulgaria During Saharan Dust Outbreaks

275

The t-test performed for the mean elemental contributions in the two groups returned a value of 0.035 for the two-tailed p-value, indicating, thus, to statistically significant differences between the two groups. Having in mind that the samples are low in number, we recognize that more robust statistical results could be obtained after collecting and analyzing more precipitations samples in different synoptic situations. The chemical analysis of all samples confirmed that the precipitation associated with dust intrusions is characterized by higher concentrations of terrigenous elements. In both types of intrusions the correlation between nssSO4 2− and Ca was relatively high, indicating similar source of origin.

4 Conclusions We have analysed the influence of Saharan dust events on the chemical composition of 26 rain samples collected at three sites in Bulgaria during 2017–2018. The samples were divided into two categories with respect to the path of the approaching dust loaded masses: group A (flow from south, south-west; direct influence) and B (flow from north, north-west; indirect influence). The analysis of composite maps for synoptic variables and aerosol optical depth highlighted the regions with dust loaded air masses approaching the country for the days of the two groups. The mean pH of the samples from both groups was 5.5. About 70% of all samples were in the neutral and acidity range. For the sampling site Sofia the mean pH of all samples (6.05) was higher than the multi-year one (5.12) indicating that, in general Saharan intrusions contribute to more alkaline character of precipitations, although for single events this value might be in the acidity range. The precipitation associated with dust intrusions from both groups is characterized by higher concentrations of terrigenous elements (Ca, Si, K). The concentrations of chloride, magnesium and sodium were significantly higher for A samples, as a consequence of the influence of the Mediterranean Sea on the air masses in this group. For the B samples higher concentrations of sulphates, nitrates and ammonium ions was found, suggesting enrichment of air masses with anthropogenic pollutants. The preliminary results, presented here, showed that the trajectory of the air masses is an important factor for the chemical composition of precipitations in Bulgaria. Additional data from sampling are needed, however, in order to extend the analysis and to perform more robust statistical estimates. Acknowledgements This study was inspired by COST CA16202 “inDust”, and was funded by the Bulgarian National Science Fund through contract N. DN-04/4-15.12.2016. The Copernicus Atmosphere Monitoring Service is acknowledged for providing analysed and forecasted model data on atmospheric chemistry. We are thankful also to model groups for providing operational dust products and maintaining archives—Barcelona Supercomputing Center, WMO SDS-WAS NAMEE Centre and the Marine Meteorology Division of U.S. Naval Research Laboratory, as well as NOAA-ARL for the HYSPLIT on-line platform. We acknowledge also the provider of satellite data observations (ESA, EUMETSAT, NASA/LANCE).

276

E. Georgieva et al.

References 1. Middleton, N., Tozer, P., Tozer, B.: Sand and dust storms: underrated natural hazards. Disasters 43, 390–409 (2019) 2. Prospero, J.M., Ginoux, P.M., Torres, O., Nicholson, S.E., Gill, T.E.: Environmental characterization of global sources of atmospheric soil dust identified with the Nimbus-7 Total Ozone Mapping Spectrometer (TOMS) absorbing aerosol product. Rev. Geophys. 40(1), 1002 (2002) 3. Goudie, A.S., Middleton, N.J.: Desert Dust in the Global System. Springer, Berlin, Heidelberg (2006) 4. Stuut, J.-B.W., Smalley, I., O’Hara-Dhand, K.: Aeolian dust in Europe: African sources and European deposits. Quat. Int. 198, 234–245 (2009) 5. Maher, B.A., Prospero, J.M., Mackie, D., Gaiero, D., Hesse, P.P., Balkanski, Y.: Global connections between Aeolian dust, climate and ocean biogeochemistry at the present day and at the last glacial maximum. Earth Sci. Rev. 99, 61–97 (2010) 6. Rogora, M., Mosello, R., Marchetto, A.: Long-term trends in the chemistry of atmospheric deposition in northwestern Italy: the role of increasing Saharan dust deposition. Tellus B 56, 426–434 (2004) 7. Klein, H., Nickovic, S., Haunold, W., Bundke, U., Nillius, B., Ebert, M., Weinbruch, S., Schuetz, L., Levin, Z., Barrie, L.A., Bingemer, H.: Saharan dust and ice nuclei over Central Europe. Atmos. Chem. Phys. 10, 10211–10221 (2010) 8. Varga, Gy., Kovács, J., Újvári, G.: Analysis of Saharan dust intrusions into the Carpathian Basin (Central Europe) over the period of 1979–2011. Global Planet 100, 333–342 (2013) 9. Pey, J., Querol, X., Alastuey, A., Forastiere, F., Stafoggia, M.: African dust outbreaks over the Mediterranean Basin during 2001–2011: PM10 concentrations, phenomenology and trends, and its relation with synoptic and mesoscale meteorology. Atmos. Chem. Phys. 1395–1410 (2013) 10. Dimitrova, M., Trenchev, Pl., Georgieva, E., Neykova, N., Neykova, R., Nedkov, R., Gochev, D., Syrakov, D., Veleva, B., Atanassov, D., Spassova, T.: Seasonal changes of aerosol pollutants over Bulgaria. In: Proceedings of the 15th International Science Conference Space, Ecology, Safety, SES 2019, SRTI-BAS, Sofia, pp. 241–252 (2019). ISSN 2603-3321 11. Hristova, E.: Chemical composition of precipitation in urban area. Bulgarian J. Meteorol. Hydrol. 22(1–2), 41–49 (2017) 12. Hristova, E., Veleva, B., Georgieva, E., Velchev, K.: Cloud and rain water chemical composition at peak Cherni Vrah, Bulgaria (in this Book) (2020) 13. Basart, S., Pérez, C., Nickovic, S., Cuevas, E., Baldasano, J.M.: Development and evaluation of the BSC-DREAM8b dust regional model over Northern Africa, the Mediterranean and the Middle East. Tellus B 64, 1–23 (2012) 14. Pérez, C., Haustein, K., Janjic, Z., Jorba, O., Huneeus, N., Baldasano, J.M., Black, T., Basart, S., Nickovic, S., Miller, R.L., Perlwitz, J., Schulz, M., Thomson, M.: An online mineral dust aerosol model for meso to global scales: model description, annual simulations and evaluation. Atmos. Chem. Phys. 11, 13001–13027 (2011) 15. WMO SDS-WAS Homepage. https://sds-was.aemet.es/. Last accessed 2020/09/10 16. Benedetti, A., Morcrette, J.-J., Boucher, O., Dethof, A., Engelen, R.J., Fisher, M., Flentjes, H., Huneeus, N., Jones, L., Kaiser, J.W., Kinne, S., Mangold, A., Razinger, M., Simmons, A.J., Suttie M., GEMS-AER team: Aerosol analysis and forecast in the ECMWF integrated forecast system. Part II: data assimilation. J. Geophys. Res. 114, D13205 (2009) 17. CAMS-ECMWF Homepage. https://apps.ecmwf.int/datasets/data/cams-nrealtime/levtype= sfc/. Last accessed 2020/09/10 18. Marécal, V., Peuch, V.-H., Andersson, C., Andersson, V, Arteta, J., Beekmann, M., Benedictow, A., Bergström, V, Bessagnet, B., Cansado, A., Chéroux, V, Colette, A., Coman, A., Curier, R.L., Denier van der Gon, H.A.C., Drouin, A., Elbern, H., Emili, E., Engelen, R.J., Eskes, H.J., Foret, G., Friese, E., Gauss, M. et al.: A regional air quality forecasting system over Europe: the MACC-II daily ensemble production. Geosci. Model Dev. 8, 2777–2813 (2015)

Precipitation Chemistry in Bulgaria During Saharan Dust Outbreaks

277

19. Xian, P., Reid, J.S., Hyer, E.J., et al.: Current state of the global operational aerosol multi-model ensemble: an update from the international cooperative for aerosol prediction (ICAP). Q. J. R. Meteorol. Soc. 145, 176–209 (2019) 20. NRL/Monterey Aerosol Page. https://www.nrlmry.navy.mil/aerosol/. Last accesses 2020/09/10 21. Stein, A.F., Draxler, R.R., Rolph, G.D., Stunder, B.J.B., Cohen, M.D., Ngan, F.: NOAA’s HYSPLIT atmospheric transport and dispersion modeling system. Bull. Amer. Meteor. Soc. 96, 2059–2077 (2015) 22. Rolph, G., Stein, A., Stunder, B.: Real-time Environmental applications and display system: READY. Environ. Model Softw. 95, 210–228 (2017) 23. Kalnay, E., Kanamitsu, M., Kistler, R., Collins, W., Deaven, D., Gandin, L., Iredell, M., Saha, S., White, G., Woollen, J., et al.: The NCEP/NCAR 40-year reanalysis project. Bull. Am. Meteorol. Soc. 77, 437–472 (1996) 24. Nabat, P., Somot, S., Mallet, M., Sevault, F., Chiacchio, M., Wild, M.: Direct and semi-direct aerosol radiative effect on the Mediterranean climate variability using a coupled regional climate system model. Clim. Dyn. 44, 1127–1155 (2015) 25. Conradie, E.H., Van Zyl, P.G., Pienaar, J.J., Beukes, J.P., Galy-Lacaux, C., Venter, A.D., Mkhatshwa, G.V.: The chemical composition and fluxes of atmospheric wet deposition at four sites in South Africa. Atm. Environ 146, 113–131 (2016)

Modelling a Composite Tsunami Scenario for Karpathos Island (Aegean Sea) Lyuba Dimova , Reneta Raykova , Gianluca Pagnoni , Alberto Armigliato , and Stefano Tinti

Abstract Karpathos is one of the biggest Greek islands, located between Crete and Rhodes in Aegean Sea. As the most of the islands in the area Karpathos is prone to earthquakes and tsunamis. The event of 9 February 1948 (M 7.1) near the eastern coast of the island caused local tsunami with damages in the area of Pigadia bay, nevertheless tsunamis also from regional sources are expected. The tsunami hazard for the Karpathos Island, focusing on the city of Karpathos and the Airport area, is modelled merging the data from simulation of tsunamis generated by three seismic sources: Eastern Hellenic Arc (EHA referring the 1303 A.D. event, Mw = 8.0); near Rhodes (hypothetical scenario earthquake, Mw = 7.3); and near the coast of Karpathos, based on the 1948, Mw = 7.3 earthquake. Numerical calculations are made using the code UBO-TSUFD on a set of nested grids. Tsunami observables, such as maximal water column height, maximum velocity flux, inundation, are computed for each individual scenario and merged to individuate the areas most exposed to tsunami. The seismic source EHA dominates in the tsunami hazard maps: moreover, the impact over the southern part of Karpathos has biggest risk since the airport and the main city of the island are located in this part. Keywords Tsunami · Numerical modelling · Composite scenario · Inundation · Karpathos Island

1 Introduction The Aegean region is characterized by high seismic activity and the numerous islands are prone to devastating earthquakes, tsunamis, landslides and in the past—volcanic eruptions. The tectonic settings in the region are complex: subduction of African L. Dimova (B) · R. Raykova Department of Meteorology and Geophysics, Faculty of Physics, Sofia University “St. Kliment Ohridski”, 5 James Bourchier, 1164 Sofia, Bulgaria e-mail: [email protected] G. Pagnoni · A. Armigliato · S. Tinti Sector of Geophysics, Department of Physics and Astronomy “Augusto Righi”, Alma Mater Studiorum—Università di Bologna, Viale Berti Pickat 8, Bologna, Italy © The Author(s), under exclusive license to Springer Nature Switzerland AG 2021 N. Dobrinkova and G. Gadzhev (eds.), Environmental Protection and Disaster Risks, Studies in Systems, Decision and Control 361, https://doi.org/10.1007/978-3-030-70190-1_19

279

280

L. Dimova et al.

plate toward north is predominant, but there is also subduction of Ionian Adriatic plate toward east, horizontal displacement along Anatolian fault, rotation of Hellenic peninsula, rifting in the Aegean area. Karpathos Island is situated in the southern part of the Aegean Sea, between the island of Crete and Rhodes, near the boundary of the main tectonic plates and volcanic arcs. It is exposed to a variety of mechanisms for tsunami generation. Historically documented tsunamis in this region are reviewed for example in [1, 2]. The devastating events are numerous, starting with earthquakes and tsunamis in 365 A.D., 1303 A.D. through more recent events in 1948 and 1956 to the nowadays events of Bodrum-Kos in 2017 and of Samos Island on October 30th, 2020. In this research we simulate tsunamis generated by 3 seismic sources in the vicinity of Karpathos Island that are not well studied: near Crete in the Eastern Hellenic Arc (EHA), with reference to the 1303 A.D. event, Mw = 8.0; near Rhodes (hypothetical scenario earthquake, Mw = 7.3); and near the south-east coast of Karpathos, based on the 1948 earthquake, Mw = 7.3. Our simulations are focused on the southern part of the island (airport region) since the main city of Karpathos and the area of the Karpathos airport are situated there, and the slopes to the north part of the island are very steep and inundations are not expected.

2 Selected Tsunamigenic Sources We select 3 seismic sources that generate tsunami inundating the coasts of Karpathos and that are not well studied. The event in 1303 A.D. is one of the biggest events in the Mediterranean and generated tsunamis that affected almost the whole eastern part of the Mediterranean. Because of this large effect the epicenter of this event is still dubious and there is no one certain location of it. We consider the parameters and location of this event given in [3]. Some authors [1] considered EHA located more toward east in southwest-northeast direction, but there are no explicit geological evidences. The hypothetical event near Rhodes is taken into account because in this area there are a number of the tsunami-generating events and some of them have the main direction of propagation toward Karpathos Island [4]. Some of the earthquakes generated tsunami events in this region are collected in the Euro-Mediterranean Tsunami Catalogue [5], like the events in 1481 (Mw = 6.5), 1609 (Mw = 7.2), 1741 (Mw = 7.3). For our hypothetical source, we selected the maximum observed magnitude and focal mechanism of the event that have largest impact toward Karpathos Island (event in 1957) proposed by Ebeling et al. [6]. The earthquake and the ensuing tsunami in 1948 are not well studied yet. The hypothesis for combined generation mechanism—earthquake and underwater landslide is explored by Ebeling et al. [6]. There are two main faults placed to the north of the island, named Karpathos I and Karpathos III [7]. Recently, [8] located the

Modelling a Composite Tsunami Scenario …

281

event from 1948 on the eastern coast of Karpathos, but in that region there are no any significant fault zone, capable to generate an event with M > 7. We associate the location of this event as a part of the Eastern Hellenic Arc, located in southeast offshore of Karpathos. All parameters and relevant references for the selected events are given in [9]. Location of the modelled events are given as simplified faults in Fig. 1.

Fig. 1 Topography and bathymetry grids: a broad region around Karpathos included in grid G1; b Karpathos Island and surrounding sea area included in grid G2; c southern part of Karpathos included in the finest grid G3; d Pigadia Bay in eastern Karpathos Island included in the finest grid G4. Black lines show the position of the considered faults. Red point shows the location of a synthetic mareogram

282

L. Dimova et al.

3 Tsunami Simulation Method Deformation of the seafloor due to the displacement along a fault is calculated by equations given in [10]. Okada presents the focal zone as a rectangular fault (specified by its location and geometry) with specific focal mechanism. Calculated seafloor deformation is used as initial conditions for the following tsunami modelling. The numerical code called UBO-TSUFD [11] is used to estimate the propagation of tsunamis and the interaction of waves with shores. Output parameters of the calculations are: the propagation field; travel times of tsunami, the field of maximum and minimum elevation; the maximum particle velocity field, the water column on land. The same procedure is used in several cases to evaluate the tsunami hazard in eastern Mediterranean [12, 13] and specifically in southern Aegean Sea [4]. The applied procedure for the tsunami simulation in the region of the Karpathos is described in details in [9]. The model UBO-TSUFD was tested and benchmarked in the framework of several EU-funded projects (TRANSFER, SCHEMA, ASTARTE) and already applied to different case studies [14, 15]. UBO-TSUFD reproduce quite well theoretical and experimental data.

4 Computational Grids Numerical tsunami simulations are carried out on rectangular grids built by making use of suitable bathymetric and topographic data. In this research are used four nested grids with different resolution. The bathymetry and topography data of different grids are compiled from several resolution data sets [16–18]. The largest domain-wide (G1) includes the eastern part of Crete, the islands of Karpathos and Rhodes and the southeastern Aegean Sea. The resolution of this grid is 500 m and it is presented in Fig. 1a. The grid G2 covers the area of Karpathos Island and it is nested in G1. The resolution of this grid is 100 m and it is shown in Fig. 1b. The finest grids G3 and G4 are located in the southern part of Karpathos with focus on the airport area (G3) and the city of Karpathos (G4). Both grids have resolution of 20 m and they are nested in G2. G3 and G4 outline the most important areas in the Island and they are presented in Fig. 1c, d. Detailed information about the used grids is given in [9].

5 Results of Tsunami Simulations 5.1 Tsunami Propagation Field The propagation field of the tsunami indicates the propagation direction of the energy, possessed by tsunami and relevant land areas reached by main flux. The propagation

Modelling a Composite Tsunami Scenario …

283

field denotes also the areas, where can be observed secondary interferences with considerable amplitudes to be explored in details. Propagation field from the nearest Karpathos event shows that one of the main directions of the energy flux is toward southeast parts of Karpathos. Significant portion of the energy flux from Rhodes event arrive to the northeast part of the island. Propagation pattern of the tsunami generated by the EHA event is given in Fig. 2. The figure shows that considerable part of the tsunami energy reaches the region of Karpathos Island. Some interference maxima are formed in the bays of the island even 20 min after. Dissected relief of the island’s shore contributes to the formation of the secondary tsunami maxima from all events.

Fig. 2 Propagation field for EHA source for different times after the event’s origin time

284

L. Dimova et al.

5.2 Travel Time Travel time maps show the arriving time of the first tsunami signal to the coasts of the Island. It is crucial to be defined the evacuation time for each area and seismic source. The tsunami waves from Karpathos event reach the southeastern coast of the Island in less than 2 min and evacuation time is very limited. The north-western part of the Island is reached a little bit later and can be useful for alert in city of Arkasa for example. Tsunamis generated by EHA event reach the most southern coast of the Island for 9 min and the airport area in the following two minutes. This time, though limited, could be sufficient to save lives. The city of Karpathos is reached in 13–14 min that is also reasonable time to put in save the people. Tsunami waves arriving from Rhodes source have travel times larger than 20 min for the area of Karpathos Island. Travel time maps for EHA and Rhodes events are given in Fig. 3.

5.3 Synthetic Mareograms The oscillations of the sea level with time at certain point are illustrated by synthetic mareograms. The comparative mareograms for 5 points in the southern part of the Karpathos are discussed in [9]. The most important conclusions are several. The southwestern part of the island in the area of Arkasa is affected mostly by the tsunami generated by EHA. The extreme south of the island (area of the airport) is affected mainly by EHA tsunamis, but amplitudes are smaller than those in the western part, since the Kasos Island creates a shadow zone for the tsunamis generated by EHA. In this area amplitudes by Rhodes and Karpathos events are similar with later oscillations from EHA event. Southeastern part of the island in the area of Kipi Afiarti is inundated by the tsunamis generated by Karpathos event but this area is reached also by considerable late interferences with similar amplitudes generated by Karpathos and EHA events. For the area of city of Karpathos the most important impact is induced by Karpathos source. Amplitudes of the later interferences are with similar amplitudes for all of the modeled sources. Figure 4 illustrates the first two hours of the tsunami amplitude oscillations due to the three sources in the area of Kipi Afiarti. Position of the mareogram is shown in Fig. 1c. First positive waves arrive from Karpathos source, later on tsunamis from EHA event. Wave periods from those two sources are quite similar 20–30 min after the earthquakes. Estimated maximum elevations for Kipi Afiarti vary between −2 and 2 m.

Modelling a Composite Tsunami Scenario …

285

Fig. 3 Travel time maps: a EHA event; b Rhodes event

5.4 Maximum Elevation Fields Maximum elevation fields show the maximal value reached in each node of the grids during the simulation. For the two nested grids we merged the results obtained from all three sources, thus we obtain the so-called composite tsunami scenario or the worst case scenario. This means that we take into account the extremum value in each cell. Figure 5 presents the maximum tsunami elevations for the region of interest. For the Airport area and the city of Karpathos the maximum values reach more than 4 m. Large part of the southern and the eastern Karpathos are exposed to waves larger than 2 m.

286

L. Dimova et al.

Fig. 4 Synthetic mareogram for the sea area in front of Kipi Afiarti (see location in Fig. 1c)

Fig. 5 Maximum elevation fields merged for all sources

5.5 Maximum Inundation Zones The inundation on the coastline is calculated when nonlinearity is included in the method. The inundation is presented as the flow depth over the land. Figure 6 shows simplified map of the inundated zones in G3 and G4. It is seen that part of the Airport and the tracks are flooded by up to 4 m water layer. Pigadia Bay is inundated mainly in the central part again reaching level up to 4 m, while the northern and the southern part remain partly flooded by the tsunamis.

Modelling a Composite Tsunami Scenario …

287

Fig. 6 Inundation zones merged for all sources

5.6 Maximum Momentum Flux Combining velocity particle field and flow parameters we can calculate wave forces on structures. Tsunami forces can be represented by two components—proportional to depth times the acceleration and another proportional to depth times the velocity squared. The maximum momentum flux is show in Fig. 7. The equation for calculating the momentum flux is as follow: M max(x, y) = max[h(x, y, t) ∗ v 2 (x, y, t)]

(1)

where M max is the maximum momentum flux, h is the maximum flow depth and v is the flow velocity. It needs to be clarified that during tsunami waves the maximum momentum flux does not surely take place at the same time with the maximum flow velocity, so that the momentum flux may not match to the same flow velocity. Maximum momentum flux does not depend on the time because it is a maximum value taken over the entire simulation time. The distribution of the momentum flux inland could be used for zoning and planning areas with larger tsunami impact. As seen from the figure, the Airport and the city of Karpathos are situated in areas where the momentum flux exceeds 70–90 m3 /s2 .

288

L. Dimova et al.

Fig. 7 Maximum momentum flux calculated for all three scenarios

5.7 Contribution of Each Scenario to the Worst Composite Scenario The contribution of all three scenarios is given by the maximum elevation field having regard the following considerations: we take into account the elevation in each node of the grid above 0 m; for each of the events, we compared the values to the other two and we took the larger one. Results show (Fig. 8) that EHA event contributes mostly to the southern and southwestern part of the island with 72% of all positive values larger than those from the other sources. Karpathos source remain with 28% Fig. 8 Contribution of all scenarios for G3

Modelling a Composite Tsunami Scenario …

289

of all positive points, mostly located from the eastern part of the Airport to the east, reaching the city of Karpathos, where this event completely dominates.

6 Discussion of the Results After reviewing the seismicity and historically documented tsunami events we have selected three major sources, referenced to events in the past. To conclude we plotted the inundation maps as layers over Google Earth Maps. Figures 9 and 10 illustrate the affected areas due to the composite scenario. The main building of the Airport is hit by the tsunami with water amplitudes between 1 and 4 m. Some of the tracks of the airport are partially inundated with smaller flow depth, around 2 m. In case of strong earthquake in the region of Eastern Hellenic Arc, there is enough time to evacuate the Airport area and to move the people in a higher safe place. The city of Karpathos and especially the central part of the coastline experience flooding around 2–4 m. In case of tsunami due to Karpathos source, the travel times show the difficulties related to the very short evacuation times. In order to solve this problem, although very expensive solution, is installing instruments (ocean bottom seismometers, pressure sensors, offshore buoys) near the hypothesized source area that one may detect the generation of the tsunami in its very early phases. An alternative could be setting up an underwater shore protection facilities, but again is a very costly decision. In parallel, proper evacuation planning and population training must be in place.

Fig. 9 Inundated zones in G3

290

L. Dimova et al.

Fig. 10 Inundated zones in G4

7 Conclusions Destructive tsunamis in this area of Karpathos Island are not so frequent events, but their tsunamigenic potential is high, since the region lies between a subduction zone to the south and a volcanic arc to the north. Therefore, numerical modelling of tsunamis is extremely important to evaluate tsunami impact, especially in areas where the historical information is insufficient. Assessment of the tsunami hazard on the coasts of Karpathos is done by a composite scenario of three modelled events. The results show that the EHA source, referred to the 1303 A.D. event has a maximum tsunami impact on the southern part of the island, whereas the Karpathos source (reference event 1948) affects mostly the eastern coastline. Modelled maximal wave amplitudes for the area of the Karpathos Airport and the city of Karpathos are 4–5 m. Inundation maps show that the central part of the Pigadia Bay is reached by the tsunami flooding. The evacuation time from Karpathos event is extremely short, while the propagation time from EHA and Rhodes events is enough for evacuation if the tsunami alert system is available. Such results could be implemented in the operational work after developing a series of tsunami scenarios incorporating different parameters and focal mechanisms. Numerical simulations of the seismically generated tsunami often underestimate the amplitudes of the observed tsunami waves. Nevertheless secondary effects like underwater landslides must not be excluded in the generation of consecutive tsunamis. Indeed such complex sequence of events may increase significantly the maximum heights of expected tsunami. Acknowledgements The first author would like to thank the Tsunami Research Team from University of Bologna for the opportunity to work with UBO-TSUFD. This paper was supported by the

Modelling a Composite Tsunami Scenario …

291

National Program “Young Scientists and Postdoctoral Students” of Ministry of Education and Science of Bulgaria and funded by BNSF # CP–06–COST-7/24.09.2020.

