195 96 172MB
English Pages 418 [414] Year 2021
Foreword
Foreword
T
he Sustainable Development Goals (SDGs) are at the core of Transforming Our World: the 2030 Agenda for
Sustainable Development, adopted by the 193 Member States of the United Nations in September 2015. The SDGs represent a commitment to addressing the tri-dimensional social, economic, and environmental issues humankind faces in the most comprehensive way in history, with the aim of achieving sustainability. China is fully committed to the 2030 Agenda for Sustainable Development. As progress has been made across the board, there have been “early harvests” of multiple SDGs. The goal to end extreme poverty will be achieved within this year, for example. China is also fully engaged in international cooperation in the SDGs, sharing knowledge and experience with and offering assistance, as much as we can, to other developing countries. The important role science and technology can play, already widely recognized, was reaff irmed in the Global Sustainable Development Report 2019 as vital to sustainable transformations as well as global development and change. The Chinese Academy of Sciences (CAS), as a member of the global scientif ic community, has worked vigorously through
i
ii
Big Earth Data in Support of the Sustainable Development Goals (2020): The Belt and Road
the science-policy-society interface to provide solutions and support, domestically and internationally, to address the new demands and challenges that humanity must address before the lofty Sustainable Development Goals can be achieved. The SDGs are gigantic, complex, and diverse systems with dynamic interactions amongst themselves. Fundamental to their implementation are effective monitoring and evaluation, where many diff iculties remain. The voluntary and nonbinding Global Indicator Framework for the Sustainable Development Goals was adopted in 2017 by the United Nations but must be further ref ined. With only ten years between now and 2030, the prospect of achieving the SDGs is not bright, not to mention the outbreak of COVID-19, which has brought unprecedented challenges to this effort. The CAS Big Earth Data Science Engineering Program (CASEarth) has studied, since 2018, six SDGs: SDG 2 (Zero Hunger), SDG 6 (Clean Water and Sanitation), SDG 11 (Sustainable Cities and Communities), SDG 13 (Climate Action), SDG 14 (Life Below Water), and SDG 15 (Life on Land), especially focusing on indicators of each of those SDGs where data and methods can be improved. CASEarth issues annual reports, Big Earth Data in Support of the Sustainable Development Goals, and leads CAS’ efforts to support the implementation of the SDGs. Big Earth Data in Support of the Sustainable Development
Foreword
Goals (2020): The Belt and Road discusses data integration, indicator development, and sustainability evaluation concerning the six SDGs through 44 case studies. Each case presents data products, methods, and models and how they can support policymaking, through f ive parts—the background, data used in the case, methods, results and analysis, and outlook— showcasing the value and prospects of applying Big Earth Dataenabled technologies and methods to the evaluation of the SDGs. In the Sustainable Development Goals Report 2019, United Nations Secretary-General António Guterres called for deeper, faster, and more ambitious responses to achieve the social and economic transformations required for the implementation of the SDGs. The report laid special emphasis on better use of data, a digital transformation while harnessing science, technology, and innovation, and more intelligent solutions. Technologies, especially data, will therefore have to play a more important role in achieving the SDGs. The United Nations “Technology Facilitation Mechanism” (TFM) is fully consistent with China’s strategy of driving development with innovation. China and many other developing countries face major challenges and pressure as we pursue the SDGs with limited capacity for data collection and processing, and for monitoring and evaluating SDG indicators, where Big Earth Data has a unique role to play.
iii
iv
Big Earth Data in Support of the Sustainable Development Goals (2020): The Belt and Road
The year 2020 marks the 75 th anniversary of the United Nations and the start of the “Decade of Action” to deliver the Sustainable Development Goals by 2030. China’s science and technology community will continue to work with other countries to advance the 2030 Agenda and contribute to achieving the SDGs as scheduled. Chinese scientists wish to present this report as part of China’s contribution as we remain committed to collaboration and sharing with the rest of the world. Finally, I would like to express my respect and appreciation for the CASEarth team of scientists led by Academician Guo Huadong for the effort they have made toward the SDGs in the spirit of science and innovation.
Bai Chunli President, Chinese Academy of Sciences Head of the Leadership Group of CASEarth November 30, 2020
Preface
Preface
A
lmost f ive years after the adoption of Transforming Our World: the 2030 Agenda for Sustainable Development by
the United Nations, the lack of data on indicators is still holding back the scientif ic evaluation of progress toward the Sustainable Development Goals (SDGs). The coronavirus pandemic since early 2020 has made the challenges countries face in realizing the SDGs all the more daunting. To support global implementation of the SDGs, the United Nations launched the “Technology Facilitation Mechanism”, encompassing three parts — the Interagency Task Team, including the 10-Member Group, the collaborative Multi-stakeholder Forum, and an online platform. One of the most important and pressing issues now is to achieve breakthroughs in data and methods for the monitoring of SDG indicators. Big Earth Data enables macroscopic, dynamic, and objective monitoring by making it possible to integrate and analyze data on the land, sea, atmosphere, and human activity to give a holistic understanding of a vast region. This technology can support policymaking by providing dynamic, multi-scale, and cyclical information on multiple SDG indicators closely related to Earth’s surface, environment, and resources.
v
vi
Big Earth Data in Support of the Sustainable Development Goals (2020): The Belt and Road
Our goal is to convert Big Earth Data to information relevant to the SDGs, construct and integrate such data to support SDG indicators, study the connections and couplings between the SDGs, and then serve as a tool for SDG-related policymaking. This year, we selected six SDGs to study, based on the advantages of Big Earth Data and features of SDG indicators. Big Earth Data can contribute to the evaluation of the six SDGs in three ways: by providing data products, new evaluation methods and models, and case studies to monitor progress and inform policymaking. In this 2020 report, 44 typical cases at local, national, regional, and global scales are presented to showcase the studies on and monitoring results of 19 SDG targets, including 43 data products, 25 methods, and 30 results that are of value to policymaking. The cases cover the topics of Mozambique’s customized agricultural monitoring system, monitoring of desert locusts in Asia and Africa, the spatial and temporal patterns of surface water, high-precision global urban impervious surface mapping for assessing sustainable urbanization, global burned area distribution, dynamic change of mangrove forests, global high-resolution forest cover, and land degradation assessment at global and local scales. They all point to the great value of Big Earth Data and related technologies and methods as new analytical tools with which we will be able to more deeply understand and make better policies for the SDGs and related issues.
Preface
This report could not have been completed without the leadership and support from the Chinese Academy of Sciences, the Ministry of Foreign Affairs, the Ministry of Science and Technology, and other ministries. We are grateful for the valuable opinions shared by leaders and experts from the National Development and Reform Commission, the Ministry of Natural Resources, the Ministry of Ecology and Environment, the Ministry of Housing and Urban-Rural Development, the Ministry of Transport, the Ministry of Water Resources, the Ministry of Agriculture and Rural Affairs, the National Health Commission, the Ministry of Emergency Management, the National Bureau of Statistics, and the National Forestry and Grassland Administration. Finally, our utmost appreciation goes to every scientist on the team for their hard work.
Guo Huadong CAS Academician Chief Scientist of CASEarth Member of the UN 10-Member Group to support the TFM for SDGs November 30, 2020
vii
Executive Summary
Five years after the adoption of Transforming Our World: The 2030 Agenda for Sustainable Development (abbreviated to 2030 Agenda for Sustainable Development), the implementation of the Sustainable Development Goals (SDGs) is still constrained, to a certain extent, by the lack of data on progress, inadequate statistical methods, the diverse issues concerning localization, and the multitude of indicators that are both intertwined and mutually restrictive. Big Earth Data— an innovative set of technologies combining big data and Earth observation—can serve as a new key to unlocking Earth’s secrets and a new engine to drive discoveries. The 2019 report Big Earth Data in Support of the Sustainable Development Goals showcases the contributions this technology can make to sustainable development. This year’s report focuses on Big Earth Data’s contributions to monitoring and evaluating six SDGs through data products, methodologies, models and policymaking support. The report covers SDG 2 (Zero Hunger), SDG 6 (Clean Water and Sanitation), SDG 11 (Sustainable Cities and Communities), SDG 13 (Climate Action), SDG 14 (Life Below Water) and SDG 15 (Life on Land).
For SDG 2, this report focuses on areas with food security vulnerability (with attention given to Africa) and a key factor for sustainable agriculture—productivity. The cases demonstrate the application of Big Earth Data to three SDG indicators/targets: production per labor unit (SDG 2.3.1), the proportion of agricultural area under productive and sustainable agriculture (SDG 2.4.1), and international cooperation and capacity building (SDG 2.a). They showcase innovations and improvements on the methodologies used to evaluate SDGs, and demonstrate applications of technology related to Big Earth Data, such as methodologies for agricultural damage assessment integrating multiple data sources and a pest forecasting model, with accuracy higher than 80%. The key f indings of the case studies on SDG 2 are summarized as: 1) the productivity of fragmented farmland in Zambia was higher than that of large-scale farmland, 2) the potential yield of wheat in Ethiopia was about 3.62 t/ha, and 3) the multiple cropping index
x
Big Earth Data in Support of the Sustainable Development Goals (2020): The Belt and Road
of rice increased signif icantly in f ive countries on the Indochina Peninsula from 2000 to 2019. Pathways for improving productivity in Zambia and Ethiopia have also progressed. Solutions for early warning of food security based on Big Earth Data and cloud services in Mozambique serve as a success story and representative case study on knowledge sharing. This report fully displays the potentially critical role of Big Earth Data in achieving SDG 2 Zero Hunger by providing data products, methodologies and support to policymaking.
With regard to SDG 6 (Clean Water and Sanitation), the report presents Big Earth Data-enabled methods for monitoring and evaluating the proportion of water bodies with good ambient water quality (SDG 6.3.2), change in water-use eff iciency over time (SDG 6.4.1), and change in the extent of water-related ecosystems over time (SDG 6.6.1). The focus is on water resources, water environments and water-related ecosystems, with cases highlighting their application in the countries involved in the Belt and Road Initiative. These countries refers to the 119 countries in Asia, Africa, Europe and Oceania that have signed the Belt and Road Initiative cooperation documents with China as of the end of January 2020. The list of countries is provided by China’s Belt and Road Portal (https://www.yidaiyilu.gov.cn/gbjg/gbgk/77073.htm, accessed on April 8, 2020). Datasets produced include the spatiotemporal distribution of surface water transparency in the countries involved in the Belt and Road Initiative (2015 and 2018), water-use eff iciency in the Lancang-Mekong River Basin, and water bodies in 86 Ramsar sites distributed in Asia, Europe and Africa (2000-2018). New methods and models include a lake clarity inversion model based on the water color index and hue angle, a crop water productivity assessment method based on spatiotemporal data fusion of multi-source remote sensing data combined with local crop growth processes, a water-use eff iciency method based on the comprehensive simulation of the water system, and an automatic water extraction algorithm for massive remote sensing big data and cloud platforms. The report’s f indings on SDG 6 reveal that the water clarity in large lakes along the Belt and Road worsened from 2015 to 2018, water consumption per 10,000 USD of GDP has decreased signif icantly in Central Asia over the past 20 years, and the internal water bodies in 50% of the studied Ramsar sites showed a signif icant change from 2000 to 2018,
Executive Summary
most of which showed growth. These outcomes can inform policymaking in developing countries on water environment monitoring and management, optimal distribution of water resources, and protection of wetlands.
