Handbook on City Logistics and Urban Freight 9781800370173, 1800370172

Providing an up-to-date and comprehensive overview of city logistics and urban freight research, this Handbook offers mu

185 61

English Pages 521 Year 2023

Report DMCA / Copyright

DOWNLOAD PDF FILE

Table of contents :
Front Matter
Copyright
Contents
Contributors
Abbreviations
Keywords and definitions
Introduction to the Handbook on City Logistics and Urban Freight
1 The challenges of freight transport in cities
2 Integrated transportation and land-use program to improve metropolitan freight system performance
Section I Modelling and Simulation
3 Overview of urban freight transport modelling
4 Estimating and forecasting urban freight origin-destination flows
5 Evaluating city logistics solutions with agent-based microsimulation
6 Freight trip generation models: using establishment data to understand the origin of urban freight traffic
Section II Logistics and Operations
7 Overview of city logistics and urban freight transport operations
8 Urban freight consolidation and delivery: state of the art
9 Towards more sustainable vehicles for the last mile? Cycle logistics as a part of the solution
10 Operations research for planning and managing city logistics systems
Section III Planning and Policy Making
11 Overview of urban freight transport planning and European suggestions
12 Land-use planning for a more sustainable urban freight
13 Assessment of innovative city logistics solutions
14 Planning for the future: urban freight transportation
Section IV Stakeholder Engagement, Public/Private Partnerships
15 Overview on stakeholder engagement
16 Participatory decision-support tools for stakeholder engagement in urban freight transport policy making
17 Living labs for transitions in urban freight transport systems
18 Urban freight transport and multi-level governance
Section V Innovation, Digitalization, and Data
19 Overview of innovations in urban freight
20 Hyperconnected city logistics: a conceptual framework
21 E-commerce and urban logistics: trends, challenges, and opportunities
22 New technologies and autonomous vehicles for urban goods distribution
Section VI Urban Freight Transport Sustainability
23 Environmentally sustainable city logistics: minimising urban freight emissions
Index
Recommend Papers

Handbook on City Logistics and Urban Freight
 9781800370173, 1800370172

  • 0 0 0
  • Like this paper and download? You can publish your own PDF file online for free in a few minutes! Sign Up
File loading please wait...
Citation preview

Handbook on City Logistics and Urban Freight

Edoardo Marcucci Valerio Gatta Michela Le Pira

HANDBOOK ON CITY LOGISTICS AND URBAN FREIGHT

RESEARCH HANDBOOKS IN TRANSPORT STUDIES This important and timely series brings together critical and thought-provoking contributions on the most pressing topics and issues in transport studies. Comprising specially commissioned chapters from leading academics, these comprehensive Research Handbooks feature cutting-edge research, help to define the field and are written with a global readership in mind. Equally useful as reference tools or high-level introductions to specific topics, issues, methods, innovations and debates, these Handbooks will be an essential resource for academic researchers and postgraduate students in transport studies and related disciplines. For a full list of Edward Elgar published titles, including the titles in this series, visit our website at www.e-elgar.com.

Handbook on City Logistics and Urban Freight Edited by

Edoardo Marcucci Professor of Transport Economics, Department of Political Sciences, Roma Tre University, Italy and Molde University College, Norway

Valerio Gatta Associate Professor of Transport Economics, Department of Political Sciences, Roma Tre University, Italy and Molde University College, Norway

Michela Le Pira Research Fellow and Lecturer of Transport Engineering, Department of Civil Engineering and Architecture, University of Catania, Italy

RESEARCH HANDBOOKS IN TRANSPORT STUDIES

© Edoardo Marcucci, Valerio Gatta and Michela Le Pira 2023 All rights reserved. No part of this publication may be reproduced, stored in a retrieval system or transmitted in any form or by any means, electronic, mechanical or photocopying, recording, or otherwise without the prior permission of the publisher. Published by Edward Elgar Publishing Limited The Lypiatts 15 Lansdown Road Cheltenham Glos GL50 2JA UK Edward Elgar Publishing, Inc. William Pratt House 9 Dewey Court Northampton Massachusetts 01060 USA

A catalogue record for this book is available from the British Library Library of Congress Control Number: 2023905631 This book is available electronically in the Geography, Planning and Tourism subject collection http://dx​.doi​.org​/10​.4337​/9781800370173

ISBN 978 1 80037 016 6 (cased) ISBN 978 1 80037 017 3 (eBook)

EEP BoX

Contents

viii x xv

List of contributors List of abbreviations List of keywords and definitions

Introduction to the Handbook on City Logistics and Urban Freight 1 Edoardo Marcucci, Valerio Gatta, and Michela Le Pira 1

The challenges of freight transport in cities Genevieve Giuliano

2

Integrated transportation and land-use program to improve metropolitan freight system performance José Holguín-Veras, Carlos Rivera-González, Benjamin Caron, Julia Coutinho Amaral, and Abdelrahman Ismael

11

35

SECTION I   MODELLING AND SIMULATION 3

Overview of urban freight transport modelling Lóri Tavasszy and Michiel de Bok

60

4

Estimating and forecasting urban freight origin–destination flows Antonio Comi and Paolo Delle Site

78

5

Evaluating city logistics solutions with agent-based microsimulation Takanori Sakai, Peiyu Jing, André Romano Alho, Ravi Seshadri, and Moshe Ben-Akiva

98

6

Freight trip generation models: using establishment data to understand the origin of urban freight traffic Ivan Sánchez-Díaz and Juan Pablo Castrellon

115

SECTION II   LOGISTICS AND OPERATIONS 7

Overview of city logistics and urban freight transport operations Eiichi Taniguchi, Russell G. Thompson, and Ali G. Qureshi