References 1. Papadopoulos, G.A., Gràcia, E., Urgeles, R., Sallares, V., De Martini, P.M., et al.: Historical and pre-historical tsunamis in the Mediterranean and its connected seas: geological signatures, generation mechanisms and coastal impacts. Mar. Geol. 354, 81–109 (2014) 2. Dimova, L., Raykova, R.: Observations and modeling of tsunamis in the Eastern Mediterranean (review). Ann. Sofia Univ. “St. Kliment Ohridski” Fac. Phys. 109, 24–41 (2016) 3. Yolsal-Cevikbilen, S., Taymaz, T.: Earthquake source parameters along the Hellenic subduction zone and numerical simulations of historical tsunamis in the Eastern Mediterranean. Tectonophysics 536–537, 61–100 (2016) 4. Dimova, L.: Tsunami radiation pattern in the southern Aegean Sea. Ann. Sofia Univ. “St. Kliment Ohridski” Fac. Phys. 111, 23–40 (2018) 5. Maramai, A., et al.: The euro-mediterranean tsunami catalogue. Ann. Geophys 57, 4 (2014) 6. Ebeling, C., Okal, E., Kalligeris, N., Synolakis, C.: Modern seismological reassessment and tsunami simulation of historical Hellenic Arc earthquakes. Tectonophysics 530–531, 225–239 (2012) 7. Basili, R., et al.: The European database of seismogenic faults (EDSF) compiled in the framework of the project share. http://diss.rm.ingv.it/share-edsf/, https://doi.org/10.6092/ingv. it-share-edsf (2013) 8. Andinisari, R., Konstantinou, K.I., Ranjan, P.: Seismotectonics of SE Aegean inferred from precise relative locations of shallow crustal earthquakes. J. Seismolog. 24(1), 1–22 (2020) 9. Dimova, L., Raykova, R., Armigliato, A., Pagnoni, G., Tinti, S.: Aggregated tsunami scenario for Karpathos Island. In: Proceedings of: 1st International Conference on Environmental Protection and Disaster RISKs, 29–30 September 2020, pp. 443–451 (2020). https://doi.org/10.48365/ envr-2020.1.40 10. Okada, Y.: Surface deformation due to shear and tensile faults in a half-space. BSSA 75(4), 1135–1154 (1985) 11. Tinti, S., Tonini, R.: The UBO-TSUFD tsunami inundation model: validation and application to a tsunami case study focused on the city of Catania. Italy Nat. Hazards Earth Syst. Sci. 13, 1759–1816 (2013) 12. Dimova, L., Raykova, R., Armigliato, A., Pagnoni, G., Tinti, S.: Modelling of earthquakeinduced tsunami in the Eastern Mediterranean region. AIP Conf. Proc. 2075(120024), 2019 (2019). https://doi.org/10.1063/1.5091282 13. Dimova, L., Raykova, R.: Tsunami radiation pattern in the Eastern Mediterranean. J. Phys. Technol. 1(2), 22–27 (2017) 14. Tonini, R., Armigliato, A., Pagnoni, G., Zaniboni, F., Tinti, S.: Tsunami hazard for the city of Catania, eastern Sicily, Italy, assessed by means of worst-case credible tsunami scenario analysis (WCTSA). Nat. Hazards Earth Syst. Sci. 11, 1217–1232 (2011) 15. Zaniboni, F., Pagnoni, G., Gallotti, G., Paparo, M.A., Armigliato, A., Tinti, S.: Assessment of the 1783 Scilla landslide–tsunami’s effects on the Calabrian and Sicilian coasts through numerical modeling. Nat. Hazards Earth Syst. Sci. 19, 1585–1600 (2019) 16. GEBCO: https://www.gebco.net/data_and_products/gridded_bathymetry_data/ 17. SRTM: https://cgiarcsi.community/data/srtm-90m-digital-elevation-database-v4-1/ 18. EMODNET: https://portal.emodnet-bathymetry.eu/

Seismic Scenario and People Exposure for Blagoevgrad Region, Bulgaria Petya Trifonova , Dimcho Solakov , Stela Simeonova, Metodi Metodiev , and Stefan Florin Balan

Abstract Present research analyses the human exposure at one of the most dangerous earthquake zones in Bulgaria-Blagoevgrad region and propose a detailed seismic scenario for the main city. Seismic hazard is modelled using GIS and overlaid with one square kilometer grid of population distribution in order to determine the population exposure in the region. We define a parameter called “population exposure index” (PEI) which has five classes: Minor, Low, Moderate, High and Major. As was expected, the seismic hazard levels of Blagoevgrad region are in the upper part of the classification scale. The total population in the Blagoevgrad region (NUTS II) is around 323,000 people. Results show that more than 130,000 people are exposed to the highest level of seismic hazard. City of Blagoevgrad gathers nearly 22% of the population in the region. A specially developed seismic scenario for the city accounting the soil conditions as well is used for detailed assessment of the people exposed to seismic hazard. The obtained values of Peak Ground Acceleration (PGA) varying between 0.29 and 0.45 g are crossed with the population living in each building to determine the levels of population exposure. Our results show that people living in 398 buildings are majorly exposed to the seismic hazard in Blagoevgrad city. Another 1465 buildings are determined as highly exposed to this threat. Delineation of these buildings might be very important for the regional authority and focusing on the prevention of possible earthquake effects. Keywords Seismic risk · Seismic scenario · Population exposure · Blagoevgrad · Bulgaria

P. Trifonova (B) · D. Solakov · S. Simeonova · M. Metodiev National Institute of Geophysics, Geodesy and Geography—Bulgarian Academy of Sciences, Acad. G. Bonchev Str., Bl. 3, 1113 Sofia, Bulgaria e-mail: [email protected] S. F. Balan National Institute for Research and Development for Earth Physics, Strada C˘alug˘areni 12, 077125 M˘agurele, Romania © The Author(s), under exclusive license to Springer Nature Switzerland AG 2021 N. Dobrinkova and G. Gadzhev (eds.), Environmental Protection and Disaster Risks, Studies in Systems, Decision and Control 361, https://doi.org/10.1007/978-3-030-70190-1_20

293

294

P. Trifonova et al.

1 Introduction The evaluated seismic hazard for Bulgaria in terms of Peak Ground Acceleration (PGA) calculated for a 475 years return period [1] shows that the highest ground motion levels are predicted for Blagoevgrad (Krupnik-Kresna) region, where the calculated PGA values exceeds 0.2 g and in some places 0.3 g (Fig. 1). The Krupnik-Kresna area which occupies the north-western part of the investigated region is considered to be one of the most dangerous earthquake zones in Bulgaria and the Mediterranean as well. The high seismic activity in this area is associated with the Struma fault system. Two of the strongest earthquakes in Europe in the twentieth century occurred in this area—the earthquakes of April 4, 1904 (MS = 7.1 and MS = 7.8). The second earthquake was felt even in Budapest. The villages of Krupnik and Simitli were completely destroyed. Serious damage has been reported in the towns of Dzhumaya (Blagoevgrad nowadays), Bansko and Razlog. At Dzhumaya (Blagoevgrad) damage was widespread and in places serious, particularly in the lower part of the town, where about 100 houses damaged by the first shock were ruined, including one mosque, three minarets, the barracks and the military hospital [2].The earthquake activated a fault line with a sub-parallel direction. There were also huge cracks in the neighbouring Brezhani graben. The subsidence at the Struma River was about 1.5–2 m, where a big step was formed. From the left bank of the Struma River the rupture was about 2 m high. In this region, where the thickness of the earth’s crust is significant (below the western part of the Rhodope massif it reaches a depth of 40–50 km), originate the deepest earthquakes in Bulgaria.

Fig. 1 Seismic hazard map in PGA for Bulgaria for a 475 years return period [1]

Seismic Scenario and People Exposure for Blagoevgrad …

295

In this research, the term risk follows the definition by the United Nations (UNDRO) and refers to the expected losses from a particular hazard to a specified element at risk in a particular future time period. Losses may be estimated in terms of human lives, buildings destroyed or in financial terms” [3]. In our case, as far as we analyse the population exposure to earthquakes, risk is connected with potential human losses (casualties) resulting from seismic hazard. One of the main parameters needed for calculating the casualties is the population exposure. It refers to the human occupancy of hazard zones, or the population present within the hazard area that would be potentially directly affected by an event [4].

2 Methodology Mathematically, the risk is a function of hazard occurrence probability, element at risk (population) and vulnerability. We accept the hypothesis that the three factors explaining the risk are multiplying each other [5] which means that if the hazard is 0 (null), then the risk is 0 (null). Theoretically, the risk is also null if nobody lives in an area exposed to hazard and if the population is invulnerable. In the present research we use the census-based resident population (i.e. night-time) and do not account for spatio-temporal variation of population distribution in urban areas. Population exposure is modelled by crossing the seismic hazard and population living in the potentially affected area. We determine a parameter called “population exposure index” (PEI) which has five classes: Minor, Low, Moderate, High and Major. The input data are: (1) seismic hazard for the territory of Bulgaria [1] re-calculated according to the USGS global Vs30 model (the time-averaged shear-wave velocity to 30 m depth) (https://earthquake.usgs.gov/data/vs30/) and (2) one sq. km. population grid of the Republic of Bulgaria resulted from the census performed in 2011 (https://www.nsi.bg/en/content/12309/population-grid-1-sqkm-census-2011). Using the population grid rather than total number of population allows direct comparison between big cities and small villages (less populated towns have the same weight as more populated ones). Similar analysis is performed by [6] but they use as input data the population density in Bulgaria which is calculated for each settlement using the total number of population living there and the area of the village/town/city. In this way, the population is artificially distributed over the entire area on the one hand and on the other hand the settlements that occupy a large area artificially reduce the value of the calculated density. First, both variables are classified in five levels using equal interval classification schemes (Tables 1 and 2): The number of levels and method for classifying the resultant map were chosen according to several criteria such as the optimum number of levels for visual representation or the number and level of errors between them.

296 Table 1 Population density levels

Table 2 Seismic hazard levels

P. Trifonova et al. Population density value (pers./sq.km)

Population density levels

0

1

1–100

2

101–1000

3

1001–10,000

4

>10,000

5

Seismic hazard and soil conditions PGA[g]*Vs30

Seismic hazard levels

≤0.10

1

0.1–0.14

2

0.141–0.18

3

0.181–0.25

4

≥0.26

5

The combination of both seismic hazard and population density provides the population exposure as an element of risk. A hypothesis is made that similar to the risk it follows the multiplicative formula that could be simplified as Eq. (1): PopE x p E = H E . Pop E .

(1)

where: PopExpE is the population exposure for the single spatial element E, one square kilometer in size. H E is the seismic hazard class for the single spatial element E determined accord. to Table 1. PopE is the population density class for the single spatial element E determined according to Table 2. As was expected, the seismic hazard levels of Blagoevgrad region are in the upper part of the classification scale (Fig. 2). The Krupnik-Kresna area is fully covered with the highest 5th level and the remaining part falls within level 4 interval limits. Only several small region in the south-east direction can be classified as having moderate level 3 values.

3 People Exposure Index in Blagoevgrad Region Calculation of the population which is exposed to seismic hazard is performed in a gridded spatial network with a single element E with dimensions 1 × 1 km.

Seismic Scenario and People Exposure for Blagoevgrad …

297

Fig. 2 Levels of the seismic hazard obtained for Blagoevgrad region according to classification scheme in Table 2

To determine the population exposure index (PEI) results obtained using Eq. 1 are also classified in five classes from 1 to 5: Minor, Low, Moderate, High and Major. Results, obtained for Bulgaria are given in the map of Fig. 3. According to those results, Sofia, Plovdiv and Blagoevgrad are the Top 3 municipalities with the major PEI level in Bulgaria. In the three districts live around 1.8 million people. A detailed breakdown for Blagoevgrad region (NUTS II classification) is given in Table 3. More than 130,000 people are exposed to the highest level of seismic hazard in the region. In the list of Table 3 are four of the major towns in this area. Spatial distribution of the results is displayed in the Fig. 4. Results obtained by [5] assessing global exposure and vulnerability towards natural hazards reveal that human vulnerability is mostly linked with country development level and environmental quality. Country development combines different indicators and parameters, but in the case of an earthquake the most important is the state of the norms for design and control over construction. The first building code in Bulgaria defining regulations and associated standards intended to control aspects of the design, construction, materials, etc. that are necessary to ensure human safety and welfare (including resistance to collapse and damage) is from 1956. It was updated several times and now Bulgaria has adopted modern standards according to the requirements of Eurocode 8.

298

P. Trifonova et al.

Fig. 3 Spatial distribution of the population exposure index obtained for the territory of Bulgaria

Table 3 Settlements with major PEI level in Blagoevgrad region Settlement name

Type

Population

Max PEI

1

Blagoevgrad

Town

69,178

5

2

Sandanski

Town

24,908

5

3

Simitli

Town

6334

5

4

Izgrev

Village

1025

5

5

Yakoruda

Town

2539

5

6

Belitsa

Town

3024

5

7

Razlog

Town

11,423

5

8

Bachevo

Village

1638

5

9

Krupnik

Village

2147

5

10

Kresna

Town

3286

5

4 Seismic Scenario for the City of Blagoevgrad From the results obtained in the region (Fig. 4) it is seen that the city of Blagoevgrad has major PEI index which means that people living there are highly exposed to the seismic hazard. Due to this, a detailed study of the city is performed to outline which are the most exposed places and how many people are living at highest risk.

Seismic Scenario and People Exposure for Blagoevgrad …

299

Fig. 4 Spatial distribution of the population exposure index (PEI) obtained for Blagoevgrad region. Each settlement is colored according to the maximum value of PEI calculated for its inner elements according to Eq. 1. Displayed are the names of municipalities

For obtaining the seismic hazard of the city of Blagoevgrad a deterministic earthquake scenario is used [7]. The scenario is defined on the base of the seismogenic potential of the causative faults defined in the SHARE project [8]. The fault system around the city of Blagoevgrad is complicated, comprising many inherited faults from the late orogenic stages. The active fault’s model of the region based on SHARE is presented in Fig. 5. The scenario considers earthquake with magnitude MW 6.5 on the closest to the city active fault (south of it, Fig. 5). The ground motion (in PGA), is calculated using Eq. 2: G M(x, y) =

6  i=1

wi gi (x, y)

(2)

300

P. Trifonova et al.

Fig. 5 Spatial pattern of seismicity (historical and instrumental earthquakes with MW > 3.0) and active faults map for the Blagoevgrad region (modified from [7])

where GM(x,y) is the evaluated ground motion at point with coordinates (x,y). gi (x,y) is the median ground motion predicted by i−th GMPE (from 6 GMPE’s for active shallow crustal tectonic regime with different weights) and wi is the corresponding weight. Results for the obtained PGA in the urban area very between 0.29 and 0.45 with the grater area being in the upper scale levels. They are presented in Fig. 6. The scenario maps account for soil amplification effects using the geotechnical zonation of the considered urban area. Representation of the soil properties is defined by the engineering parameter Vs30 —average shear-wave velocity in the upper 30 m of the soil/rock profile. The Vs30 values for the city of Blagoevgrad are based on the results published in [9].

5 Distribution of Buildings and Population in the City of Blagoevgrad According to the last census performed in 2011, the total number of population in the Blagoevgrad city is 69,178 people. They are living in 5994 residential buildings. The distribution of buildings, by number of floors and total floor area—TFA (m2 ) is given in Table 4. Most of the buildings are low-rise constructions with up to 3 floors but there are also 806 buildings which are above 6 floors. These are the homes of 73% of the people. According to the cadaster information most of the buildings were built in the period 1966–1977. The distribution of buildings by typologies shows that main part of the building stock are masonry structures, unreinforced masonry with wooden beams without reinforced concrete belts, not framed by columns. There is also a significant

Seismic Scenario and People Exposure for Blagoevgrad …

301

Fig. 6 Deterministic earthquake scenario for the city of Blagoevgrad in Peak Ground Acceleration (PGA). Obtained PGA very between 0.29 and 0.45 g with the grater area being in the upper scale levels, colored in red Table 4 Distribution of buildings (by pieces and TFA) in different levels of damage Residential buildings

1–3 floors

4–6 floors

7–10 floors

Above 10 floors

Number of buildings

4074

1483

423

14

Number of buildings (%)

67

24

7

2

Total floor area

(m2 )

Total floor area (%)

739,450

1,576,673

791,471

67,348

23

50

25

2

302

P. Trifonova et al.

number of masonry structures, unreinforced masonry with reinforced concrete slabs, beams and belts, not framed or framed by columns.

6 People Exposure in the City of Blagoevgrad Similar to the calculation of the people exposure in Blagoevgrad region we implemented the same methodology which crosses the seismic hazard and population living in the potentially affected area. According to the results obtained for the regional assessment which are described in the previous chapter seismic hazard and population density of Blagoevgrad follows within the major PEI category. That is why for the detailed assessment of the exposure inside the city we did not make a re-classification of the input variables. Population exposure is obtained for every single building using Eq. 3: PopE x p b = H b . Pop b . . . .

(3)

where: PopExpb is the population exposure for a single building. H b is the seismic hazard for the site of the building. Popb is the population for a single building (residents). Thus, we obtain a list of the population exposure for all residential buildings in the city of Blagoevgrad. To perform a quality assessment of the results we classified the results in the three upper categories: Moderate, High and Major (Fig. 7). Results of the assessment are summarised in Table 4. Majorly exposed are 21,820 people living in 398 buildings. All of those building have more than 3 floors. High levels of exposure have 1466 buildings with 32,284 residents. Both categories encounter for 75% of the Blagoevgrad citizens (Table 5).

7 Conclusion Determination of the population exposure index (PEI) in one of the most seismically active zone in Bulgaria–Blagoevgrad region shows that less than 30% of the settlements are characterised with Moderate (3) values of PEI. These are 85 small villages where the population is between 2 and 99 people. For all 14 towns, the obtained PEI values are High (4) and Major (5). Purpose of the calculated index is to give a qualitative information about the size of negative consequences which may occur in case of a strong earthquake. For quantitative analysis, measures of the exposure should be combined with the specific vulnerability of the exposed elements to seismic hazard, to estimate the quantitative risks and to determine the potential loses.

Seismic Scenario and People Exposure for Blagoevgrad …

303

Fig. 7 Population exposure in the city of Blagoevgrad. In red are marked places where population is majorly exposed to the effects of a potential earthquake

Table 5 Categories of the population exposure obtained for Blagoevgrad

Population exposure type

Number of buildings

Number of residents

Moderate

4129

15,074

High

1466

32,284

Major

398

21,820

An usable and realistic seismic scenario for the city of Blagoevgrad is proposed for detailed assessment of the people exposed to seismic hazard. The obtained values of Peak Ground Acceleration (PGA) varying between 0.29 and 0.45 g are crossed with the population living there to determine the levels of population exposure. The population of the Blagoevgrad city numbers about 70,000 people living in 5994 residential buildings. Most of them are low-rise constructions with up to 3 floors

304

P. Trifonova et al.

but there are also 806 buildings which are above 6 floors. We consider a deterministic earthquake scenario with magnitude MW 6.5 on the closest to the city active fault, located south of it. When combining the earthquake scenario with the distribution of residents in the city we obtain the population exposure as a parameter of the seismic risk. Our results show that people living in 398 buildings are majorly exposed to the seismic hazard in Blagoevgrad city. Another 1465 buildings are determined as highly exposed to this threat. In total, 75% of the Blagoevgrad citizens are largely exposed to a potential earthquake’s effect. The obtained results are intended to serve as input for developing of detailed earthquake damage scenarios for the settlements and can be used for disaster preparedness and emergency reaction planning. Acknowledgements Present work is supported by Contract No D01-282/17.12.2019 (Project "National Geoinformation Center (NGIC)" financed by the National Roadmap for Scientific Infrastructure 2017–2023. Development of the input seismic database is financed by the National Science Program “Environmental Protection and Reduction of Risks of Adverse Events and Natural Disasters”, supported by the Ministry of Education and Science of Bulgaria (Agreement № DO-230/06-12-2018).

References 1. Solakov et al.: Seismic risk management for buildings. Acad. Publ. house Prof. M. Drinov, Sofia. p. 195 (2019) 2. Watzof, S.: The earthquakes in Bulgaria. Report on the earthquakes felt in 1904. Central Meteorological Institute, Sofia (in Bulgarian, abstract in French) (1905). 3. Burton, I., Kates, R.W., White, G.F.: The Environment as Hazard, 2nd edn. Guilford Press. New York/London, 290 pp. (1993) 4. Freire, S., Aubrecht, C.: Integrating population dynamics into mapping human exposure to seismic hazard. Nat. Haz. Earth Syst. Sci. \textbf{12}, 3533–3543 (2012) https://doi.org/10. 5194/nhess-12-3533-2012 5. Dao, H., Peduzzi, P.: Global Risk And Vulnerability Index Trends per Year (GRAVITY), Phase IV: Technical annex and multiple risk integration, UNDP/BCPR, Geneva, Tech. Rep., 31 pp. (2003) 6. Solakov, D., Metodiev, M., Simeonova, S., Trifonova, P.: Population exposure index – an element of seismic risk assessment. In: 10th Congress of the Balkan Geophysical Society. Sofia (2019). https://doi.org/10.3997/2214-4609.201902659 7. Solakov, D., Simeonova, S., Raykova, P., Rangelov, B.: Earthquake scenarios for the city of Blagoevgrad. In: XXth International Multidisciplinary Scientific GeoConference Surveying, Geology and Mining, Ecology and Management – SGEM 2020, Albena, Bularia

Seismic Scenario and People Exposure for Blagoevgrad …

305

8. Basili, R., Kasteli,c V., Demirciogl,u M., Garcia Moreno, D., Nemser, E., Petricca, P., Sboras, S., Besana-Ostman, G., Cabral, J., Camelbeeck, T., Caputo, R., Danciu, L., Domac, H., Fonseca, J., García-Mayordomo, J., Giardini, D., Glavatovic, B., Gulen, L., Ince, Y., Pavlides, S., Sesetyan, K., Tarabusi, G., Tiberti, M., Utkucu, M., Valensise, G., Vanneste, K., Vilanova, S., Wössner, J.: The European Database of Seismogenic Faults (EDSF) compiled in the framework of the Project SHARE (2013). https://diss.rm.ingv.it/share-edsf/, doi: https://doi.org/10.6092/INGV. IT-SHARE-EDSF 9. Rangelov, B., Solakov, D., Dimovsky, St., Kisyov, A., Georgieva, B.: Mapping and digitalization of the ground conditions for the seismic hazard assessment. In: 19th “Days of Physics 2020”, Tech. Univ., Sofia, 28–30 of May, 2020

Informatics, Remote Sensing, High Performance Computing and GIS for Environmental Monitoring and Management

Forecasting the Propagation of HF Radio Waves Over Bulgaria Rumiana Bojilova and Plamen Mukhtarov

Abstract A new methodology for forecasting the propagation of HF radio waves by reflection from the ionosphere over Bulgaria in the absence of ionosonde data is presented. The proposed methodology contains three main parts. Based on the long-term ionosonde data an empirical model of the critical E region frequency (foE) has been built; the latter depends on the season, local time and the level of solar activity described by the solar radio flux at 10.7 cm wavelength (F10.7). The critical frequency of the F2-layer (foF2) and the maximum usable frequency at a propagation of 3000 km (MUF3000) are obtained by means of the proposed empirical relationships between these two critical frequencies and the Total Electron Content (TEC). Based on these three ionospheric characteristics a modeled electron density profile is compiled by using the method of Di Giovani-Radicella (Giovanni and Radicella in Adv Space Res 10:27–30, 1990 [1]). The constructed in this way electron density profile allows calculating the lowest and maximum usable frequency at a given distance up to 500 km according to the theory of radio wave propagation in the ionospheric plasma, namely the equivalence theorem and the secant law, as well as the law of reduction of the group velocity of propagation depending on the ionosphere electron density. Keywords HF radio waves · Electron density profile · Empirical model

1 Introduction Radio communication and broadcasting systems can be divided into either controlled by the ionosphere, as in HF sky-wave systems, or simply affected by it, as in transionospheric radio signals, used by the Global Navigation Satellite Systems (GNSS) for communication and navigation systems. In the former case, the ionosphere is actually an inflexible part of the system; while in the latter case, the ionosphere is the largest source of errors, i.e. errors in GNSS positioning, timing and navigation. In R. Bojilova (B) · P. Mukhtarov National Institute of Geophysics, Geodesy and Geography, Bulgarian Academy of Sciences (NIGGG-BAS), Acad. Georgi Bonchev str., Bl. 3, 1113 Sofia, Bulgaria e-mail: [email protected] © The Author(s), under exclusive license to Springer Nature Switzerland AG 2021 N. Dobrinkova and G. Gadzhev (eds.), Environmental Protection and Disaster Risks, Studies in Systems, Decision and Control 361, https://doi.org/10.1007/978-3-030-70190-1_21

309

310

R. Bojilova and P. Mukhtarov

both instances however, an account of the ionosphere is at least beneficial to system design and operation [2]. The present study is directed to the first case systems and particularly for not very long-range ones because mainly the propagation of HF radio waves (HF is usually taken as the frequency band of 2–30 MHz) by reflection from the ionosphere over Bulgaria will be considered. The basic aim of the radio wave propagation forecasting is to improve communication systems and give recommendation about the future reliability of frequency bands propagated via the ionosphere. The impact of the ionosphere on the radio wave propagation has been well known since the past several decades and is described by the relationship between plasma frequency and signals which propagate within that medium. A wide range of telecommunication systems have been developed and many have been fielded for operational use. While many of the systems received their impetus through military necessity, the utility of telecommunications is evident in virtually all aspects of human activity [3]. In civilian society the HF communication is used predominantly by radio amateurs and for international broadcasts by different government organizations. The use of HF communication is preferred due to its relative simplicity, its capability to provide long range communication at low power without repeater base stations, its ease of development and its low cost. The disadvantage is low throughput of HF channel and significant influence of propagation environment (ionosphere variability) on the quality of transmission. It is known that to provide accurate prediction method of HF radio systems different ionospheric models are needed. The models considered are for the ionospheric characteristics and electron density profile models, a propagation model, transmission loss and characteristics of the radio noise models [4]. This means that for producing different ionospheric models, particularly over a limited region, long-term ionosonde measurements of the main ionospheric characteristics over the considered region have to be available. The basic aim of the present study is to offer a new methodology for forecasting the propagation of HF radio waves by reflection from the ionosphere over Bulgaria in the absence of data from a vertical sounding station.