In terms of SDG 11 (Sustainable Cities and Communities), the report focuses on three indicators: public transportation (SDG 11.2.1), urbanization (SDG 11.3.1), and protection of natural and cultural heritage (SDG 11.4.1). These indicators were monitored and evaluated in Asia, Europe and Africa with the support of Big Earth Data. Several data products were independently generated: three phases (2015, 2017 and 2019) of urban road network products with a resolution of 10 m in the Belt and Road region; two phases of global impervious surface products with a resolution of 10 m in 2015 and 2018; forest disturbance and man-made interference index datasets for World Natural Heritage sites from 1997 to 2019; two phases of normalized urban construction land, night light and gridded population data in 2000 and 2015; and an urban built-up area dataset for 1, 000 cities in Asia, Europe and Africa (including China) with a resolution of 30 m from 1990 to 2015. Modeling methods were also developed, including a road network extraction method based on deep learning and multi-source remote sensing data. Several indicators of SDG 11 were modif ied and expanded to further improve the indicator system, such as proposing a new urbanization intensity indicator for World Cultural Heritage sites while also establishing a quantitative assessment method for man-made pressure indicators for World Natural Heritage sites. The model used for SDG 11.3.1 was improved to solve the coupling problem between population distribution and land expansion data. There is a deliberate focus in this report on three aspects of interconnection in Asia, Europe and Africa: land use, heritage site protection and urban spatial patterns. Global and local dynamic processes for sustainable development were monitored and evaluated for cities in Asia, Europe and Africa using Big Earth Data. The Global Sustainable Development Report 2019 outlined four levers of change and six entry points for the sustainable development goals; in that context, this report takes aim at the “science and technology” lever and the “urban and peri-urban development” entry point. In this way, both data and technical support can be provided for urban sustainable development at the regional scale.
xi
xii
Big Earth Data in Support of the Sustainable Development Goals (2020): The Belt and Road
This report takes two specif ic objectives for SDG 13 (Climate Action): reducing the loss of climate-related disasters (SDG 13.1) and climate change impact and response (SDG 13.2). SDG 13 index monitoring and progress assessments were carried out using Big Earth Data. In terms of climate change-related disasters, high-resolution spatial distribution datasets of different disaster types (flood, landslide, f ire and dust storm) were added based on previous statistical data. The results show that Asia and Africa were the most seriously affected by natural disasters from 2015 to 2019; the global burned areas in 2015 and 2019 were similar, though South America’s burned area changed signif icantly; and the problem of dust storm pollution in most parts of Central Asia shows an improving trend. In terms of the impact of and response to climate change, this report analyzes multiple aspects through Big Earth Data and provides decision support for future action. It was found that the high-risk freeze-thaw disasters in Asia showed an increasing trend, from 17% of the total area in 2001 to 20% in 2019. The analysis of various meteorological conditions revealed that a signif icant increase in precipitation on the Arabian Peninsula and other regions from 2018 to 2019 was an important reason for the serious disasters caused by African desert locusts in Africa and Asia. Long-term time-series observations showed that Asia experienced a signif icant increase in precipitation from 1961 to 2018. The results indicate that the central region is getting wetter, and the frequency of meteorological drought events is gradually increasing, but the average area, duration and severity of droughts are gradually decreasing. The report includes a prediction of the trend of global forest carbon budget in the next decade, f inding that the carbon sequestration capacity of northern hemisphere forests will be enhanced in the next ten years, while the carbon sequestration capacity of southern hemisphere forests will be signif icantly weakened. It is estimated that the navigable mileage of merchant ships in the Northeast Passage will increase slowly in 2030 and 2040, but the navigable mileage will increase signif icantly around 2050.
Regarding SDG 14 (Life Below Water), the report focuses on three indicators: the prevention and substantial reduction of all types of marine pollution (SDG 14.1), the sustainable management and protection of marine and coastal ecosystems (SDG 14.2), and the protection of coastal and marine areas (SDG 14.5). The report presents three case studies
Executive Summary
where Big Earth Data technology is applied to the dynamic monitoring and integrated assessment of various indicators at local levels with spatiotemporal data fusion and simulation modeling, among other methodologies. The topics include: a eutrophication assessment of the South China Sea and its surrounding waters; spatiotemporal patterns of coastal aquaculture ponds on the Indochina Peninsula and their impact on offshore chlorophyll-a concentration; monitoring the changes of water environments in the ocean around the Port of Colombo; monitoring the extent and dynamic change in mangrove forests along the Maritime Silk Road; and conservation and utilization of coastlines along the Maritime Silk Road. Several f indings were revealed. Compared with historical data, the nutrient pollution emissions in the Johor Strait, the Manila Bay and the Bay of Bengal have not decreased in recent years, and they have even increased in the latter. Coastal aquaculture ponds are an important terrestrial nutrient pollution source for offshore eutrophication. A case study on the impact of the construction of the Port of Colombo found that water environment responses including increases in chlorophyll-a and suspended sediment occurred in the initial stage of port construction but returned to the pre-construction state within four months. In studies of the coastal zone, 68.4% of mangrove habitats in the Asian countries have continued to decline. In the past 30 years, the length of the coastline in the study area has generally shown a stable growth trend, though the coastlines along Southeast Asian countries, have changed more.
For SDG 15 (Life on Land), the report focuses on six aspects, namely forest protection and restoration, land degradation and restoration, habitats of endangered species, biodiversity protection, mountain ecosystem protection, and key basic datasets to monitor progress toward SDG 15. Seven indicators are used to assess as a whole as well as representative areas: the forest area as a proportion of total land area (SDG 15.1.1), the proportion of biodiversity conservation sites (SDG 15.1.2/15.4.1), net permanent forest loss (SDG 15.2.1), the proportion of degraded land (SDG 15.3.1), the mountain green cover index (SDG 15.4.2), and the Red List Index (SDG 15.5.1). Big Earth Data was fully leveraged to develop the datasets, models, methodologies and conclusions established in this report. These include a dataset of global-scale 30 m forest cover (2019), global land degradation and restoration (2018), global land cover change, global potential net
xiii
xiv
Big Earth Data in Support of the Sustainable Development Goals (2020): The Belt and Road
primary productivity, water use eff iciency of terrestrial ecosystems, and changes in major ecological elements of global arid ecosystems. For models and methods, the report establishes a rapid global forest cover production framework based on machine learning and cloud computing, a protection area vulnerability assessment methodology, a grassland degradation evaluation system considering ecology and production factors, and a grid-scale mountain green cover index calculation model. The key conclusions include the f inding that China’s Land Degradation Neutrality (LDN) trend continue to improve from 2015 to 2018 (contributing nearly one-f ifth of the global total), and most parts of Central Asia have not completed the LDN target. Finally, valuable suggestions are proposed for preventing and controlling land degradation in Mongolia and protecting Asian elephants and Siberian tiger habitats. This research provides strong support for dynamically monitoring and evaluating SDG 15 indicators.
This report is being published at the beginning of the Decade of Action for achieving the United Nations SDGs. The studies presented in each chapter make the case for the role of Big Earth Data in realizing the SDGs by demonstrating science-based approaches to overcoming the constraints on monitoring and assessing progress toward SDG targets. Moreover, the report emphasizes action by connecting research f indings to policy recommendations.
—
—
Spatial distribution and yield levels of major crops in Northeast Eurasia Spatial distribution of the yield gap of wheat in Ethiopia (1 km)
—
Cultivated land productivity dataset and its application in Northeast Eurasia
Ethiopian wheat production potential
Capacity building contributing to food security governance in Mozambique
SDG 2.a
SDG 2.4
Thematic maps of paddy rice planting on the Indochina Peninsula 2000–2019
Spatiotemporal patterns of rice paddy dynamics on the Indochina Peninsula
—
A systematic agricultural monitoring and food security early warning solution together with the capacity building activities provided by CropWatch Cloud promoting “leapfrog” development of Mozambique’s food security governance capabilities
Promote a pathway toward sustainable agriculture and doubling agriculture production by 2030 in Ethiopia
Analysis of differences in cultivated land use patterns and yield levels between Northeast China and neighboring countries
Provide spatial data support and scientif ic reference for regional paddy rice production
The development trend of desert locusts was analyzed to support pest prevention in severely damaged countries
Methods and algorithms for monitoring desert locusts by coupling multi-source data and an insect dispersal model
Desert locust migration paths and core breeding areas in Asia and Africa from 2018 to 2020
Monitoring desert locusts in Asia and Africa
—
—
Methods for agricultural productivity assessments by classes of farmland size
SDG 2.3
Decision support
Spatial distribution of classes of farmland size and agricultural productivity in Zambia (500 m-1 km resolution)
Methods and models
Spatial distribution of classes of farmland size and agricultural productivity in Zambia
Data products
Cases
Targets
List of Cases on Big Earth Data for SDGs
List of Cases on Big Earth Data for SDGs xv
Provide decision support for the Lancang-Mekong River Basin management institutions and Association of Southeast Asian Nations (ASEAN) centers Provide decision support to implement the Convention on Wetlands and Biodiversity
Analysis of water-use eff iciency in the Lancang-Mekong River Basin based on the comprehensive simulation of the water system Spatiotemporal parameters for a global water NDVI threshold suitable for a variety of optical sensors and massive remote sensing data; no sample migration is required
Water resource utilization eff iciency evaluation in Central Asia 2000-2019
The f irst dataset of water-use eff iciency on the scale of the Lancang-Mekong River Basin Changes in water area in Ramsar sites in the countries involved in the Belt and Road Initiative 2000-2018 Road network dataset with full coverage of the 65 countries in Asia, Europe and Africa in 2015, 2017 and 2019 Population monitoring data along rural roads in the area along the China-Pakistan Railway in 2014 and 2019
Water use eff iciency in Central Asia
Water-use eff iciency in the Lancang-Mekong River Basin
Changes in water bodies in Ramsar sites
Road network changes and road connectivity assessment in the 65 countries in Asia, Europe and Africa
Analyzing rural accessibility along the China-Pakistan Railway
SDG 11.2
SDG 6.6
—
—
—
Crop water productivity on a f ine (f ield) scale (2016-2019)
SDG 6.4
—
Crop water productivity assessment based on spatiotemporal data fusion of multi-source remote sensing data combined with local crop growth processes
Crop water productivity assessment of a representative irrigation district in Morocco
SDG 6.3
—
—
Provide decision support for the United Nations Central Asia water crisis coordination agency and negotiation on water resources of cross-border rivers in Central Asia of United Nations
—
A lake clarity inversion model based on the water color index and hue angle
The f irst temporal and spatial distribution dataset of surface water transparency in the countries involved in the Belt and Road Initiative (2015 and 2018)
Spatial and temporal patterns of surface water clarity in the countries involved in the Belt and Road Initiative
Decision support
Methods and models
Data products
Cases
Targets
(continued)
xvi Big Earth Data in Support of the Sustainable Development Goals (2020): The Belt and Road
SDG 13.1
SDG 11.4
SDG 11.3
Targets
Spatial distribution disaster dataset based on Earth observation data
Natural disaster impact evaluation in Asia, Europe and Africa, 20152019
Normalized urban construction land data, night light data and population grid data
Study on spatial and temporal distribution and protection countermeasures of cultural heritage in Asia, Europe and Africa Forest disturbance and human disturbance data of natural heritage sites in Asia, Europe and Africa
Remote sensing data for urban environments
Assessment of land use changes in large cities in Asia, Europe and Africa
Human disturbance monitoring and integrated pressure analysis in World Natural Heritage sites in Asia, Europe and Africa
Urban built-up land for cities with a population over 300,000 in Asia, Europe and Africa in 1990, 1995, 2000, 2005, 2010 and 2015
Global high-precision urban impervious surface remote sensing datasets with 10 m resolution in 2015 and 2018
Data products
Urban land use eff iciency and sustainability analysis for cities in Asia, Europe and Africa
High-precision global urban impervious surface mapping for global urbanization assessment
Cases
Analysis of sustainable development trends of representative cases in Asia, Europe and Africa
—
Compared to statistical database products, Earth observation products can provide more accurate and timely disaster area assessment
Data and evaluation reports provided to UNESCO and national institutions in cooperation with the countries involved in the Belt and Road Initiative
Report on urban development in the eco-environment
Provide time-series data to support the construction of sustainable cities and human settlements in regions in Asia, Europe and Africa
Provide data support and decision support for the sustainable development of global urban areas
Decision support
Establish a human disturbance index that can quantitatively assess the human disturbance state of natural heritage sites
Propose new indicators of urbanization intensity