141

8

Urban freight consolidation and delivery: state of the art Maria Björklund and Britta Gammelgaard

160

9

Towards more sustainable vehicles for the last mile? Cycle logistics as a part of the solution Philippe Lebeau, Bart Cok, Clarissa Kees, and Cathy Macharis v

178

vi  Handbook on city logistics and urban freight

10

Operations research for planning and managing city logistics systems Teodor Gabriel Crainic, Jesus Gonzalez Feliu, Nicoletta Ricciardi, Frédéric Semet, and Tom Van Woensel

190

SECTION III   PLANNING AND POLICY MAKING 11

Overview of urban freight transport planning and European suggestions Francesco Russo and Antonio Comi

225

12

Land-use planning for a more sustainable urban freight Laetitia Dablanc

246

13

Assessment of innovative city logistics solutions Paolo Delle Site

267

14

Planning for the future: urban freight transportation Daniel Haake

287

SECTION IV  STAKEHOLDER ENGAGEMENT, PUBLIC/PRIVATE PARTNERSHIPS 15

Overview on stakeholder engagement Michael Browne and Anne Goodchild

16

Participatory decision-support tools for stakeholder engagement in urban freight transport policy making Michela Le Pira, Edoardo Marcucci, Valerio Gatta, Matteo Ignaccolo, and Giuseppe Inturri

311

327

17

Living labs for transitions in urban freight transport systems Hans Quak, Nina Nesterova, and Giacomo Lozzi

346

18

Urban freight transport and multi-level governance Lisa Hansson

365

SECTION V   INNOVATION, DIGITALIZATION, AND DATA 19

Overview of innovations in urban freight M. Jaller, A. Pahwa, C. Otero-Palencia, and E. Pourrahmani

382

20

Hyperconnected city logistics: a conceptual framework Teodor Gabriel Crainic, Walid Klibi, and Benoit Montreuil

398

21

E-commerce and urban logistics: trends, challenges, and opportunities Valerio Gatta, Edoardo Marcucci, and Michela Le Pira

422

22

New technologies and autonomous vehicles for urban goods distribution Daniela Paddeu

444

Contents 

vii

SECTION VI   URBAN FREIGHT TRANSPORT SUSTAINABILITY 23 Index

Environmentally sustainable city logistics: minimising urban freight emissions Alan McKinnon

463 483

Contributors

André Romano Alho, Singapore-MIT Alliance for Research and Technology, Singapore Julia Coutinho Amaral, Rensselaer Polytechnic Institute, USA Moshe Ben-Akiva, Massachusetts Institute of Technology, USA Maria Björklund, Linköping University, Sweden Michael Browne, University of Gothenburg, Sweden Benjamin Caron, Rensselaer Polytechnic Institute, USA Juan Pablo Castrellon, Chalmers University of Technology, Sweden and Universidad Nacional de Colombia, Colombia Bart Cok, Vrije Universiteit Brussel, Belgium Antonio Comi, University of Rome Tor Vergata, Italy Teodor Gabriel Crainic, Université du Québec à Montréal & CIRRELT, Canada Laetitia Dablanc, University Gustave Eiffel/IFSTTAR, France Michiel de Bok, TU Delft, the Netherlands Paolo Delle Site, Niccolò Cusano University, Italy Jesus Gonzalez Feliu, Excelia Business School, CERIIM, France Britta Gammelgaard, Copenhagen Business School, Denmark Valerio Gatta, University of Roma Tre, Italy and Molde University College, Norway Genevieve Giuliano, University of Southern California, USA Anne Goodchild, University of Washington, USA Daniel Haake, HDR/Cambridge Systematics, Inc., USA Lisa Hansson, Molde University College, Norway José Holguín-Veras, Rensselaer Polytechnic Institute, USA Matteo Ignaccolo, University of Catania, Italy Giuseppe Inturri, University of Catania, Italy Abdelrahman Ismael, Rensselaer Polytechnic Institute, USA M. Jaller, University of California, Davis, USA Peiyu Jing, Massachusetts Institute of Technology, USA viii

Contributors 

ix

Clarissa Kees, Vrije Universiteit Brussel, Belgium Walid Klibi, Kedge Business School & CESIT, France Philippe Lebeau, Vrije Universiteit Brussel, Belgium Michela Le Pira, University of Catania, Italy Giacomo Lozzi, Roma Tre University, Italy Cathy Macharis, Vrije Universiteit Brussel, Belgium Edoardo Marcucci, University of Roma Tre, Italy and Molde University College, Norway Alan McKinnon, Kuehne Logistics University, Germany Benoit Montreuil, Georgia Institute of Technology, USA Nina Nesterova, Breda University of Applied Sciences, the Netherlands C. Otero-Palencia, University of California Davis and Universidad del Norte, USA Daniela Paddeu, University of the West of England, UK A. Pahwa, University of California, Davis, USA E. Pourrahmani, University of California, Davis, USA Hans Quak, TNO and Breda University of Applied Sciences, the Netherlands Ali G. Qureshi, Kyoto University, Japan Nicoletta Ricciardi, Sapienza, Università di Roma & CIRRELT, Italy Carlos Rivera-González, Rensselaer Polytechnic Institute, USA Francesco Russo, Mediterranea University of Reggio Calabria, Italy Takanori Sakai, Tokyo University of Marine Science and Technology, Japan Ivan Sánchez-Díaz, Chalmers University of Technology, Sweden Frédéric Semet, Université Lille, CNRS, Inria, Centrale Lille, UMR 9189 CRIStAL, F-59000 Lille, France Ravi Seshadri, Technical University of Denmark, Denmark Eiichi Taniguchi, Kyoto University, Japan Lóri Tavasszy, TU Delft, the Netherlands Russell G. Thompson, The University of Melbourne, Australia Tom Van Woensel, Eindhoven University of Technology, the Netherlands