2 Empirical Model for FoE Prediction This study uses the data from vertical sounding of the ionosphere from the ionosonde station Sofia at the National Institute of Geophysics, Geodesy and Geography, Bulgarian Academy of Sciences for the period of 1995–2014, as well as the data for solar and geomagnetic activity from the National Oceanic and Atmospheric Administration, USA. The basis of the model idea for foE prediction is connected with the stable dependence of the E-region maximum electron density in day time conditions   Nm E cm−1 = 1.24.104 f oE 2 [MHz]

(1)

Forecasting the Propagation of HF Radio Waves Over Bulgaria

311

from the cosine of the solar zenith angle, which is due to the fact that the electron density in the E region with sufficient accuracy obeys Chapman’s theory [5]. Figure 1 shows the dependence of the monthly medians of the maximum electron density NmE on the solar zenith angle cosχ for four years: high solar activity in 1999, declining phase in 2003, low solar activity in 2008 and rising phase in 2011. The wanted dependence with sufficient accuracy is represented by a second degree polynomial calculated by a least squares method and shown by solid line in the plots. It is obvious that this dependence is different in the presented four years which is due to the solar cycle impact on the ionizing radiation. Pancheva and Mukhtarov [6] demonstrated that the dependence of N mE on the solar zenith angle varies not only on the solar activity, but also the season and is different before local noon compared to that after noon. 20

16

1999 NmE*1.24e-4[cm-1]

NmE*1.24e-4[cm-1]

16 12 8 4

12

8

4

0

0 0 20

0.2

0.4 0.6 cosChi

0.8

0

1 16

2008 NmE*1.24e-4[cm-1]

16 NmE*1.24e-4[cm-1]

2003

12 8 4 0

0.2

0.4 0.6 cosChi

0.8

1

0.4 0.6 cosChi

0.8

1

2011

12

8

4

0 0

0.2

0.4 0.6 cosChi

0.8

1

0

0.2

Fig. 1 Solar zenith dependence of the critical frequency foE during different solar activity conditions; 1999 (upper left plot) shows high solar activity; 2003 (upper right plot) declining phase; 2008 (bottom left plot) low solar activity, and 2011 (bottom right plot) rising phase

312

R. Bojilova and P. Mukhtarov

The most ionospheric modelers use the sunspot number (number of dark spots on the solar disc) and the solar radio flux at 10.7 cm wavelength (F10.7) as solar indices, since both can be observed from the ground, long data records exist and they can be predicted. These indices together with their 6-month predictions are regularly published by [7]. Some preliminary experiments were performed with both indices and the results revealed that particularly for the considered last solar cycle F10.7 describes better the ionospheric variability. Thus, in the present study F10.7 is used as a proxy for the solar activity. It is known however that the ionosphere behaves differently at the rising and declining phase of the solar cycle at one and the same F10.7 [8–10]. To include this ionospheric feature in the model an additional parameter kF10.7 is used which describes the linear rate of change of F10.7. The idea for describing the solar activity by such two parameters, i.e. by the level of solar activity and its tendency, was introduced for the first time by [11] in modeling the monthly median critical frequency of the ionospheric F-region, foF2, above Sofia. It was found that the inclusion of the kF10.7 in the monthly median foF2 model decreased its mean standard deviation by ~0.5 MHz. Later this idea was successfully applied in building of global background TEC model reported by [12]. In this study the smoothed time series of F10.7 is used obtained by a sliding 11-month window and denoted as sF10.7. Usually the seasonal dependence the electron density in the E region is described only by both annual and semiannual components. In this study however the seasonal components with periods shorter than 6 months have also some contribution even though they are weaker than annual and semiannual ones. Based on the above mentioned clarifications the empirical model of the maximum electron density N mE separately for for-noon N mEfn and afternoon N mEan conditions can be described by the following functions: Nm E f n (s F10.7, k F10.7, month, cos χ)   = a0 f n + a1 f n s F10.7 + a2 f n s F10.72 + a3 f n k F10.7 + a4 f n k F10.72 + a5 f n s F10.7k F10.7 ⎛ ⎞



4  2π 2π month + b2k f n sin k month ⎠ b1k f n cos k × ⎝ b0 f n + 12 12 k=1   (2) × c0 f n + c1 f n cos χ + c2 f n cos χ 2 Nm Ean (s F10.7, k F10.7, month, cos χ )   = a0an + a1an s F10.7 + a2an s F10.72 + a3an k F10.7 + a4an k F10.72 + a5an s F10.7k F10.7 ⎞ ⎛



4  2π 2π month + b2kan sin k month ⎠ b1kan cos k × ⎝b0an + 12 12 k=1   (3) × c0an + c1an cos χ + c2an cos χ 2

The expression in the first right hand bracket of the above formulas (2) and (3), i.e. the Taylor series expansion up to degree of 2, represents the solar activity term which modulates the seasonal and day time behavior of the ionosphere. The seasonal

Forecasting the Propagation of HF Radio Waves Over Bulgaria

313

term (expression in the second right hand bracket) includes 4 sub-harmonics of the year, i.e. annual, semiannual, 4- and 3-month components; it modulates the day time behavior of the ionosphere. The expression in the last, third right hand bracket is again a Taylor series expansion up to degree of 2, describing the solar zenith angle day time variability of the E-region shown in Fig. 1. The empirical foE model described by (2) and (3) contains 162 constants and they are determined by the least squares fitting techniques, i.e. at minimizing the root mean square error of the model relative to the data. Figure 2 presents the comparisons between the model foE results (thin line with dots) and the observations (thick line) for two years with different solar activity; 2000 for high solar activity (upper plot) and 2007 for low solar activity (bottom plot). The root mean square error (RMSE) calculated for the whole data set is 0.083 MHz, i.e. this is a completely satisfactory result. Data Model

2000

foE[MHz]

4

3

2

1 0

24

48

72

96

120

144

168

192

216

240

264

288

Months Data Model

2007

foE[MHz]

4

3

2

1 0

24

48

72

96

120

144

168

192

216

240

264

288

Months

Fig. 2 Comparison between the model data (thin line with dots) and observations (thick line) for 2000 (upper plot, high solar activity) and 2007 (bottom plot, low solar activity)

314

R. Bojilova and P. Mukhtarov

3 Empirical Model for Calculating FoF2 and MUF3000 Based on TEC Data The reason for building such a model is the fact that the main part of the TEC is formed by the region of the electron density profile around the maximum of the F region and it is reasonable to expect a well-defined relationship between the maximum electron density N mF2 [cm−1 ] = 1.24 104 (foF2[MHz])2 and the TEC [13]. The regressions shown in Figs. 3 and 4 indicated that the wanted dependence between foF2 or MUF3000 on TEC can be described as a second degree polynomial. f oF2 ≈ a f (month, U T )T EC 2 + b f (month, U T )T EC + c f (month, U T ); MU F3000 ≈ am (month, U T )T EC 2 + bm (month, U T )T EC + cm (month, U T )

(4)

The constants of the above models are determined by least squares fitting techniques based on the foF2 and MUF3000 data from the ionosonde station Sofia and TEC data from the Center for Orbit Determination of Europe (CODE) for the closest to Sofia point (42.5° N, 25° E) for the period of 1999–2014. The comparison between the model and measured foF2 and MUF3000 values reveals that the mean error is practically zero while the RMSE for foF2 is e 0.55 MHz and for MUF3000 is 2.1 MHz. For practical purposes this error can be considered as permissible. Further, a mathematical procedure for obtaining foF2 and MUF3000 values from vertical TEC measurements has been proposed very recently by [14]. The mathematical relationships between F2-layer characteristics and vertical TEC values have been derived using South-African co-located ionosonde and GNSS stations but the same procedure can be applied successfully in each region where such data are simultaneously available. Evaluating the error of this transformation the authors found RMSE for foF2 to be 0.754 MHz, while for MUF3000 is 2.782 MHz; hence both errors are larger than the obtained ones in this study. Figure 5 presents the comparison between the measured (red line) and reconstructed by TEC (blue line) ionospheric characteristics foF2 (left panel) and MUF3000 (right panel) for the period of time 21–31 October 2003. The considered time interval was selected due to the registered extremely strong geomagnetic storms, known as Halloween geomagnetic storms, in the period of 29–30 October 2003. It can be seen from Fig. 5 that the coincidence between the measured and modeled values is satisfactory not only in the quiet days of the month but also in the disturbed ones. The mean foF2 error for the considered time interval is 0.2 MHz while that for MUF3000 is 1.55 MHz; the corresponding RMSE for foF2 is 0.73 MHz and for MUF3000 is 3.47 MHz. The RMSE values of the two ionospheric parameters in the above discussed example do not differ significantly from the errors for whole interval of years (1999–2014).

Forecasting the Propagation of HF Radio Waves Over Bulgaria

315

Fig. 3 Regression dependences of foF2 on TEC for the calendar months January, March, May and September for 10 UT (left column panels) and 22 UT (right column panels) obtained from all data (1999–2014)

316

R. Bojilova and P. Mukhtarov

Fig. 4 The same as Fig. 3 but for regression dependences of MUF3000 on TEC

Forecasting the Propagation of HF Radio Waves Over Bulgaria

317

Fig. 5 (left panel) Comparison between the reconstructed by TEC hourly values of foF2 (blue line) and the measured ones (red line) for the period of 21–31 October 2003; (right panel) the same as the left panel but for MUF3000

The use of the data from CODE for producing an empirical model for the foF2 prediction based on the TEC data is caused by the available long enough time period when there simultaneous ionosonde and TEC data. However, the data from CODE arrive with a delay of one day. We note that there is a possibility to use data arriving in a real time from a single GNSS receiver operating in the territory of Sofia, denoted as station SOFI. The methodology for calculating the vertical TEC from the raw data has been reported by [15].

4 Model Electron Density Profile for a Given Time After defining the critical frequencies based on the above described methodology, the construction of the electron density profile has been performed according to [1]. For day-time conditions: h ≤ hm E ; N (h) = Nm E sech2 h > hm E ;



hm E − h 2B Eb







h m F2 − h m E h − hm E sech2 Nm E − Nm F sech2 2B F B Et



h h − h m F2 − h m F2 mE sech2 + Nm F2 − Nm E sech2 2B Et BF

N (h) =

(5)

318

R. Bojilova and P. Mukhtarov

and respectively for night-time conditions: f oE = 0;



N (h) = Nm F2 sech

2

h m F2 − h Bf



(6)

This model is based on the representation of the electron density profile with hyperbolic secant functions [13]. The values of the three critical frequencies are sufficient to calculate the model electron density profile for heights up to the height of the F-region maximum; the latter is calculated by the given below formulas. The maximum electron concentration of both the E and F layers are determined directly by the critical frequencies. The height of the E-layer maximum is assumed to be fixed 120 km. The model uses the empirically determined by [16] dependence of the height of the F2-layer maximum hmF2 on the parameter M3000F2 = MUF3000F2/foF2, the ratio foF2/foE and the correction value Dm according to the given below formulas. The parameter Bf , related to the half-thickness of the ionospheric F-layer, is expressed by an empirical dependence of the derivative of the electron profile dN/dh, which in turn is expressed by the values of foF2 and M3000F2. The half-thicknesses of the E-layer (BEb -below and BEt -above the maximum of the E-region) take fixed values.   Nm E cm−2 = 1.24.104 f oE 2 [MHz]   Nm F2 cm−2 = 1.24.104 f oF22 [MHz] h m E = 120 km; B Eb = 7.5 km; B Et = 7 km h m F2 = 1470

(7)

Mf − 176 M3000F2 + Dm

M3000F2 =

MU F3000 f oF2

B f = 0.04774.1011 

f oF22 d N dh

0.0196M3000F22 + 1 1.2967M3000F22 − 1 d N = 1.109 exp−3.467 + 0.857 lg f oF22  + 2.02 lg(M3000F2) dh M f = M3000F2

(8)

Forecasting the Propagation of HF Radio Waves Over Bulgaria

319

5 Calculation of the Radio-Path Parameters The task for the calculation of radio paths contains the determination of the frequency range at which a radio communication at a given distance between the two radio communication points can take place. The frequency range is limited by the lowest and maximum usable frequencies (LUF and MUF respectively), which depend on the distance of the radio communication and the state of the ionosphere at the respective time. The calculation methodology is based on the theory of radio wave propagation without taking into account the influence of Earth’s magnetic field. We note that this simplification does not reduce the accuracy of the calculation. Upon entering the ionosphere, radio waves suffer the so-called magnetic-ion splitting, i.e. they break up into two separate waves that propagate at different speeds. Traditionally, the data from the ionograms take into account ionospheric characteristics (e.g. critical frequencies) of the reflections of the ordinary wave. The propagation of an ordinary wave is not affected by the presence of the Earth’s magnetic field [17, 18]. The geometry of the ionospheric reflection is illustrated schematically in Fig. 6. In the case of vertical radio-wave propagation, the virtual reflection height of the ordinary wave h vert depending on the frequency f is obtained according to the formula (9) below. Fig. 6 Basic geometrical ratios at oblique reflection of radio signals from the ionosphere h'

ϕ0

z0

r

θ a

320

R. Bojilova and P. Mukhtarov

h ver t (

z0 

f) = 0

80.8

dz 1 − 80.8 Nf(z) 2

N (z 0 ) =1 f2

(9)

where N(z) is the altitude electron density profile. The height z0 is that height at which the denominator of the sub-integral function is canceled, which means the so-called “full internal reflection” when the radio wave changes its direction of propagation. When the electron density profile has a maximum at altitude of z0 then the integral tends to infinity for the corresponding frequency. The ratio between the frequencies and virtual heights in the cases of vertical and oblique propagation is given by the equivalence theorem and the secant low: h ob ( f ob ) = h ver t ( f ver t ) sec φ0 f ob = f ver t sec φ0

(10)

The secant theorem shown in the formula reflects the following regularity. An oblique ray with a frequency f ob incident on the ionosphere at an angle ϕ 0 is reflected by the ionosphere at the same height as a vertical ray with a frequency f vert . Based on Fig. 6 it follows: tgφ0 =

a sin(θ ) h ver t + a(1 + cos(θ ))

(11)

where a is the radius of the Earth (about 6370 km), the angle θ is half of the central angle between the two endpoints of the radio path. If the length of the radio path on the Earth’s surface is denoted by r, then its angle θ in radians can be calculated by: θ [rad] =

r 2a

(12)

The calculation of a given radio-path begins with calculating the model ionogram h vert (f vert ) by numerical integration of the model electron density profile. The left panel of Fig. 6 shows examples of electron density profiles at noon (solid line) and midnight (dash line) conditions (electron density is expressed in plasma frequencies) while the right panel presents the corresponding model vertical ionograms. After calculating the model vertical ionograms, the reflection frequencies of the respective virtual heights at oblique propagation are determined according to the secant theorem. As a result the so-called “oblique ionogram” h’ob (f ob ) dependence at a given distance of the radio-path is found. Figure 8 shows the oblique ionograms for three distances based on the electron density profiles presented in the left panel of Fig. 7; the midnight ionograms can be seen in the left panel while the noon time ones in the right panel.

Forecasting the Propagation of HF Radio Waves Over Bulgaria 400

400

Night 00LT Noon 12LT

Night 00LT Noon 12LT

300 Virtual height [km]

300 Height [km]

321

200

200

100

100

0

0 0

1

2 3 4 5 Plasma frequency [MHz]

6

0

7

1

2 3 4 5 Frequency [MHz]

6

7

Fig. 7 (Left panel) model electron density profiles at noon (solid line) and midnight (dash line) conditions; (right panel) the corresponding model vertical ionograms 500

500

Oblique propagation 00LT Distance 0-100 km 250 km 500 km

400

Virtual height [km]

Virtual height [km]

400

Oblique propagation 12LT Distance 0-100 km 250 km 500 km

300

200

MUF 500km MUF 250km MUF 0- 100km

100

300

200

MUF 500km MUF 250km MUF 0- 100km

100

LUF 500km LUF 250km LUF 0- 100km

LUF 500km LUF 250km LUF 0- 100km

0

0 0

1

2

3

4

5

Frequency [MHz]

6

7

8

0

1

2

3

4

5

Frequency [MHz]

6

7

8

Fig. 8 Oblique ionograms based on the model profiles shown in the left panel of Fig. 7 at radio communication distances 0–100 km (dots), 250 km (crosses) i 500 km (diamonds) for midnight (left panel) and noon time (right panel) conditions

Figure 8 illustrates also the determination of the minimum and maximum usable frequencies at a given distance of the radio communication distance. At distances up to 100 km the reflection is practically vertical. The main condition is the reflection from the F region to be secured. The reflections from the E region (these are the frequencies lower than the area where the ionogram is interrupted, corresponding to foE) are practically not used due to the strong absorption of the radio waves. Figure 9 presents the diurnal variability (LT dependence) of the calculated LUF (dash line) and MUF (solid line) for the considered in Fig. 8 three radio-path distances for a given date of 09 June 2014 according to the above presented methodology. At

322

R. Bojilova and P. Mukhtarov 10

Frequency [MHz]

8

LUF 0-100 km LUF 250km LUF 500km MUF 0-100 km MUF 250 km MUF 500 km

6

4

2

0 0

3

6

9

12

LT

15

18

21

24

Fig. 9 Diurnal variability (LT dependence) of the LUF (dash line) and the MUF (solid line) calculated for three radio-path distances and for a given date of 09 June 2014

nigh conditions there is no foE, the electron density and respectively the radio wave absorption is low. Actually, the lowest usable frequency at night conditions is not determined by the state of the ionosphere but by technical considerations. Most of the shortwave transmitters have a frequency band from 1.5 to 30 MHz but the frequencies below 2 MHz fall within the field of broadcasting. Due to this reason the LUF at night conditions is accepted to be 2 MHz.

6 Conclusion The methodology proposed in the present paper is intended for servicing users who carry out long-distance radio communications, as radio amateurs and government organizations. The methodology is suitable for organizing automatic data processing. In order to make a forecast for the propagation of radio waves at a given hour of the day, only three ionospheric characteristics are needed: foE, foF2 and MUF3000. In the presence of working vertical ionosonde station the measured values of the above mentioned ionospheric characteristics can be used. A significantly new element of

Forecasting the Propagation of HF Radio Waves Over Bulgaria

323

this study is offering a methodology for forecasting the propagation of HF radio waves by reflection from the ionosphere over Bulgaria in a case when such equipment is absent. Then two types of new empirical models are provided: (i) for predicting foE that depends on the solar activity, and (ii) for estimating foF2 and MUF3000 based on the TEC data for the same or the closest point to already not working ionosonde station. Summarizing it is worth noting that the presented new methodology for forecasting the propagation of HF radio waves by reflection from the ionosphere significantly increases the possibility for countries where the vertical sounding stations no longer operate. The International GNSS Service (IGS) has become a powerful TEC measurement tool to investigate the global and regional ionospheric structures owing to its continuous, easy operation, and worldwide distributed receivers. The IGS offers low cost information characterized by its accuracy, high temporal and spatial resolution, and availability. Acknowledgements The present work is supported by the Bulgarian Ministry of Education and Science under the National Research Programme “Young scientists and postdoctoral students” approved by DCM № 577/ 17.08.2018. The presentation of the results is financed by Contract No D01-282/17.12.2019—Project “National Geoinformation Center (NGIC)” funded by the National Roadmap for Scientific Infrastructure 2017–2023 of Bulgaria.

References 1. Di Giovanni, G., Radicella, S.M.: An analytical model of the electron density profile in the ionosphere. Adv. Space Res. 10(11), 27–30 (1990) 2. Barclay, L. (ed.): Propagation of Radio Waves. The Institution of Engineering and Technology, UK (2003) 3. Goodman, J.M.: Operational communication systems and relationships to the ionosphere and space weather. Adv. Space Res. 36, 2241–2252 (2005) 4. Hanbaba, R.: Performance prediction methods of HF radio systems. Annali Di Geofis. 41(5–6), 715–742 (1998) 5. Chapman, S.: The absorption and dissociative or ionizing effect of monochromatic radiation in an atmosphere on a rotating Earth. Proc. Phys. Soc. 43, 26–45 (1931) 6. Pancheva, D., Mukhtarov, P.: A diurnal asymmetry of the monthly median E-region critical freouency. Adv. Space Res. 22(6), 771–774 (1998) 7. Space Weather Prediction Center: https://www.swpc.noaa.gov/ 8. Huang, J-.N.: The hysteresis variation of the semi-thickness of the F2-layer and its relevant phenomena at Kokubunji, Japan. J. Atmos. Terr. Phys. 25, 647–658 (1969) 9. Gopal Rao, M.S.V., Sambasiva Rao, R.: The hysteresis variation in the F2-layer parameters. J. Atmos. Terr. Phys. 31, 1119–1125 (1969) 10. Apostolov, E., Alberca, L., Pancheva, D.: Long-term prediction of the foF2 on the rising and falling parts of the solar cycle. Adv. Space Res. 14(12), 47–50 (1994) 11. Pancheva, D., Mukhtarov, P.: A single-station spectral model of the monthly median F-region critical frequency. Annali Di Geofis. 39(4), 807–818 (1996) 12. Mukhtarov, P., Pancheva, D., Andonov, B., Pashova, L.: Global TEC maps based on GNSS data: 1. Empirical background TEC model. J. Geophys. Res. Space Phys. 118(7), 4594–4608 (2013)

324

R. Bojilova and P. Mukhtarov

13. Stankov, S.M., Jakowski, N., Heise, S., Muhtarov, P., Kutiev, I., Warnant, R.: A new method for reconstruction of the vertical electron density distribution in the upper ionosphere and plasmasphere. J. Geophys. Res. Space Phys. 108(A5) (2003) 14. Pignalberi, A., Habarulema, J.B., Pezzopane, M., Rizzi, R.: On the development of a method for updating an empirical climatological ionospheric model by means of assimilated vTEC measurements from a GNSS receiver network. Space Weather 17, 1131–1164 (2019) 15. Andonov, B.: Vertical total electron content and receiver bias calculations for Balkan Peninsula GNSS stations. C. R. Acad. Bulg. Sci. 70(12), 1719–1729 (2017) 16. Dudeney, J.R., Kressman, R.I.: Empirical models of the electron concentration of the ionosphere and their value for radio communications purposes. Radio Sci. 21(3), 319–330 (1986) 17. Cherny, F.B.: Radiowave propagation. M. Sov. Radio, 311–353 (in Russian) (1972) 18. Rawer, K.: Wave Propagation in the Ionosphere, vol. 5. Springer Science & Business Media (2013)

Wildfire Risk Reduction Based on Landscape Management Nina Dobrinkova, Carlos Trindade, Craig Hope, Chuck Bushey, Alexander Held, Ciaran Nugent, Georgios Eftychidis, Adrián Cardil, George Boustras, and Evangelos Katsaros

Abstract Wildfires occur traditionally in the southern situated countries of Europe. However, in the last few years northern states are experiencing wildland fires (vegetation fires) which sustain their propagation for more than few hours. Reasons for N. Dobrinkova (B) Institute of Information and Communication Technologies—BAS, acad. G. Bonchev str. Bl. 2, 1113 Sofia, Bulgaria C. Trindade Municipality of Mafra, Praça do Município, 2644-001 Mafra, Portugal C. Hope South Wales Fire and Rescue Service, Business Park, Forest View, Llantrisant, Pontyclun C72 8LX, UK e-mail: [email protected] C. Bushey International Association of Wildland Fire, Billings, MT, USA e-mail: [email protected] A. Held European Forest Institute, Platz der Vereinten Nationen 7, 53113 Bonn, Germany e-mail: [email protected] C. Nugent Forest Service, Department of Agriculture, Food and the Marine, Johnstown Castle Estate Co., Wexford Y35 PN52, Ireland e-mail: [email protected] G. Eftychidis Center for Security Studies (KEMEA), 4, P. Kanellopoulou str., 101 77 Athens, Greece e-mail: [email protected] A. Cardil Technosylva Inc., Parque Tecnológico de León, 24009 León, Spain e-mail: [email protected] G. Boustras · E. Katsaros Centre of Excellence in Risk and Decision Science (CeRiDeS), Diogenis Str 6, 2404 Nicosia, CY, Cyprus e-mail: [email protected] © The Author(s), under exclusive license to Springer Nature Switzerland AG 2021 N. Dobrinkova and G. Gadzhev (eds.), Environmental Protection and Disaster Risks, Studies in Systems, Decision and Control 361, https://doi.org/10.1007/978-3-030-70190-1_22

325

326

N. Dobrinkova et al.

this new situation have been discussed during the CMINE Wildfire Task Group meetings (DRIVER+ project with a mandate of one year—2019). In our paper we will summarize our main findings, which refer to two basic topics—landscape management and climate change. There are more and more parcels of land which have become abandoned because of numerous reasons, but the outcome is the same. The absent people are not cutting the trees, and the people do not bring their cattle or sheep to graze the grass and shrubs. This vegetation grows every spring and lies down on the ground every autumn as potential fire propagation fuel waiting for an ignition. Climate change is another global issue which is creating dangerous weather conditions with extreme high temperatures during the summer season plus mild winters, which is leading to wildfire occurrence in some parts of Europe year round. In our paper we will describe two case studies from Portugal and South Wales, having the same conclusions—no land management strategy and extreme weather are the best conditions for life threatening wildfires. Keywords Wildfires · Land management · Climate change

1 Overview of the Wildfires and Their Reoccurrence Pattern in Europe Wildfires are a serious and increasing threat throughout Europe. They occur as mega fires in the South and unprecedented fires in Northern Europe. Reason for this phenomenon is the decline of rural economies and agroforestry mosaics in Europe that create more continuous and dense forest landscapes. The European Union in the last 20 years has invested over 103 million euro in 56 wildfire related projects. The type of projects varied from Large Scale Integrated Projects (e.g., FIRE PARADOX, FUME) to smaller projects and individual Marie Sklodowska Curie Grants (e.g., FIRESCAPE, GRADIENT). Other research projects which emphasized the demonstration of effective forest fire management, were funded under the LIFE program (e.g., ENERBIOSCRUB, MONSERRAT), or under the Civil Protection Mechanism (e.g., PREDICATE, WUIWATCH). EU funding also targeted coordination actions between research institutions (e.g., PHOENIX, FORESTERRA), and cooperative actions among neighboring countries (e.g., HOLISTIC) Fig. 1 [1]. In general, most projects have been concentrated on research in Europe, particularly around the Mediterranean Basin, including non-EU countries from this area, but research was also carried out in other parts of the world. Projects were divided into 6 thematic areas corresponding to the sequence of forest fire risk management activities. The areas most addressed by EU research on forest fire were fire prevention, fire suppression and fire science. Less attention has been dedicated to research topics related to post-fire recovery and fire detection [1].

Wildfire Risk Reduction Based on Landscape Management

327

Fig. 1 The type of projects that EU has funded on the wildfire topic [1]

A more than three times increase in the period between January 1st and Dec. 9th, 2019 compared to the average number of wildfires for 2008–2018 decade has been recorded within the European Union EFFIS system as shown on Fig. 2 [2].

Fig. 2 Number of fires in the period Jan. 1st–Dec. 9th, 2019 (red) and average number of fires between 2008 and 2018 (blue) by EFFIS source [2]

328

N. Dobrinkova et al.

The summer of 2018 illustrates that climate change is driving the fire regimes faster than expected. Climate change is more visible as it generates drought stress in abandoned rural areas, where dense vegetation is available for burning during wildfires. This phenomenon is not only affecting the South, but all of Europe as presented in the EU Annual Report on Forest Fires for 2018 [3]. According to the report, wildfires destroyed nearly 178 000 hectares (ha) of forests and land in 2018. In 2017 fire season destroyed 1.2 million hectares of forests and land in Europe based on the EU Annual Report on Forest Fires for 2017 [4]. The 2017 report was about megafires as single events, the 2018 report was about globalization of such events. The first Pan-European fire season was during the summer months of 2018. The novelty is not that fires are happening in Northern Europe, in fact, 2007, 2011 and 2014 have been notable fire seasons in these countries. What makes the difference is the fact that during the summer of 2018 Peloponnesus and Algarve burned on the same day as the forests of Sweden, Latvia, the agroforestry mosaic from Denmark and a range of ecosystems of central Europe in between. This all happened during July 23, 2018 [5].

2 Case Studies 2.1 Pedrógão Grande (Portugal June 2017) Case Study Portuguese territory has 70% of its surface area occupied by sylvan spaces, which include forests (35%), shrubs lands and pastures (31%), unproductive land (2%), and inland waters and wetlands (2%). The sylvan spaces that currently exist in the Portuguese territory result from the long evolutionary process always linked to the human presence. Portugal today has a forest sector with peculiar characteristics, it has one of the smallest public forest areas of the world, only 3% of the total area are public forest systems with highly productive potential and the possibility of adapting to many forest species. The Portuguese forest has quite diverse in its composition native species (especially Quercus sp., with 36% of the total, and maritime pine, with 30%), there are also exotic species such as eucalyptus (with 26% of the total wooded surface) present. The forest area has increased significantly between the nineteenth and late twentieth centuries, due to public policies and private actions of forestry plantations. In the public area, more than 1 million hectares were forested. However, with the increase in the number of wildfires and burnt areas since 1995 the forestry area has failed to increase and has even decreased slightly (−4.6%) in 2010. The forest and the associated resources contribute annually to the Portugal economy with 982 million euro, not including the related value of the recreation, landscape and ecosystems services (as water, carbon retention, etc.). Traditional

Wildfire Risk Reduction Based on Landscape Management

329

forestry industries (forestry, hunting, fishing and industries) create around 80,000 jobs, mostly in regions with demographic and economic difficulties. Despite of the high number of owners and the small size of the forest property, the goods produced this way support an important and integrated industrial chain, based on natural resources, supporting itself a strong export sector. According to an estimate for 2001 the actual annual economic output was 1.3 million euro, i.e. 344 euro per ha per year. Therefore, forest and forestry in Portugal have great importance for the economy. Portugal, in the European context and even internationally is a country specialized in the forestry sector, revenue being an important contribution to GDP. This is bigger than the European average figures. The Portugal land cover map present 35% of the country with forest (this gives first place to the state among other countries in Europe) Fig. 3. The human activity was one fundamental element for the Portuguese landscape formation. On the most difficult landscapes, domestic animals ingested the vegetation for production of meat, milk and wool. The local population used fire to clear, heat

Fig. 3 Forest coverage in Portugal. Source Maria Caetano, Cristina Igreja, Filipe Marcelino e Hugo Costa, “Ocupação/uso do solo em Portugal Continental 1995/2010” presentation—2017 May

330

N. Dobrinkova et al.

or cook food. Forest areas (forest and scrubland) were instruments of the agricultural success and livestock activities. They were objects of management for the goods and services they produced and, because of their value, were fostered, cared for and protected. In the 1960s the traditional rural society foundations in Portugal began to weaken. Littoral industry, big city services and the reconstruction of Europe attracted many Portuguese people to immigrate or leave their villages and move to the larger cities. During this period the depopulation of the rural Portugal areas had begun. During that time the earlier managed plant communities which had degree of adaptation to fire, became artificially substituted by the industrial forest industry with tree species not accustomed to fire. The Portuguese farmers and Shepard’s, who has historically been using fire like a tool to clean up and do firefighting were now missing. The strong agricultural community in the country, who had doing the farming, cutting the scrublands, and their animals (goats and sheep) grazing the vegetation no longer existed. “Natural fuel breaks” were no longer in place and the vegetation was growing unrestrained. In April 1974 Portugal experienced the “democratic revolution,” resulting in stronger forest control by the state. Special legislation was introduced for forested areas, preventing landowners, shepherds and farmers from using the land. With the end of the previous regime an exponential increase of forest fires started. In 1980 wildfire management was removed from the jurisdiction of the Portuguese foresters and this service was delegated to the Portuguese firefighters. During the time period of the 1960s–1990s massive depopulation of the rural areas in Portugal was taking place. Fire as a tool to manage land and vegetation has disappeared from the population’s knowledge. Fire use has been kept away and its worth as an important tool for landscape modelling forgotten. The result was the great accumulation of forest canopy, modifying the vegetation and thus becoming more prone to more intense fires, of larger size and with greater difficulties of extinction, especially during years unusually dry, associated with global warming. The end result was the large Pedrógão Grande Fire which was actively burning during 17–21 June 2017 Fig. 4. An intense heat wave preceded the fires, with many areas of Portugal registering temperatures exceeding 40 °C (104 °F). During the night of 17–18 June, a total of 156 fires erupted across the country, particularly in mountainous areas 200 km (120 mi) north-northeast of Lisbon. The fires began in the Pedrógão Grande municipality before spreading dramatically causing a firestorm where 66 people died and more than 200 people were injured [6]. The fire was caused mainly by lack of prevention that generated accumulation of fuel which in turn generated conditions for megafire development.