—
—
Propose a rapid global impervious surface extraction method based on multi-phase ascending and descending Sentinel-1/2A data using the texture and land surface phenology of this data; provide online calculation tools for SDG 11.3.1 supported by a Big Earth Data cloud service platform
Methods and models
(continued)
List of Cases on Big Earth Data for SDGs xvii
SDG 13.2
SDG 13.1
Targets
Decision support system for freezethaw disaster evaluation in High Mountain Asia
Provide data support for analyzing the contributing factors, impacts and growth trends of the early 2020 desert locust upsurge Provide decision guidance for the planning of Arctic navigation paths and channel development Accomplish the dynamic monitoring and impact tracking of typical drought events, and provide decision support for the government on measures of disaster prevention and reduction
More accurately estimate the roles of forest carbon sinks in regulating global carbon balance and maintaining the global climate Development of an online calculation system for the identif ication and evaluation of freeze-thaw disasters Nonlinear artif icial neural network with multiple inputs, including the Fengyun-3C VSM product, MODIS NDVI, latitude, longitude, and DEM, for accurate soil moisture at the regional scale Optimal route algorithm for the Northeast Passage Propose an evaluation model for the dynamic monitoring of meteorological drought events based on the threedimensional clustering algorithm in the space of longitude-latitude-time
Forest carbon budget dataset with long-term time series Freeze-thaw disaster assessment and evolution datasets with high temporal and spatial resolution in High Mountain Asia, 1980–2020 Fengyun satellite-based land surface soil moisture in the breeding areas of the African desert locust Optimal route products for the Northeast Passage Long-term standardized drought index SPEI data that comprehensively considers both precipitation and evaporation
Global forest carbon budget and climate change
Risk assessment of freeze-thaw disasters in High Mountain Asia
Contributing factors and growth trend analysis for the desert locust upsurge in early 2020
Evaluation of navigation capacity in the Northeast Passage
Meteorological drought monitoring over arid Central Asia
Deep learning-based dust storm intensity extraction using multisource data
Datasets of dust storm intensity distribution and dust dry deposition
Factors and temporal variation of emissions from dust sources in Central Asia over the last 40 years —
Provide support for dust storm early warning for disaster management organizations in Central Asia
—
Automatic production method based on Big Earth Data and artif icial intelligence technology
Global burned area product with the current highest spatial resolution
Global burned area distribution and changes
Decision support
Methods and models
Data products
Cases
(continued)
xviii Big Earth Data in Support of the Sustainable Development Goals (2020): The Belt and Road
1990, 2000, 2010 and 2015 mangrove datasets of some countries’coastal zone Dataset on land expansion and shoreline change in port cities in 1990, 2000, 2010 and 2015
Monitoring the extent and dynamic change in mangrove forests of some coastal countries
Conservation and utilization of the coastline
SDG 14.2
SDG 14.5
Urban port construction land expansion analysis model and shoreline protection analysis method
Provide a scientif ic and quantitative basis for further promoting shoreline protection and ecological restoration in the coastal zone
Mangrove conservation, restoration and management, as well as ecological and environmental protection of the sea and its surroundings for some countries
Water environment parameters near the Port of Colombo (2013-2020, 30 m)
Precise mangrove extraction methods based on machine learning and analysis of mangrove dynamics
Objectively reflect the characteristics of water environment changes before and after the port project, and provide a scientif ic basis for promoting the recovery of the marine environment in the port
Quantitative inversion of important water environment parameters such as chlorophyll-a and suspended sediment near the port
Monitoring changes in water environments in the ocean around the Port of Colombo
SDG 14.1
Provide data and decision support for policy development related to improving eutrophication of coastal waters and reducing marine nutrient salt pollution in aquaculture countries
A quantitative method for evaluating the effect of spatial distribution of nearshore aquaculture ponds on offshore chlorophyll-a
Temporal and spatial distribution of nearshore aquaculture ponds and marine chlorophyll-a monitoring products for representative aquaculture countries on the Indochina Peninsula in 2000, 2010 and 2015
Provide results and recommendations on eutrophication assessment in the South China Sea and the surrounding waters
Spatiotemporal patterns of coastal aquaculture ponds on the Indochina Peninsula and their impact on offshore chlorophyll-a concentration
Decision support
Methods and models A new method for eutrophication evaluation based on inorganic nitrogen, inorganic phosphorus and dissolved oxygen
Data products Eutrophication f ield survey data in the South China Sea and surrounding waters in 2019
Cases
Eutrophication assessment of the South China Sea and surrounding waters
Targets
(continued)
List of Cases on Big Earth Data for SDGs xix
SDG 15.3
SDG 15.2
SDG 15.1
Targets
Quantifying the achievement of the zero net land degradation in Central Asia and providing its spatial distribution map —
—
—
Map of forest cover in nine Southeast Asian countries, 2014-2018 Datasets on global land degradation/improvement in 2018 Datasets of yearly land degradation index in Central Asia at 500 m spatial resolution for 2000-2019
Spatiotemporal dynamics of forest cover in Southeast Asia
Global Land Degradation Neutrality (LDN) progress report
Monitoring and assessing land degradation in f ive Central Asian countries
The evaluation index system for grassland degradation considering ecological and production factors in North Africa
—
Dataset of grassland degradation during 2007-2019 with 250 m spatial resolution Datasets of land degradation and restoration of Mongolia in 1990-2000, 2000-2010 and 2010-2015 with 30 m spatial resolution
Assessment of grassland degradation in Mediterranean Africa
Dynamic monitoring and control measures of land degradation and restoration in Mongolia
—
Ecological vulnerability assessment of protected areas with Big Earth Data
Dataset of ecological vulnerability of protected areas among “Three Seas and One Lake” (TSOL) from 2001 to 2015.
Assessment of vulnerability in protected areas of the “Three Seas and One Lake” international basins, 2001–2015
Analyzing the driving forces and land degradation and restoration in Mongolia, and providing land degradation control suggestions for typical regions
Report on SDG 15.3.1 progress at the national scale, providing decisionmaking support for UN agencies
—
—
—
—
Global forest cover map at 30 m spatial resolution for 2019, with accuracy no less than 85%
Global/regional forest cover (2019)
Decision support
Methods and models
Data products
Cases
(continued)
xx Big Earth Data in Support of the Sustainable Development Goals (2020): The Belt and Road
SDG 15
—
1 km global land cover 20002018, the accuracy improved 5% than current products Estimations of climate potential productivity (CPP) and water use eff iciency (WUE) for global terrestrial ecosystems Global annual SPEI (Standardized Precipitation Evaporation), precipitation, and air temperature data, which span from 2000 to 2015 with 0.5° spatial resolution
Global land cover changes in the past 20 years
Potential productivity and water use eff iciency for global terrestrial ecosystems
Global assessment of the impact of population and climate change on the ecosystem material supply capacity of drylands
—
—
Forest types and proportions in Northeast Asia in 1990-2010 with 30 m spatial resolution
Monitoring forest type of Amur tiger habitat —
—
Annual forest loss of the Asian elephant habitat in 13 range states (2001-2018), with spatial resolution better than 30 m
Assessment of forest changes in Asian elephant habitats
SDG 15.4
SDG 15.5
To construct the grid-based mountain green cover index calculating model
High spatial resolution mountain green cover index datasets of the Belt and Road economic corridor
Time-series monitoring and decision support of the Mountain Green Cover Index in the Belt and Road economic corridor
Methods and models
Data products
Cases
Targets
—
—
—
Provide support for the National Forestry and Grassland Administration of China and the Northeast China Tiger and Leopard National Park to better restore the Amur tiger’s habitat
To provide the evaluation report for the IUCN to make global protection strategies
To propose the protection suggestions for the mountain ecosystem of the Belt and Road economic corridor
Decision support
(continued)
List of Cases on Big Earth Data for SDGs xxi
Contents
i
Foreword
v
Preface
ix
Executive Summary
xv
List of Cases on Big Earth Data for SDGs
Chapter 1
Challenges to implementing the SDGs / 2
Introduction
Big Earth Data / 3 Big Earth Data in Support of SDGs / 4
Background / 8 Main Contributions / 9 Chapter 2
SDG 2 Zero Hunger
Case Study / 11 Spatial distribution of classes of farmland size and agricultural productivity in Zambia / 11 Monitoring desert locusts in Asia and Africa / 18 Spatiotemporal patterns of rice paddy dynamics on the Indochina Peninsula / 25 Cultivated land productivity dataset and its application in Northeast Eurasia / 32 Ethiopian wheat production potential / 40
Contents
Capacity building contributing to food security governance in Mozambique / 46 Summary / 55
Background / 58 Main Contributions / 59 Chapter 3
SDG 6 Clean Water and Sanitation
Case Study / 60 Spatial and temporal patterns of surface water clarity in the countries involved in the Belt and Road Initiative / 60 Crop water productivity assessment of a representative irrigation district in Morocco / 67 Water-use eff iciency in Central Asia / 74 Water-use eff iciency in the Lancang-Mekong River Basin / 81 Changes in water bodies in Ramsar sites / 90 Summary / 97
Background / 100 Main Contributions / 101 Chapter 4
SDG 11 Sustainable Cities and Communities
Case Study / 103 Road network changes and road connectivity assessment in the 65 countries in Asia, Europe and Africa / 103 Analyzing rural accessibility along the China-Pakistan Railway / 113
xxiii
xxiv
Big Earth Data in Support of the Sustainable Development Goals (2020): The Belt and Road
High-resolution urban impervious surface mapping for global urbanization assessment / 119 Urban land use eff iciency and sustainability analysis for cities in Asia, Europe and Africa / 125 Assessment of land use changes in global large cities / 133 Study on spatial and temporal distribution and protection countermeasures of cultural heritage in Asia, Europe and Africa / 140 Human disturbance monitoring and integrated pressure analysis in World Natural Heritage sites in Asia, Europe and Africa / 151 Summary / 159
Background / 162 Main Contributions / 163 Chapter 5
SDG 13 Climate Action
Case Study / 165 Natural disaster impact evaluation in Asia, Africa, Europe and Oceania, 2015-2019 / 165 Global burned area distribution and changes / 174 Factors and temporal variation of emissions from dust sources in Central Asia over the past 40 years / 179 Global forest carbon budget and climate change / 185
Contents
Risk assessment of freeze-thaw disasters in High Mountain Asia / 191 Contributing factors and growth trend analysis for the desert locust upsurge in early 2020 / 197 Evaluation of navigation capacity in the Northeast Passage / 203 Meteorological drought monitoring over arid Central Asia / 209 Summary / 216
Background / 220 Main Contributions / 221 Chapter 6
SDG 14 Life Below Water
Case Study / 223 Eutrophication assessment of the South China Sea and surrounding waters / 223 Spatiotemporal patterns of coastal aquaculture ponds on the Indochina Peninsula and their impact on offshore chlorophyll-a concentration / 232 Monitoring changes in water environments in the ocean around the Port of Colombo / 238 Monitoring the extent and dynamic change in mangrove forests of some coastal countries / 244 Conservation and utilization of the coastline / 251 Summary / 258
xxv
xxvi
Big Earth Data in Support of the Sustainable Development Goals (2020): The Belt and Road
Background / 262 Main Contributions / 264 Chapter 7
SDG 15 Life on Land
Case Study / 266 Global/regional forest cover (2019) / 266 Assessment of vulnerability in protected areas of the “Three Seas and One Lake” international basins, 2001-2015 / 271 Spatiotemporal dynamics of forest cover in Southeast Asia / 276 Global Land Degradation Neutrality (LDN) progress report / 284 Monitoring and assessing land degradation in f ive Central Asian countries / 289 Assessment of grassland degradation in Mediterranean Africa / 296 Dynamic monitoring and control measures of land degradation and restoration in Mongolia / 304 Time-series monitoring and decision support of the Mountain Green Cover Index in the Belt and Road Economic Corridors / 311 Assessment of forest changes in Asian elephant habitats / 319 Monitoring forest type of Amur tiger habitat / 325 Global land cover changes in the past 20 years / 331 Potential productivity and water use eff iciency for global terrestrial ecosystems / 339 Global assessment of the impact of population and climate change on the ecosystem material supply capacity of drylands / 347 Summary / 355
Contents
Chapter 8
Summary and Prospects / 358
References / 362 Acronyms / 381
xxvii
2 Challenges to implementing the SDGs 3 Big Earth Data
4 Big Earth Data in Support of SDGs
Chapter 1
Introduction
In 2015, the United Nations Sustainable Development Summit adopted a plan of action titled Transforming Our World: The 2030 Agenda for Sustainable Development, which proposed 17 SDGs covering economic, social and environmental aspects. These goals represent the direction of national development and international cooperation needed to address the world’s most pressing environmental and societal issues. In the almost f ive years since the Agenda’s adoption, the monitoring and evaluation of its implementation have been constrained by a lack of data, varying capacities, and indicators that are both intertwined and mutually restrictive. Scientif ic and technological innovation is a solution to this pressing issue, and this report focuses specif ically on “Big Earth Data”, a set of technologies and methods being practiced at the intersection of big data and Earth observation. The CAS Big Earth Data Science Engineering Project (CASEarth) has released annual scientif ic, evidence-based monitoring results for six SDGs — SDG 2 (Zero Hunger), SDG 6 (Clean Water and Sanitation), SDG 11 (Sustainable Cities and Communities), SDG 13 (Climate Action), SDG 14 (Life Below Water), and SDG 15 (Life on Land) — by drawing on Big Earth Data’s strengths in multi-scale, near-real-time processing and system integration. This is a concrete contribution to implementing the SDGs.