Abbreviations

2E-LRTW two echelon-location routing problem with time windows 2E-VRP two echelon-vehicle routing problem ADA aggregate-disaggregate-aggregate simulation approach autonomous delivery robot ADR analytic hierarchy process AHP artificial intelligence AI Akaike information criterion AIC Alliance for Logistics Innovation through Collaboration in Europe ALICE analysis of variance ANOVA Atelier Parisien d’Urbanisme APUR American Trucking Research Institute ATRI American Time Use Survey ATUS autonomous vehicle AV automated vehicle monitoring AVM B2B business-to-business B2C business-to-consumer business as usual BAU battery electric vehicle BEV Business Improvement Districts BID battery swap stations BSS Board of Zoning Appeals BZA C2C consumer-to-consumer connected and autonomous vehicle CAV CB cargo-bicycles Central Business District CBD community-based platforms CBP courier, express and parcel CEP CINASPIC Constructions et Installations Nécessaires Aux Services Publics ou d’Intérêt Collectif city logistics analysis and simulation support system CLASS compressed natural gas CNG carbon dioxide CO2 CP collection-points CRLC Columbus Region Logistics Council CVRS computerised vehicle routing and scheduling D2D door-to-door DC distribution center DLC delivery to/from local carriers DNA deoxyribonucleic acid DSS decision support system x

Abbreviations 

E2EVRP electric two echelon-vehicle routing problem E5 5% ethanol mix in petrol EEP exclusive or private platforms EFV electric freight vehicle EICSP Equipement d’Intérêt Collectif et Services Publics ELECTRE ELimination Et Choix Traduisant la REalité EOQ economic order quantity EPA Environmental Protection Agency EU European Union EVRPTW electric vehicle routing problem with time windows FAC Freight Advisory Committee FAST Act Fixing America’s Surface Transportation Act FASTGS freight and service activity trip generation software FCD floating car data FDM freight demand management FELU freight-efficient land uses FG freight generation FHWA Federal Highway Administration FIS freight-intensive sector FMaaS freight mobility as a service FMCSA Federal Motor Carrier Safety Administration FMTP Florida Freight Mobility and Trade Plan FQP freight quality partnership FREMIS freight market interactions simulation FRP facility location problem FT freight transport FTA freight trip attraction FTG freight trip generation FTP freight trip production functional urban area FUA GA genetic algorithms GDP Gross Domestic Product greenhouse gas GHG GLIM generalized linear models green location routing problem GLRP GPS global positioning system GVW gross vehicle weight GVZ GüterVerkehrsZentrum HCL hyperconnected city logistics HDT heavy-duty trucks HDV heavy-duty vehicle HGV heavy-goods vehicle ICE internal combustion engine ICT information and communication technology IoT Internet of Things IRR internal rate of return

xi

xii  Handbook on city logistics and urban freight

ISIC international standard industrial classification of all economic activities IT information technology ITF International Transport Forum ITS intelligent transportation system JKP joint knowledge production KIPDA Kentuckiana Regional Planning and Development Agency KPI key performance indicator KSIC Korean standard statistical classification kWh kilowatt hours LAET Laboratoire Aménagement Economie Transport LCV light commercial vehicle LEFV low emission freight vehicle LGV light-goods vehicle LLL Logistics Living Lab LPG liquid petroleum gas LRP location routing problem LRPTW location routing problem with time windows LSP logistics service providers LTG large traffic generator LUTI land use transport interaction MaaS mobility as a service MAC marginal abatement cost MACBETH measuring attractiveness by a categorical based evaluation technique MAMCA multi-actor multi-criteria analysis MAS multi-agent simulation MAS-ADP multi-agent simulation-adaptive dynamic programming Mass-GT multi-agent simulation system for goods transport MATSim multi-agent transport simulation MAVT multi-attribute value theory multiple classification analysis MCA MCC municipal consolidation center MH micro-hubs Maritime Industrial Zoning Overlay District MIZOD MnDOT Minnesota Department of Transportation multinomial logistic regression MNL MORPC Mid-Ohio Regional Planning Commission MPO Metropolitan Planning Organization MTB multi-tenant building MTC Metropolitan Transportation Commission MVP minimum viable product NACE nomenclature des Activités Économiques dans la Communauté Européenne (classification of economic activities in the European community) NAF nomenclature d’activités française (French classification of activities) NAICS North American industry classification system NHFN National Highway Freight Network NHS National Highway System

Abbreviations 

xiii

NHTS National Household Travel Survey NHTSA National Highway Safety Administration NIC national industrial classification NMFN National Multimodal Freight Network non-FIS non-freight-intensive sectors NOx nitrogen oxides or nitrogenous oxides NPV net present value NS Norfolk Southern Railroad NSFP National Freight Strategic Plan Orientation d’Aménagement et de Programmation OAP O/D origin-destination Ohio Department of Transportation ODOT on-demand delivery platforms ODP Organisation for Economic Co-operation and Development OECD off-hours delivery OHD ordinary least square OLS physical internet PI Plan Local de Mobilité PLM Plan Local d’Urbanisme PLU particulate matter PM particulate matter smaller than 10 microns PM10 particulate matter smaller than 2.5 microns PM2.5 parts per million Ppm Planning and Operations Language for Agent-based Regional Integrated POLARIS Simulation public-private partnerships PPP PROMETHEE Preference Ranking Organization METHod for Enrichment Evaluations pollution routing problem PRP research & development R&D roadway autonomous delivery robots RADR radio frequency identification RFID Rickenbacker Infrastructure Coordinating Committee RICC reinforcement learning RL root mean square error RMSE Regional Plan Association RPA Swedish Office of Statistics SCB social benefit-cost analysis SCBA SCM supply chain management SCOT Schéma de Cohérence Territoriale SDRIF Schéma Directeur de la Région Ile-de-France SEA strategic environmental assessment SELA Southeast Los Angeles SEM structural equation model SIS service-intensive sector SKU stock keeping unit SMILE Strategic Model for Integrated Logistic Evaluations