Wildfire Risk Reduction Based on Landscape Management

331

Fig. 4 Pedrógão Grande Fire location in Portugal [7]

3 Brecon Beacons National Park (South Wales April 2017) Case Study Historically agriculture in Wales has been a major part of the economy. Wales is a largely rural country that forms part of the United Kingdom. Wales is mountainous and typically has a mild, wet climate. This results in only a small proportion of the land area being suitable for arable cropping, but grass for the grazing of livestock is present in abundance. As a proportion of the national economy, the importance of agriculture has become much reduced in recent years, and a higher proportion of the population now lives in the towns and cities in the south of the country. Tourism has become an increasingly important form of income in the countryside and on the coast. Arable cropping is limited to the flatter parts and elsewhere dairying and livestock farming predominate. Arable crops and horticulture are limited to southeastern Wales, the Welsh Marches, and the northeastern part of the country, the coastal fringes and larger river valleys. Dairying takes place on improved pasture in lowland areas and beef cattle and sheep are grazed on the uplands and more marginal land. Much of the land at higher elevations is extensive sheep-walk country and is grazed by hardy Welsh Mountain sheep. Large areas are grass and scrub lands, nearly 15% of Wales is covered by trees, the majority of this being pine plantations that were originally planned to be used as pit props for the mining industry. As with other parts of the United Kingdom, farming has been under great economic pressure, leading to declines in the number of people permanently employed on the landscape and increasing the role of part-time farming. Early farmhouses have been changed into Bed-and-Breakfast guest houses, or converted into self-catering accommodation, and farmers have diversified into the tourism-related industry and other activities.

332

N. Dobrinkova et al.

South Wales is a loosely defined region of Wales bordered by England and the Bristol Channel to the east and south. It has a population of around 2.2 million, almost three-quarters of the whole of Wales, including 400,000 in Cardiff, 250,000 in Swansea and 150,000 in Newport. This area is the most fire prone for grass and brush fires based on South Wales Fire and Rescue Service (SWFRS) annual report 2018–19 [8]. Warm wet summers lead to an abundance of annual wildfire fuels such as bracken and Molinia grasses. Less grazing animals and changes in land use have led to an accumulation of greater dead and dormant vegetation (fuel loading). The numbers of sheep in Wales 2017 are more than 10 million, whereas in the period 2018–19 were 8.56 million head, down 473,013 head, a 5.2% decrease year-on-year. This decrease of sheep and other farmed animals has allowed an increase in the grass and brush load making the areas vulnerable in case of ignition. Wildfires can occur as soon as the herbaceous vegetation becomes dormant, usually November to May with most fires occurring in March, April and May and coinciding with Easter holidays. Between the 8th and 9th April 2017 an area of 800 ha was burnt by a deliberate fire in the Brecon Beacons National Park, South Wales. It was once heavily grazed and a drinking water catchment for the South Wales valleys is now becoming very vulnerable to large fires. This fire spread for about 24 h in the park zone causing huge damages. Normally this type of fast progressing fires burning hectares is unusual. However official statistics from Global Wildfire Information System (GWIS) per country shows that since 2001 the United Kingdom has been experiencing wildfires of approx. 25 ha or larger see Fig. 5 [9]. South Wales has been collecting its own wildland fire statistics since 2010. The data base is maintained by the South Wales Fire and Rescue Service (SWFRS). The summary of fire investigation results indicates that most of the grass wildfires in the South Wales are deliberate. In rare cases there

Fig. 5 GWIS statistics about UK wildfires—25 ha or larger since 2001 [9]

Wildfire Risk Reduction Based on Landscape Management

333

Fig. 6 Grass fires statistics 2010–2018 in South Wales source by annual report 2018–19 SWFRS [9]

are also some accidental fires, but most of these were also probably human caused. One of the reasons for such fires are that large areas of scrubland and forest are very close to residential properties. Between 2000 and 2008 there were over 55,000 recorded grassfires and nearly 550 forest fires in South Wales, this equates to eight times more per unit area than in the United Kingdom as a whole see Fig. 6. In order to reduce the wildfire risk in 2007 SWFRS started the “Wildfire Project”. The aim of this project was to reduce the number of deliberate wildfires in South Wales. The two parts to the project covered operational tactics and equipment including community safety. Fire breaks of 800 m have been done in Wildland-Urban Interface (WUI) areas see Fig. 7. SWFRS estimate an annual cost in their service area of around £7 m due solely to wildfires [9]. Traditionally sheep and farming animals were grazing this type of grasslands, and their reduction on the landscape is the most probable reason for such fires to easily spread. Solutions in legislative economic stimulus of the farmers to increase landscape grazing may most probably solve the problem.

4 Wildfire Paradox In both case studies, Portugal and the United Kingdom, the main reason for fire occurrence and spread is the continuous fuel bed of dead and dormant vegetation which has been accumulated due to a lack of land management. The emergency services refer to this as the “Wildfire Paradox.” The process in which the lack of forest and grassland management result in an increase in fire suppression efforts which drives further a change in fire regimes that burn at increasingly greater fire behavior. The “Paradox” is given by the fact that a rapid suppression of all wildfires contributes to creating increasingly homogeneous and greater fuel loads. The effect on the midterm is a decrease in the number of wildfires. However, given extreme weather conditions, the fires spread rapidly, becoming larger, and burning at higher intensities from the combustion of the accumulated fuel loads available on the continuous landscape.

334

N. Dobrinkova et al.

Fig. 7 WUI in South Wales is common thus fire breaks are part of the prevention measures of the local fire brigade [9]. Source SWFRS

The “Wildfire Paradox” can be explained by the “Wildfire Generation” concept. Different generations are identified to explain the wildfire evolution into large wildfires burning with high intensity (i.e. megafires). In the first place wildfire became a threat due to the fuel continuity (1st generation). With time, fuel continuity increased leading to faster fires (2nd generation) and more intense fires (3rd generation). The third generation wildfires challenge fire suppression agencies as they spread fast and with a higher intensity over dense and continuous vegetation across the landscape. These fires are typically a problem from Southern countries but are now also expanding to Northern Europe [5]. Climate change impacts on temperate regions landscapes have observed this during 2018. Summers in temperate countries are traditionally wet, have all of a sudden been transformed into long, hot and dry periods. Large fuel loads have became available in a landscape without a previously defined fire regime. In these landscapes, fires do not burn as in 1st generation, but rather as 2nd and 3rd generation; fast and with high intensity. Northern countries, with large extensions of wildlands and forests have now a fire regime of increasing intensity. But fire services have not had the opportunity to approach the problem and learn or develop a need for changing tactics and strategy. They are going from fire regimes of small and rare fires directly into megafires. The only option seems to fight flames, but doing that there is a high risk of losing the big picture. In such a scenario, a defensive strategy will help fighting flames, however flames are only the visible part of a deeper problem. It is not a problem of fighting flames, it’s a landscape problem.

Wildfire Risk Reduction Based on Landscape Management

335

Southern Europe is already in the 3rd generation, with two added weaknesses, the wildland-urban interface and the simultaneity of fire events. These two phenomena demand an exhaustive amount of resources and often cause the collapse of the emergency services. The more resources are deployed, the larger is the “Wildfire Paradox.” Fires in the wildland-urban interface are the 4th generation and simultaneity of fire events is the 5th generation (e.g. 3rd and 4th generation fires). The immediate response is the international cooperation and support to the wildfire suppression operations. But this is not the solution, it is only temporary containment. Once the 5th generation of wildfires take place, the success of fire suppression depends mainly upon the likelihood and duration of extreme weather events. And that is uncertainty at its best. So, economy and landscape depend on uncertain weather events. If we are keen to fight the real problem, this should be fought from the very beginning: acting on the fuel load availability and having a better understanding of fire ecology (response of vegetation and wildlife to fire) and the role of fire as a natural disturbance. Landscape management and planning should integrate wildfire risk. Fires are nothing neither strange nor exceptional. They are part of the natural lifecycle and cannot disappear. Fire seasons that before happened only in the south are now a reality across Europe. First, to reduce the fire intensity and spread rates, it is necessary to reduce the load of dead fuel. Second, to reduce fuel continuity it is necessary to restore a mosaic structure recreating a heterogeneous landscape that includes agriculture. In that sense, bioeconomy shall become a priority, as it contributes to reduce emergencies and the risks associated with climate change. The lessons learned fighting wildfires on the past can be of great help. All are pointing at the recovery of the mosaic landscape in order to keep fire generations at the minimum. Taking a defensive strategy will only contribute to enlarge the problem, and forest will no longer be an ally but an enemy, then megafires will happen more often. The evidence of failing to see this big picture is the focus on the demanding daily emergency. The current trend under the context of climate change is the spread of increasing fire occurrence and behavior currently happening in Southern Europe to more northern regions of Europe. Those forests that are not adapted to the new climate are experiencing renovation through disturbances. It is urgent that we face the current situation in a manner that helps forests, and societies to adapt and change, rather than take the hard strategy of preserving the current forests. We should aim to create the landscape of the future, resilient against large emergencies, protected from climate change and from megafires. Forest management and the associated bioeconomy are tools to take action. The fire suppression services shall be well trained and provided with the appropriate resources. However, their efficiency relies upon the preparation of the landscape to fight the flames on it. Without prepared landscapes there is no possible suppression of the 6th generation fires, or ‘fire storms’. They were seen in Portugal, Chile, Canada and USA in 2017, and again in 2018. The years to come will bring more.

336

N. Dobrinkova et al.

5 Steps for Improvements Portugal spend 25 million euro in wildfire prevention and 75 millions in wildfires combat until 2016 (average/year). The aerial support each year increase. The aeroplanes and helicopters, and the number and burned areas in Portugal also increase. However heavy equipment does not solve the wildfire occurrence problem. Thus municipalities started to experiment and implement new tactics. The municipality of Mafra, Portugal is a good example implementing prescribed fire burning into a mosaic pattern see Fig. 8. Such initiatives on a land management scale can and will bring good results if people work together with the responsible authorities. Existing laws are static in the framework that there are now new social and climatic conditions. The new technologies can bring added value to the land management if the fuel models represent correctly the vegetation world that can burn. Such basic system with appropriate calibration could be the Canadian fire weather index Fig. 9. This basic schema adapted to the land management daily routine can give good estimation to the land managers, how dangerous every day is.

Fig. 8 Mosaic prescribed fire burned areas in Mafra municipality

Wildfire Risk Reduction Based on Landscape Management

337

Fig. 9 Algorithm used by the Canadian fire weather index giving dynamic fire danger for the land managers. Source https://cwfis.cfs.nrcan.gc.ca/background/summary/fwi

6 Conclusions The population must be proactive and do its best to maintain the land within their neighbourhoods. State reactions in most cases of rapid wildfire spread show that there is missing coordination between all first responders in the field. So, working together must cover national and all other levels up to the community representatives’ special regulations and coordination cooperation initiatives which in cases of major disasters will be helpful and should be in place. The scientific community can be a pillar on simulations about potential scenarios for wildfire spread. This can help and support the civil protection decision making process. The gaps between the research community and operational response can be overcome only if all stakeholders learn to work together and the responsibilities are clear. Acknowledgements This work has been supported by the NNP-OS contract number D01322/18.12.2019, the Bulgarian National Scientific Fund project number DFNI DN12/5 “Efficient Stochastic Methods and Algorithms for Large-Scale Problems” and the DRIVER+ project funded by the European Union’s 7th Framework Programme for Research, Technological Development and Demonstration under Grant Agreement (GA) N° #607798 (CMINE network). .

338

N. Dobrinkova et al.

References 1. Forest Fires—Sparking Firesmart policies in the EU.https://ec.europa.eu/info/sites/info/files/ 181116_booklet-forest-fire-hd.pdf 2. EFFIS Service for EU Countries about Seasonal Fire Trends. https://effis.jrc.ec.europa.eu/sta tic/effis.statistics.portal/seasonal-trend/EU 3. Annual Forest Fires report of EU for 2018. https://ec.europa.eu/jrc/en/publication/forest-fireseurope-middle-east-and-north-africa-2018 4. Annual Forest Fires report of EU for 2017. https://ec.europa.eu/commission/news/annual-rep ort-forest-fires-europe-2018-sep-20_en 5. Castellnou, M.: Why and How Forest Fires are Becoming a European Problem? https://www. efi.int/news/why-and-how-forest-fires-are-becoming-european-problem-2018-08-09 6. Rego, F.C., Oliveira, T. Fernandes, et al.: Economia da Floresta e Ordenamento do Território Conselho Económico e Social; Lisboa (2017) 7. Reuters News Pedrógão Grande Map. https://www.straitstimes.com/world/europe/portugal-for est-fire-it-does-not-seem-real-it-is-out-of-this-world 8. South Wales Fire and Rescue Service (SWFRS) Annual Report 2018–19: https://www.fire.nsw. gov.au/gallery/files/pdf/annual_reports/annual_report_2018_19.pdf 9. Global Wildfire Information System (GWIS) Per Country. https://gwis.jrc.ec.europa.eu/static/ gwis_stats/

Numerical Weather Prediction for the Bulgarian Antarctic Base Area and Sensitivity to the SST Variable Boriana Chtirkova , Elisaveta Peneva , and Gergana Georgieva

Abstract The weather forecast of good quality is essential for the humans living and operating in the Bulgarian Antarctic base (BAB), located on the Livingston Island coast at 62.64◦ S and 60.36◦ W. The numerical weather prediction models in southern high latitude regions still need improvement as the user community is limited, little test cases are documented and validation data are scarce. In this study, we suggest several ways to improve the local weather forecast model skill by modifications of the land cover and ocean temperature. We tested the sensitivity of the numerical weather prediction modelling system based on the Weather Research and Forecasting (WRF) model, configured for the BAB area, to the Sea surface temperature (SST) of the ocean around the island. The model configuration is described and details on the model performance are given. Several experiments with SST coming from different sources are performed, as well as experiments where the SST is scaled linearly. The conducted sensitivity experiments show that all of the considered meteorological variables are affected by the sea surface temperature, the most prominent differences being observed in the 2 m temperature field. With a uniform rise in SST, the corresponding tendencies are: an increase of the 2 m temperature, a decrease of the sea level pressure and an increase of the average wind speed. For the BAB region, the best results with unmodified SST data are obtained when using SST from the Copernicus Marine Service ocean model. Keywords Numerical weather prediction · Antarctica · Sea surface temperature

1 Introduction The Bulgarian Antarctic base “St. Kliment Ohridski” (BAB) is a national scientific facility, located on the coast of Livingston Island, South Shetland Islands, at 12– 15 m above sea level. An average of 25 people work there during the austral summer, B. Chtirkova (B) · E. Peneva · G. Georgieva Department of Meteorology and Geophysics, Faculty of Physics, Sofia University “St. Kliment Ohridski”, 5 James Bourchier Blvd., 1164 Sofia, Bulgaria e-mail: [email protected] © The Author(s), under exclusive license to Springer Nature Switzerland AG 2021 N. Dobrinkova and G. Gadzhev (eds.), Environmental Protection and Disaster Risks, Studies in Systems, Decision and Control 361, https://doi.org/10.1007/978-3-030-70190-1_23

339

340

B. Chtirkova et al.

usually from late November until early March. The weather in the region is mostly influenced by extratropical cyclones, which move west to east and tend to diffuse over land, due to friction. Thus, the weather is highly variable with intense storms and strong wind events occurring regularly. Weather forecast of good quality is essential for the activities in BAB. A modelling system for weather prediction with high resolution in the BAB area based on the Weather Research and Forecasting (WRF) model is developed and validated in [2, 4]. The purpose of this work is to seek ways to improve prediction skill of this modeling system that could be implemented in operational mode. We have performed sensitivity experiments towards the land cover and the Sea Surface Temperature (SST) input variable and analyzed the obtained results. The impact of SST on the weather forecast is more peculiar than that of land, because the SST is not only influenced by the heat fluxes between the atmosphere and ocean, but also by various fluxes and mixing processes in between ocean layers. Numerical modeling of the processes of interaction between the sea surface and the planetary boundary layer (PBL) have been a long term object of study. Tuleya and Kurihara [12] investigate the SST impact on the formation of tropical cyclones and demonstrate that increasing the SST with 4 K may result in a lower surface pressure by 7.6 hPa, compared to the case without altering the SST. While studying the effect of SST on the characteristics of Mediterranean cyclones with WRF, Miglietta et al. [8] reach the conclusions that increasing the SST leads to the following effects: deepening of the pressure minimum (as in [12]); the maximum wind speed at 10 m increases quasilinearly with SST; the maximum accumulated precipitation increases linearly with SST. Senatore et al. [10] conduct a study with different SST datasets for the Mediterranean sea with WRF. Their conclusions show that the use of different datasets in long-term simulations leads to the same results, but the simulations of specific events are distinguished as a result of the SST representation. On the basis of WRF simulations in the Yellow Sea and Eastern Chinese Sea regions, Bai et al. [1] demonstrate that the SST front is closely connected to the regulation of the marine atmospheric boundary layer (MABL). According to them, the various wind directions activate different MABL regulation mechanisms.

2 Model Configuration and Validation The modelling system is based on the Weather Research and Forecasting model, version 4.0, developed by the National Center for Atmospheric Research (NCAR) and the National Centers for Environmental Prediction (NCEP). The numerical model uses a staggered Arakawa C-grid [11] and the nesting is performed in a ratio 9:3:1 km. The domain configuration is shown in Fig. 1: it is centered over BAB with coordinates 62.64◦ S and 60.36◦ W and consists of three nested domains d01, d02 and d03. The grid configuration is achieved using a Lambert conformal conic projection with standard parallels at 60◦ S and 30◦ S. The outermost domain—d01, with a resolution

Numerical Weather Prediction for the Bulgarian Antarctic …

341

Fig. 1 Three domains coverage—d01, d02 and d03. Background image from NASA Visible Earth—Blue Marble

of 9 km, has horizontal dimensions of 999 km in both directions and covers the northern part of the Antarctic Peninsula. It is a parent domain to the first nested domain—d02, with a horizontal resolution of 3 km and dimensions of 342 km in both directions. The finest domain—d03, nested in d02, has a horizontal resolution of 1 km, and covers the area of Livingston Island and its neighbouring small islands; the horizontal dimensions of d03 are 129 km in west–east direction and 111 km in south–north direction. The domain configuration is made so that there are no high mountains or complex relief near the domain borders.

2.1 Topography and Land Use Data The land surface boundary condition of the modelling area is assembled via the WRF Preprocessing system program—geogrid, which interpolates topography and land use type data into the model grid. The topography data is taken from the GMTED2010 dataset, developed by the United States Geological Survey (USGS) and the National Geospatial-Intelligence Agency (NGA) and has a horizontal resolution of 225 m. The fitted in the finest domain topography is shown in Fig. 2a. A comparison with regional maps such as the map from [7] show that this dataset does not represent accurately Livingston Island. The mountain range Tangra mountains, reaching heights up to 1700 m, is represented as a flat surface with elevation of 50 m. Figure 2-a represents the topography grid in the finest domain, which has a number of points in the xdirection i max = 129 and in the y-direction— jmax = 111, each grid point covers an area of 1 km2 . The WRF model can perform mainly with two land use datasets, which cover the entire globe—USGS and MODIS. The USGS data is based on satellite advanced very-high-resolution radiometer (AVHRR) data, collected in the period April 1992–

342

B. Chtirkova et al.

Fig. 2 Topography (a) and land use type data from MODIS (b) and USGS (c) of the finest domain— d03 with a resolution of 1 km, i max = 129, jmax = 111. The coastlines on the image are taken from the Natural Earth Database (https://www.naturalearthdata.com/) and have a horizontal resolution of 10 m

March 1993. They contain 24 land use types and have a resolution of 1 km. The MODIS data (Moderate resolution imaging spectroradiometer) are gathered by NASA satellite missions in the period 2001–2005. They are made up of 20 land use type categories and their resolution reaches 500 m. A visual comparison between the MODIS and USGS land use type data is given in Figure 2b and c. According to both datasets, the entire domain area is described with only 2 land use types— snow/ice and water. The land use type is presented in numerical modelling through the following parameters: albedo α (%), soil moisture availability M (%), surface emissivity,  (%), roughness length z 0 (m), thermal inertia λT (J m−2 K−1 s−1/2 ) and surface heat capacity C (J m−2 K−1 ). The values of these parameters differ throughout the seasons and are used to describe the energy, momentum, water and heat fluxes. They slightly differ between the two datasets, mainly in the parameters z 0 and λT , which are slightly higher within the MODIS data, but this distinction should not result in large computational differences. Comparing Fig. 2b and c, one can conclude that the coastal line, formed by the USGS data is not as continuous as the MODIS one. There is a slight displacement of the grid between the two datasets but this problem is eliminated with a manual choice of grid point to represent BAB. Having taken this into account, and the fact the MODIS data is more recently collected, the authors conclude it is better suited for the modelling system. However, the description of the whole Livingston Island as covered with snow and ice may not still be accurate in the recent years. Experiments from [4] demonstrate that using a different land use type with a lower heat capacity and a lower surface albedo, significantly improves the 2 m temperature forecast. For a vertical coordinate in WRF, version 4.0, one can choose a terrain following (TF) coordinate or a hybrid vertical coordinate (HVC). In the present study, hybrid η-levels are used, unevenly distributed from the surface up to isobaric level 50 hPa. In order to determine the optimal number a vertical levels, a sample model run of a 72 h forecast with a different number of levels has been performed. All WRF experiments in the present study are run on the Sofia University Parallel Computer Center cluster PHYSON.1 Three different vertical level configurations have been tested—with 35 vertical levels (close to the WRF minimum number), 50 and 70 levels. The η-levels 1 PHYSON

computer cluster: http://physon.phys.uni-sofia.bg.

Numerical Weather Prediction for the Bulgarian Antarctic …

343

from the three configurations are distributed with equal density near the surface but their distribution difference manifests after certain height. Hence, the description of the atmospheric state near the surface is similar, but the configuration with the least densely distributed levels in height fails to represent adequately high altitude phenomena, such as the polar jet stream. All simulations are performed using the same number of computer cores, 88 in this case, on the PHYSON cluster. For an optimal configuration, performed for a minimal time, the 50 η-levels configuration is chosen.

2.2 Model Parameterization Schemes Even though the high horizontal and vertical resolution of recent numerical models allows us to describe smaller scale phenomena, other physical processes on scales smaller than the model grid still need to be parametrized. The parametrization schemes and their combinations has a profound effect on numerical forecasting, especially in larger time scales and therefore the schemes should be chosen carefully. The subgrid processes that are or may be parametrized in WRF comprise of microphysics, convection, turbulence in the planetary boundary layer (PBL), interactions between the atmosphere and the surface layer and the longwave and shortwave radiation. The choice of parametrization schemes for the present study has been made through a literature review. The Antarctic Mesoscale Prediction System (AMPS) produces numerical forecasts for the Antarctic region, made through a modified PolarWRF [9]. The physics parametrization in the present configuration are chosen to be coherent and done in accordance with AMPS. The following schemes are used: • • • • • • •

Boundary layer: Mellor-Yamada-Janjic (Eta) TKE scheme Surface layer: Monin-Obukhov (Janjic Eta) scheme Land-surface interactions: Unified Noah Land Surface Model Microphysics: WSM 5-class scheme Long-wave radiation: RRTMG longwave radiation scheme Short-wave radiation: Goddard shortwave radiation scheme Convection: Kain-Fritsch (new Eta).

For a more detailed description of each scheme, the reader is referred to [11] or [3]. The Kain-Fritsch convection parametrization scheme is not used in the finest domain, because its horizontal resolution of 1 km can resolve convective processes. The regional models need suitable atmospheric initial and lateral boundary conditions. They are taken from the GFS 0.25 Degree Historical Archive (NCEP, NWS, NOAA, U.S. 2015) and the lateral boundary conditions are updated every 3 h of the simulation. The sea surface temperature is also taken as a time varying surface boundary condition and is updated every 3 h into the simulation. The GFS model analysis in 0 UTC is taken as an initial condition for each of the three domains, while the lateral boundary conditions are given only to the outermost domain. Two-way nesting is performed, which means that the forecast in the parent domains is affected by the solutions in the finer domains.

344

B. Chtirkova et al.

2.3 Model Validation The above described modelling system is validated in [4]. The authors consider three test cases with rapidly changing weather in the recent years records from the GFS model analysis and in-situ observations at BAB: 16–19 December 2016, 26 February–1 March 2020 and 25–28 January 2020. Each simulation is run over 3 days, starting at 0 UTC. The starting date is chosen so that the rapid change of weather happens at least 24 h into the forecast. The model configuration has been validated against measurements from an automatic meteorological station at BAB, synoptic measurements in the nearby meteorological stations2 and ERA-5 climatic hourly reanalysis data [6]. The in-situ measurements come from an automatic meteorological station Davis Vantage Vue for the 2016 and 2017 test cases, and an automatic station assembled by MeteoRocks3 for the 2020 test case. The validated meteorological variables are temperature at 2 m, surface pressure, wind speed and wind direction. The total number of synoptic stations in the largest domain is 18, two of which lie in the finest domain. The synoptic observations from “Base Arturo Prat” with WMO index 89057 and coordinates 62.3◦ N, −59.41◦ E, are used in the comparison as indicative of the weather pattern in the region. They are in agreement with the observations from BAB in all test cases.

3 SST Sensitivity Experiments Planning The results from the model validation in [4] show a general negative bias of the 2 m temperature at BAB. Although in situ observations of that kind tend to increase the temperature around noon, the forecasted temperature curve is entirely below the observed one, even in the experiment with the lower thermal capacity of the land surface. One of the possible reasons for this could be the unrealistic representation of the Sea Surface Temperature, as the BAB is located on the beach. This was the motivation to test different sources of data for SST as model input. The default configuration of WRF model sets SST as initial condition which is not modified during the integration. We have tested the varying SST by activating the key sst_update = 1. An alternative to the GFS SST variable is to use the data from the operational global ocean model of Copernicus Marine Environment Monitoring Service (CMEMS [5]). The hourly data are distributed in a grid with resolution 1/12 degree. Comparing SST data between the GFS and the CMEMS operational ocean model for the three test cases, one can find significant differences between the two SST fields, together with different tendencies for the fields’ evolutions in time. Figure 3 demonstrates the difference in SST fields from the two models. The coldest areas of the ocean surface, reaching temperatures under 1 ◦ C are shown by the GFS to be near the icy land areas. The CMEMS data show a significant temperature gradient, 2 Ogimet: 3 The

https://www.ogimet.com. MeteoRocks project: https://meteo.rocks/page/aboutus.

Numerical Weather Prediction for the Bulgarian Antarctic …

345

Fig. 3 SST comparison between GFS and CMEMS data for 6 UTC on 26.02.2017 (a, b) and 6 UTC on 28.02.2017 (c, d).

orientated northwest-southeast, the cold part of which is propagating northwards like a front. The difference in time between Fig. 3a, b and c, d is 48 h, during which SST in the BAB bay has dropped less than 0.5 ◦ C according to GFS and more than 1.5 ◦ C as shown by CMEMS. Another drastic difference between the two datasets from the 2020 test case can be seen in Fig. 4. In this case, the difference between the area average values of the two fields is larger—more than 2.5 ◦ C. The tendencies of the two models, however, are similar to those in Fig. 3. The GFS model shows generally lower SSTs, which again do not change considerably in time, while the CMEMS model shows higher temperatures, which fall drastically with the cold front passage in both cases. The SST in the BAB bay has fallen by ∼1.5 ◦ C, while the SST around Rozhen peninsula (the southernmost part of Livingston island) has dropped by more than 2.5 ◦ C, as seen in Fig. 4. The comparison between SST and sea water temperature at 3 m depth at the points shown in Fig. 4, can be made through the assumption of well-mixed water near the coast. The observed values at 3 m depth are closer to the temperatures, modeled by CMEMS. The in situ measurements also acknowledge the drop in temperature, caused by the front—the water temperature in the northern of the two points has fallen from 2.9 to 2.4 ◦ C in two days.