Challenges to implementing the SDGs The Global Indicator Framework for the Sustainable Development Goals and Targets was adopted in 2017 by the United Nations as a preliminary, voluntary system for the Member States to monitor progress in implementing the SDGs. The Framework, subject to regular ref inement and updates, faces the following problems. (1) Lack of data. The previous situation where there were neither evaluation methods nor data has improved for all indicators in the past f ive years. For 46% of indicators, however, there are methods but no data; for those where both are available, the results are measured largely in a statistical way without the support of spatial distribution information. Spatial data that is objective and accurate at varying scales is necessary for achieving the SDGs. Specif ically, data collected scientif ically can be used to assess changes in the natural
Chapter 1 Introduction
environment regularly and quantitatively. They can accurately identify the spatial position of disasters and predict their future trends, in such cases as extremely high temperatures and heatwaves, higher frequency of f ires, ocean acidif ication, increased eutrophication, continued land degradation, reduced biodiversity and increased environmental impact on agricultural production. (2) Imbalance in capacities. Developing countries are constrained by the level of their economic growth and carrying capacity of resources and environment in their abilities to collect and analyze timely, quantitative data. The lack of data has rendered invisible such serious issues as high ratios of stunting, inadequacy of urban housing and public space, weak disaster resilience, lack of access to safe drinking water, and overuse of forests. Those who take advantage of Big Earth Data can collect objective data at global, regional, and other scales in a timely, accurate, and comprehensive way while improving their compatibility and comparability, so that “no one is left behind” on data essential to achieving the SDGs. (3) Intertwined, but mutually restrictive indicators. SDG indicators are wide-ranging, long-term, and intertwined. These diverse, complex indicators at multiple tiers come together to form a coherent, feasible whole. The creation of methods and models for objective and effective monitoring and evaluation based on compatible, quantif iable data is an urgent issue for which a solution must be found.
Big Earth Data The United Nations launched the Technology Facilitation Mechanism (TFM) to address the above-mentioned issues and challenges through Science, Technology, and Innovation (STI) by pooling the collective wisdom of the scientif ic community, business community and other stakeholders. Big Earth Data refers to the use of big data in the f ield of Earth science with spatial attributes, especially the massive Earth observation data generated by space technology (Guo et al., 2016). Such data is mainly produced at a large spatial scale by scientif ic devices, detection equipment, sensors, socio-economic observations and computer simulation processes. Similar to other types of big data, Big Earth Data is massive, multisource, heterogeneous, multi-temporal, multi-scaled and non-stationary. Moreover, it has strong spatiotemporal and physical correlations, and the data generation methods and sources are controllable. Big Earth Data science is interdisciplinary, encompassing natural
3
4
Big Earth Data in Support of the Sustainable Development Goals (2020): The Belt and Road
sciences, social sciences and engineering. It systematically studies the correlation and coupling of the Earth system based on data analysis. Earth is observed and studied as a whole by simultaneously employing big data, artif icial intelligence and cloud computing to understand the complex interactions and development processes between Earth’s natural system and the human social system. Big Earth Data can therefore make an important contribution to the realization of SDGs. In 2018, CAS launched the CASEarth, an endeavor to accelerate the transformation from Earth data systems and sharing to Digital Earth. It promotes the sharing of data, knowledge and experience across the world, and supports scientif ic discovery, technological innovation and policymaking. Big Earth Data science is a solution to crosssectoral, multidisciplinary collaboration (Guo, 2020a). It is an innovation under the TFM that can support the achievement of SDGs (Guo, 2020b).
Big Earth Data in Support of SDGs The CASEarth Big Earth Data system can support the implementation of SDGs by converting Big Earth Data to relevant information, providing policymaking support, constructing and integrating an index system, and studying the relationships and couplings between various SDG targets from the perspective of the Earth system. It can also support the monitoring and evaluation of SDG indicators through data-sharing platforms and cloud infrastructure by providing data, online calculations and visualizations. Currently, CASEarth shares a total of 8 PB of data, 3 PB of which is updated annually. It can provide 1 PF of high-performance computing and big data processing in the cloud. The Big Earth Data system’s full capabilities, from data-to-information visualization to numerical simulation, can support dynamic monitoring and macro-level policymaking for SDGs.
Chapter 1 Introduction
CASEarth approaches SDGs from the following four aspects: (1) Constructing a Big Earth Data infrastructure for SDGs to provide data products to close the gap of missing data and facilitate data sharing. (2) Creating methodologies and a technical system for achieving SDGs. (3) Providing data for monitoring SDG indicators from Earth science satellites. (4) Issuing annual reports in the series of Big Earth Data in Support of the Sustainable Development Goals to showcase the latest progress. This year’s report presents 44 representative cases concerning six SDGs and discusses methods and pathways by which Big Earth Data can be used for eff iciently and accurately assessing the implementation of SDGs, and as timely, scientif ic evidence to support policymaking.
5
81 91
Background Main Contributions
111 Case Study 551 Summary
Chapter 2
SDG 2 Zero Hunger
8
Big Earth Data in Support of the Sustainable Development Goals (2020): The Belt and Road
Background SDG 2 (Zero Hunger) aims to end hunger, achieve food security and improve nutrition and promote sustainable agriculture. China’s National Plan on Implementation of the 2030 Agenda for Sustainable Development lists it, together with ending poverty, as the top goal of sustainable development by 2030. However, the realization of the goal of zero hunger at a global scale still faces enormous challenges. Climate change has led to an increase in the frequency of extreme climate events and changes in temperature and precipitation, thus affecting the availability of regional water resources and land resources. These, in turn, change the capacity of agricultural production in different regions and the layout and structure of agriculture, and increase the uncertainty of agricultural production. In Africa, where rain-fed cropland is dominant, the vulnerability of agricultural production is more signif icant, and Africa has become the region most sensitive to global climate change (Thornton et al., 2009). The Global Ecological Environment Remote Sensing Monitoring Report of 2019 and the State of Food Security and Nutrition in the World 2018 show that both globally and in Africa, the population suffering from hunger and malnutrition has been increasing for three years. Food production per capita was reduced to the lowest level since 2014 in Africa, showing that food security has deteriorated over the past f ive years (FAO, 2019). Assessing the status of Zero Hunger, tracking progress toward the goal and monitoring changes in its different indicators have become the common responsibility and obligation of the international scientif ic community. The United Nations Off ice for South-South Cooperation released Good Practices in South-South and Triangular Cooperation for Sustainable Development in 2016 and 2018, which summarized the efforts and effects of achieving SDG 2 in different countries and regions under the support of the United Nations South-South Cooperation and Tripartite Agreement projects. It also emphasized the key role of stimulating cooperation, knowledge sharing and capacity building in accelerating the achievement of SDG 2. However, the data and methodologies for evaluating different indicators of SDG 2 have not been evenly prepared, and most of the evaluations are based on traditional statistical and census data. A complete assessment of the progress toward Zero Hunger must employ new ideas and innovative approaches. The development of space science, digital technology, Big Earth Data and artif icial intelligence can help monitor indicators of SDG 2 on a large scale. This chapter focuses on indicators reflecting food production and national action, targeting Africa, rice, and pests and disease. Africa is a key region with great vulnerability in global food security, rice is one of the most important food crops in the world, and pests and disease are an urgent issue affecting both of the former. To strengthen governance for the goal of Zero Hunger in
Chapter 2 SDG 2 Zero Hunger
some countries involved in the Belt and Road Initiative and some neighboring countries, Big Earth Data techniques must be used. It is through this innovation and improvement of methodologies that technology can be applied toward building a community with a shared future for humankind.
Main Contributions Focusing on three indicators/targets covering information on food production and national action (Table 2.1), this chapter promotes two sets of evaluation methods, including a comparison method for analyzing agricultural productivity at different f ield scales and a method for estimating crop damages from pests and disease by fusing multi-source data and a pest forecasting model. Five sets of products were produced for monitoring indicators at regional and national scales. The spatiotemporal analysis was further applied to showcase the progress on Zero Hunger, helping to support policies for sustainable crop production systems (Table 2.1). Table 2.1 Cases and their contributions Indicators
Cases
Contributions
SDG 2.3.1 Volume of production per labor unit in farming, pastoral, and forestry sectors, by classes of enterprise size
Spatial distribution of classes of farmland size and agricultural productivity in Zambia
Data product: Spatial distribution of classes of farmland size and agricultural productivity in Zambia (500 m-1 km resolution) Method and model: Methods for agricultural productivity assessments by classes of farmland size
Monitoring desert locusts in Asia and Africa
Data product: Desert locust migration paths and core breeding areas in Asia and Africa from 2018 to 2020 Method and model: Methods and algorithms for desert locust monitoring by coupling multi-source data and an insect dispersal model Decision support: The development trend of desert locusts was analyzed to support pest prevention in severely damaged countries
Spatiotemporal patterns of rice paddy dynamics on the Indochina Peninsula
Data product: Thematic maps of paddy rice planting on the Indochina Peninsula 2000-2019 Decision support: Provide spatial data support and scientif ic reference for regional paddy rice production
Cultivated land productivity dataset and its application in Northeast Eurasia
Data product: Spatial distribution and yield levels of major crops in Northeast Eurasia Decision support: Analysis of differences in cultivated land use patterns and yield levels between Northeast China and neighboring countries
SDG 2.4.1 Proportion of agricultural area under productive and sustainable agriculture
9
10
Big Earth Data in Support of the Sustainable Development Goals (2020): The Belt and Road
(continued) Indicators
Cases
Contributions
SDG 2.4.1 Proportion of agricultural area under productive and sustainable agriculture
Ethiopian wheat production potential
Data product: Spatial dataset of the yield gap of wheat in Ethiopia (1 km) Decision support: Promote a pathway toward sustainable agriculture and doubling agriculture production by 2030 in Ethiopia
SDG 2.A Increase investment, including through enhanced international cooperation, in rural infrastructure, agricultural research and extension services, technology development and plant and livestock gene banks in order to enhance agricultural productive capacity in developing countries, in particular least developed countries
Capacity building contributing to food security governance in Mozambique
Decision support: A systematic agricultural monitoring and food security early warning solution together with the capacity building activities provided by CropWatch Cloud, promoting “leapfrog” development of Mozambique’s food security governance capabilities
Chapter 2 SDG 2 Zero Hunger
Case Study Spatial distribution of classes of farmland size and agricultural productivity in Zambia Target SDG 2.3: By 2030, double the agricultural productivity and incomes of small-scale food producers, in particular women, indigenous peoples, family farmers, pastoralists and f ishers, including through secure and equal access to land, other productive resources and inputs, knowledge, f inancial services, markets and opportunities for value addition and non-farm employment.