xiv  Handbook on city logistics and urban freight

SNI SPM SRADDET

standard industrial classification suspended particulate matter Schéma Régional d’Aménagement, de Développement Durable et d’Egalité des Territoires SSA Shared Situational Awareness SSP shipping and shopping platforms State DOT State Department of Transportation SULP sustainable urban logistics plan SUMP sustainable urban mobility plan triple access planning TAP triple access system TAS traffic analysis zone TAZ total cost of ownership TCO Trans-European Transport Network TEN-T Transport for London TfL traffic impact analysis TIA transhipment point TP Transportation System Management Operations TSMO Tokyo Sky Tree Town TSTT transit-oriented development TOD unmanned aerial vehicles UAV urban consolidation center UCC urban freight lab UFL urban freight system UFS urban freight transport UFT urban freight transport plan UFTP Urbanisée de Grands Services Urbains UGSU ULLTRA-SIM Urban Logistics Land-use and Traffic Simulator United Nations Framework Convention on Climate Change UNFCCC United States US United States Department of Transportation USDOT very-heavy-goods vehicle VHGV vehicle kilometers traveled VKT vehicle miles traveled VMT volatile organic compounds VOC vehicle routing problem VRP vehicle routing problem with multiple time windows VRPMTW VRPTW vehicle routing problem with time windows WHO World Health Organisation WTT well-to-tank WTW wheel to wheel ZEV zero emission vehicles ZINB zero-inflated negative binomial

Keywords and definitions

ADA: Aggregate-Disaggregate-Aggregate simulation approach, applied in different strategic multimodal freight transport demand models in North-West Europe (definition by Tavasszy and de Bok, Chapter 3). agent-based modelling: modelling systems composed of autonomous, interacting agents to simulate the dynamics of complex systems and complex adaptive systems (definition by Sakai et al., Chapter 5). autonomous delivery robots (ADRs): small autonomous or automated (e.g., connected to a central control unit/system) robots, which can travel on the pavement or pedestrian areas, and avoid obstacles (e.g., other road users, physical constraints such as stairs or other objects) autonomously (definition by Paddeu, Chapter 22). Autonomous Vehicles (AVs): vehicles that can completely or partially autonomously drive themselves on existing roads using their own internal intelligent system. Also known as “selfdriving vehicles” or “driverless vehicles” (definition by Paddeu, Chapter 22, and Gatta et al., Chapter 21) commodity-based models: models representing goods transport demand in terms of goods quantity (e.g., weight, volume) (definition by Comi and Delle Site, Chapter 4). cost-benefit analysis: methodology aiming to provide the economic value of a project (definition by Delle Site, Chapter 13). crowdshipping: it is one of the most promising solutions that foresees an integration of passenger and freight mobility. In line with the sharing economy, as a declination of the “crowd sourcing” concept applied to the field of logistics, crowdshipping implies delivering goods using the crowd making use of modern information communication technologies (definition by Gatta et al., Chapter 21) cycle logistics: cycle logistics involves getting unneeded motor vehicles off the roads by using more cycles for goods transport in urban centres (definition by Lebeau et al., Chapter 9). decarbonization: process of reducing carbon emissions from human activity (definition by McKinnon, Chapter 23). delivery-based models: models representing goods transport demand in terms of deliveries (e.g., loading and unloading operations) (definition by Comi and Delle Site, Chapter 4). drones: pilotless aircraft that can potentially be used in a wide range of applications, and depending on their degree of automation, they can fully autonomously fly themselves, making decisions about complex tasks in an uncertain environment, or can be remotely controlled by a human pilot/operator. Also known as unmanned autonomous vehicles (UAVs) (definition by Paddeu, Chapter 22). xv

xvi  Handbook on city logistics and urban freight

E-commerce: buying and selling of goods and services via communications platforms on the internet (definition by Giuliano, Chapter 1, and Jaller et al., Chapter 19). e-delivery trips: trips performed for delivering items bought online (definition by Comi and Delle Site, Chapter 4). emerging technologies: emerging technologies relating to city logistics represent digital innovative technologies which make city logistics more efficient, safer and more environmentally friendly, including Information and Communication Technology (ICT), Intelligent Transport Systems (ITSs), Internet of Things (IoT), big data, robotics, Artificial Intelligence (AI) and Autonomous Vehicle (AV) (definition by Taniguchi et al., Chapter 7). engagement: communication and/or partnership between groups of stakeholders either directly or through existing networks (definition by Browne and Goodchild, Chapter 15). environmental sustainability: managing the physical environment in a responsible way that minimizes damage to human and ecological systems now and in the future (definition by McKinnon, Chapter 23). establishment: a space where commercial activity takes place. Typical establishments in urban environments are retailers, restaurants, cafés, offices, schools, hospitals and other service activities, while suburban establishments include other activities, such as manufacturing, warehousing and wholesaling (definition by Sanchez-Diaz, Castrellon-Torres, Chapter 6). ex-ante policy evaluations: quantitative appraisal of the impact of alternative policies performed before implementation to support decision-making (definition by Le Pira et al., Chapter 16) external costs: externalities are related to social welfare and to the economy. The idea is first to measure the damages to society which are not paid for by its main actors; second, to translate these damages into a monetary value; and third, to explore how these external costs could be charged to the producers and consumers. Indeed, if the market takes into consideration the private costs, policy-makers should try to take account of the external costs (definition by Lebeau et al., Chapter 9). freight demand: demand derived from all activities required to move goods between locations of production and consumption (definition by Giuliano, Chapter 1). freight externalities: the uncompensated costs of the freight system, including air pollution, GHG emissions, noise and safety (definition by Giuliano, Chapter 1). Freight-Efficient Land Uses (FELUs): the land-use patterns that minimize the social costs (private plus external costs) associated with both the supply chains and the economic activities that consume and produce goods, at all stages of production and consumption; including reverse and waste logistics (definition by Holguin-Veras et al., Chapter 2) freight spatial planning: the planning of the location of freight facilities such as warehouses and distribution centres. Freight spatial planning happens at various scales, from regional (freight master plans, or inclusion of freight within regional master plans) to local (logistics land use zoning) (definition by Dablanc, Chapter 12).