346

B. Chtirkova et al.

Fig. 4 SST comparison between GFS and CMEMS data for 12 UTC on 25.01.2020 (a, b) and 12 UTC on 27.01.2020 (c, d). The white dots and the numbers beside them represent the water temperature at 3 m depth in ◦ C for the corresponding day from in situ measurements at the dots’ locations

By default, the WRF SST does not vary in time but remains constant throughout the length of the simulation, unless set up otherwise and provided suitable periodic boundary conditions. After the analysis of the two model fields, and taking into consideration previous numerical experiments, the following simulations are planned: numerical forecast with GFS SST data, where (1) SST does not vary in time; (2) SST varies in time; SSTs vary in time and their value at each point is (3) reduced by 3 K, (4) reduced by 1 K, (5) increased by 1 K, (6) increased by 3 K; and (7) a numerical simulation with CMEMS SST data, which varies in time. The validation of the 2017 test case shows large discrepancies between the modeled and observed meteorological variables during the cold front passage. In an attempt to improve the numerical forecast and to determine the reasons for these discrepancies, the following simulations are performed: (1) a simulation with a 12hour spin-up of the model—during the spin-up process, the GFS analysis is used as boundary conditions, not the forecast; (2) a model free run without any update of the boundary conditions after the initialization.

Numerical Weather Prediction for the Bulgarian Antarctic …

347

4 Assessment of the SST Impact on the Weather Forecast Accuracy The WRF model validation results demonstrate that the model predicts lower than the observed 2 m temperature. Since the BAB grid point is coastal, it is clear that it will be strongly influenced by the SST. For this reason different experiments are conducted, these include higher GFS SSTs, as well as CMEMS SSTs, which also tend to be higher than the ones by GFS. For symmetry, experiments with lower altered SSTs are also performed. We will first examine the test cases from 2016 and 2020 for which WRF performs well in forecasting the phenomena, associated with the front passage. After this, we will look into the 2017 test case, where a different approach has been applied. Figure 5 displays the time series of the temperature at 2 m and the wind speed at 10 m according to each SST experiment for the 2016 test case. The graphical comparison shows higher air temperatures values with higher SSTs and accordingly lower air temperatures with lower SSTs, furthermore the lower SSTs result in larger temperature amplitudes. The unmodified curves, corresponding to GFS data with and without time evolving SSTs, are very close, while the CMEMS data result in larger amplitudes. The differences in the 10 m wind speed do not appear to be linear—in the first 36 forecast hours, the highest peaks result from the data with lowest SST, while in the next 36 h the wind speed in this case is the lowest among all simulations. The forecasts for the 2020 test case, illustrated on Fig. 6, confirm our conclusions for the 2 m temperature, although one can notice a slightly larger distancing between the GFS data forecast curves with and without time evolution. Following the 10 m wind speed curve, the highest peaks are observed with the lowest SSTs, while the minimums are lower with the highest SSTs. The forecast statistics for the 2 m temperature are given in Table 1. In the 2016 and 2020 test cases, the best overall results are

Fig. 5 A comparison between the temperature at 2 m (a) and wind speed at 10 m (b) between WRF simulations with different SST fields, plotted against BAB measurements (black dots), ERA5 reanalysis data (brown pentagons) and measurements from a SYNOP station nearby (dark red squares) for the 2016 test case

348

B. Chtirkova et al.

Fig. 6 Same as Fig. 5 but for the 2020 test case Table 1 Forecast statistics of the temperature at 2 m against measurement data, ERA-5 reanalysis data and the WRF run with unmodified evolving in time SST data from GFS Temperature (◦ C) BAB measurements BIAS

RMSE MAE

ERA-5 reanalysis BIAS

RMSE MAE

WRF-GFS SST BIAS

RMSE MAE

WRF simulation for the 2016 test case WRF-GFS SST without update

−1.45

1.69

1.45

−0.18

0.98

0.62

−0.01

0.07

0.05

WRF-GFS SST with update

−1.44

1.69

1.44

−0.17

0.98

0.62

–

–

–

WRF-GFS SST with update (−3 K)

−2.37

2.47

2.37

−1.11

1.52

1.21

−0.95

1.14

0.95

WRF-GFS SST with update (−1 K)

−1.75

1.93

1.75

−0.48

1.05

0.65

−0.32

0.39

0.32

WRF-GFS SST with update (+1 K)

−1.15

1.51

1.23

0.12

0.99

0.71

0.28

0.34

0.29

WRF-GFS SST with update (+3 K)

−0.70

1.19

0.94

0.58

1.15

1.00

0.75

0.81

0.75

WRF-CMEMS SST with update

−1.30

1.62

1.34

−0.02

0.97

0.66

0.14

0.25

0.20

WRF simulation for the 2017 test case WRF-GFS SST without update

−1.11

3.76

3.23

−1.56

3.82

2.75

0.10

0.23

0.13

WRF-GFS SST with update

−1.21

3.86

3.31

−1.67

3.95

2.83

–

–

–

WRF-CMEMS SST with update

−1.33

4.10

3.45

−1.79

4.25

2.98

−0.12

0.61

0.35

WRF—free run

0.39

1.61

1.37

−0.02

1.47

1.18

1.65

3.37

2.62

WRF—with 12-hour spin-up

−1.17

3.84

3.27

−1.42

3.64

2.44

0.03

0.29

0.17

WRF simulation for the 2020 test case WRF-GFS SST without update

−1.95

2.11

1.95

−0.38

0.64

0.47

0.12

0.17

0.14

WRF-GFS SST with update

−2.07

2.23

2.07

−0.50

0.75

0.57

–

–

–

WRF-GFS SST with update. (−3 K)

−2.57

2.76

2.57

−1.00

1.25

1.02

−0.50

0.56

0.50

WRF-GFS SST with update (−1 K)

−2.23

2.40

2.23

−0.67

0.90

0.71

−0.16

0.21

0.17

WRF-GFS SST with update (+1 K)

−1.86

2.02

1.86

−0.29

0.57

0.43

0.22

0.26

0.22

WRF-GFS SST with update (+3 K)

−1.32

1.55

1.34

0.25

0.44

0.34

0.75

0.83

0.75

WRF-CMEMS SST with update

−1.89

2.08

1.89

−0.32

0.60

0.44

0.18

0.24

0.20

Numerical Weather Prediction for the Bulgarian Antarctic …

349

Table 2 Forecast statistics of the sea level pressure against measurement data, ERA-5 reanalysis data and the WRF run with unmodified evolving in time SST data from GFS Pressure (hPa) BAB measurements BIAS

RMSE MAE

ERA-5 reanalysis BIAS

RMSE MAE

WRF-GFS SST BIAS

RMSE MAE

WRF simulation for the 2016 test case WRF-GFS SST without update

0.58

1.10

0.81

−0.82

1.27

1.14

0.00

0.03

0.02

WRF-GFS SST with update

0.57

1.10

0.81

−0.82

1.27

1.14

–

–

–

WRF-GFS SST with update (−3K)

0.70

1.20

0.88

−0.70

1.22

1.11

0.13

0.17

0.14

WRF-GFS SST with update (−1 K)

0.62

1.13

0.83

−0.78

1.25

1.13

0.05

0.07

0.05

WRF-GFS SST with update (+1 K)

0.53

1.07

0.80

−0.86

1.29

1.15

−0.04

0.06

0.05

WRF-GFS SST with update (+3 K)

0.42

1.01

0.75

−0.98

1.37

1.22

−0.16

0.18

0.16

WRF-CMEMS SST with update

0.55

1.07

0.79

−0.85

1.27

1.14

−0.02

0.05

0.04

WRF simulation for the 2017 test case WRF-GFS SST without update

0.74

2.20

1.85

−1.19

2.53

1.71

−0.02

0.04

0.03

WRF-GFS SST with update

0.76

2.21

1.86

−1.16

2.52

1.71

–

–

–

WRF-CMEMS SST with update

0.74

2.21

1.84

−1.19

2.52

1.69

−0.02

0.10

0.08

WRF – free run

0.34

5.49

5.09

−1.60

5.29

4.69

−0.44

5.76

5.26

WRF—with 12-hour spin-up

0.58

2.19

1.86

−1.31

2.48

1.75

−0.17

0.30

0.18

WRF simulation for the 2020 test case WRF-GFS SST without update

−0.29

0.73

0.62

−1.39

1.44

1.39

−0.02

0.04

0.03

WRF-GFS SST with update

−0.27

0.72

0.60

−1.37

1.42

1.37

–

–

–

WRF-GFS SST with update. (−3 K)

−0.13

0.69

0.56

−1.23

1.28

1.23

0.14

0.16

0.14

WRF-GFS SST with update (−1 K)

−0.22

0.71

0.58

−1.32

1.37

1.32

0.05

0.06

0.05

WRF-GFS SST with update (+1 K)

−0.33

0.75

0.64

−1.43

1.48

1.43

−0.06

0.08

0.07

WRF-GFS SST with update (+3 K)

−0.57

0.90

0.77

−1.67

1.72

1.67

−0.30

0.33

0.30

WRF-CMEMS SST with update

−0.29

0.73

0.61

−1.39

1.44

1.39

−0.02

0.04

0.03

obtained via the highest (modified) SSTs. Comparing the RMSE of the three unmodified cases with the BAB observations, the CMEMS time-varying SST experiment shows better results then the GFS (time-evolving and non-time-evolving) SST. The forecast statistics calculated against the unmodified WRF run (using GFS SST data without evolution in time) are given in order to distinguish the differences among all simulations. One can spot an almost symmetric linear dependency between the elevated and reduced SST values, as a SST field increase of 3 K results in a 2 m temperature increase at sBAB of 0.75 K. The sea level pressure statistics, presented in Table 2, differ to a smaller extent. Comparing the modified forecasts, one can notice a decrease in the simulated pressure field, when the SSTs are higher and vice versa. The differences in the wind speed and direction at 10 m between the different simulations are shown in Tables 3 and 4 respectively. With higher SSTs, lower wind speeds are forecast and vice versa. A lowering of the SST field with 3 K may result in a 1 m s−1 wind speed difference and a deviation in wind direction of almost 30◦ , according to the model output for the 2016 test case. This may be due to a ratio change

350

B. Chtirkova et al.

Table 3 Forecast statistics of the wind speed against measurement data, ERA-5 reanalysis data and the WRF run with unmodified evolving in time SST data from GFS Wind speed (m/s) BAB measurements BIAS

RMSE MAE

ERA-5 reanalysis BIAS

RMSE MAE

WRF-GFS SST BIAS

RMSE MAE

WRF simulation for the 2016 test case WRF-GFS SST without update

−0.99

2.69

2.21

0.58

1.48

1.18

−0.03

0.33

0.17

WRF-GFS SST with update

−0.96

2.67

2.18

0.61

1.39

1.10

–

–

–

WRF-GFS SST with update (−3 K)

−0.83

2.67

2.19

0.74

1.69

1.43

0.13

1.13

0.87

WRF-GFS SST with update (−1 K)

−0.90

2.70

2.17

0.67

1.65

1.38

0.06

0.72

0.46

WRF-GFS SST with update (+1 K)

−1.03

2.72

2.20

0.53

1.42

1.12

−0.07

0.57

0.41

WRF-GFS SST with update (+3 K)

−1.03

2.82

2.16

0.53

1.32

1.09

−0.08

1.01

0.75

WRF-CMEMS SST with update

−1.02

2.70

2.21

0.55

1.41

1.09

−0.06

0.56

0.35

WRF simulation for the 2017 test case WRF-GFS SST without update

−4.64

8.32

6.49

−1.29

4.56

3.37

0.07

0.45

0.26

WRF-GFS SST with update

−4.70

8.36

6.51

−1.36

4.61

3.36

–

–

–

WRF-CMEMS SST with update

−5.20

8.61

6.62

−1.87

4.84

3.55

−0.51

1.48

0.82

WRF—free run

−0.83

8.32

7.34

2.53

6.23

4.96

3.89

5.45

4.48

WRF—with 12-hour spin-up

−4.67

8.40

6.39

−1.12

4.52

3.19

0.02

1.71

0.93

WRF simulation for the 2020 test case WRF-GFS SST without update

−0.06

3.92

3.12

−0.32

3.01

2.51

−0.18

0.43

0.34

WRF-GFS SST with update

0.12

3.95

3.13

−0.14

3.04

2.49

–

–

– 0.60

WRF-GFS SST with update. (−3 K)

0.50

4.15

3.33

0.24

3.26

2.73

0.38

0.83

WRF-GFS SST with update (−1 K)

0.27

4.04

3.18

0.01

3.13

2.55

0.14

0.48

0.35

WRF-GFS SST with update (+1 K)

−0.05

3.99

3.21

−0.31

3.05

2.55

−0.18

0.57

0.41

WRF-GFS SST with update (+3 K)

−0.33

3.81

2.87

−0.59

2.81

2.24

−0.45

1.00

0.75

WRF-CMEMS SST with update

−0.29

3.91

3.02

−0.55

3.01

2.45

−0.41

0.72

0.58

between the surface temperatures of land and sea, which may have induced a breezelike circulation. The 2020 test case forecast simulations show the same differences in the wind speed field, but the forecasts differ less in wind direction—up to 13◦ . The lowest RMSE value of the wind speed against ERA-5 data in the 2016 case is observed with the simulation of lowering the SST with 1 K, while for the 2020 case the simulation with best results is the one, where the SST is increased by 1 K. The 2017 test case is prone to special attention—Fig. 7 shows the 2 m temperature and 10 m wind speed curves. One can see that the WRF forecasts the temperature drop and the wind speed maximum with a delay of about 13 h. Besides, the forecast drop in temperatures reaches lower values than the observed, while the simulated wind speed only reaches values of about 15 m s−1 , while the observed wind speed goes above 25 m s−1 . Since no significant improvement is observed via altering the SSTs, two additional simulations are performed - a forecast, preceded by a 12-hour model spin-up and a model free run with the GFS model analysis as an initial condition. The SST in the model is set as time evolving in the spin-up experiment while in the free run it only enters once as an initial condition. The 12-hour spin-up simulation

Numerical Weather Prediction for the Bulgarian Antarctic …

351

Table 4 Forecast statistics of the wind direction against measurement data, ERA-5 reanalysis data and the WRF run with unmodified evolving in time SST data from GFS Wind direction (deg) BAB measurements

ERA-5 reanalysis

WRF-GFS SST

BIAS

RMSE

MAE

BIAS

RMSE

MAE

BIAS

RMSE

MAE

WRF-GFS SST without update

−5.52

33.46

26.59

−1.09

36.99

27.28

2.50

20.68

6.47

WRF-GFS SST with update

−6.15

35.14

27.48

−3.60

35.61

26.76

–

–

–

WRF-GFS SST with update (−3 K) −12.18 40.53

30.48

−4.02

37.14

26.40

−0.42

27.33

13.64

WRF simulation for the 2016 test case

WRF-GFS SST with update (−1 K) -7.84

37.32

29.97

2.76

34.70

26.07

6.36

28.81

10.63

WRF-GFS SST with update (+1 K)

−4.76

32.40

25.09

−1.80

42.15

28.81

1.80

19.07

8.37

WRF-GFS SST with update (+3 K)

−13.19 41.77

35.35

−11.24 38.74

29.34

−2.64

23.87

12.47

WRF-CMEMS SST with update

−5.62

26.47

−3.66

28.51

−0.06

15.13

6.97 2.05

32.83

41.79

WRF simulation for the 2017 test case WRF-GFS SST without update

−33.01 119.28

110.08

−53.35 100.91

83.71

−0.29

3.97

WRF-GFS SST with update

−32.18 119.22

109.55

−53.07 99.88

83.46

–

–

–

WRF-CMEMS SST with update

−35.53 121.54

112.31

−54.79 106.24

89.83

-6.72

23.34

9.66

WRF—free run

10.01

47.07

29.76

78.64

52.42

22.83

70.09

55.23

WRF—with 12-hour spin-up

−33.43 121.45

52.28

112.40

−44.52 97.48

78.21

−4.38

18.36

7.84

WRF-GFS SST without update

−53.41 58.77

54.14

−16.14 26.80

20.97

−0.51

3.43

1.91

WRF-GFS SST with update

−52.90 58.27

53.68

−15.63 26.76

21.16

–

–

–

WRF-GFS SST with update. (−3 K) −55.95 61.67

56.34

−18.68 30.42

23.16

−3.04

6.37

4.33

WRF-GFS SST with update (−1 K) −53.78 59.03

54.33

−16.51 27.39

21.78

−0.88

3.74

2.39

WRF-GFS SST with update (+1 K)

−53.11 58.10

53.57

−15.84 25.95

21.11

−0.21

6.87

3.58

WRF-GFS SST with update (+3 K)

−52.13 58.41

52.63

−14.85 26.15

20.62

0.78

12.75

6.44

WRF-CMEMS SST with update

−54.31 59.25

54.73

−17.04 26.77

21.10

−1.40

4.59

2.97

WRF simulation for the 2020 test case

Fig. 7 Same as Fig. 5 but for the 2017 test case

352

B. Chtirkova et al.

is similar to the original forecasts and performs better in the initial forecast hours, when it shows higher temperatures and higher wind speeds. The ranges of the 2 m temperature, simulated in the free run, are closer to the observed ones, but the forecast curve moves behind the events. The free run simulation does not perform well in forecasting the wind speed at 10 m. The earlier illustrated sudden temperature drop in CMEMS SSTs (Fig. 3) only deepens the temperature minimum and this results in s higher RMSE value, compered to the other simulations in Table 1. According to the forecast statistics for the 2017 test case in Table 2, the sea level pressure is best forecasted in the 12-hour spin-up simulation, which pinpoints the negative impact of the coarse initial condition on high resolution numerical models. The necessity of lateral boundary conditions is demonstrated by the large differences in sea level pressure, simulated in the model free run. The comparison between the wind speed and direction in the 2017 test case against ERA-5 data shows best results in the 12-hour spin-up simulation. The forecast statistics for the wind directions are most promising in the free run with RMSE values more then 2 times lower than in the other simulations. This may imply that the wind field around BAB is influenced by some local circulation and the boundary conditions, incoming from the GFS, worsen the forecast in the 2017 test case.

5 Relation Between the Meteorological Variables and the Modification of SST The symmetrical planning of our numerical experiments for the 2016 and 2020 test cases enables us to construct the dependencies of some of the meteorological variables against the applied modification of SST. This approach deals entirely with model data, and thus the analysis cannot be affected by observational errors. In order to quantify the SST field modification, let us introduce the variable SST , which only takes discrete values of −3, −1, 0, 1 and 3 K. These values correspond to the five simulations, in which the SST has been equally modified in each sea point of the model grid. When SST = 0, the SST field consists of the unmodified values, provided by the GFS model. In this section, we will only consider the temperature at 2 m, sea level pressure and wind speed at 10 m in the 24th forecast hour in the finest domain—d03. This hour has been chosen to be sufficiently ahead in time for the forecasted field not to be disturbed by the initial condition, and to properly represent the model characteristics. For the 2016 test case the 24th forecast hour corresponds to 17.12.2020 0 UTC, and for the test case in 2020 this is 26.01.2020 0 UTC. Let us introduce the variables Tmin , Pmin and Vmin , which represent the minimum value of the whole model domain d03 in the 24th hour of the variables 2 m temperature, sea level pressure and 10 m wind speed respectively. In the same way, we can introduce Tmax , Pmax and Vmax as the corresponding maximums in the field at that moment. The aggregate of all field values will be represented by the aerial average values T ,

Numerical Weather Prediction for the Bulgarian Antarctic …

353

Fig. 8 Relations between SST and the field characteristics of T, P and V for the 24th forecast hour for the 2016 test case (left column) and the 2020 test case (right column)

P and V . The values at the BAB point at the 24th forecast hour will be denoted as TB AB , PB AB and VB AB . Figure 8 presents the relations between the different field characteristics and the imposed modification of SST for the 2016 and 2020 test cases. As anticipated, the 2 m temperature increases when we rise the SST, as Tmin rises more steeply than Tmax . The temperature value at BAB TB AB is the least affected by the numerical modification, because it characterizes only the temperature over land, while the other characteristics are representative for the whole domain, which is mostly composed of sea grid points. Consequently, a change in the SST has a greater impact on the 2 m temperature over water than over land. Looking through the change in the sea level pressure characteristics, given in Fig. 8c and d, one can see that an increase in SST leads to lower pressure values in both test cases. The value of Pmax is reduced more by the increase compared to Pmin .

354

B. Chtirkova et al.

The fact the curves P and PB AB have similar behavior illustrates that the sea level pressure field is homogeneously affected throughout the model domain. The 10 m wind speed characteristics, given in Fig. 8e and f, show a non-linear increase of Vmin . This sudden “jump” suggests that there is a value of V, above which additional processes are unlocked, which significantly increase the wind speed at the points of null speed. The decrease in Vmax and VB AB , but increase in V shows that the wind speed is differently affected in different grid points. However, its average value slightly increases with rising SST. The conducted sensitivity experiments show that all of the considered meteorological variables are affected by the sea surface temperature, the most prominent differences being observed in the 2 m temperature field. With an uniform rise in SST, the corresponding tendencies are: an increase of the 2 m temperature, a decrease of the sea level pressure and an increase of the average wind speed. For the BAB region, the best results with unmodified data are obtained when using SST data from the CMEMS operational ocean model [5].

6 Conclusion and Outlook The performed numerical experiments show that the results are rather influenced by the choice of the SST field as initial and boundary conditions for the weather forecast at the Bulgarian Antarctic base area. The examination of the time evolution of the different SST fields coming from the operational global models GFS and CMEMS indicates that the SST field is difficult to forecast and may vary rapidly of more than 2.5 K in the range of 48 h. Based on the model validation and the availability of modeled and measured water temperature data in the BAB bay, different experiments with different SSTs have been planned. The conducted experiments with the regional WRF show that increasing the SST leads to a rise in the 2 m temperature in the whole model grid, while the sea level pressure decreases linearly. The relationship between wind speed at 10 m and SST is more complex - with increasing the SST, the maximum wind speed decreases and the minimum suffers a non-linear increase. The different numerical experiments in the BAB region show that an increase of SST with 3 K leads to an increase of the 2 m temperature at BAB with 0.75 K, which reduces the RMSE. Overall, the numerical model underestimates the temperature, which is supposedly due to an inappropriate representation of the land surface cover. In order to increase the forecast skill in the summer season, a revision of the land use type is advisory. The simulations with model spin-up show slightly improved results. Optimum results with unmodified data are obtained, when using evolving in time CMEMS SST as model input. The authors recommend regular in situ measurements of the SST in the BAB bay, which will allow to calculate the global forecast SST bias in the area. The BAB modelling system configuration is modified to use the CMEMS global ocean forecast analysis and forecast as surface boundary condition in operational mode.

Numerical Weather Prediction for the Bulgarian Antarctic …

355

References 1. Bai, H., Hu, H., Yang, X.Q., Ren, X., Xu, H., Liu, G.: Modeled MABL responses to the winter Kuroshio SST front in the East China Sea and Yellow Sea. J. Geophys. Res.: Atmos. 124(12), 6069–6092 (2019). https://agupubs.onlinelibrary.wiley.com/doi/abs/10.1029/2018JD029570 2. Chtirkova, B.: A study on the impact of sea surface temperature on the weather forecast for the Bulgarian Antarctic base on Livingston Island. Master’s thesis (2020) 3. Chtirkova, B., Peneva, E.: The impact of SST on the weather forecast quality in the Bulgarian Antarctic Base area on Livingstone Island (2020). https://doi.org/10.5194/egusphere-egu20209347 4. Chtirkova, B., Peneva, E., Georgieva, G.: A modelling system for numerical weather prediction in the Bulgarian Antarctic Base area. In: Proceedings of the 1st International Conference on Environmental Protection and Disaster Risks, 29 Sept–1 Oct 2020, Sofia, Bulgaria (2020) 5. Copernicus: Copernicus Marine Environment Monitoring Service (2020). https://marine. copernicus.eu/ 6. Hersbach, H., Bell, B., Berrisford, P., Hirahara, S., Horányi, A., Muñoz-Sabater, J., Nicolas, J., Peubey, C., Radu, R., Schepers, D., Simmons, A., Soci, C., Abdalla, S., Abellan, X., Balsamo, G., Bechtold, P., Biavati, G., Bidlot, J., Bonavita, M., Chiara, G., Dahlgren, P., Dee, D., Diamantakis, M., Dragani, R., Flemming, J., Forbes, R., Fuentes, M., Geer, A., Haimberger, L., Healy, S., Hogan, R.J., Hólm, E., Janisková, M., Keeley, S., Laloyaux, P., Lopez, P., Lupu, C., Radnoti, G., Rosnay, P., Rozum, I., Vamborg, F., Villaume, S., Thépaut, J.N.: The ERA5 global reanalysis. Q. J. R. Meteorol. Soc. 146(730), 1999–2049 (2020). https://doi.org/10.1002/qj. 3803 7. Ivanov, L.L.: Antarctica: Livingston Island and Smith Island. Scale 1:100000 topographic map. Manfred Wörner Foundation (2017) 8. Miglietta, M.M., Moscatello, A., Conte, D., Mannarini, G., Lacorata, G., Rotunno, R.: Numerical analysis of a Mediterranean ‘hurricane’ over south-eastern Italy: sensitivity experiments to sea surface temperature. Atmos. Res. 101(1–2), 412–426 (2011). https://doi.org/10.1016/j. atmosres.2011.04.006 9. NCAR/UCAR: The Antarctic Mesoscale Prediction System (AMPS) (2020). https://www2. mmm.ucar.edu/rt/amps/ 10. Senatore, A., Mendicino, G., Knoche, H.R., Kunstmann, H.: Sensitivity of modeled precipitation to sea surface temperature in regions with complex topography and coastlines: a case study for the mediterranean. J. Hydrometeorol. 15(6), 2370–2396 (2014). https://doi.org/10. 1175/JHM-D-13-089.1 11. Skamarock, W.C., Klemp, J.B., Dudhia, J., Gill, D.O., Liu, Z., Berner, J., Wang, W., Powers, J.G., Duda, M.G., Barker, D.M., Huang, X.Y.: A description of the advanced research WRF model version 4. UCAR/NCAR (2019). 10.5065/1DFH-6P97. https://opensky.ucar.edu/ islandora/object/opensky:2898 12. Tuleya, R.E., Kurihara, Y.: A note on the sea surface temperature sensitivity of a numerical model of tropical storm genesis. Mon. Weather Rev. 110(12), 2063–2069 (1982). https://doi. org/10.1175/1520-0493(1982)1102063:ANOTSS2.0.CO;2

Water Resources, Human Activities and Management

Value Eco-Innovation as a Basis for Clean Production Through Ecodesign in the Bulgarian Food Industry Silviya Topleva , Tsvetko Prokopov , and Donka Taneva

Abstract In this paper we propose a model for implementation of value ecoinnovations of clean production through ecodesign in the SMEs of the food industry in Bulgaria. The clean production is a preventative approach to managing environmental aspects. The paper presents the essence of clean production through ecodesign, the methods for its achievement and outlining the possible options and barriers for SMEs from the food industry in Bulgaria to implement the relevant value eco-innovations. The ecological footprint of the food industry makes the need for an integrated implementation of clean production practices from raw material extraction to packaging and waste disposal of final consumer. Keywords Clean production · Value eco-innovation · Ecodesign · Food industry

1 Introduction The deepening of the consumer culture, the intensive production of food products and the exploitation of natural resources face the food industry to the challenge to manage and constantly reduce its environmental aspects in the conditions of constantly growing requirements for ensuring high quality and food safety. The striving to retain market positions require reengineering of traditional business models with a focus on eco-innovation for clean production and sustainable development. The modern high-technological innovative environment of digital and ecological transformations highlights the network creation of high value added. The idea of value added is to generate and obtain a higher level of welfare that goes beyond the utility, productivity and efficiency of the products, services and processes. The value added is causally related to the improvement of environmental aspects through ecodesign S. Topleva (B) · T. Prokopov · D. Taneva University of Food Technologies—Plovdiv, Plovdiv, Bulgaria T. Prokopov e-mail: [email protected] D. Taneva e-mail: [email protected] © The Author(s), under exclusive license to Springer Nature Switzerland AG 2021 N. Dobrinkova and G. Gadzhev (eds.), Environmental Protection and Disaster Risks, Studies in Systems, Decision and Control 361, https://doi.org/10.1007/978-3-030-70190-1_24

359

360

S. Topleva et al.

of processes, products and services. The innovative creation and generation of high value added for stakeholders through clean production and ecodesign takes place within cross-sectoral network partnership structures. The value eco-innovations of clean production determine the achievement of highly ecological supply chains in the food industry. The clean production is a preventative approach to managing environmental aspects. More broadly, the clean production through ecodesign concerns the technological and technical aspects of product manufacturing [1]. In the practice of clean production, ecodesign innovations occupy a leading position. In the food industry in Bulgaria, the clean production is not yet widespread. The food industry could be seen as the cross-point of the relationship between nature and respect attributed to the human race. In this sense, the clean production of food products expresses simultaneously the preservation of nature and the improvement of the quality of life of the people. The priority for clean production of food products is the reduction of the consumption of energy and water resources and the establishment of a waste management system at each phase of the life cycle. The clean production achievement is not a one-dimensional business activity, but an integrated, complex process of management and redesign of the environmental aspects of the company’s production, distribution, and consumer functions. The eco-innovation project management and the ecodesign of process and products are at the basis of this complex practice. The food industry is characterized by cumulative growing environmental aspects throughout all the life-cycle phases of the product. The production stage requires a high consumption of water and energy resources. The high quantity of wastes marks every preserved product from the raw material procurement to the disposal of the packaging by the final consumer. The structure of the food industry traditionally builds by a number of small and medium-sized enterprises (SMEs) operating in conditions of monopolistic competition. According to the National Statistical Institute, for 2018 in Bulgaria has a total of 31,272 enterprises in the manufacturing sector, as SMEs among them are 30,985. The added value that generate these enterprises amounted to 3,655,355 thousand Euros, as for the entire sector is 7,767,101 thousand Euros [2]. The SMEs in food industry are often family businesses that combine tradition and innovation. Moreover, the enhancement of the competitiveness of small and medium enterprises faces them to the challenge to develop innovative production solutions. The participation of small and medium-sized enterprises from the food industry in cross-sectoral network consortia deepens and expands their innovation activity through the mechanisms of technology transfer and the diffusion of collective innovative production and product solutions. The innovations of clean production created within cross-sectoral network structures, offer to the small and medium enterprises incentive to be both innovative, competitive and profitable [3]. The implementation of clean production practice in food industry SMEs goes beyond the scope of their innovation and investment activity. SMEs often do not have the methodology and resources to implement the practices of clean production in their technology and business processes. They face the challenge to optimize

Value Eco-Innovation as a Basis for Clean Production …

361

resource use in the production process, to meet regulatory and consumer demands for environmentally-friendly products and, at the same time—the inability to achieve them. The purpose of the paper is to propose a model for implementation of ecoinnovations of clean production in the SMEs of the food industry in Bulgaria. The achievement of this objective requires presentation of the essence of clean production, the methods for its achievement and outlining the possible options and barriers for SMEs from the food industry in Bulgaria to implement the relevant eco-innovations.