Background Zambia is located in southern Africa at an altitude between 1,000 m and 1,500 m above sea level. Most areas of the country are on plateau terrain, with higher altitudes in the northeast and lower in the southwest. Agricultural land in the country is generally fertile, with suff icient water supply thanks to hydrological infrastructure. Various crops are suitable for agricultural production over more than half of the land area. The average annual rainfall is 800-1,000 mm. However, the percentage of land shared by cultivated farmland is low. At the same time, Zambia faces many challenges in terms of food security, climate change and natural disasters. With an increasing population, rapid urbanization, economic growth and changing diets, there is escalating competition for water, food and energy over the region. However, baseline data such as cropland extent maps are out of date. Therefore, accurate and timely scientif ic data is urgently needed to support sustainable development in this region in terms of water resources, agricultural resources and energy under environmental stress. At the same time, large commercial farms are commonly found in Zambia, and there are signif icant differences between commercial farms and small-holder farmers in various aspects such as agricultural practices and management. The effects of farmland use patterns on agricultural development (Collier and Dercon, 2014) and productivity (Jombo et al., 2017) under the state and tribal dual land management systems are not yet clear. In order to guide the policies of the Zambian government’s planning on agricultural developments, it is urgent to understand the spatial patterns of productivity and natural endowments. Earth observation provides the opportunity to study and monitor large-scale agriculture, water resources and climate change on a near-real-time basis. With the support of Big Earth Data from the CASEarth and cloud computing technology, the CropWatch team worked closely with scientists from Zambia
11
12
Big Earth Data in Support of the Sustainable Development Goals (2020): The Belt and Road
to generate a 10 m resolution cropland product on a near-real-time basis for 2018. The classes of farmland size and their productivity were also analyzed to guide the planning of agricultural development. This will provide scientif ic support to achieve SDG 2, Zero Hunger, in the country.
Data used in the case Multi-source satellite data covering Zambia from 2017 to 2019, including GF-1, Sentinel-2, Landsat-8, Sentinel-1 SAR, and MODIS NPP since 2014, with spatial resolutions of 16 m, 10 m, 10 m, 30 m and 500 m, respectively. The second version of the US National Center for Environmental Prediction (NCEP) Climate Forecast System (CFSv2) data at 0.25° resolution. In situ measurements of cropland samples shared by GEOWIKI, and validation datasets from the NASA farmland data product (Laso Bayas et al., 2017; Xiong et al., 2017). Field data from local teams was also acquired. Apart from this, in situ samples using GVG APP and UAV were collected by the CASEarth. Global Food Security-support Analysis Data (GFSAD), Cropland Extent 2015 Africa 30 m dataset (Xiong et al., 2017), the European Space Agency Climate Change Initiative Land Cover, Global Land Cover map at 300 m for 2017 (ESACCL-LC-L4-300), the Sentinel-2 Prototype (ESACCI-LC_ S2_Prototype) map for Africa at a 20 m resolution for 2016, and other existing land cover datasets.
Method In situ cropland samples were collected in Zambia using the GVG mobile app together with researchers from the Zambia Ministry of Agriculture. Multi-source remote sensing data including Sentinel-1/Sentinel-2, GF-1 and Landsat-8 was used for cropland mapping along with groundtruth data supplemented by UAV aerial imagery and other public cropland samples. Spectral features, shape and phenological indicators were jointly applied as inputs to a random forest classif ier to generate the updated cropland data for 2017-2019. Based on independent sample data, a confusion matrix accuracy evaluation method was used for farmland. The University of Zambia independently verif ied the accuracy of the product to quantitatively evaluate its quality. Highprecision cropland data enables Zambia to take reasonable, science-based action in agricultural planning and land management. It also helps Zambia achieve sustainable development goals, addressing the No Poverty and Zero Hunger goals in particular. Relying on the improved Zambia 30 m resolution cropland dataset, the classes of farmland size were categorized by the integration of spatial aggregation, zonal statistics, cluster analysis, a decision tree and expert knowledge in cooperation with the Ministry of Agriculture. The farmland productivity was also analyzed by classes of farmland size to identify the spatial variation patterns
Chapter 2 SDG 2 Zero Hunger
in productivity of different farmland scales. The possible reasons for the spatial variations were analyzed by comparing the agro-climatic conditions (radiation, temperature and precipitation) in order to provide suggestions for future planning of agricultural production.
Results and analysis 1. Distribution of the classes of farmland size Remote sensing monitoring shows that Zambia has 8.8782 million ha of cropland (Figure 2.1), N
Legend Non-cropland Cropland
0
150
300
600 km
(a) Cropland map N
Legend Positive cropland Negative cropland
Positive non-cropland Negative non-cropland
0
150
300
600 km
(b) Accuracy assessments
Figure 2.1 Cropland map of Zambia for 2017-2019 and its accuracy assessments
13
14
Big Earth Data in Support of the Sustainable Development Goals (2020): The Belt and Road
accounting for 11.8% of the country’s land area and only 20.3% of Zambia’s arable land, indicating that Zambia still has vast potential in terms of agricultural development in the future to use more cropland for agricultural production. Independent accuracy verif ication by the University of Zambia and the Ministry of Agriculture shows that the overall accuracy (OA) of the new cropland data reached 85.6%. Both large-scale commercial farmlands and fragmented farmland plots scattered by small-holder farmers were accurately extracted. The results of spatial aggregation analysis show that Zambia is dominated by fragmented and medium-scale farmland at 1 km resolution, accounting for 47% and 35% of the country, respectively. Although Zambia has a large number of commercial farms compared with neighboring countries, the proportion of the area of large-scale farmland is still relatively low at only 18% (Figure 2.2). Spatially, fragmented arable land is distributed in western and northern Zambia while large-scale farmland is mainly distributed in the central and southeastern regions (Figure 2.2). There are still some scattered farmlands in National Reserves, but the overall proportion is extremely low, indicating that there are few farming activities in national parks and that nature reserves are effectively protected. N
Legend Large-scale farmland accounts for 18% Medium-scale farmland accounts for 35% Fragmented farmland accounts for 47%
0
150
300
600 km
Figure 2.2 Classes of farmland size in Zambia and their proportions
2. Farmland productivity by classes of farmland size The productivity of farmland in Zambia is generally low in the southwest and relatively high in the north (Figure 2.3). The productivity of farmland in fragmented farming areas is generally higher
Chapter 2 SDG 2 Zero Hunger
than that of large-scale farmlands. At the same time, the variation range of farmland productivity in fragmented farming areas is also greater than that in large-scale farms. The farmland productivity of large-scale farmland presents a normal distribution, while that of fragmented farmland areas shows a partial normal distribution (Figure 2.4).
N
Legend NPP/ [gC/(m2·a)]
1,350
0
150
300
600 km
(a) Large-scale farmlands N
Legend NPP/ [gC/(m2·a)]
1,350
0
150
300
600 km
(b) Fragmented farmlands
Figure 2.3 Land productivity by different classes of farmland size (average over 2014-2018)
15
Big Earth Data in Support of the Sustainable Development Goals (2020): The Belt and Road
16
20,000
Large-scale farmland
Medium-scale farmland
Fragmented farmland
18,000
Number of Grids/(1 km2)
16,000 14,000 12,000 10,000 8,000 6,000 4,000 2,000 0
0
200
400
600
800
1,000
1,200
1,400
1,600
2
NPP/[gC/(m ·a)]
Figure 2.4 Histograms of land productivity by classes of farmland size (average over 2014-2018)
An analysis of the annual average precipitation, temperature and photosynthetically active radiation in the past 15 years revealed that the precipitation presents a decreasing gradient pattern from south to north. The interannual variation of precipitation in the central and southern regions is larger, while the photosynthetically active radiation and temperature present a gradient increase from south to north. The interannual variations of temperature and PAR are larger in the central and southern regions. A multi-factor analysis method f inds that the high availability of water and radiation in northern Zambia results in higher land productivity, but the farmland is more fragmented. In southern Zambia, such as Livingstone and Lusaka provinces, large-scale farms are commonly observed. However, the weather conditions fluctuate greatly from year to year, and the unstable agro-meteorological conditions are generally detrimental to crop production. At the same time, fragmented farmland is dominated by a large number of small-holder farmers. Due to the limited amount of farmland occupied by each farmer, the common use of manure may be another reason for the higher productivity of fragmented farmland.
Chapter 2 SDG 2 Zero Hunger
Highlights Fragmented farmland and medium-scale farmland are dominant in Zambia, while large-scale farmland only accounts for 18%. Z ambia’s large-scale, high-productivity farmland is spatially dislocated, mainly distributed in the central and southwestern regions. Due to unstable weather conditions, productivity is generally low. Irrigation facilities should be appropriately strengthened to ensure the region’s outputs in terms of agricultural production. Fragmented arable land is distributed in western and northern Zambia, with abundant water resources and high land productivity. Reclaimed arable land should be extended.
Outlook The crop production of Zambia has shown rapid growth in the past ten years, making it a net grain exporter. Based on the methods supported by Big Earth Data, the spatial patterns of the classes of farmland size were identif ied and compared with the spatial variations of land productivity as well as the limiting factors such as radiation, temperature and precipitation. Zambia’s large-scale, high-productivity farmlands are spatially dislocated. Large-scale farmland is mainly distributed in the central and southeastern regions. Due to unstable weather conditions, productivity is generally low. Irrigation facilities should be appropriately strengthened to ensure the region’s outputs in terms of agricultural production. Fragmented farmland is distributed in western and northern Zambia, with abundant water resources and high land productivity. However, the interannual fluctuation of agro-meteorological conditions under climate change scenarios will further increase. In recent years, Zambia has been continuously affected by El Niño. In the future, Zambia could optimize the agricultural development policies over different regions according to local conditions such as farmland size, land productivity and natural conditions. The strengthening of the crop varieties and seeds also needs to be considered in order to adapt to the changing climate and to mitigate the adverse effects on agricultural productivity. With the support of Big Earth Data, more reasonable agricultural development policies could be implemented to increase the agricultural outputs, and Zambia might become an important food supplier in southern Africa in the future.
17
18
Big Earth Data in Support of the Sustainable Development Goals (2020): The Belt and Road
Monitoring desert locusts in Asia and Africa Target SDG 2.4: By 2030, ensure sustainable food production systems and implement resilient agricultural practices that increase productivity and production, that help maintain ecosystems, that strengthen capacity for adaptation to climate change, extreme weather, drought, flooding and other disasters and that progressively improve land and soil quality.