Keywords and definitions  xvii

freight transport modelling: modelling freight transport demand and supply to forecast the impacts of the changes in infrastructures, services and regulations (definition by Sakai et al., Chapter 5). freight trip generation: amount of freight vehicle trips required to transport the cargo generated by commercial activities (definition by Sanchez-Diaz, Castrellon-Torres, Chapter 6). FREMIS: Freight Market Interactions Simulation, a conceptual model of freight transport demand with dynamic agent interaction (definition by Tavasszy and de Bok, Chapter 3). functional urban area (FUA): a city and its commuting zone. Functional urban areas therefore consist of a densely inhabited city and a less densely populated commuting zone whose labour market is highly integrated with the city (definition by Russo and Comi, Chapter 11). GoodTRIP: simulation tool to evaluate different concepts of freight distribution of consumer demand (definition by Tavasszy and de Bok, Chapter 3). governance: the act or process of governing or overseeing direction of an entity (such as a country or an organization) (definition by Haake, Chapter 14). Hyperconnected City Logistics: combining City Logistics with the Physical Internet (PI) to create an open, global and multimodal logistics system founded on universal physical, digital, operational, business and legal interconnectivity enabled through world standard encapsulation, protocols and interfaces (definition by Taniguchi et al., Chapter 7). ICT: ICT (Information and Communication Technology) is a set of technological tools and resources used to transmit, store, create, share or exchange information. It is an umbrella term that includes any communication device, encompassing radio, television, cell phones, computer and network hardware, satellite systems and so on (definition by Russo and Comi, Chapter 11). implementation: the process of putting a plan into action: making real world changes based on a plan (definition by Haake, Chapter 14). instant delivery services: these provide on-demand delivery within two hours (and often less) by either private individuals, independent contractors, or employees – by connecting shippers, couriers and receivers via a digital platform (definition by Dablanc, Chapter 12). joint knowledge production: scientists, policymakers and sometimes other societal actors cooperate in the exchange, production and application of knowledge (definition by Quak et al., Chapter 17). last-mile distribution: the last leg in supply-chain encompassing flow of goods from a distribution hub (depot/transfer facility where the last goods handling is done) to the final destination (definition by Lebeau et al, Chapter 9, and Jaller et al., Chapter 19). living labs: user-centred, open innovation ecosystems based on a systematic user co-creation approach, integrating research and innovation processes in real-life communities and settings (definition by Quak et al., Chapter 17). logistics land use: the use of land that includes a logistics activity, in most cases an activity happening in one form or another of a warehouse (definition by Dablanc, Chapter 12).

xviii  Handbook on city logistics and urban freight

logistics service provider (LSP): provider of logistics services, for example haulers and thirdparty logistics service providers (definition by Björklund and Gammelgaard, Chapter 8). logistics sprawl: the tendency of warehousing development to move away from inner urban areas toward more suburban and exurban areas. It has multiple impacts, including an increase in freight vehicle-miles traveled within urban regions (definition by Dablanc, Chapter 12). Mass-GT: open source Multi-Agent Simulation System for Goods Transport, that simulates the logistic choices of freight agents with an application in the Netherlands (definition by Tavasszy and de Bok, Chapter 3). MATSim: an open source simulation framework for large scale implement of agent-based traffic simulation, including traffic flow simulation and re-planning decisions of agents (definition by Tavasszy and de Bok, Chapter 3). microsimulation: the set of techniques that employ computers to imitate the operation of various kinds of real-world facilities or processes on a micro scale (definition by Sakai et al., Chapter 5). multi-criteria analysis: methodology providing the ranking of a set of projects on the basis of multiple assessment criteria (definition by Delle Site, Chapter 13). multi-level governance: emphasis on how networks can be governed effectively and solve complex problems, taking into account both vertical and horizontal public-private actor relations (definition by Hansson, Chapter 18). municipal consolidation centre (MCC):  goods from different providers are consolidated in a logistics facility that enables consolidation, before the goods are distributed to the municipality’s daycare centres, offices and schools (definition by Björklund and Gammelgaard, Chapter 8). omnichannel retailing: The term omnichannel retailing reflects the fact that retailers are able to interact with customers through countless channels-websites, physical stores, kiosks, direct mail and catalogs, call centers, social media, mobile devices, gaming consoles, televisions, networked appliances, home services, and more (definition by Gatta et al., Chapter 21). participatory policy-making: the process of involving stakeholders and citizens in defining and deciding the policies to implement in a participated way, i.e. decisions should be shared by the largest possible number of decision-makers and stakeholders (definition by Le Pira et al., Chapter 16). POLARIS: Planning and Operations Language for Agent-based Regional Integrated Simulation, with application to large-scale agent-based simulation of freight movements with passenger and freight market interactions with an application in the United States (definition by Tavasszy and de Bok, Chapter 3). policy acceptability: a positive or negative attitude towards a policy scheme before it has been implemented (definition by Le Pira et al., Chapter 16). policy instruments: a set of techniques by which governmental authorities wield their power in attempting to ensure support and effect (or prevent) social change (definition by Hansson, Chapter 18).