2 Clean Production Model Tools Methodology There is an abundance of scientific literature on the concepts for business models. In general, it outlines four pillars: value proposition, customer focus to increase revenue, an infrastructure network of partners and stakeholders to build value added, cost and revenue-related financial aspects [4]. The essence of modern business models is the creation of value added for the customer [5]. The consumers in food industry also support the creation of value added for SMEs [6]. The environmental challenges position innovation as a main source of value added for the firm’s stakeholders [7, 8]. The economic model of added value focuses on the client and the portfolio of his needs, which is associated with transformations in the strategic orientation of companies and in the process and project implementation of their activities. In the light of environmental aspects, the business model implies the value-added creation for stakeholders through the use of existing company’s resources and capabilities [9]. The resource-based view, dynamic capabilities view, and strategic entrepreneurship are the main drivers of innovative sustainable business models innovations [9]. The resource-based view is associated with the effective management of the company’s resources. The dynamic capabilities view relies on the company’s potential to create value added through eco-innovation. The strategic entrepreneurship involves the company’s capabilities to combine its strengths with the favorable opportunities of the external environment. Thus, the unity of the company’s resources, competencies and capabilities becomes the source of innovative green business modeling of corporate sustainability through implementation of clean production. To achieve sustainable business growth through clean production is necessary the correct application of appropriate methods. The clean production methodology through ecodesign typically affects the technological processes and technical characteristics of the products and environment. The methodology for implementation of clean production through ecodesign usually concerns the technological processes and technical characteristics of the products and the environment. The basis of clean production through ecodesign is Out-of-the-box thinking, inspiration from nature and 5 × R (see Fig. 1). The clean production is based on the reuse and recycling of by-products and waste, the use of renewable energy sources, optimization of the consumption of energy and water resources. Thus, the clean production is becoming an expression of the essence

362

S. Topleva et al.

Fig. 1 The principles 5 × R as a basis for clean production through ecodesign. Source Boye and Arcand [1]

Re-think (product and funcon) Re-cycle/recover (materials, energy)

Re-use (product)

Re-duce (energy and materials)

Re-place (hazardous chemicals)

of the bio-based economy. The clean production means that the by-products and the waste from the production of one product become a raw material for other products and production processes. The recycling and reuse of resources and raw materials as a basis for clean production are a new area for eco-innovation in the food industry [10]. The recycled food content is a sensitive issue for both producers and consumers. The studies in the field are still experimental and at the sensory analysis level [11]. In light of this idea, the clean production of food products helps to relieve the raw material dependence of the sector and contributes to enhancing the food security of mankind. The eco-innovation of clean production through ecodesign of food products is not limited to the production process. The ecodesign covers the packaging of the product, its distribution, storage and completion of the life cycle. At the same time, the value eco-innovations imply a complete organizational transformation of the dominant business models. The focus is on creating and delivering high value added for the customer and society by achieving sustainable production of eco-efficient products. The complex nature of clean production through ecodesign of processes and products imposes the need for integrated application of environmental and economic methodologies to achieve sustainability in the food industry. The main methods for innovative implementation of clean production through ecodesign in the supply chain in the food industry are Cradle-to-Cradle certification, LCA, MET Matrix, corporate social responsibility, stakeholder approach and lean process methodology. Cradle-to-Cradle certification is applied for products, which are result of clean production, recycling and reuse of waste products with a high content of the biologically active substances. The adequate transformation of conventional technological

Value Eco-Innovation as a Basis for Clean Production …

363

processes into clean production requires a careful analysis of the environmental aspects of the product life cycle phases. LCA is a product-oriented tool for analyzing and quantifying the products’ environmental impact of cradle-to-grave [1, p. 3]. LCA is an internationally recognized tool whose application is regulated by ISO standards. The Life Cycle Assessment supports the definition and calculation of the cumulative impact of the product or process on the environment, the formulation of alternatives for improving the environmental aspects of products and processes, development of company and sector analysis, quantitative substantiation of the change in the ecological footprint of a product or process [12, p. 2]. Thus, LCA is often perceived as a source of clean production. The Life Cycle Assessment and accompanying methods diagnose, summarize and indicate where are the most prominent environmental aspects of the food product [13]. They do not provide strategies for the environmental development, innovations, redesign of the product or process and clean production. This requires the application of the MET Matrix method. The MET Matrix method allows to investigate and to operationalize the specific environmental aspects of the product through input materials, energy and the level of emitted toxicity [14]. The method combines qualitative and quantitative analysis. Based on the results of the MET Matrix analysis, the products are grouped into five groups: • • • • •

Type A: Raw material intensive product Type B: Manufacture intensive product Type C: Transportation intensive product Type D: Use intensive product Type E: Disposal intensive product.

The application of the MET Matrix method guides manufacturers in choosing a specific strategy to improve the environmental aspects of the product and the implementation of clean production. Corporate social responsibility is a crosspoint of community interdependence and selfishness, which determines responsible business behavior and environmental aspects management of the activities of small and medium enterprises in the food industry. Thus, CSR is becoming one of the main tools for sustainability achievement. However, the CSR reporting systems do not provide a comprehensive framework for the company’s environmental performance assessment [15] and do not assure an adequate tool for introduction of clean production through ecodesign, especially in the food industry. The application of environmental standards and sustainability reports are fragments of the efforts to achieve eco-innovative clean production through ecodesign of processes and products. In the context of a high-technological networked economy, responsible business behavior integrates the interests of all stakeholders. The satisfaction of CSR requirements for the purposes of clean production depends on the application of a stakeholder approach. The stakeholder approach focuses on involving stakeholders in the

364

S. Topleva et al.

company’s external effects management and revealing valuable opportunities ecoinnovation for them. The model of clean production through ecodesign, as an open social innovation in which all stakeholders, in the supply chain and the value chain, interact, is both a source and a result of green social change of process and products [16]. The basis of the process-oriented approach to the management of value ecoinnovations of clean production through ecodesign is the transformation of all processes of creation, production and distribution of products with a focus on the customer. The customer is placed at the beginning of the eco-innovative supply chain. One of the most common ways to implement process management is the lean methodology. The content of the lean methodology is expressed in the efforts to reduce the time for execution of the customer request and continuous improvement (kaizen) of all processes through standardized tasks, 5S, production in small volumes, waste elimination, inventory reduction, integration of all stakeholders, preventive maintenance, teamwork, quality control, value flow mapping, training and coaching, process synchronization [17]. The 5S methodology aims to reduce waste and increase productivity through Sort, Set in order, Shine, Standardize, Sustain. Eco-innovations are the basis of lean management of business activity in industrial production. The Lean methodology aims to reduce defective products, establish a single production flow, reduce waste and increase value, focus on the customer, reduce costs, increase quality and productivity [17]. The effective management of the waste products in the process lean methodology achieves both the improvement of the ecological aspects of the production and the optimization of the costs of the enterprise. The efforts to reduce resource consumption and the emissions, as well as the processing and secondary use of materials and raw materials through ecodesign and clean production, help reduce operating costs. The responsible business conduct goes beyond the imperatives of business efficiency and increasingly focuses on eco-efficiency and eco-balances. The clean production is increasingly recognized a guiding principle in manufacture development strategies, especially in the food industry.

3 Results and Discussion The clean production of food products determines the prevention and reduction of environmental aspects of food throughout all phases of the life cycle. The clean production also affects the engineering characteristics of the technological processes [18, 19]. The strategic green resource management is at the core of clean production. The eco-efficient products have reduced energy and resource consumption and minimal waste. It stimulates the development of innovations. The clean production is an incremental green innovation that allows simultaneous cost and resource input optimization, improvement of quality and of applied resources and technologies.

Value Eco-Innovation as a Basis for Clean Production …

365

The innovation model of clean production is among the basic business case for sustainability. The innovative business models for sustainability can be summarized in three aspects: (1) business case for sustainability generates corporate sustainability and sustainable development [20]; (2) eco-innovation is the result of the company’s economic rationality [21]; and (3) eco-innovation is encouraged if it brings value for the company [22]. The business model for implementation of clean production is a proactive eco-innovation, according to the Wilson’s classification [23]. The food industry is very resource intensive. According to Eurostat data, electricity consumption by industry, transport activities and households/services (GWh) in Bulgaria varies between 7818 and 8910 GWh for the period 2010–2016, which is above the electricity consumption of comparable countries such as Estonia (2095–2152 GWh), Croatia (3478–3291), Latvia (1590–1667 GWh), Slovenia (5487–6057 GWh). Drying and packaging are also resource intensive. The packaging of food products is related to food quality and safety. At the same time, it is also a communication tool between manufacturers and consumers. The possibilities of digital marketing contribute to the implementation of clean production at the stage of market presentation and advertising of food products [24]. The implementation of value innovations of clean production through ecodesign in the food industry determines the achievement of sustainability of the sector. It is therefore important to make a distinction between short-term market success based on digital and green marketing and comprehensive long-term sustainability, which increases the resource efficiency of the sector [25]. The nature of production technology in the food industry is also characterized by high waste intensity. Animal and vegetal wastes, according to Eurostat data in Bulgaria, vary between 731,091 and 972,685 T for the period 2010–2016. For comparison, these wastes for Estonia are 280,338–151,405 T, Croatia—119,502– 614,474 T, Latvia—166,304–143,395 T, Hungary—808,058–734,568 T, Slovenia— 264,045–267,438 T (Eurostat). The majority of countries are trying to reduce the amount of generated animal and vegetable waste. The trend in Bulgaria is the opposite, which also calls for more intensive measures for the implementation of clean production. A positive trend in Bulgaria is the increase in the share of recyclable wastes—1,708,337–2,049,372 T (Eurostat). The trend in other European countries in the region is similar. But, the waste generated by the Bulgarian industry for the period 2010–2016 is characterized by sustainability and even slight growth: 3,306,468– 3,469,171 T (Eurostat). In the other European countries in the region, the trend is similar. The optimization of waste management in the production processes is supported by the project management of the corresponding innovations [26]. Significantly the achievement of sector sustainability is a function of waste reduction. On this background, Environmental protection expenditure in Bulgaria for the period 2010–2013 is 0.37–0.66% of GDP (Eurostat), bringing the country closer to other European countries. The share of Total environmental investments for 2010–2013 in Bulgaria is 0.46–0.34% of GDP (Eurostat), whereby the country also does not differ from other European countries.

366

S. Topleva et al.

In terms of content, the ecoindustrial synergy of clean production is based on recycling design, design for recovered production, and design for release [27, pp. 156– 157]. The idea is to incorporate recyclable materials with extended life and multiple uses into the production process and to design the waste-free end of product life cycle. The achievement of clean production is a function of the use in the manufacturing process of materials and components with extended life, which can be recovered and reused, the use of recycled materials and/or raw materials that are recyclable, designing waste-free end of product life cycle, i.e. the transformation of waste materials (by-products) into production raw materials. The achievement eco-industrial synergy depends on the establishment of crosssectoral network consortia for innovative ecodesign of processes, products and services. Traditionally, the matrix of the structure of network cooperation includes educational and research centers, business organizations, government institutions and civil society organizations. The educational and research institutions are at the heart of innovative network partnerships. The social capital of academic communities helps to intensify their network participation and the transfer of innovative developments into entrepreneurial business initiatives [28]. The aim is the development, implementation and diffusion of value ecoinnovations of clean production through ecodesign. The source of eco-innovative proactivity in the network partnership is the construction of eco-industrial parks. Shared eco-innovation proactivity within a networked cluster consortium achieves a reduction in the use of natural resources in the production process and the amount of waste. The time for transferring by-products into productive raw materials and useful products is shortening. The possibility of direct application of by-products reduces harmful emissions and waste. The scarcity of resources, the need for energy efficiency, the optimization of waste production intensity and the rising food quality and safety requirements determine the need for the implementation of clean production in SMEs in food industry. In addition to improving the environmental aspects of the product, clean production could also be motivated by the striving to reduce costs, increase confidence, open up new market segments, increase fairness and responsibility. Thus, environmental, economic and social factors intertwine with the application of clean production. The functional innovations for clean production implementation help to achieve ecoefficiency. In the light of this analysis, clean production is becoming the main source for the food industry sustainability. The implementation of eco-innovations often encounters a number of constraints and barriers, including inertia in companies [29, 30]. The business model for the clean production implementation is generally applicable to all interested small and medium-sized enterprises in food industry, but it still requires some limitations. They are primarily related to the innovative nature of the model and hence the need for investment. Often, it is difficult for SMEs to allocate funds for green innovation, even if they are not so complex. Value eco-innovation of clean production through ecodesign of processes and products requires solid investments that are not the strength of small and medium enterprises [31]. The construction of eco-industrial parks also

Value Eco-Innovation as a Basis for Clean Production …

367

requires huge investments. Unfortunately, there is no such eco-industrial park in Bulgaria yet. The self-organization of enterprises is not sufficient for the implementation of high-technological value eco-innovations. The achievement of clean production through ecodesign determines the need for state support through environmental, industrial and fiscal policies. Furthermore, the production volume of SMEs usually does not allow the negotiation of lower prices for ecological raw materials. Thus, the risk of unprofitable manufacture increases. The consumer prices would also be uncompetitive, threatening the market realization of the product. Therefore, the eco-innovative business model of clean production is better suited to implementation in innovative food start-ups, SMEs that are a part of green cluster networks and eco-oriented companies with previous green innovations. The dimensions and application of clean production in the Bulgarian food industry at this stage are hampered by the immaturity of the economic forms in the country, the lack of free capital in the enterprises for voluntary investments in environmental projects (rather it meets regulatory requirements), the lack of experience and insufficient methodological knowledge. The benefits and contributions of clean production in SMEs can help to overcome these difficulties.

4 Conclusion The eco-innovation of clean production through eco-design is an integral part of efforts to achieve high quality and food safety from raw material extraction to enduser consumption. The clean production is not solely a technological innovation, but a holistic business model that involves integrating environmental requirements into corporate strategy, indicators, tools and process and industrial design. The ecological footprint of the food industry makes the need for an integrated implementation of clean production and eco-design practices from raw material extraction to packaging and waste disposal of final consumer. Thus, functional food products can be created to combine the requirements of high quality, safety, eco-efficiency and health. Theoretically, the business model for implementation of eco-innovations of clean production in the activities of small and medium enterprises in the food industry, as a source of high value added, complements scientific knowledge with the complex application of vertically and horizontally integrated mix of methods and approaches to sustainability achievement. The integrated application of the selected methods contributes to the deepening of the research on the eco-innovative reengineering of the technical, technological and organizational processes in the company and the product portfolio. The sustainability achievement through value eco-innovations of clean production requires a deepening of applied research. In this regard, the subject of future research should be the approbation of the eco-innovative business model of clean production in a small and medium-sized company of food industry in Bulgaria. The

368

S. Topleva et al.

investigation of this business case suggests: first, study on firm internal environment and market factors for the implementation of value eco-innovations of clean production and, if necessary—reengineering of the corporate activities organization with a focus on strategic management of sustainable corporate development; second, monitoring of the process of implementation of integrated innovative business model for value eco-innovations of clean production in terms of deepening the digitalization of manufacturing and economy; and third, a study of the effects of the implementation of the business model on stakeholders with a focus on acquired value added, consumer satisfaction and sustainable development. The effects of the implementation of value eco-innovations of clean production in the SMEs in food industry can provoke accelerated diffusion and the creation of new eco-innovations. Thus, the complex clean production will be recognized as a major source of value added and sustainability in high-technological bio-based economy.

References 1. Boye, J., Arcand, Y.: Current trends in green technologies in food production and processing. Food Eng. Rev. 5(1), 1–17 (2013) 2. National Statistical Institute: http://www.nsi.bg 3. Birkin, F., Cashman, A., Koh, S. C. L., & Liu, Z.: New sustainable business models in China. Bus. Strategy Environ. 18(1), 64–77(2009) 4. Ballon, P.: Business modelling revisited: the configuration of control and value. J. Policy, Regulation Strategy Telecommun., Info. Media 9(5), 6–19 (2007) 5. Wirtz, B.: Business Model Management Design, Instruments, Success Factors, Gabler, Wiesbaden, (2011) 6. Tardivo, G., Thrassou, A., Viassone, M., et al.: Value co-creation in the beverage and food industry. British Food J. 119(11), 2359–2372. https://doi.org/10.1108/BFJ-02-2017-0119 (2017) 7. Pohle, G., Chapman, M.: IBM’s global CEO report 2006: Business model innovation matters. Strategy Leader. 34(5), 34–40 (2006) 8. Teece, D. : Business models, business strategy and innovation. Long Range Plan. 43(2/3) 172–194 (2010) 9. Schneider, S., Spieth, P.: Business model innovation: towards an integrated future research agenda. Int. J. Innov. Manag. 17(01), (2013) 10. Dalhammar, C.: Industry attitudes towards ecodesign standards for improved resource efficiency. J. Clean. Prod. 123, 155–166 (2016) 11. Prokopov, T., Chonova, V., Slavov, A., et al.: Effects on the quality and health-enchancing properties of industrial onion waste powder on bread. J. Food Sci. Technol. 55(12), 5091–5097 (2018) 12. Williams, A.: Life Cycle Analysis: A Step by Step Approach. Illinois Sustainable Technology Center (2009) 13. Calderón, L.A., Iglesias, L., Laca, A., et al.: The utility of life cycle assessment in the ready meal food industry. Resour. Conserv. Recycl. 54, 1196–1207 (2010) 14. Ostad-Ahmad-Ghorabi, H., Pamminger, R.: Ecodesign for Sustainable Development. Education and Culture Lifelong Learning Progrmme Erasmus (2007) 15. Munif, M.: Corporate accountability in the context of sustainability—a conceptual framework. EuroMed J. Bus. 8(3), 243–254 (2013). Available at: https://doi.org/10.1108/EMJB-05-20130026

Value Eco-Innovation as a Basis for Clean Production …

369

16. Santoro, G., Ferraris, A., Vrontis, D.: Open social innovation: towards a refined definition looking to actors and processes. Sinergie 36(105). https://doi.org/10.7433/s105.2018.02. (2018) 17. Pettersen, J.: Defining lean production—some conceptual and practical issues. TQM J. 21, 127–142 (2009) 18. Dewberry, E., Goggin, P.: Spaceship ecodesign. Co-Design 12–17 (1996) 19. Graedel, T.E., Allenby, B.R.: Industrial Ecology. Prentice Hill, New Jersy (1995) 20. Dyllick, T., Hockerts, K.: Beyond the business case for corporate sustainability. Bus. Strategy Environ. 11(2), 130–141 (2002) 21. Eden, S.: Using sustainable development: the business case. Glob. Environ. Change 4(2), 160–167 (1994) 22. van Marrewijk, M.: Concepts and definitions of CSR and corporate sustainability, between agency and communion. J. Bus. Ethics 44(2/3), 95–105 (2003) 23. Wilson, I.: What one company is doing about today’s demands on business. In: Steiner, G.A. (ed.) Changing Business-society Interrelationships. Los Angeles Graduate School of Management, UCLA, Los Angeles (1975) 24. Kulova, I., Mihailov, M.: Digital marketing—the key to successful electronic business, scientific works of University of food technologies. In: Proceedings of the 65th Anniversary Scientific Conference with International Participation “Food Science, Engineering and Technology— 2018”, Organized in Cooperation with the Foundation “Scientific Research”, pp. 205–210 (2018) 25. Mastroberardino, P., Calabrese, G., Cortese, F., Petracca, M.: Sustainability in the wine sector: an empirical analysis of the level of awareness and perception among the Italian consumers. Br. Food J. 122(8), 2497–2511 (2019). https://doi.org/10.1108/BFJ-07-2019-0475 26. Teneva, A., Nikolova-Alexieva, V., Tuntova, A.: Maintaining steady cost projection on wastewater engineering projects in Bulgaria. In: International Conference on High Technology for Sustainable Development (HiTech) (2018). https://doi.org/10.1109/hitech.2018.8566388 27. Madu, C.: Competing on Quality and Environment. Chi Publishers (2003) 28. Kalar, B., Antoncic, B.: Social capital of academics and their engagement in technology and knowledge transfer. Sci. Public Policy 43(5), 646–659 (2016) 29. Chesbrough, H.: Business model innovation: opportunities and barriers. Long Range Plan. 43(2/3), 354–363 (2010) 30. Doz, Y., Kosonen, M.: Embedding strategic agility: a leadership agenda for accelerating business model renewal. Long Range Plan. 43(2/3), 370–382 (2010) 31. Murovec, N., Erker, R., Prodan, I.: Determinants of environmental investments: testing the structural model. J. Clean. Prod. 2012(37), 265–277 (2012). https://doi.org/10.1016/j.jclepro. 2012.07.024

Porewater Nutrient and Oxygen Profiles and Sediment-Water Interface Fluxes Under Extreme Organic Loading in Different Sedimentary Habitats in Sozopol Bay (SW Black Sea): A Laboratory Experiment Stefania Klayn , Dimitar Berov , and Ventzislav Karamfilov Abstract Coastal benthic sediments play an important role in regulating water column nutrient concentrations and primary production via nutrient regeneration and exchanges at the sediment-water interface. This study aimed to characterize 3 the porewater concentrations and diffusive benthic fluxes of NH4 + , NO3 − , PO− 4 , and O2 in some of the most common shallow sedimentary habitats (fine and coarse sands, seagrass beds, and unvegetated patches within the seagrass beds) along the Bulgarian coast, and their changes under organic loading, through a laboratory experiment. Ammonium was the dominant form of nitrogen in porewaters, and its concentration generally increased under organic loading in most sediment types. Nitrate concentrations were high in the overlying water, and decreased with depth within the sediments, becoming depleted at ~3 cm with the development of anoxic conditions. Phosphate concentrations were low, and tended to increase with depth by the end of the experiment in most sediment types and especially under organic loading. Nutrient fluxes were dominated by a release of NH4 + to the water column in all sediment types, and a parallel uptake of NO3 − by the sediments; both fluxes increased under organic loading, possibly indicating stimulation of nitrate reduction within the 3 sediments. The PO− 4 fluxes were smaller, and the sediments mostly acted as a source for phosphorus under organic loading. O2 was taken up from the overlying water in all treatments and sediment types, and this flux increased under organic loading, probably in relation to the decomposition of the organic matter and spontaneous chemical oxidation of sulphide ions, released during sulphate reduction within the sediments. The study contributes towards the understanding of nutrient cycling and the role of the benthic compartment in Black Sea coastal soft-bottom habitats. Keywords Benthic fluxes · Sediment-water interface · Pore waters · Nutrient recycling · Coastal zone · Black Sea

S. Klayn (B) · D. Berov · V. Karamfilov Institute of Biodiversity and Ecosystem Research, Bulgarian Academy of Sciences, 2 Major Yurii Gagarin Street, 1113 Sofia, Bulgaria © The Author(s), under exclusive license to Springer Nature Switzerland AG 2021 N. Dobrinkova and G. Gadzhev (eds.), Environmental Protection and Disaster Risks, Studies in Systems, Decision and Control 361, https://doi.org/10.1007/978-3-030-70190-1_25

371

372

S. Klayn et al.

1 Introduction Coastal ecosystems, located at the interface of land and sea, and home to a large and continuously growing proportion of the global human population [1], are subjected to various natural and anthropogenic pressures. Yet, these are some of the most diverse ecosystems, with a wide variety of habitats, and an important part of global biogeochemical cycles, acting as a buffer for the open ocean from anthropogenic nutrient inputs [2]. Eutrophication resulting from increased nutrient inputs to the coastal zone, and consequent deterioration in water quality, continue to be reported worldwide, despite legislative and management measures taken in recent years to reduce loads [3, 4]. Particulate nutrients deposited in coastal sediments may either recycle back into the water column or become retained or transformed in the seabed, depending on physicochemical and biological factors, and particularly the oxygen concentration in the bottom water and advective transport [5, 6]. The benthic compartment can thus act as either a source or a sink for nutrients, and can exert control on the nutrient levels in the overlying water column, especially in shallow environments. The release of nutrients can sometimes lead to secondary benthic-driven eutrophication, particularly via sediment resuspension or diffusion from the sediments to water columns depleted of nutrients from high levels of primary productivity in summer [6, 7]. Nutrient porewater profiles have been used as an indication for biogeochemical reactions taking place within the sediments, in order to characterize the role of the benthic compartment in biogeochemical cycles (e.g. [8–10]). Fluxes at sedimentwater interface, linking the benthic with the pelagic compartment, also represent mineralization processes within the sediments and nutrient regeneration, and are a crucial factor affecting nutrient balance and primary productivity in the water column, and from there—coastal water quality [11]. While many benthic flux studies have focused on compact, non-permeable silts, sandy sediments, with their varying topography and sometimes deep pore water penetration, can also provide conditions for efficient remineralization of organic matter and nutrient exchange. Additionally, sands are some of the most common and widely distributed benthic substrate types, and therefore likely play a significant role in biogeochemical processes, acting as filters where organic matter inputs from the water column are quickly decomposed [12, 13]. Along the Bulgarian Black Sea coast, nutrient fluxes at the sediment-water interface have been studied primarily in silty sediments. Doncheva and Shtereva [14] registered the highest nutrient fluxes in front of Kamchia River, attributable to the riverine inputs. Doncheva [15] also found very high sediment-water fluxes in Varna Lake and Varna Bay, especially in summer, which were enough to support hypereutrophic conditions in the lake, and eutrophic—in the bay. However, there are few studies on these processes in the southern Black Sea, where Burgas Bay—the other large bay along the Bulgarian coast, with potentially high water residence times, and which acts as a source of nutrients to the neighbouring coastal waters—is located [16–19].