Background Food security has always been a top concern in the international community. In the context of climate change, the scope and prevalence of pests have noticeably expanded and increased. Locusts are a major migratory pest worldwide, and since 2018, the abnormal climate has caused desert locusts to reproduce freely in the southern Arabian Peninsula and gradually sweep across the Horn of Africa and Southwest Asian countries. So far, desert locusts have seriously endangered the food security of many countries including Pakistan, Ethiopia, Kenya, Somalia and Yemen. The Food and Agriculture Organization of the United Nations (FAO) has issued an early warning to the world, hoping that countries will be highly alert to locust plagues and adopt multi-country joint prevention and control measures to prevent serious food crises from invading pests. Traditional visual hand-check single-point monitoring methods and limited-site meteorological prediction methods can only obtain information on the occurrence and development of pests at sampling points, and cannot meet the needs of large-area monitoring and timely prevention and control of pests at the regional scale. Remote sensing can eff iciently and objectively monitor the occurrence and development of pests temporally and spatially on a large scale. The rapid development of Earth observation technology in recent years has provided effective technical means for largescale, rapid monitoring of locusts and is highly effective in scientif ic prevention and control of pests. In addition, continuously encrypted meteorological station data and regional meteorological parameter products formed by the coupling of remote sensing and meteorological data provide a richer source of information for the dynamic monitoring of locust occurrences. This case uses multi-source data, combined with continuous in-depth research on the biological characteristics of migratory pests, their development and diffusion processes, and environmental factors to build a pest monitoring model. Through big data analysis and processing on the Digital Earth Science Platform, quantitative analysis was carried out on the temporal and spatial distribution of the desert locust reproduction and migration raging in the Horn of Africa and
Chapter 2 SDG 2 Zero Hunger
countries in Southwest Asia. The results of this intercontinental monitoring were provided to FAO to support the joint prevention and control of multiple countries and ensure the safety of agricultural production and regional stability.
Data used in the case Remote sensing data: MODIS in Africa and Asia since 2000 (spatial resolution: 500 m, https://ladsweb.modaps.eosdis.nasa.gov/search/), Landsat (spatial resolution: 30 m, https:// earthexplorer.usgs.gov/), Sentinel (spatial resolution: 10 m, https://scihub.copernicus.eu/), Planet (spatial resolution: 3 m), Worldview (spatial resolution: 0.5 m) in representative areas, greenness (http://iridl.ldeo.columbia.edu/maproom/Food_Security/Locusts/Regional/ greenness.html) and rainfall (data source: https://sharaku.eorc.jaxa.jp/GSMaP/). Meteorological data: long-term meteorological data from international meteorological stations from 2000 to present, tropical cyclone data from 2018 to present, and meteorological forecast products (data source: http://data.cma.cn/). Basic geographic information: global land use (spatial resolution: 10 m and 30 m), DEM, crop planting areas in Africa and Asia, and administrative division data. Other data: ground survey data released by the FAO of the United Nations (https://locusthub-hqfao.hub.arcgis.com/), crop calendar (http://www.fao.org/agriculture/seed/cropcalendar/ welcome.do), and so on.
Method An indicator system for monitoring desert locusts with remote sensing was established with quantitative remote sensing extraction and time-series analysis of indicators closely related to the reproduction, development and migration of locusts. This includes pest sources, hosts, and environment. A habitat suitability model was constructed to extract locusts’ core breeding areas on a large scale by integrating multi-source data, e.g., global land use, ground surveys, GIS analysis, geo-statistics, and spatiotemporal data fusion methods and algorithms. Then, locust migration paths were monitored by coupling Earth observation data, meteorological data, crop calendar and planting data with an insect dispersal model. Finally, locust damage distribution and area were extracted based on the vegetation growth curve analysis of the past 20 years along with pest dispersal mechanisms. Also, more detailed monitoring of locust damage was conducted in some severely affected countries, such as damaged vegetation types (crop, grass and shrub), the spatial distribution and total damaged area.
19
20
Big Earth Data in Support of the Sustainable Development Goals (2020): The Belt and Road
Results and analysis In 2018, the Red Sea coast became a core breeding area for desert locusts due to hurricanes and other climatic conditions. Since then, locusts have been breeding and spreading in Yemen, Oman and Somalia. In January 2019, swarms invaded eastern Saudi Arabia and southern Iran. From February to June, locusts on the India-Pakistan border and the spring breeding area of the Arabian Peninsula continued to increase. The unusually long summer monsoon from June to October caused the locusts in the summer breeding area on the India-Pakistan border to continuously hatch and swarm. Then, , the swarms on the India-Pakistan border began to reproduce for three generations and moved to spring breeding areas in southern Iran and northern Oman. At the same time, swarms in the Horn of Africa continued to grow and migrate to southern Somalia and northeastern Kenya due to the influence of climate. From January to February 2020, desert locusts broke out in Ethiopia and Kenya in the Horn of Africa and invaded Uganda and Tanzania, continuing to reproduce along the southeastern coast of Iran. The locusts on the India-Pakistan border then entered the next round of spring reproduction (Figure 2.5).
0
Core breeding area
Breeding time
500
Invading time
1,000 km
Migration path
Figure 2.5 Core breeding area and the migratory path of desert locusts (June 2018-February 2020)
Chapter 2 SDG 2 Zero Hunger
0
Core breeding area
Breeding time
Invading time
500
1,000 km
Migration path
Figure 2.6 The core breeding area and the migratory path of desert locusts (March-July 2020)
At the beginning of March 2020, locust swarms in Kuwait continued to spread to southeastern Iraq, and swarms on the eastern coast of Saudi Arabia spread to the west coast of the United Arab Emirates. At the same time, swarms in southern Ethiopia began to migrate north. In mid-March, immature locust swarms were found on the coast of the Red Sea in southeastern Egypt. At the end of the month, immature locust swarms appeared on the east coast of Djibouti. In April, there was heavy precipitation in eastern Africa and locusts continued to reproduce in spring and mature into groups. Swarms in Ethiopia and Somalia continued to move north, swarms in the northern Arabian Peninsula spread to central Iran, and the density of locusts on the border between Iran and Pakistan continued to increase. In May, locust eggs continued to hatch and reproduce, and by mid to late May, swarms began to migrate from spring breeding areas such as Kenya, Ethiopia and western Pakistan to summer breeding areas such as central Sudan, southwestern Saudi Arabia and the India-Pakistan border. At the end of the month, swarms migrated eastward from the IndiaPakistan border to northern India. From mid to late June to July, locusts in spring breeding areas such as Kenya, Ethiopia and Somalia migrated west or northwest to central Sudan and northeast to the India-Pakistan border for summer breeding. At the same time, locusts from southern Iran migrated eastward into western Pakistan, while locusts from northern India continued to reproduce and spread eastward (Figure 2.6).
21
Big Earth Data in Support of the Sustainable Development Goals (2020): The Belt and Road
22
As of April 2020, desert locusts in Pakistan harmed 2.4174 million ha of vegetation area, seriously affecting wheat crops. The locusts harmed 2.6315 million ha of vegetation in Somalia, mainly cropland, shrubland and grassland. As of May 2020, desert locusts have harmed 5.3977 million ha of vegetation area in Ethiopia. The Great Rift Valley, known as the granary of Ethiopia, has suffered serious losses. The harmed area in Kenya was 6.1045 million ha, mainly located in the Rift Valley Province and the Eastern Province. From February 2019 to May 2020, desert locusts have invaded 20 provinces in Yemen, with a harmed vegetation area of 2.0652 million ha
N Khyber Pakhtunkhwa 2020.02
Lucky Marwat
Dera Ismail Khan 2020.02
2020.01
Okara 2020.02
Bahawalpur
2020.02
Dalbandin
Khuzdar
2020.01
Cholistan
2020.02
2020.02
Nara
Turbat
Tharkarpar
2019.112020.02
0 Core breeding area
Breeding time
Invading time
200
400 km
Migration path
0
200
400 km
Oct.-Dec. 2019 Jan.-Mar. 2020 2020年4月 Apr. 2020 2020年4月 2020年4月 2019年10-12月 2019年10-12月 2019年10-12月 2020年1-3月 2020年1-3月 2020年1-3月
(a) Migratory path
(b) Damage monitoring
Figure 2.7 Desert locust migratory path and damage monitoring in Pakistan (June 2019-April 2020)
N Tigray
Amhara
Afar
Dire Dawa Harar Jijjiga Warder Ogaden Kebridehar Teltele Yabello
0
200
Gode
Core breeding area Breeding time Invading time Migration path
400 km
(a) Migratory path
0
Jun.-Sep. 2019 Oct.-Dec. 2019
200
Jan.-Feb. 2020 Apr.-May. 2020
(b) Damage monitoring
Figure 2.8 Desert locust migratory path and damage monitoring in Ethiopia (June 2019-May 2020)
400 km
Chapter 2 SDG 2 Zero Hunger
23
(Figure 2.7-Figure 2.11). Being verif ied by ground survey data provided by FAO, the locust plague monitoring accuracy was higher than 80%.
N Mandera
Moyale Tukana Marsabit West Pokot Kaped Samburu Laikipia Isiolo Meru
Wajir
Garissa
Kericho
Core breeding area Breeding time Invading time Migration path 150 300 km
Kajiado
0
0
Jan.-Sep. 2019 Oct.-Dec. 2019
200
400 km
Jan.-Feb. 2020 Apr.-Ma.y 2020
(b) Damage monitoring
(a) Migratory path
Figure 2.9 Desert locust migratory path and damage monitoring in Kenya (January-May 2020)
N
Bosaso Borama
Berbera
Hadaftimo Iskushuban
Burao Laascaanood Garowe Boheley
Gaalkacyo Dhuusa Mareeb Huddur
Beled Weyne
Garbaharey
Core breeding area Breeding time Invading time Migration path 0
(a) Migratory path
150
300 km
0
Jun.-Sep. 2019 Oct.-Dec. 2019
150
Jan.-Mar. 2020 Apr. 2020
(b) Damage monitoring
Figure 2.10 Desert locust migratory path and damage monitoring in Somalia (June 2019-April 2020)
300 km
24
Big Earth Data in Support of the Sustainable Development Goals (2020): The Belt and Road
N
Sa'dah
AIJawf
Suq Abs
Marib Sana'a
AlHudaydah Dhamar Ta'tzz Abyan
Wadi Hadramaet Ahabwah Atag Abyan
Aden
0
Core breeding area Breeding time Invading time Migration path 150 300 km
0
Feb.-May. 2019
(a) Migratory path
Jun.-Dec. 2019
150
300 km
Jan.-May. 2020
(b) Damage monitoring
Figure 2.11 Desert locust migratory path and damage monitoring in Yemen (February 2019-May 2020)
Highlights T he core breeding areas and migration paths of desert locusts were monitored from 2018 to 2020 in Asia and Africa, especially in severely damaged countries, including Pakistan, Ethiopia, Kenya, Somalia and Yemen. T he results were adopted by FAO to support multi-country joint pest prevention and control to ensure agricultural production.
Outlook In terms of technological innovation, this case used an international shared remote sensing dataset to conduct systematic research on the extraction of large-scale desert locust breeding areas, longterm quantitative monitoring of locust migration paths and quantitative monitoring of locust plagues through big data analysis and processing on the Digital Earth Science Platform. Then, remote sensing monitoring of the desert locust plague in Africa and Asia was continued to update the damage data. The research results can contribute to the protection of agricultural production and food security and provide important informational support for emergency response to the locust plague. In terms of application and promotion, the FAO adopted the monitoring results of the core breeding area and the migratory path of desert locusts in Asia and Africa from 2018 to 2020, as well as the plague monitoring results in the key countries (Pakistan, Ethiopia, Kenya, Somalia and Yemen). The results provided informational support for multi-country joint prevention and control of pests to sustain agricultural production.