Keywords and definitions  xix

private and public sectors: organizations run by individuals or private companies, and organizations controlled by governments, respectively (definition by Browne and Goodchild, Chapter 15). restocking trips: trips of goods vehicles with the purpose of renewing inventory of shops and warehouses (definition by Comi and Delle Site, Chapter 4). risk analysis: qualitative or quantitative analysis of the uncertainty on the achievement of a project’s objectives (definition by Delle Site, Chapter 13). shopping trips: trips of passengers with shopping purpose (definition by Comi and Delle Site, Chapter 4). SimMobility Freight: urban freight simulator integrated in the agent-based urban transportation simulation platform SimMobility with application to Singapore (definition by Tavasszy and de Bok, Chapter 3). SMILE: Strategic Model for Integrated Logistic Evaluations that simulates distribution chains and logistic decision making with an application to The Netherlands (definition by Tavasszy and de Bok, Chapter 3). stakeholder behavioural analysis: analysis of stakeholder preferences and behaviour towards policy change via ad hoc modelling and simulation tools (definition by Le Pira et al., Chapter 16). stakeholders: people, groups and organizations interested in the implementation of a project or affected by the impacts of urban freight. Examples include public policy makers, freight transport companies, shippers and receivers (definition by Delle Site, Chapter 13, and Browne and Goodchild, Chapter 15). SULP: Sustainable Urban Logistics Plan (SULP) is a useful tool supporting local public decision-makers and stakeholders in “governing” city logistics measures and enhancing freight distribution processes towards economic, social and environmental sustainability and efficiency. The plan involves strategies, measures and rules that can be adopted with a cooperative approach among different actors for reaching common objectives aimed at an overall urban sustainability (definition by Russo and Comi, Chapter 11). SUMP: Sustainable Urban Mobility Plan (SUMP, a European acronym) is a strategic plan designed to satisfy the mobility needs of people and businesses in cities and their surroundings for a better quality of life. It builds on existing planning practices and takes due consideration of integration, participation and evaluation principles (definition by Russo and Comi, Chapter 11). sustainability: the ability of a system to be economically viable, environmentally efficient, and socially equitable (definition by Jaller et al., Chapter 19). TAS: future sustainable urban accessibility can be achieved through the transport system (physical mobility), the land-use system (spatial proximity) and the telecommunications system (digital connectivity); together constituting a Triple Access System (TAS) (definition by Russo and Comi, Chapter 11). TOD: transit-oriented development (TOD) is a planning and design strategy that consists in promoting urban development that is compact, mixed-use, pedestrian- and bicycle-friendly

xx  Handbook on city logistics and urban freight

and closely integrated with mass transit by clustering jobs, housing, services and amenities around public transport stations (definition by Russo and Comi, Chapter 11). tours: route travelled by goods vehicles with the purpose of collecting and/or delivering in multiple locations (definition by Comi and Delle Site, Chapter 4). TransTools: forecasting model system for passenger and freight transport in Europe (definition by Tavasszy and de Bok, Chapter 3). Transportation planning: the confluence of many different disciplines coming together in the first stages of the development of plans, policies and legislative activities, funding and project development. Transportation planning is defined as improving coordination between land use and transportation system planning; providing cooperative interaction between planning, design and operation of transportation services; maintaining a balance between transportationrelated energy use, clean air and water, and encouraging alternative modes of transportation that will enhance efficiency while providing high levels of mobility and safety (definition from https://www.ite.org/technical-resources/topics/transportation-planning/; Haake, Chapter 14). trip-based models: models representing goods transport demand in terms of vehicle-trips (definition by Comi and Delle Site, Chapter 4). ULLTRA-SIM: Urban Logistics Land-use and Traffic Simulator that simulates logistics facility locations, urban logistics chains, and truck flow with an application to the Tokyo Metropolitan Area (definition by Tavasszy and de Bok, Chapter 3). urban consolidation centre (UCC): a logistics facility located relatively close to the city or city area it serves. The facility enables consolidation of shipments from different shippers and carriers within the same vehicle (definition by Björklund and Gammelgaard, Chapter 8). urban freight demand models: quantitative models to assist in the tasks of estimating demand for travel and the impacts of proposed land use or transport system changes. They can be delivery-, vehicle- or commodity-based (definition by Sanchez-Diaz and Castrellon-Torres, Chapter 6). urban freight policy: set of regulations and practices aimed at managing freight movements in relation to the corresponding impact on emissions, accessibility, safety, parking and traffic in metropolitan areas (definition by Giuliano, Chapter 1, and Browne and Goodchild, Chapter 15). urban warehouse: a warehouse located in an urban area. An urban warehouse can take many forms, from large logistics facilities (logistics hotels, vertical logistics parks) to small logistics facilities (micro-hub, small cross-dock terminal) (definition by Dablanc, Chapter 12). vehicle emissions: exhaust emissions of air pollutants and greenhouse gases (definition by McKinnon, Chapter 23).