Porewater Nutrient and Oxygen Profiles and Sediment-Water …

373

This study aims to quantify the porewater nutrient concentrations and the interfacial diffusive fluxes in different sedimentary habitat types in Sozopol Bay, and their changes under experimental organic loading, through a laboratory experiment simulating an extreme eutrophication event.

2 Materials and Methods 2.1 Study Area and Field Sampling Sozopol Bay is a semi-enclosed bay located in the southern part of the Bulgarian Black Sea, within the larger Burgas Bay. The bay is part of the NATURA 2000 protected area BG0000146 “Plazh Gradina-Zlatna Ribka”. The majority of the area is marine, and includes three seagrass meadows, different soft bottom biotopes, and reefs with macroalgae and black mussels. The area is strongly influenced by different anthropogenic pressures from the town of Sozopol, extensive tourism, and fisheries industries in the area. Sampling was carried out during the summer of 2019. 3 replicate core samples (internal diameter 10.5 cm, sampling depth ~20 cm) were collected by Scuba divers from 4 different sedimentary habitats in Sozopol Bay, representing the most common substrates in the area (Fig. 1). Station 1 has coarse sandy substrate (mean grain size 0.5–1 mm according to the Folk and Ward classification [20]; station 2—fine sand (mean grain size 0.125– 0.25 mm), station 3—seagrass meadow (Zostera marina and Zostera noltei), with predominantly fine sandy substrate mixed with a little silt; and station 4—an adjacent unvegetated patch within the seagrass meadow (“no Zostera”), with similar substrate, occasionally mixed with mollusk shells and fragments. All stations are located at 4– 4.5 m depth. The main characteristics of the sampling sites are presented in Table 1. One of the coarse sand samples was disturbed during handling and transportation, leaving only 2 for that habitat type.

2.2 Laboratory Experiment In the laboratory, samples were placed in a gently aerated aquarium, and left to acclimate in near-natural conditions for 1 month. Distilled water was added weekly to maintain water level and salinity. After this period, one replicate per habitat type was used to measure sediment parameters (water content, porosity) and organic matter content. Nutrient (N–NH4 , N–NO3 , P–PO4 ) and oxygen concentrations were measured in the water layer immediately overlying the sediment surface, and in the pore waters within the sediments in each of the remaining cores. Pore waters were sampled each cm to 10 cm depth, and oxygen was measured each cm until the

374

S. Klayn et al.

Fig. 1 Study area and sampling stations. 1—coarse sand, 2—fine sand, 3—seagrass meadow (Zostera), 4—unvegetated patch within seagrass meadow (no Zostera). Stations 3 and 4 are adjacent, located within several metres of each other, and so share the same coordinates

Table 1 Surface sediment characteristics (top 1 cm) at the sampling stations, with the average values for each sample shown in brackets. %TOM—% total organic matter content Station

Habitat type

%TOM

Porosity (mL cm−3 )

Bulk density (g cm−3 )

Water content (%)

1

Coarse sand

1.38 (1.35)

0.34 (0.34)

1.09 (1.17)

23.73 (22.60)

2

Fine sand

1.28 (1.36)

0.33 (0.19)

1.13 (0.65)

22.33 (22.34)

3

Zostera

1.47 (1.45)

0.58 (0.41)

1.55 (1.13)

27.17 (26.66)

4

No Zostera

1.17 (1.33)

0.46 (0.37)

1.56 (1.24)

22.52 (22.91)

Bulk density is derived from dry sediment weight

detection of persistent anoxic conditions. The nutrient concentrations were measured according to [21]. The oxygen concentrations were measured with a microelectrode (MC100 Microcell O2 m). Organic matter was determined as weight loss on ignition at 520 °C. One of the two replicates per habitat type was loaded with 6 g dried and finely ground green algae deposited on the sediment surface, or 0.069 g cm−2 — about double the normal surface organic matter content in a typical seagrass bed in

Porewater Nutrient and Oxygen Profiles and Sediment-Water …

375

the study area (0.03–0.04 g cm−2 ) [22]. Effectively, this simulates organic loading from an extreme eutrophication event. The other replicate served as control. The sole remaining coarse sand sample was loaded with organic matter. The experiment continued for 39 days, after which the pore water nutrients and oxygen, and the sediment parameters in each sample were measured again according to the same protocol.

2.3 Flux Calculations Diffusive fluxes at the sediment-water interface were calculated according to Fick’s first law, corrected with the porosity and tortuosity for the particular sedimentary matrix, and expressed in mmol m−2 day−1 (nutrients) and mg m−2 day−1 (oxygen), with positive values representing uptake by the sediments, and negative—release to the water column: F =−

φ D0 dC t θ2 dx

φ—sediment porosity at the interface; calculated from the sediment samples D0 — diffusion coefficient of the solute in seawater without the presence of the sediment matrix, corrected for average temperature (25 °C for the start and 21 °C for the end of the experiment) and salinity (16.5) at the study site and during the experiment [23] (Table 2). θ —sediment tortuosity (dimensionless), expressing the influence of the sediment matrix on the diffusion, and calculated based on the porosity as θ 2 = 1 − 2 ln(φ). dC—difference in concentration of the solute between the pore water at a particular depth and in the seawater immediately overlying the sediment surface d x—distance (cm) which the ion has to migrate from that depth to the sediment surface t—time, as number of seconds in a day (86,400). Table 2 Specific diffusion coefficients for the solutes in seawater at 21 °C (D21) and 25 °C (D25)

Solute

D21

D25

O2

0.195

0.216

NH4

0.176

0.191

NO3

0.170

0.185

PO4

0.054

0.060

376

S. Klayn et al.

3 Results 3.1 Porewater Nutrient and Oxygen Profiles Porewater oxygen concentrations sharply decreased, from 7 to 8 mg L−1 in the water overlying the sediments, to 0 at the 1–2 cm in most sediment types; only the sediments from the seagrass bed had higher oxygen penetration depth (3–4 cm) (Fig. 2). By the end of the experiment, all sediment types had become anoxic below 2 cm depth in both control and treated conditions. Ammonium was the main form of dissolved inorganic nitrogen in the pore waters. Its concentrations also increased with depth in the sediment as the latter became more reduced. Under organic loading, the porewater NH4 + concentrations increased, but the shape of the vertical profile did not change. No change was observed in the seagrass sediments (Fig. 2). The concentrations of nitrates were relatively high in the overlying water—the highest of any of the measured nutrients; within the pore waters, NO3 − decreased from the superficial sediments downwards, becoming depleted at about 3 cm depth in most sediment types and experimental conditions. There were no obvious changes in porewater nitrate profiles between treatments in the different sediment types, with the exception of the sediments from the unvegetated seagrass patch (higher values in both treatments, and depletion only at 10 cm in the treated sample) (Fig. 2).

Fig. 2 Porewater nutrient and oxygen profiles in the different sedimentary habitat types under control conditions (green circles) and organic matter loading (orange triangles) at the start and the end of the experiment. 0 cm indicates the water immediately overlying the sediments

Porewater Nutrient and Oxygen Profiles and Sediment-Water …

377

3 The porewater PO− 4 concentrations were low, and increased with depth. By the end of the experiment, phosphate concentrations in pore waters had increased slightly, especially in the fine and coarse sand, and to a certain extent in the unvegetated seagrass sediment, in both control and treated samples. In the seagrass sediments, there was no change in porewater phosphates in the control, and an overall decrease under organic loading (Fig. 2).

3.2 Sediment-Water Interface Fluxes All sediment types acted as a source of NH4 + throughout the experiment, and the magnitude of the flux increased under organic loading—sometimes as much as tenfold, e.g. in the fine sand (Table 3). By contrast, all were a sink of NO3 − , and in most cases its uptake also increased under organic loading (with the exception of the 3 sediments from the unvegetated patch in the seagrass bed). The PO− 4 fluxes were Table 3 Diffusive fluxes at the sediment-water interface of oxygen (O2 ), ammonium (NH4 ), nitrate (NO3 ), and phosphate (PO4 ) in different sediment types from Sozopol Bay at the start of the experiment and 39 days after extreme organic loading Habitat type Coarse sand

Fine sand

Zostera

No Zostera

Solute

Start

End

Control

Organic loading

Control

Organic loading

O2

–

13.22

–

50.42

NH4

–

−34.82

–

−208.13

NO3

–

126.16

–

180.62

PO4

–

0.00

–

1.33

O2

12.12

12.03

28.17

91.12

NH4

−197.71

−125.54

−535.98

−1357.14

NO3

113.51

115.31

145.61

453.99

PO4

−0.70

0.00

0.00

−9.18

O2

35.18

28.66

30.54

40.99

NH4

−161.48

−20.97

−161.17

−100.00

NO3

246.18

40.21

243.11

283.45

PO4

0.98

4.69

−0.79

−1.18

O2

20.57

22.23

20.72

14.32

NH4

−239.90

−235.93

−44.51

−392.96

NO3

73.06

111.18

101.48

94.62

PO4

0.00

0.00

0.53

−1.43

Fluxes are in mmol m−2 d−1 for the nutrients, and in mg m−2 d−1 for the oxygen. Positive values indicate solute uptake by the sediments, and negative—release to the water column. There are no values for control conditions in coarse sediments, because the sample was disturbed during transport and handling

378

S. Klayn et al.

the smallest in magnitude; three of the sediment types subjected to organic loading had started releasing them to the overlying water at the end of the experiment, and only the coarse sands were taking them up.

4 Discussion The exchange of nutrients between sediments and water is a complex phenomenon which depends on multiple physical, chemical and biological factors, such as temperature, dissolved oxygen, redox potential, benthic organism activities and organic matter. Our results are comparable to other experimental studies that found that benthic nutrient regeneration generally increases with organic loading [24]. The high fluxes of ammonium and nitrates observed in our experiment reflect the strong early diagenesis of organic matter in all sediment types. This is also supported by the increased consumption of O2 by the sediments in the organic loading treatment at the end of the experiment—especially high in the fine sands, which also exhibited the largest differences in nitrogen fluxes under organic loading. Spontaneous chemical oxidation of sulphide ions, released during sulphate reduction within the sediments, likely also contributes to the O2 consumption by the sediments. Porewater NH4 + is probably mainly derived from the bacterial mineralization of the organic matter, which is supported by its general increase in concentrations under organic loading. The observed release of ammonium by the sediments and consumption of nitrates could be related to stimulated reduction of nitrates to ammonium within, in anoxic conditions, since by the end of the experiment, most sediment types had become anoxic below 2 cm depth—also consistent with the general increase of ammonium concentrations with depth in nearly all sediment types, and the parallel depletion of nitrates. In our experiment, NH4 + dominated the exchange of nitrogen between the sediment and the water column, and was the main form of nitrogen in sediment pore waters, which has also been observed in other similar studies [10, 25]. The increased consumption of nitrates by the sediments, as well as their depletion below 3 cm depth in all sediment types except the unvegetated seagrass patch, could indicate that denitrification is occurring at these depths [10], although [26] observed decreased denitrification rates under high organic loading. The greatest differences in nutrient fluxes induced by the organic loading were observed in the fine sands, although all sediment types exhibited them. There were also differences in solute fluxes between the start and the end of the experiment in the controls. 3 The PO− 4 porewater concentrations and fluxes in all sediment types and conditions during our experiment were much smaller than those observed in Varna Lake during the summer, where significant stratification and bottom hypoxia induce a P release from the sediments [15]. No hypoxic conditions were observed in the water layer overlying the sediments during our experiment, but by the end most sediments had 3 become reduced at depths below 2 cm, and a release of PO− 4 to the water column

Porewater Nutrient and Oxygen Profiles and Sediment-Water …

379

was present. Higher nitrate concentrations in the overlying water have been linked to slower phosphate release from reduced sediments [9], which could explain the smaller P fluxes observed. The effects observed in our experiment could in part be due to the characteristics of the organic matter used for the loading. The amount and quality of organic matter reaching the sediments are known to influence the spatial variability and seasonal differences in sediment nutrient fluxes [27]. The algal organic matter used in this experiment is relatively refractory; an eutrophication-induced bloom is more likely to result in accumulation of phytoplankton-derived organic matter, which is more easily decomposed. The seagrass sediments are also naturally enriched in organic matter from the decomposition of seagrass biomass and epiphytes. This study made apparent the high variability typical for coastal sediments, even at very small scales, which makes the interpretation of the results more difficult. The fluxes at the start of the experiment, before any treatment was applied, sometimes differed greatly between replicates of the same sediment type (e.g. Zostera). More replication should be considered in future studies to account for that. The two-point study design also could be improved by including more intermediary measurements, which would allow to trace the immediate effects of the organic loading on the different sedimentary habitats. Additionally, benthic faunal activities and behaviour may strongly influence nutrient cycling and fluxes by redistributing organic material, modifying sediment redox conditions, and creating chemical gradients and interfaces for solute exchange [28]. There are no studies on the effects of bioturbation on solute concentrations and fluxes in the Bulgarian Black Sea, and extrapolation from other seas is unlikely to be meaningful, since these processes are highly context-dependent even on a very local scale, and could be influenced by even small variations in environmental conditions [29]. Although this study did not take into account the macrofaunal communities in the different habitats, other studies have found that bioturbation and bioirrigation can alter near-surface porewater nutrient concentrations toward bottom water values [30]. Indeed, the ammonium and phosphate porewater profiles in our experiment are very similar to those in the sandy stations with high bioturbation potential observed by Gogina et al. [30], suggesting a possible mixing and transport effect. Finally, our experiment only considered diffusive fluxes across the sedimentwater interface. In natural conditions, and especially for permeable sandy sediments, advection (e.g. through sediment resuspension) can also be an important mechanism, sometimes exceeding diffusive fluxes, and should therefore also be taken into account [6, 31]. Our study area is probably advection-dominated—shallow, characterized by coarser sandy sediments, with potentially high physical exchange between bottom and interstitial water—and the solute fluxes are likely driven primarily by hydrological processes, with diffusion and bioturbation/biodeposition playing a secondary role [32]. Dense benthic macrophyte cover such as seagrass beds could also alter the porewater nutrient distribution and fluxes by capturing nutrients from the water column or from the sediments, and by increasing oxygen penetration in the sediments through their root systems [33]. Our results support this conclusion—oxygen penetration

380

S. Klayn et al.

was highest in the seagrass sediments, there was generally little influence of organic loading on the nutrient profiles, and the sediment-water interface fluxes were more stable than in the other sediment types. This suggests that seagrass beds likely play a significant role in coastal water nutrient balance and water quality regulation.

5 Conclusions Nutrient cycling in coastal marine sediments represents important ecosystem functions such as nutrient regeneration for primary producers and regulation of the ability to remove excess (natural or anthropogenic) nitrogen and phosphorus, which confer resilience to coastal zones through mitigating eutrophication. Although limited in scope, this study adds to the current knowledge of the influence of sediment type on the nutrient cycling and balance in the coastal Bulgarian Black Sea, and contributes towards our understanding of the role of coastal benthic habitats in water quality regulation. Future experiments will attempt to quantify the in situ fluxes in the main habitat types in Sozopol Bay, and compare them with previous in situ measurements from an impacted period before the construction of a wastewater treatment plant in the area. The experiments will also take into account the resident macrofaunal communities and their activities. This will allow a more realistic characterization of the sediment-water interface fluxes, and the effects of the interactions of multiple processes in coastal Black Sea habitats. Such results are especially valuable given the high context dependency and natural variability of these processes, and the consequent difficulty of extrapolating between regions. Acknowledgements This study was financed by the Bulgarian National Scientific Program “Environmental Protection and Reduction of the Risk of Adverse Events and Natural Disasters” Approved by Council of Ministers Decision No 577/17.08.2018 and funded by the Ministry of Education and Science (Agreement No D01-230/06-12-2018), WP1.4.

References 1. Lehmköster, J., Schröder, T., der Maribus Zukunft, E.O.: Coasts—A Vital Habitat Under Pressure. Maribus, Hamburg (2017) 2. Marchant, H.K., Lavik, G., Holtappels, M., Kuypers, M.M.M.: The fate of nitrate in intertidal permeable sediments. PLoS ONE 9, (2014). https://doi.org/10.1371/journal.pone.0104517.X 3. Paerl, H.W.: Cultural eutrophication of shallow coastal waters: coupling changing anthropogenic nutrient inputs to regional management approaches. Limnologica 29, 249–254 (1999). https://doi.org/10.1016/S0075-9511(99)80009-7.X 4. Kelly, J.R.: Nitrogen effects on coastal marine ecosystems. In: Hatfield, J.L., Follett, R.F. (eds.) Nitrogen in the Environment. pp. 271–332. Academic Press, San Diego (2008). https://doi.org/ 10.1016/B978-0-12-374347-3.00010-X.X

Porewater Nutrient and Oxygen Profiles and Sediment-Water …

381

5. Ekeroth, N., Blomqvist, S., Hall, P.O.J.: Nutrient fluxes from reduced Baltic Sea sediment: effects of oxygenation and macrobenthos. Mar. Ecol. Prog. Ser. 544, 77–92 (2016). https://doi. org/10.3354/meps11592.X 6. Pearce, A.R., Chambers, L.G., Hasenmueller, E.A.: Characterizing nutrient distributions and fluxes in a eutrophic reservoir, Midwestern United States. Sci. Total Environ. 581–582, 589–600 (2017). https://doi.org/10.1016/j.scitotenv.2016.12.168.X 7. Pitkänen, H., Lehtoranta, J., Räike, A.: Internal nutrient fluxes counteract decreases in external load: the case of the estuarial Eastern Gulf of Finland, Baltic Sea. AMBIO: J. Hum. Environ. 30, 195–201 (2001). https://doi.org/10.1579/0044-7447-30.4.195.X 8. Boers, P., de Bles, F.: Ion concentrations in interstitial water as indicators for phosphorus release processes and reactions. Water Res. 25, 591–598 (1991). https://doi.org/10.1016/00431354(91)90131-9.X 9. Gao, Z., Zheng, X.-L., Li, W., Song, H.: Determination of nutrient fluxes across the sedimentwater interface in a nitrate-rich reservoir. In: 2nd International Conference on Bioinformatics and Biomedical Engineering, pp. 3319–3322. IEEE, Shanghai, China (2008). https://doi.org/ 10.1109/ICBBE.2008.1159.X 10. Zhang, L., Wang, L., Yin, K., Lü, Y., Zhang, D., Yang, Y., Huang, X.: Pore water nutrient characteristics and the fluxes across the sediment in the Pearl River estuary and adjacent waters, China. Estuar. Coast. Shelf Sci. 133, 182–192 (2013). https://doi.org/10.1016/j.ecss.2013.08. 028.X 11. DiDonato, G.T., Lores, E.M., Murrell, M.C., Smith, L.M., Caffrey, J.M.: Benthic nutrient flux in a small estuary in northwestern Florida (USA). GCR 18 (2006). https://doi.org/10.18785/ gcr.1801.02.X 12. Huettel, M., Røy, H., Precht, E., Ehrenhauss, S.: Hydrodynamical impact on biogeochemical processes in aquatic sediments. Hydrobiologia 494, 231–236 (2003) 13. Santos, I.R., Eyre, B.D., Huettel, M.: The driving forces of porewater and groundwater flow in permeable coastal sediments: a review. Estuar. Coast. Shelf Sci. 98, 1–15 (2012). https://doi. org/10.1016/j.ecss.2011.10.024.X 14. Doncheva, V., Shtereva, G.: Preliminary studies of chemical composition of pore waters from sediments along the Bulgarian Black Sea shelf. In: Presented at the Fifth International Conference on Marine Sciences and Technologies—BLACKSEA2000 (2000) 15. Doncheva, V.: Nutrients in pore water from surface sediment layer along the eutrophication gradient (Varna Lake—Varna Bay case study). C. R. l’Académie Bulgare Sci. 63, 547–554 (2010) 16. Hiebaum, G.: Transformation of carbon in the phytoplankton and bacterioplankton communities of the Burgas bay (in Bulgarian) (1990) 17. Hiebaum, G., Karamfilov, V.: Regime shifts in the annual dynamics of primary production and chlorophyll-a concentrations in the coastal zone of the Bourgas Bay (Western Black Sea). In: Large-Scale Disturbances (Regime Shifts) and Recovery in Aquatic Ecosystems: Challenges for Management Towards Sustainability, pp. 143–158. Bulgarian Academy of Sciences, Varna, Bulgaria (2005) 18. Berov, D., Deyanova, D., Georgieva, I., Gyosheva, B., Hiebaum, G.: Cystoseira sp.-Dominated macroalgal communities in the SW Black Sea (Burgas Bay, Bulgaria). Current state and possible long-term effects of eutrophication. C. R. l’Académie Bulgare Sci. 65, 821–830 (2012) 19. Miladinova, S., Marinov, D., Krastev, V., Marinski, J.: Multi-compartment water quality assessment of Port Burgas and Burgas Bay. In: Stylios, C., Floqi, T., Marinski, J., Damiani, L. (eds.) Sustainable Development of Sea-Corridors and Coastal Waters, pp. 95–102. Springer International Publishing, Cham (2015) 20. Folk, R.L., Ward, W.C.: Brazos River bar: a study in the significance of grain size parameters. J. Sediment. Petrol. 27, 3–26 (1957) 21. Grasshoff, K. (ed.): Methods of seawater analysis. Verlag Chemie, Weinheim and New York (1976) 22. Klayn, S.: Macrozoobenthic communities as an indicator of the ecological state of benthic habitats along the Bulgarian Black Sea coast (Burgas Bay) (2019)

382

S. Klayn et al.

23. Boudreau, B.P.: Diagenetic Models and Their Implementation: Modelling Transport and Reactions in Aquatic Sediments. Springer, Berlin Heidelberg (1997) 24. Kelly, J., Berounsky, V., Nixon, S., Oviatt, C.: Benthic-pelagic coupling and nutrient cycling across an experimental eutrophication gradient. Mar. Ecol. Prog. Ser. 26, 207–219 (1985). https://doi.org/10.3354/meps026207.X 25. Chowdhury, M., Bakri, D.A.: Diffusive nutrient flux at the sediment-water interface in Suma Park Reservoir, Australia. Hydrol. Sci. J. 51, 144–156 (2006). https://doi.org/10.1623/hysj.51. 1.144.X 26. Sloth, N.P., Blackburn, H., Hansen, L.S., Risgaard-Petersen, N., Lomstein, B.A.: Nitrogen cycling in sediments with different organic loading. Mar. Ecol. Prog. Ser. 116, 163–170 (1995) 27. Cowan, J.L.W., Boynton, W.R.: Sediment-water oxygen and nutrient exchanges along the longitudinal axis of Chesapeake Bay: seasonal patterns, controlling factors and ecological significance. Estuaries 19, 562–580 (1996). https://doi.org/10.2307/1352518.X 28. Renz, J.R., Powilleit, M., Gogina, M., Zettler, M.L., Morys, C., Forster, S.: Community bioirrigation potential (BIPc), an index to quantify the potential for solute exchange at the sedimentwater interface. Mar. Environ. Res. 141, 214–224 (2018). https://doi.org/10.1016/j.marenvres. 2018.09.013.X 29. Gammal, J., Norkko, J., Pilditch, C.A., Norkko, A.: Coastal hypoxia and the importance of benthic macrofauna communities for ecosystem functioning. Estuaries Coasts 40, 457–468 (2017). https://doi.org/10.1007/s12237-016-0152-7.X 30. Gogina, M., Lipka, M., Woelfel, J., Liu, B., Morys, C., Böttcher, M.E., Zettler, M.L.: In search of a field-based relationship between benthic macrofauna and biogeochemistry in a modern brackish coastal sea. Front. Mar. Sci. 5, 489 (2018). https://doi.org/10.3389/fmars.2018.004 89.X 31. Ospina-Alvarez, N., Caetano, M., Vale, C., Santos-Echeandía, J., Bernárdez, P., Prego, R.: Exchange of nutrients across the sediment–water interface in intertidal ria systems (SW Europe). J. Sea Res. 85, 349–358 (2014). https://doi.org/10.1016/j.seares.2013.07.002.X 32. Mermillod-Blondin, F.: The functional significance of bioturbation and biodeposition on biogeochemical processes at the water–sediment interface in freshwater and marine ecosystems. J. North Am. Benthological Soc. 30, 770–778 (2011). https://doi.org/10.1899/10121.1.X 33. Glud, R.N., Berg, P., Stahl, H., Hume, A., Larsen, M., Eyre, B.D., Cook, P.L.M.: Benthic carbon mineralization and nutrient turnover in a Scottish sea loch: an integrative in situ study. Aquat. Geochem. 22, 443–467 (2016). https://doi.org/10.1007/s10498-016-9300-8.X

Remote Sensing and Modelling of the Mopang Oil Pollution Near the Bulgarian Black Sea Coast Irina Gancheva

and Elisaveta Peneva

Abstract This work investigates the extend of oil pollution released by the sunken cargo ship Mopang, located in the Bourgas bay on the Western Black Sea shelf. We have analysed the available Synthetic Aperture Radar (SAR) data from the Sentinel-1 mission for the years 2017 and 2018 and identified the surface features which could be referred to oil pollution, originating from Mopang for the given timeframe. To detect the oil leaks an adaptive threshold algorithm is used and the detections are visualized cumulatively in order to estimate the continuity and intensity of the leak throughout the period. The radar acquisitions from both the ascending and descending pass of the Sentinel-1A and B satellites with oil detections visible for three dates are plotted together with the surface currents in attempt to study the evolution of the leak and its dependence on the marine conditions. The possibility to simulate the dispersion of oil pollution on the surface with a Lagrangian particle model is tested for one of the dates. Three seeding scenarios are run: (1) release from a shape, such as the one of the morning detection; (2) release from a point source; and (3) continuous release from a point source for the entire simulation period. The numerical simulations are performed with the OpenDrift trajectory model and the results after 12-h run are validated against the satellite images. Keywords Oil pollution in Black Sea · SAR · Lagrangian particle modelling

1 Introduction Oil pollution is a serious threat for the marine ecosystems: it might enter the water system through anthropogenic activities or through natural sources such as crude oil sleeks from the sea bottom. The ecosystems in semi-enclosed basins like the Black Sea are especially vulnerable to oil pollution as the water exchange in such seas is limited. I. Gancheva (B) · E. Peneva Faculty of Physics, Sofia University “St. Kliment Ohridski”, 5 James Bourchier blvd., 1164 Sofia, Bulgaria e-mail: [email protected] © The Author(s), under exclusive license to Springer Nature Switzerland AG 2021 N. Dobrinkova and G. Gadzhev (eds.), Environmental Protection and Disaster Risks, Studies in Systems, Decision and Control 361, https://doi.org/10.1007/978-3-030-70190-1_26

383

384

I. Gancheva and E. Peneva

Fig. 1 Position of the sunken ship in Bourgas bay, marked with a star

Over 80% of the Bulgarian shoreline is protected area by Natura 2000 as it provides a nesting area for various protected bird species, which makes it particularly vulnerable to chemical contamination [1]. However, oil pollution is common in the Bulgarian waters due to accidental spills or routine waste discharges from cargo ships. In this work we study another, rather uncommon source of oil pollution for the region—a sunken ship with significant amount of engine fuel. In the beginning of August 2018 the sunken ship Mopang began releasing some of its engine fuel, which attracted serious public attention and concerns, as the ship remains are located in direct vicinity of the island St. Ivan, about 10 km away from the busy tourist town Sozopol, Fig. 1. The accident was reported by citizens to the Maritime Safety Agency in Bourgas which then initiated cleaning up the tanks and pumping out the remaining fuel. The activity was compleated in the summer of 2019. Mopang was an US build cargo steamship of the Liberty ships type, build in 1920 and sunk in June 1921 after hitting a mine. The sea depth in the region where the shipwrecks are located is between 20 and 33 m [2]. The capacity of the engine tanks is to hold about 600 tons of fuel and it was estimated to have ~100 to 150 tons inside [3]. The potential of the Synthetic Aperture Radar (SAR) as an information source for the automatic detection of oil pollution released in the sea has been acknowledged and investigated years ago and nowadays this method is widely used for the operational monitoring [4–6]. On radar images oil appears as dark formation, as it smoothens the natural sea roughness and thus decreases the backscattering properties of the sea surface. Other natural phenomena, such as local surface currents, eddies or algal blooms, have the same appearance on radar data and might cause false detections in case of an automatization of the oil detection process. Simulation of the dispersion of oil pollution on the sea surface is important practice to support the sea operations to limit the leak, especially in the case of more severe pollution events. The model calculations consider the local meteorological and marine conditions and could define the affected areas, thus optimizing the cleanup activities.