Chapter 2 SDG 2 Zero Hunger
Spatiotemporal patterns of rice paddy dynamics on the Indochina Peninsula Target SDG 2.4: By 2030, ensure sustainable food production systems and implement resilient agricultural practices that increase productivity and production, that help maintain ecosystems, that strengthen capacity for adaptation to climate change, extreme weather, drought, flooding and other disasters and that progressively improve land and soil quality.
Background Rice is one of the main food crops worldwide and the Indochina Peninsula is one of the main rice-producing areas. With suitable climatic conditions and abundant water resources, it is possible to intensively produce rice and plant multiple crops in this region. The high yields and huge carrying capacity of rice-producing areas have in turn supported population growth. The excess rice production in Thailand and Vietnam has enabled these countries to export rice to the rest of the world. Therefore, long-term monitoring of rice cultivation in this region is of great signif icance for assessing the sustainability of the international rice supply. SDG 2.4.1 is def ined as the percentage of agricultural area under productive and sustainable agriculture. Remote sensing is an important means for mapping the spatial distribution of paddy rice and many scholars have carried out research in this f ield. However, paddy rice-specif ic maps are still lacking, though some efforts at paddy rice mapping have been made on the sub-continental scale (Bridhikitti and Overcamp, 2012). Existing global paddy rice maps are generally produced using statistical approaches, but these efforts generally have limited spatial resolution and temporal coverages. There is still no operational paddy rice mapping approach for larger scale or global efforts (Dong and Xiao, 2016). Due to its short growing period, rice can be planted multiple times a year on the Indochina Peninsula. Compared with the statistical data of rice planting area only, a comprehensive expression of the total area of paddy rice planting using rice paddy area and the multiple cropping index (MCI) is more meaningful to accurately describe the sustainability of the rice industry. This case study is an effort to thematically map paddy rice on the Indochina Peninsula from 2000 to 2019 using Big Earth Data to provide data support for monitoring and evaluating SDG 2.4.1. The changes in the total area of paddy rice planting, rice paddy area and paddy MCI were analyzed to evaluate the sustainability of the rice industry on a national scale for the f ive countries on the Indochina Peninsula. For two representative regions, the main causes of change in paddy rice planting patterns were analyzed.
25
26
Big Earth Data in Support of the Sustainable Development Goals (2020): The Belt and Road
Data used in the case 2000-2019 MODIS Normalized Difference Vegetation Index (NDVI) time-series data, 250 m spatial resolution. 2000-2019 MODIS Normalized Difference Water Index (NDWI) time-series data, 500 m spatial resolution. Copernicus Global Land Service (CGLS) global land cover in 2015, 100 m resolution. 2000-2019 Landsat series (Landsat-5/Landsat-8) data covering the Indochina Peninsula, 30 m resolution.
Method To ensure accurate thematic products for paddy rice, it was necessary to extract the cultivated land distribution every f ive years on the Indochina Peninsula from 2000 to 2019. Based on CGLSLC100 land cover products, a classif ication sample database was constructed. Then the samples were f iltered based on the time series vegetation index of Landsat data, and a land cover/use classif ication feature dataset was constructed. Finally, the cultivated land was extracted with high accuracy every f ive years from 2000 to 2019 using object-oriented segmentation and a random forest classif ication algorithm. Based on the distribution of cultivated land, the automatic extraction algorithm of paddy rice MCI was developed in Google Earth Engine. First, the time-series MODIS NDVI/NDWI data were reconstructed. Next, the time-series NDVI data were smoothed, f iltered and compared with the rice samples of different regions for similarity. Lastly, non-paddy rice was removed because, in the rice sowing period, the paddy rice f ields have high NDWI values. This method was applied to the f ive countries on the Indochina Peninsula to extract the spatial distribution and MCI of paddy rice every f ive years from 2000 to 2019. This case analyzed the spatiotemporal dynamics of rice planting patterns on the Indochina Peninsula. According to the characteristics of rice planting and the def inition of SDG 2.4.1, the def inition of a paddy rice planting pattern includes three sub-indicators: total area of paddy rice planting, rice paddy area and paddy MCI: (1) Total area of paddy rice planting: the total paddy rice planting area within a calendar year in the study area. (2) Rice paddy area: the area of cultivated land where paddy rice has been planted within a calendar year in the study area. (3) Paddy MCI: in the study area, the total area of paddy rice planting divided by the rice paddy area times 100%. Based on the spatial distribution of paddy rice and paddy MCI on the Indochina Peninsula from
Chapter 2 SDG 2 Zero Hunger
2000 to 2019, the three sub-indicators were statistically analyzed for the f ive countries, and the long-term evolution of the rice planting pattern was analyzed. In addition, intersection processing was utilized among the multitemporal paddy rice distribution and paddy MCI to extract the change information from 2000 to 2019. To analyze the change information, two types of paddy rice planting change were def ined. (1) Change in rice paddies: For each rice paddy, a change from non-paddy to paddy was def ined as “paddy increase”, a change from paddy to non-paddy was def ined as “paddy decrease”, and paddy remaining as paddy was def ined as “paddy unchanged”. (2) Change in paddy MCI: By comparing the paddy MCIs at different times, the rice paddies were divided into three types: MCI increased, MCI decreased and MCI unchanged.
Results and analysis 1. Rice paddy distribution and paddy MCI mapping on the Indochina Peninsula in 2015 Due to the lack of relevant products for comparison, this case study validated the paddy-specif ic data in two ways. (1) First, a large number of samples were randomly distributed throughout the Indochina Peninsula. By analyzing the sample time series of MODIS NDVI/NDWI and Landsat images in 2015 one by one, the features of each sample were determined, including paddy or non-paddy and paddy MCI. Based on these samples, the accuracy of paddy-specif ic data in 2015 was evaluated. Results showed that the OA of rice paddy distribution was higher than 92%, and that of paddy MCI was higher than 86%. (2) Comparing the statistical results of the paddy-specif ic data in 2015 with the statistical data of the total rice planting area issued by the f ive countries in the same year, results showed that for all f ive countries, the paddy-specif ic data had an error of less than 10%. The distribution of rice paddies on the Indochina Peninsula in 2015 is shown in Figure 2.12. Rice paddies on the Indochina Peninsula during the study period were mainly distributed in the regions of the Red River Delta and Mekong River Delta in Vietnam, the Khorat Plateau and Chao Phraya Delta in Thailand, and the Irrawaddy Delta in Myanmar. 2. Changes in paddy rice planting patterns on the Indochina Peninsula from 2000 to 2019 The changes in paddy-specif ic data in the f ive countries on the Indochina Peninsula from 2000 to 2019 are shown in Figure 2.13. The rice paddy area in Vietnam decreased slightly, but the paddy MCI increased greatly, from 178 to 191, which made the total area of paddy planting increase slightly. Both the rice paddy area and the total area of paddy planting in Thailand showed a signif icant decrease, and the paddy MCI changed slightly, from 121 to 125. The change characteristics of Myanmar and Cambodia were similar, for which the rice paddy area, the total area of paddy planting and the paddy MCI showed an increasing trend. The paddy MCI increased
27
28
Big Earth Data in Support of the Sustainable Development Goals (2020): The Belt and Road
from 109 to 121 and from 112 to 124, respectively, in Myanmar and Cambodia. In Laos, both the rice paddy area and the total area of paddy planting showed growth trends with slight increases, while the paddy MCI was slightly decreased. The spatial distribution of paddy planting pattern changes on the Indochina Peninsula from 2000 to 2019 are shown in Figure 2.14. The several main rice paddy distribution areas on the Indochina Peninsula demonstrated different change characteristics: In the Red River Delta, the paddy MCI changed little, but the rice paddy area decreased to a certain extent; In the Mekong River Delta, the paddy MCI increased with a relatively high proportion; In the Khorat Plateau, the decrease in paddy MCI was mainly due to the decrease in the rice paddy area; Both in the Chao Phraya Delta and Irrawaddy Delta, the paddy MCI showed an obvious increase. 3. Discussion of the paddy planting pattern changes This case selected two representative areas for specif ic discussion of paddy planting pattern changes, the Mekong River Delta in Vietnam and Rohingya settlements in the Rakhine State of Myanmar, as shown in Figure 2.15. N
Paddy area as percent of land area/% Myanmar
11.06
Thailand
16.44
Laos Cambodia Vietnam
3.29 19.47 13.87
Paddy area
Non Paddy area
Legend Paddy area
0
245
490
Non paddy area
Figure 2.12 Rice paddy distribution in the f ive countries on the Indochina Peninsula in 2015
980 km
Chapter 2 SDG 2 Zero Hunger
50,000
128
40,000
120
30,000
112
20,000
104
10,000 0
2000
2005
2010 2015 Year (a) Cambodia
2019
96
10,000
128
10,000
128
8,000
120
8,000
120
6,000
6,000
112
4,000 2,000 0
2000
2005
2010 2015 Year (b) Myanmar
112
4,000
104
104
2,000 0
2019
2000
2005
2010 Year (c) Laos
2015
2019
96
15,000
128
100,000
200
12,000
120
75,000
192
112
50,000
184
104
25,000
176
9,000 6,000 3,000 0
2000
2005
2010 2015 Year (d) Thailand
2019
0
96
Total area of paddy planting
2000
Paddy f ield area
2005
2010 2015 Year (e) Vietnam
168
2019
Paddy Ma
Figure 2.13 Changes in the triple-paddy planting pattern sub-indicators in the f ive countries on the Indochina Peninsula (2000-2019) N
Legend
N
Paddy unchanged Paddy minus Paddy increased
0
250
500
(a) Changes in rice paddy distribution
1,000 km
Legend MCI unchanged MCi minus MCI increased
0
250
500
1,000 km
(b) Changes in paddy MCI
Figure 2.14 Spatial distribution of paddy planting pattern changes on the Indochina Peninsula
29
Big Earth Data in Support of the Sustainable Development Goals (2020): The Belt and Road
30
N
N
Legend MCI minus MCI increased MCI unchanged
0
25 50
(a) Mekong River Delta
100 km
Legend Paddy minus Paddy increased Paddy unchanged
0 5 10
20 km
(b) Rohingya settlements in Myanmar
Figure 2.15 Representative areas of paddy planting pattern changes on the Indochina Peninsula (2000-2019)
The change in paddy planting patterns in the Mekong River Delta is shown in Figure 2.15(a). Comparing the distribution of paddy MCI in 2000 and 2019, it was found that the proportion of triple-cropping of paddy rice and two-year seven-cropping of paddy rice in this region increased signif icantly, with the proportion of triple-cropping rising from 15.9% to 32.78% and the proportion of two-year seven-cropping rising from 1.33% to 2.81%. In addition, the paddy MCI in this region rose from 185 in 2000 to 214 in 2019. Preliminary analysis showed that climate warming and the increase in precipitation were the main influencing factors of the paddy planting pattern changes in this region. Figure 2.15(b) shows the change in paddy planting patterns in Rohingya settlements in Rakhine. Comparing the Rohingya settlements with other areas in Rakhine and neighboring Bangladesh, the proportion of the rice paddy decrease was signif icantly higher, reaching more than 80%. Preliminary analysis showed that human activity is the main factor of the paddy planting pattern changes in this region. In 2017, the Rohingya crisis broke out and continues to this day. During this period, more than 700,000 Rohingya refugees have fled from Rakhine to Bangladesh. The decline in the agricultural population and continued unrest resulted in a large number of rice paddies being abandoned.