Introduction to the Handbook on City Logistics and Urban Freight Edoardo Marcucci, Valerio Gatta, and Michela Le Pira

The Handbook on City Logistics and Urban Freight (HCLUF) you are holding in your hands is characterized by three very specific features that make it, at least from the editors’ perspective, unique. In fact, the HCLUF is not only comprehensive in terms of topics investigated and interdisciplinary in the way those topics have been discussed, interpreted, and reported, but also systematically planned, executed, and coordinated. This Handbook is comprehensive in terms of topics investigated since it delves into all those characterizing aspects that are typically treated in various books but never in a single one. In more detail, HCLUF takes a deep dive into modeling, operations, planning, stakeholder engagement, and innovation. Furthermore, HCLUF adopts an ambitious methodological approach in dealing with the topics examined. In fact, the multidisciplinary approach characterizing this Handbook and differentiating it substantially from similar books present on the market is essential when dealing with such a complex subject. The editors believe that taking different perspectives enriches the overall knowledge needed to develop effective and efficient intervention policies capable of modifying the functioning of the entire system so as to pursue those objectives that, at first sight, might appear contradictory, such as economic growth, sustainable environment, and social inclusion. In fact, HCLUF has benefitted from the contributions coming from colleagues and practitioners with heterogeneous methodological backgrounds, such as business administration, economics, engineering, geography, management, operation research, political science, statistics, and urban planning. Finally, HCLUF was systematically planned, executed, and coordinated, thus resulting in a highly structured book comprising a hierarchical organization within each section where the authors, acting as ideal weavers, have drawn the methodological weft through the thematic warp, thus producing a high-quality cloth/book. This is testified and illustrated in what follows. On 26 June 2019, Edoardo Marcucci received an invitation from Edward Elgar Publishing to consider editing a handbook on city logistics and urban freight. While this was considered a nice event, pleasing the ego of one of the authors of the book you are now holding in your hands, since an internationally highly qualified publishing house was asking him to consider such an endeavor, at the same time, given the knowledge he had acquired in similar experiences in the past, he was also aware of the effort needed to perform this activity. Naturally inclined to cooperate, Edoardo informed both Valerio Gatta, a colleague at Roma Tre University, and Michela Le Pira, a colleague at the University of Catania, about this opportunity. He asked them both about their opinion and interest in getting involved. It took some time to make a decision, especially since, back then, there were several edited books appearing on the market that were covering some relevant issues concerning urban freight transport and city logistics.

1

2  Handbook on city logistics and urban freight

The decision to be taken was quite simple: either pass on the offer or attempt to produce something new with a clear and distinct added value. After three weeks of investigation on what was then available on the book market and many critical discussions online, they agreed that there was a niche of the market not yet covered that allowed them, at least in principle, to provide a valuable contribution to the literature. Even if the invitation was accepted, they did not communicate their decision to Edward Elgar immediately since they had in mind to develop the overall architecture of the book first. In fact, the strategy they followed was simple. They started off imagining the ideal/ best possible (from their perspective) book one might want to read and use in class as a “handbook” which, according to the Cambridge University Press online dictionary, is “a book that contains instructions or advice about how to do something or the most important and useful information about a subject” (https://dictionary​.cambridge​.org​/it ​/dizionario​/inglese​/ handbook). In other words, they asked themselves what a HCLUF should look like and tried to write it down. The intent was to: (1) provide an up-to-date and comprehensive overview of City Logistics and Urban Freight Transport research; (2) systematically focus on: (a) main critical issues; (b) most appropriate theories; (c) suitable methodologies; (d) representative case studies; and (3) adopt a distinct yet coordinated thematic perspective throughout the book. Moreover, they had it clear in their minds that the HCLUF was to be conceived also for educational purposes and had to reach a wide and composite audience made up of researchers, students, and practitioners. They only decided to sign the contract six months after receiving the invitation, when all the preliminary actions had been taken and there was, at least in principle, room for producing an innovative and valuable contribution that Edward Elgar Publishing also found convincing. In fact, producing a convincing proposal was not an easy task. They performed the actions reported below to provide Edward Elgar Publishing with a formal proposal for a HCLUF. They searched for relevant literature and defined a draft structure based on seven themes that encompassed both general aspects of urban freight transport and private/public issues. They followed a procedure based on the following steps: (1) search Scopus for documents containing “urban freight” or “city logistics”; (2) associate keywords which emerged from the search with topics related to City Logistics and Urban Freight; (3) define the structure of the book based on seven macro-areas (and 15 topics arising from a Scopus search); and (4) outline a preliminary list of potential authors to be involved in line with the results of the Scopus search. The process that finally led them to the current format of the HCLUF was tortuous and there were many revisions, re-considerations, and restructurings of the table of contents. They were, in fact, facing different and often contrasting problems. The list of issues they wanted to address was the following (with the respective order changing several times during long discussions online): 1 – Comprehensiveness; 2 – Heterogeneity; 3 – Harmony; 4 – Synthesis; 5 – Ease of use. The Handbook had to include all the most important issues pertaining to this research field (1). It had to account for the different perspectives and disciplinary nuances characterizing the researchers investigating Urban Freight Transport (UFT) (2). The Handbook also had to cover all the relevant themes in a harmonious way so as not to overemphasize one approach with respect to others, avoiding proposing a biased perspective to the reader (3). The Handbook had a maximum length set by Edward Elgar Publishing, so there was a clear upper limit to the length (4). Finally, they considered that it was extremely important to produce a Handbook