Remote Sensing and Modelling of the Mopang Oil Pollution Near …

385

The use of Lagrangian particle tracer models is a reliable method to simulate the movement of pollution particles in fluids. In the current study we use the module for oil simulations of the OpenDrift model [7] for tracing the movement of oil particles released from Mopang. The satellite data used in this study is from the Sentinel constellation, which is part of the Copernicus program for Earth observation. The program is managed by the European commission, together with the European Space Agency and all data it delivers is distributed free to the scientific and business community [8]. We have analyzed Sentinel-1 radar data and use an adaptive threshold algorithm for detection of surface features which can be related to oil pollution on the sea surface in direct vicinity of the sunken ship Mopang. All available acquisitions for the years 2017 and 2018 were reviewed in order to understand the extend of the pollution and its intensity. Radar data reveals slicks on most of the summer time images with varying intensity and distribution. Additionally, the images from the ascending and descending passes of Sentinel1A and B respectively at 4 am and pm on the same day give possibility to visualize the oil dispersion and relate it with the surface currents at acquisition time in order to study the evolution of the pollution dispersion. A simulation test with the Lagrangian particle model OpenDrift is performed for one of these dates. We have considered three seeding scenarios as an initial condition of the oil particles position. The satellite image after 12 h gives possibility to validate the model simulation drift in order to evaluate which seeding scenario is probable and to verify the credibility of model calculations.

2 Methods The satellite data used in this study comes from the Sentinel-1 mission and is processed and visualized with the software SNAP (Sentinel Application Platform). The automatic detection of oil leaks from radar imagery is done using an adaptive threshold algorithm [9] and the detected areas are cumulatively plotted according to the month of acquisition (Fig. 2). For images, acquired on the same day from ascending and descending pass, the detected oil areas are plotted separately on Fig. 3 together with the surface currents of the acquisition time in order to track the development of the slick. The ocean currents data is acquired from the Copernicus Marine Environment Monitoring Service—CMEMS Black Sea physics reanalysis product [10]. The adaptive threshold algorithm calculates the mean backscatter value for pixels in a large window and applies a user defined detection threshold value for estimating each pixel of the area [5]. If the backscatter value of the pixel is below the threshold, it is detected as dark spot. If it is above the threshold, it is estimated as background pixel. Moving the window of averaging and estimating the backscatter for each pixel gives information if a given spot has lower reflectance, which can be caused by oil spreading on the ocean surface, damping the capillary waves. Clustering of the dark

386

I. Gancheva and E. Peneva

July

September

August

October

Fig. 2 Cumulative plot of all detections for 2017 and 2018 for the summer months from July until October. All oil leak detections from 2017 are plotted with red and those from 2018 in black. The coordinates of the ship wrecks are denoted with a dot

spots and defining a minimum cluster size, provides the user with a mask, showing all detected dark spots having larger dimension compared to the defined minimum [5]. The main drawback of the adaptive threshold is the notable amount of false detection of look-alike objects, caused by other phenomena. These could be minimized by further investigation of the detected areas. A classification according to shape, size and orientation can significantly reduce the amount of false alarms, especially for the areas in direct vicinity of the shoreline [5, 11]. The capability of the SNAP ocean tool for oil spill detection for the Black Sea region has been demonstrated in previous studies and is estimated as good [12]. Some of its major drawbacks is detection in areas close to the shoreline, which underlines the importance of additional analysis and classification of the results in this study. Due to this the initial parameters for detection were fine-tuned and additional inspection of the surface reflectance and texture of the image in the area where the ship is positioned was done for each acquisition. This procedure gave confidence to clearly distinguish between oil trace and look-alike objects detected by the algorithm. Next step in this study is to investigate the propagation of the oil pollution, released from the sunken ship Mopang by simulating the movement of oil droplets with a Lagrangian particle model. It enables a simple tracking of tracers by which we can determine the position of oil particles at selected time.

Remote Sensing and Modelling of the Mopang Oil Pollution Near …

387

7 July 2018, 4 am (left) and 4 pm (right)

6 August 2018, 4 am (left) and 4 pm (right)

29 September 2018, 4 am (left) and 4 pm (right)

Fig. 3 Oil detections, plotted in navy blue for three different days—7 July, 6 August and 29 September 2019. Acquisition, taken at 4 am is on the left and from 4 pm is on the right. The speed of the surface currents is color-coded, units—[cm/s] and the arrows indicate the direction. The position of the ship is marked with a star

The simulations are conducted with the open source model OpenDrift which is developed at the Norwegian Meteorological Institute and is Python-based [13]. OpenDrift is a generic framework for tracing objects, drifting in the ocean and more specific functionalities, depending on the purposes are implemented in the available trajectory modules [7]. For our purposes we use the module OpenOil which is developed specifically for simulating the fate of oil in marine environment and includes a number of features aiming description of the processes with maximum accuracy. OpenOil requires input for the meteorological and marine conditions which are necessary as forcing data for the correct calculation of advection and transformation of oil particles. For the simulations done in this study we use hourly values for the ocean surface current velocity from the Copernicus Marine Services [10] and surface wind speed at 10 m height from the European Center for Medium-Range Weather Forecast (ECMWF) atmospheric reanalysis—ERA5 [14]. In OpenOil the horizontal drift of oil droplets is parametrized with the assumptions that particles drift with the ocean current and are subject of Stokes drift, depending on their depth. Oil elements on the sea surface get an additional moving factor from

388

I. Gancheva and E. Peneva

the wind stress and Stokes drift. The horizontal movement is scaled by taking into consideration the vertical transport, which is implemented in the calculations by different factors. One of the major agents for the vertical mixing is the presence of strong wind and waves, which leads to intensive mixing of oil and water and entrainment and sinking of oil droplets in the water column [15]. The oil properties, such as its density, viscosity and the size of the oil droplets together with the ocean stratification profile, determine the buoyance of oil particles. The model has implemented a broad library of different oil types, which includes specific information about their densities and viscosities, which is then considered for the weathering and emulsification of oil in marine environment. For the particular example of oil seeps, originating from ship wrecks, which is the subject of our study, after experimenting with different oil types we have selected a generic heavy fuel oil for the simulations, as it has similar properties to the mazut fuel, which was widely used as ship fuel at that time and which we assume was in the ship tanks.

3 Results and Discussion 3.1 Mopang Oil Detections from 2017 to 2018 In order to estimate the outreach of the oil pollution originating from the ship wrecks of Mopang all available acquisitions from Sentinel-1 for the years 2017 and 2018, which include the region of interest were processed. Both time periods reveal similar results and conclusions. Figure 2 is a cumulative plot, showing the detections for both years only for the summer months July, August, September and October. The purpose is to check if the leaks differ in number and intensity during the suggested period of Mopang oil release. The detections done in 2017 are pictured in red and from 2018 in black. In September 2017 there were no significant detections related to Mopang. Acquisitions from the rest of the year were processed as well. Occasional surface features, which could be referred to Mopang were visible in few days spread throughout the year without significant accumulation related to particular month or season. The small colored areas which are not clustered with the main detection slick (see the position of the ship with dot) are caused by false detections due to the small scale of the event and inaccuracy of the detection algorithm. The majority of the detections for both 2017 and 2018 are during the summer months. This observation can be explained with the calmer sea conditions being favourable for the oil to remain on the sea surface and not mix with the ambient water after emerging from the ship tanks. During 2018 there are more oil detections compared to 2017, as visible in Fig. 2. July and August 2018 show similarly intensive oil releases, which could be related to the Mopang leaks. In the Bulgarian media it was stated, that part of the ship collapsed due to its age and poor state in the beginning

Remote Sensing and Modelling of the Mopang Oil Pollution Near …

389

of August 2018, leading to the oil leakages, but our investigation disproves it as the leak was evident during most part of the summer during the previous year as well [16]. Due to the small scale of the event there are number of false detections deteriorating the clear assignment of the oil detections on the cumulative plots. They remain visible despite the fine-tuning of the variables which can be changed in the detection algorithm. This illustrates the challenges in front of the complete automatization of the process in the case of small scale events close to the shoreline.

3.2 Propagation of Oil with Surface Currents Every six days the Sentinel-1A and B pass over the area of interest, acquiring images at 4 am in descending and 4 pm in ascending orbit. In favorable meteorological conditions, such as low sea waves and moderate winds, the oil slick originating from the sunken ship is visible on both acquisitions, which provides a unique opportunity to study the evolution of the surface dispersion over time depending on the surface currents, wind speed and direction. Three selected cases are presented in Fig. 3: the position of the oil slick is plotted on the map together with the surface current speed at the moment of acquisition. The vectors indicate the direction of the current and their length—the speed. The grid resolution of the reanalysis data for the surface currents is ca. 3 km (1/36° in meridional and 1/27° in zonal direction), thus the current speed closer to the shoreline cannot be determined. Comparing the left and right column images, one can conclude that the current direction affects the shape of the oil slick after 12 h: on 6th of August 2018 the slick is moved to the west due to the north-eastern current; similarly on 27 of September 2018 the slick is displaced from the southeastern current. An interesting case is the 7th of July—morning-to-afternoon the current changed to the opposite direction (from east to west), and the slick on second position is most likely new released Mopang oil. The plots in Fig. 3 reveal that the position of the oil slick is in the low current speed area and it disperses mainly along when the speed is higher than 10 cm/s or across the stream lines in the case of weak current. It is noticeable that the oil leaks are of similar size and form throughout the same day, however they are usually visible as discontinuous patches. This, together with the moderate speed of the surface currents, which varies between 5 and 25 cm/s for the areas of oil propagation, suggests that the oil detections in the afternoon are not of the same oil which was visible in the morning. This statement is also in tact with the presence of more intensive breeze circulation in the afternoon hours with typical wind speeds between 2 and 4 m/s, which would lead to entrainment of oil particles in the water column. The assumption that during the summer months a continuous release of oil from the ship tanks was visible is also intact with the results of the simulation, ran for 7 July 2018 with the trajectory model OpenOil presented in Sect. 3.3.

390

I. Gancheva and E. Peneva

3.3 Simulation of the Propagation of Oil Leaks with OpenOil As test case to simulate the propagation of oil leaks with the model OpenOil the date 7 July 2018 at 4 am to 4 pm UTC was chosen (Fig. 3). The reason is that both detections are of particular and different form and can clearly be referred as originating from the ship wrecks. The aim is to simulate the distribution of oil droplets with starting time 4 am, when the first Sentinel-1 acquisition is taken and compare the results with the real detection from the second acquisition at 4 pm. The good overlapping of the form and position of tracers from model calculations with the actual detection from the satellite image would prove the credibility of the Lagrangian particle trajectory method for predicting the spread of oil in marine environment for the particular case we observe. The speed and direction of propagation of the surface currents at 4 am (morning acquisition time), 8 am, 12 pm and 4 pm (afternoon acquisition time) are shown in Fig. 4. These are the time steps in which we track the position of the oil elements after starting the simulation. The plotted area covers the entire Bourgas bay. Figure 5 displays the wind barbs in 2 h steps again for the timeframe of the simulation—12 h at the coordinates where the ship wrecks are located (the source of 4 am

8 am

12 pm

4 pm

Fig. 4 Surface current speed (in [cm/s]) and direction for the day, when model calculations are done—7 July 2018 from 4 am until 4 pm in 4 h’ time steps. The data is from the CMEMS—Black Sea physics reanalysis product. The position of Mopang is located with a purple star

Remote Sensing and Modelling of the Mopang Oil Pollution Near …

391

Wind speed on 7 July 2018 in 2 hours’ time step

Fig. 5 Wind barbs indicating wind speed in [m/s] and direction for the coordinates, where the ship wrecks are located at the day of model calculations—7 July 2018. Start time is 4 am, end time 4 pm in 2 h’ time steps. The data is from ECMWF climate reanalysis ERA5

the data is ERA5). In the beginning of the period a calm weather is observed followed by weak eastern wind. The model performance depends on the initial conditions for the oil slick tracers position. We did three simulations with different initial conditions, reflecting the oil release treatment. Each simulation ran for 12 h with 12 time-steps and 300 tracers. The number 300 tracers is chosen as optimal for the size of the oil slick and the horizontal resolution of the marine data. Their distribution during the oil dispersion is indicative for accumulation of oil droplets in certain areas and their general movement pattern. All simulation results are being illustrated in 4-h time steps from the time of the morning overpass of Sentinel-1, until the afternoon overpass together with the oil detection in the afternoon in order to verify the reliability of the calculations. The forcing data for meteorological conditions is the same for all experiments. The major drawback for the process is that the resolution of the surface current speed data doesn’t reach the shoreline, but ends approx. 3 km in water, as it is visible on the surface current plot on Fig. 4. OpenOil extrapolates the values of the current direction and speed, which however doesn’t take into account the local circulation specifications and unpredictable changes in the direction of the current flow. This limitation should be considered at the final evaluation of the environmental impact of the oil leak on the coastal ecosystem. During the simulation experiments different values for all variables, needed for the description of the behavior of oil in marine environment, were tested. The oil entrainment rate is implemented in the model as calculated by Li et al. [17]. The diffusion rate of oil in the marine environment can be regulated with the two drift parameters for current and wind uncertainties. Different values for both of them were tested, resulting in broad dispersion of the tracers at their maximum and drifting solely with the speed of the forcing data at their minimum. A middle range value for the parameters was chosen, which according to our observations most realistically describes the behavior of the tracers considering the meteorological conditions at the time of simulation.

392

I. Gancheva and E. Peneva

The processes of evaporation, dispersion and emulsification are taken into consideration and their impact on the final results depend mostly on the type of oil, which is chosen. Tests with different oil types have shown little change in the evaporation and entrainment rate, since the simulation is run for 12 h, which is rather short period of time. The here presented simulations are done with a generic type of heavy fuel oil. Main limitation for the exact description of oil type for the simulations is that the ship wrecks were under water for almost 100 years, which has caused significant impact on the chemical composition of the oil product. Nevertheless after finishing the simulations we are positive that the chosen variables and oil type accurately describe the drifting behavior of oil and reveal the tendency of the dispersion mechanisms. Three runs of the model are performed with different initial shape of the oil slick: (1) the tracers follow the shape of the oil-slick from the satellite image; (2) the oil is released at once as a point source; (3) the oil is continuously released by a point source at the ship location. The results from the three runs are summarized in Figs. 6, 7 and 8. For the first simulation all 300 tracers are distributed within the boundaries of the form of the detected oil leak at 4 am UTC on 7 July 2018 and are released for drifting with forcing data for the same time. The results of the simulation after 4, 8 and 12 h from the beginning are shown on Fig. 6 together with the detected oil leak at 4 pm on the same day. The time steps reveal that the exact shape of the oil leak is lost shortly after release due to diffusion and entrainment, however the tracers keep the elongated form, similar to the initial. In the end of the simulation the tracers are located in the small bay between Sozopol and Chernomoretz, identifying that a longer drift would spread them along the entire coastline of the region and indeed there are reports from tourists for notable oil amount on the near beaches. After 12 h running time 54 elements of the initial 300 are stranded and the active tracers are more widely dispersed. The second simulation is done with a point source with small diameter, which released all 300 tracers at once at 4 am UTC on 7 July 2018. The results are displayed on Fig. 7 in the same manner at three time steps, 4, 8 and 12 h after beginning of the integrations together with the satellite acquisition at 4 pm, so that we are able to estimate the credibility of the simulations. The plots are similar to the ones in Fig. 6, but the cloud of tracers has more circular form. After 12 h simulation time none of the tracers was stranded and they haven’t reached the coast in Chernomoretz village area and the Sozopol beaches. This type of source shows significantly less dispersion of the tracers. The third simulation is done as well with a point source with small diameter, which this time starts releasing the tracers at the same time as the previous simulations at 4 am, and continuously releases the entire amount of 300 tracers throughout the time of the simulation, which is 12 h. Figure 8 illustrates the propagation of the tracers during the time of integration. The cloud of tracers moves in the form of a plume originating from the Mopang location. It moves at slower rate than in the other two cases and does not reach the

Remote Sensing and Modelling of the Mopang Oil Pollution Near …

Simulation run – 4 hrs

393

Simulation run – 8 hrs

Simulation run – 12 hrs

Fig. 6 Results of the simulation of oil dispersion, seeded from a source with the shape of the oil leak detection at 4 am on 7 July 2018 are plotted in green. The initial detection is plotted in navy blue. Simulation results are presented after 4, 8 and 12 h propagation after start. The oil detected from the acquisition at 4 pm is plotted with dark red. The exact coordinates of the ship are denoted with a yellow star

coast within 12 h integration. However, the probable coast landing is Chernomoretz area. As expected, the longer the tracers were in the sea water, the further they are from the source and the more dispersed they are between each other. The tracers close to the coordinates of the sunken ship keep a propagation in a more straight line as those released earlier. The propagation of the release is further to the West, compared to the previous which kept more South to South-West direction. For the running time of this simulation no tracers were stranded.

394

I. Gancheva and E. Peneva

Simulation run – 4 hrs

Simulation run – 8 hrs

Simulation run – 12 hrs

Fig. 7 Results of the simulation of oil dispersion, seeded from a point source which released all 300 tracers simultaneously at 4 am on 7 July 2018 are plotted with green. Simulation results are presented after 4, 8 and 12 h propagation after start. The oil detected from the acquisition at 4 pm is plotted in dark red. The exact coordinates of the ship are denoted with a yellow star

Finally, a comparison of the position of the tracers in the end of the 12 h simulation time with the actual oil detection from the afternoon Sentinel-1 overpass should serve as orientation for the credibility of the model calculations. The simulation done with tracers seeded in the same shape as in the morning acquisition (Fig. 6) reveals stronger diffusivity among ensemble members and a generally similar prolonged form, which however is shifted away from the coordinates of the source and has drifted towards the coastline. This behaviour is expected as the simulation assumes that the oil leak is separated from the source right after release and no more oil is further leaking. Thus the seeded elements disperse in the environment.

Remote Sensing and Modelling of the Mopang Oil Pollution Near …

Simulation run – 4 hrs

395

Simulation run – 8 hrs

Simulation run – 12 hrs

Fig. 8 Results of the simulation of oil dispersion, seeded from a point source which releases its tracers continuously from 4 am on 7 July 2018 until 4 pm on the same day are plotted in green. Simulation results are presented after 4, 8 and 12 h propagation after start. The oil detected from the acquisition at 4 pm is plotted with dark red. The exact coordinates of the ship are denoted with a yellow star

The dispersion of tracers, released simultaneously from point source (Fig. 7) shows least similarities with the acquisition at 4 pm. Such dispersion scenario appears more plausible in the case of sudden ship incident, which releases its fuel. In this case the ensemble of released tracers keep a circular shape, moving towards the coastline. This behaviour doesn’t comply with the satellite observations. The final experiment scenario of continuous oil release (Fig. 8) has as a form, vaguely similar of the detected oil patch at 4 pm, even though the direction of propagation is different. Here the low resolution of the forcing data close to the shoreline

396

I. Gancheva and E. Peneva

should be taken into consideration and the possibility that the exact current circulation might differ from the extrapolated values used in OpenOil. The three figures Figs. 6, 7 and 8 suggest that the simulations fail to reproduce exactly the detected by satellite shape of the oil slick on 7th of July at 4 pm. In order to understand the reason we have performed sensitivity experiments for the wind speed and horizontal diffusivity coefficient. Our results (not shown here) indicate that the wind speed is very important and a good correlation is obtained when moderate south wind is taken as input in the model. Our conclusion is that more sensitivity experiments should be performed in order to fine tune the model and the input global atmospheric model data should be compared with the local meteorological conditions.

4 Conclusions In this study we demonstrate the capacity for reliable detection of small scale oil slicks near the shoreline from radar satellite data, using images from Sentinel-1 processed with an adaptive threshold algorithm. As a test case is used the particular case of the oil leakage from the sunken ship Mopang, which has been under water since 1921, located in direct vicinity of the Bulgarian coastline. For the purposes of our research we process the available Sentinel-1 data-sets from 2017 and 2018 and we demonstrate that oil spills, which could be identified as released from the ship tanks are visible as early as February 2017 and throughout the summer seasons of 2017 and 2018. The oil pollution visible during the summer period of 2017 is of the same order as that in 2018, which leads to the conclusion that the leak started before 2017. Both in July and August 2018 and August 2017 similarly intensive leaks are identified. The spatial extend of the detected oil patches varies between 0.17 and 2.4 km2 . The findings of our study suggest that oil leaks from the tanks of sunken ships might be a common problem and regular monitoring of such areas using radar data and knowing the exact position on the ship wrecks might be beneficial for preventing significant oil releases and thus additional pollution load on the marine ecosystem. The availability of SAR satellite images with detected oil slick in the morning and afternoon for three dates (7th July, 6th August and 29th September 2018) gives an unique opportunity to investigate the evolution of the spill in terms of shape and extension and to evaluate the surface current impact on dispersion. It is visible that in general the shape and extension don’t change significantly over the day, however they change their direction and orientation relative to the surface current. The slicks propagate along or across the surface current stream, away of the areas with large current velocity.

Remote Sensing and Modelling of the Mopang Oil Pollution Near …

397

In the final part of the study we simulate the dispersion of oil particles with the open source trajectory model OpenOil for one of the days when Sentine-1A and B overpassed our area of interest within the same day—7th of July 2018. The elements are seeded according to three different scenarios—embedded in a shape such as the oil detection at 4 am, seeded at once from a point source with small diameter and continuously seeded from a point source throughout the entire time of the simulation. All simulation experiments were done for 12 h running time as this is the time difference between the Sentinel overpass. The main limitation of the approach is the resolution of the forcing meteorological data, as the observed leak is very close to the shoreline and the surface current data ends approx. 3 km away from the coast. In this case OpenOil extrapolates the values, which process however cannot take into consideration local circulation phenomena. According to the findings of the simulation tests we conclude that the model fails to describe precisely the propagation of oil droplets as none of the simulation scenarios predicts correctly the exact form and propagation direction of the oil leak detected at 4 pm on 7th of July 2018. The third seeding experiment with a continuous oil leak results in a shape slightly similar to the expected detection, however it has different spatial orientation, as it follows the extrapolated surface currents in southwestern direction and the detected oil slick is elongated towards northwest from the Mopang location. This finding suggests that forcing the model with more precise meteorological data especially about the wind direction from local measuring stations could deliver more accurate predictions. Considering these model simulations and taking into account the forms and orientations of oil detections within the same day, it can be speculated that in the Mopang case we observe continuous engine fuel leak throughout the day and the oil patches observed on the afternoon acquisitions are not of the same oil, which was visible in the morning. This would suggest that the oil originating from Mopang which was released in the marine ecosystem is of a significant amount and is more intensive than visible on the occasional radar observations. Nevertheless the exact impact of the oil pollution on the coastal area cannot be estimated as the detection algorithm is not applicable at the shoreline and 2–3 km away from it. However radar detections from 19th July and 12th August 2017 and 1st July and 6th August 2018 spread through a large area in the Bourgas bay and propagate very close to the coast, increasing the probability of oil reaching recreation tourist areas and directly impacting the marine habitat in the coastal shelf zone. Acknowledgements The research presented in this study is done entirely with open access data. The satellite, meteorological and marine data are distributed by the Copernicus program, managed by the European Commission. This simulations are performed with the OpenDrift model, which is developed by the Norwegian Meteorological institute with contributions from the scientific community.

398

I. Gancheva and E. Peneva

References 1. National Emergency Plan for Protection of the Black Sea against Oil Spill Pollution (in Bulgarian)—Hacionalen avapien plan za bopba c nefteni pazlivi v Qepno mope (2011) 2. Wreck site, SS Mopang (in Bulgarian). https://www.wrecksite.eu/wreck.aspx?17360. Last accessed 2020/10/09 3. Xinhuanet news website. https://www.xinhuanet.com/english/2018-08/13/c_137387725.htm. Last accessed 9 Oct 2020 4. Brekke, C., Solberg, A.H.: Oil spill detection by satellite remote sensing. Remote Sens. Environ. 95(1), 1–3 (2005) 5. Topouzelis, K.N.: Oil spill detection by SAR images: dark formation detection, feature extraction and classification algorithms. Sensors 10, 6642–6659 (2008) 6. Fingas, M., Brown, C.: Review of oil spill remote sensing. Marine Pollut. Bull. 83(1), 9–23 (2014) 7. Dagestad, K.F., Röhrs, J., Breivik, Ø., Ådlandsvik, B.: OpenDrift v1. 0: a generic framework for trajectory modelling. Geosci. Model Dev. 11, 1405–1420 (2018). https://doi.org/10.5194/ gmd-11-1405-2018 8. Sentinel-1 technical information. https://sentinel.esa.int/web/sentinel/missions/sentinel-1 9. Solberg, A.H., Brekke, C., Husoy, P.O.: Oil spill detection in Radarsat and Envisat SAR images. IEEE Trans. Geosci. Remote Sens. 45(3), 746–755 (2007) 10. Product user manual for the Black Sea Physics Reanalysis Product. https://marine.copernicus. eu/documents/PUM/CMEMS-BS-PUM-007-004.pdf. Last accessed 9 Oct 2020 11. Capizzi, G., Sciuto, G.L., Wo´zniak, M., Damaševicius, R.: A clustering based system for automated oil spill detection by satellite remote sensing. In: International Conference on Artificial Intelligence and Soft Computing (pp. 613–623). Springer, Cham (2016) 12. Gancheva, I, Peneva, E.: Verification of the SNAP ocean-tool for oil spill detection for the Bulgarian Black sea region. In: AIP Conference Proceedings (vol. 2075, No. 1, p. 120009). AIP Publishing LLC (2019) 13. OpenDrift model github repository. https://opendrift.github.io/. Last accessed 6 Oct 2020 14. ECMWF atmospheric reanalysis of the global climate. https://cds.climate.copernicus.eu/cds app#!/dataset/reanalysis-era5-single-levels?tab=overview. Last accessed 9 Oct 2020 15. Röhrs, J., Dagestad, K.F., Asbjørnsen, H., Nordam, T., Skancke, J., Jones, C., Brekke, C.: The effect of vertical mixing on the horizontal drift of oil spills. Ocean Sci. 14, 1581–1601 (2018). https://doi.org/10.5194/os-14-1581-2018 16. News website 24 hours (in Bulgarian). https://www.24chasa.bg/novini/article/7023835. Last accessed 9 Oct 2020 17. Li, Z., Spaulding, M.L., French-McCay, D.: An algorithm for modeling entrainment and naturally and chemically dispersed oil droplet size distribution under surface breaking wave conditions. Mar. Pollut. Bull. 119(1), 145–152 (2017)