Chapter 2 SDG 2 Zero Hunger
Highlights A dataset of rice paddy distribution and planting patterns with a resolution of 500 m in the Indochina Peninsula was produced every f ive years from 2000 to 2019, which can provide spatial data support for SDG 2.4.1 and tackle the problem of missing data. T he temporal and spatial dynamics of rice paddy planting patterns were analyzed in f ive countries on the Indochina Peninsula. From 2000 to 2019, the multiple cropping index of rice paddy planting in the f ive countries showed a signif icant increase. T aking the Mekong Delta in Vietnam and the Rohingya settlements in Rakhine State of Myanmar as examples, the impact of climate warming and human activity on rice paddy planting patterns were illustrated.
Outlook This case study produced an internationally shared, paddy-specif ic dataset on the Indochina Peninsula from 2000 to 2019, proposed an SDG 2.4.1 evaluation index system suitable for paddy rice, and analyzed the temporal and spatial changes in the paddy planting patterns on the Indochina Peninsula. Analysis at the country level demonstrates the differences in paddy planting pattern changes in different countries, and the main influencing factors of the changes were analyzed for two representative regions. This case provided important data support for the realization of SDG 2.4. In the future, based on the paddy-specif ic data, and combined with multi-source data such as the population, economy, policy, night lighting, urban development and climate change, the causes of paddy planting pattern changes on the whole Indochina Peninsula and sub-regions will be explored more deeply. The method will be further improved and applied to other major paddy rice-producing areas in the world. The paddy-specif ic products will be released and updated annually. In addition, the abundance of high-resolution data, especially SAR data, can overcome the diff iculty in obtaining optical images in paddy rice-producing areas, and greatly improve the accuracy of rice distribution products, which will also be an important development direction of this case in the future.
31
32
Big Earth Data in Support of the Sustainable Development Goals (2020): The Belt and Road
Cultivated Land productivity dataset and its application in Northeast Eurasia Target SDG 2.4: By 2030, ensure sustainable food production systems and implement resilient agricultural practices that increase productivity and production, that help maintain ecosystems, that strengthen capacity for adaptation to climate change, extreme weather, drought, flooding and other disasters and that progressively improve land and soil quality.
Background Cultivated land is a basic resource for human survival, and crop production capacity, as the most important feature of cultivated land, is the foundation of food security (Duan et al., 2012) and sustainable agriculture. China is one of the most important food producers in the world. In 2015, China’s cereal production accounted for 22.8% of the world’s total, and in the future, China’s crop production will continue to grow. After entering the 21st century, China has paid more attention to coordinating the use of international and domestic markets and resources, and has formulated a “go global” strategy for agriculture. Also, China has strengthened agricultural cooperation with some countries involved in the Belt and Road Initiative and some neighboring countries. The next decade will be an important stage for China to transform from a major agricultural product trading country to a major agricultural investment country. SDG 2.4.1 is def ined as the proportion of agricultural area under productive and sustainable agriculture. This goal proposes that the current production system, and policies and institutions that support global food security are insuff icient. Prompt and accurate crop yield estimation can provide important information for decision-makers and be used to improve crop management and control the market to maintain regional and even world food security. Regarding SDG 2.4.1, Northeast Eurasia was selected as the study area and the main work includes: (1) researching crop classif ication and yield estimation methods suitable for the study area; (2) observing crop types and yields in the study area over the past 20 years based on Big Earth Data; and (3) analyzing the temporal and spatial variations of cultivated land use patterns and cultivated land productivity levels in the study area over the past 20 years and the differences between China and neighboring countries. Aiming at SDG 2.4.1 (proportion of agricultural area under productive and sustainable agriculture), a long-term time series of f ield crop planting types and yield levels were used for quantitatively evaluating the target.
Chapter 2 SDG 2 Zero Hunger
Data used in the case Landsat images from April to October in 2000, 2005, 2010, 2015 and 2019, respectively, provided by the United States Geological Survey (USGS), with a spatial resolution of 30 m. 2000-2019 Normalized Difference Vegetation Index (NDVI) data, provided by USGS, with a spatial resolution of 30 m and a time resolution of eight days. Products of six-hour solar radiation data from April to October in 2000, 2005, 2010, 2015 and 2019, respectively, provided by the National Oceanic and Atmospheric Administration (NOAA). Meteorological data from April to October in 2000, 2005, 2010, 2015 and 2019, perspectively, provided by the European Centre for Medium-Range Weather Forecasts (ECMWF).
Method The Google Earth Engine JavaScript API was used to perform radiometric calibration, geometric correction, cloud/snow/shadow mask, minimum cloud image synthesis and other operations on a long-term time-series of Landsat images. The pixel-level minimum cloud cover image synthesis method was adopted to make full use of the image information of the same season in the study area and overcome the influence of clouds. Then the full Landsat data of the target year was selected. The Landsat image data of the study area for the entire growth period of 2000, 2005, 2010, 2015 and 2019 were obtained and preprocessed as image mosaics on Google Earth Engine, where radiation correction and cloud removal were automatically performed. The crop classif ication process analyzed the impact of crop classif ication characteristics on a parcel-oriented classif ication method and solved the problem of the number of samples and mixed pixels in crop classif ication, improving the parcel-oriented method of classif ication. First, highresolution remote sensing images of the study area were obtained along with f ield boundaries through methods of image segmentation and manual intervention. Then the images of the crop planting area were obtained according to the recognition results of the planted/non-planted area. Different classif ication features and their combinations for crop classif ication were analyzed using Maximum Likelihood Classification (MLC), Support Vector Machine (SVM) and Neural Network (NN) supervised classif ication methods to classify crops and select the most suitable method for the research area. Then, based on this method, the planted area was mapped. Yield estimation was based on crop growth models and data assimilation algorithms from three different perspectives, including model mechanism supplementation, multi-model coupling and model algorithm improvement. First, the World Food Studies (WOFOST) model was calibrated based on a large number of f ield campaigns in the study area over many years. Then the crop classif ication methods were applied using the Google Earth Engine JavaScript API to precisely
33
Big Earth Data in Support of the Sustainable Development Goals (2020): The Belt and Road
34
estimate the crop yield with the improved CASA-WOFOST coupled model and the Ensemble Kalman Filter (EnKF) method. Finally, the accuracy of the simulation results was evaluated against f ield observation data.
Results and analysis 1. Datasets of cultivated land types and yields in Northeast Eurasia from 2000 to 2019 The datasets of crop types and yields in the study area from 2000 to 2019 are shown in Figure 2.16 and Figure 2.17. 2. Interannual change and difference analysis of crop yields The interannual changes and differences in crop yields in Northeast Eurasian countries from 2000 to 2019 are shown in Figure 2.18. From 2000 to 2019, China’s soybean and rice yields increased, peaked in 2015, and then slightly decreased in 2019, while maize yields increased from 2000 to 2019. Mongolian yields were generally low and relatively high in 2010. In Russia, the soybean yield fluctuated greatly from 2000 to 2019, and reached the highest in 2015. The rice yield and the maize yield f irst rose and then declined. The crop yields of the Democratic People’s Republic of Korea (DPRK) and the Republic of Korea (ROK) were both low from 2000 to 2010, and the crop yields in 2015 and 2019 had a clear upward trend compared with previous years. 3. Analysis of average crop yields in Northeast Eurasian countries from 2000 to 2019 The differences in average crop yields across countries in Northeast Eurasia from 2000 to 2019 are shown in Figure 2.19. Russia had the highest average crop yield, followed by China, and Mongolia had the lowest average yield from 2000 to 2019. 4. Comparative analysis of the production level of cultivated land in the countries of Northeast Eurasia from 2000 to 2019 To analyze the differences in the level of cultivated land productivity and interannual changes in the study area more accurately, the yields of soybean, rice and maize in the study area were normalized, and the three crops were compared for analysis. The normalization method used the 95% lower-side value and 5% lower-side value of the normalized images as the maximum and minimum yields. 1) Differences in the level of cultivated land productivity and interannual changes in different countries The interannual changes in the level of cultivated land productivity in the countries of Northeast Eurasia are shown in Figure 2.20. The productivity of China’s cultivated land increased from 2000 to 2019, peaking in 2015 and slightly declining in 2019. Mongolia rose f irst, peaked in 2010 and showed a downward trend after 2010. The cultivated land productivity of Russia was stable from 2000 to 2019. The productivity of cultivated land in DPRK and ROK was low from 2000 to 2010, and there was a clear upward trend in 2015 and 2019 compared with previous years.
Chapter 2 SDG 2 Zero Hunger
N
N
(a)2000
(a) 2000
N
N
(b)2010 (b) 2010
N N
(c)2019 (c) 2019 Legend Soybean Maize
Rice Others
0
500
1,000 km
Figure 2.16 Spatial distribution of crop types in Northeast Eurasia (2000-2019)
Legend Maize/(kg/ha) >7,000 6,000-7,000 Rice/(kg/h) >8,000 7,000-8,000 Soybean/(kg/ha) >2,500 2,000-2,500
0
500
1,000 km
5,000-6,000
4,000-5,000
90.00%. In addition, 300 m × 300 m blocks were randomly selected at the global scale over continents in order to validate the global impervious surface products in 2015 and 2018. The percentage of impervious surfaces in the blocks was calculated and then verif ied with visual interpretation using high-resolution remote sensing imagery. There were 4,000 blocks selected for 2015 and 4,000 selected for 2018. The results showed that the average R2 was 0.84 in 2015 and 0.79 in 2018. By using the two verif ication methods, random point and block percentage, the results showed that the products were highly accurate, which demonstrates the effectiveness of the proposed methods for extracting urban impervious surfaces on a global scale based on SAR and optical image fusion. Figure 4.11 shows the 10 m impervious surface for 2018 in China and local detailed maps in 2018. In summary, our proposed methods and production can provide high-resolution spatial data with high accuracy to support decision-making for the Sustainable Development Goals. 2. Global urbanization monitoring and evaluation In this study, the LCRPGR indicator was calculated for each continent (Figure 4.12). The overall distribution of LCRPGR is roughly between 0 and 1.5. The indicator values of Asia, Oceania and South America were mainly between 0.5 and 1, while the values of Europe and North America were between 1 and 1.5. The overall indicator for Africa was low, with a value of less than 0.5. The study reveals that global urbanization tends to be steadily under development, but population urbanization is slightly faster than land urbanization for most regions. This is especially apparent in Africa, where population urbanization is far greater than land urbanization due to economic and geographic environmental factors. This presents a great challenge in achieving the 2030 Agenda for Sustainable Development.
121
Big Earth Data in Support of the Sustainable Development Goals (2020): The Belt and Road
122
In South America and Oceania, land consumption basically meets the urbanization needs of the population. The gap in the development level among most countries is relatively small, while the urbanization developments in Europe and the United States are the most stable, with land consumption meeting the needs of population urbanization. According to the LCRPGR indicators on six continents, although global urbanization is steadily under development, the LCRPGR values are outside of [0.5, 1.5] for nearly 50% of countries according to national statistics.
・
N
Legend PSA ISA
0
500
1,000
2,000 km
NANHAI ZHUDAO
Figure 4.11 ISA 2018: Spatial distribution and local detailed maps of impervious surfaces in Chinese cities
0
10
20
30
40
50
60
N
0
3,050
6,100
Figure 4.12 Spatial distribution of global LCRPGR at the continental scale from 2015 to 2018
1.09
0.98
0.66
0.63
0.55
0.44
No dala
Legend LCRPGR by Continent
12,200 km
Chapter 4 SDG 11 Sustainable Cities and Communities
123
>2.5 (2-2.5] (1.5-2] (1-1.5] (0.5-1] (0-0.5]