Introduction  3

that could easily be used, in line with alternative course organizations and delivery methods (e.g., online vs. offline) in a course at either a masters or undergraduate level (5). Given the clear objectives to pursue and considering the constraints, the editors, after long discussions, opted for a five-part structure to the book preceded by two keynote chapters and followed by a concluding one. The two opening chapters were conceived so to set the scene for all the others. It was deemed necessary to have two distinct chapters so as to provide not only an overview of both UFT challenges but also to clarify the land-use management/planning holistic UFT vision one needs to have in order to account for the inherent synergies and correlations characterizing the various initiatives deployed in this sector. The five-part structure was considered appropriate given the five most important issues characterizing UFT, namely modeling, operations, planning, stakeholder engagement, and innovation. To be faithful both to the heterogeneity and harmony principles, the editors opted for a hierarchical organization of each section where they foresaw an overview chapter that had the task of illustrating the main issues pertaining to the section and pointing out the most relevant issues to be discussed in more detail by the remaining three chapters in each section. For each chapter, the functional structure is always the same and was formalized by providing the authors with a semi-structured template of what should be discussed in each type of article (i.e., keynote, overview, and detailed chapter). In fact, each article, even if of different width and depth, followed the same functional structure. The keynote articles were structured as follows: (1) short introduction to the specific issue discussed; (2) detailed illustration of the specific theoretical tenets characterizing the specific discipline used to treat UFT-related issues and a sketch of the interconnections with other disciplines; and (3) critical reporting of the limitations characterizing the current theoretical approaches and future research directions. The overview chapters included: (1) an introduction to the specific topic, clarifying its relevance to UFT; (2) a detailed discussion of current approaches to the specific topic and the interconnections with the other three overview chapters; and (3) a report on the current limitations and prospect of future research directions. The standard chapter template foresaw: (1) introduction to the specific issue; (2) illustration of the main methods used; (3) report and discussion of specific and relevant case studies where the methods used were applied; and (4) discussion of the most important limitations and future research paths to be pursued in the near future. After approximately two and a half months of discussions back and forth, the editors agreed on the architecture of a book that pretty much resembles the current one. At that point, they had to find the most appropriate authors to involve in the writing of the book. While they had several authors in mind, based on the long-standing experience most of the editors have on UFT, in order to acquire an objective and unbiased representation of who were the best authors to ask to participate with respect to each specific chapter, they conducted formal keyword research on Scopus. This helped matching the tentative title of the chapter with the author(s) who had produced the most cited/relevant papers on the subject. This was the original plan that had to be confronted with the constraints posed by the acceptance of the various authors invited to participate. This process was not linear and there were some issues to be solved that had to do with the timing of the invitations, the sometimes long response time, and the necessarily sequential management of the invitations. Notwithstanding the process was somewhat cumbersome, they were able, in approximately six months, to get all the relevant chapter/author matches nailed down.

4  Handbook on city logistics and urban freight

Once they had the architecture of the book settled and matched with authors, the editors started working on coordinating the various steps. They organized one-to-one meetings with the authors responsible for the keynote chapters. Once this task was completed, they moved on to the organization of a meeting with the authors responsible for the overview chapters (i.e. section leaders) that were informed about the editors’ strategic vision of how the Handbook should be organized. Finally, they arranged five different meetings with all the authors who had agreed to write a chapter for the Handbook. To ensure a systematic presentation of the various chapter types (i.e., keynote chapters, overview chapters, and section chapters) throughout the book, the editors thought it appropriate to define a template for each one of them that was presented during the kick-off meeting when the original group of authors met for the first time. While the editors had a clear vision of how the Handbook should be organized, what should be included, how it should be presented, the sequence of the chapters and the different level of depth each chapter should have, they always adopted a participatory approach, keeping the structure open to suggestions and being ready to include any suggested improvement which might arise from the frank and open-minded discussions they had with the authors who accepted the invitation to contribute to the Handbook. To provide a sense of “group effort”, the editors also invited the overview chapters’ authors to organize “section meetings” online to discuss what should be reported in each chapter so to minimize overlap while incentivizing complementarity and interrelatedness. To keep all this together, a tentative timeline for each milestone was illustrated (see Table I.1). Once all the clarifications were spelled out to all the participants, the editors set a date for the preparation of the first draft of the chapters that were peer-reviewed among participants and editors, with a focus kept at the section level. This process took approximately seven months. After the first round of reviews, the authors had three months to provide the section leaders the revised version. This additional round was needed to ensure the cohesion and representativeness of the overall section, making clear the linkages both among the chapters within the section and the correlations with other sections. The overall duty to ensure coordination, minimize overlap, and achieve harmonious descriptions of the relevant heterogeneous issues was in the hands of the editors. Table I.1   Tentative timeline 2-pager using Edward Elgar template and following chapter structure Editors’ coordination meetings with section leaders

July 2020 (31 July 2020 deadline for 2-pager submission) First week of September 2020

Section leaders’ meetings with section contributors

September 2020

Chapter writing

October–December 2020 (31 December 2020 deadline for chapter submission)

Chapter two-round reviews by the editors with the guidance of section leaders

January–mid September 2021 (15 September 2021 deadline for final chapter submission)

Final complete handbook delivered to the publisher

31 October 2021

Expected publication date

Mid-2022

Introduction  5

At the end of the second round of reviews, each author received some suggestions to explicitly tag the interlinkages among the different chapters, both within and across sections, so as to favor the navigation within the Handbook by the reader. Finally, the editors asked each author to provide a set of keywords characterizing the specific contribution, along with a definition of each keyword used. All the definitions provided were checked in detail and, whenever reasonable, a process of homogenization was performed so to provide a minimum commonly accepted number of keyword/definition combinations. A similar process was developed with respect to acronyms so to provide the book with a general list of the acronyms used in the HCLUF. Figure I.1 reports the structure of the HCLUF and summarizes its content as follows. The first keynote chapter, by Genevieve Giuliano (Chapter 1), sets the scene by presenting the main challenges connected with UFT. It discusses the growth of urban freight and its causal factors, as well as the state of knowledge regarding the related major negative externalities. It concludes by describing the constantly growing pattern of goods movement, as well as the big changes that are in motion and the important role that technological innovations and climate change mitigation efforts will have in the near future (some of which are already taking place now). The second keynote chapter, by José Holguín-Veras et al. (Chapter 2), illustrates and discusses a holistic UFT and land-use management and planning process to improve metropolitan freight system performance and focuses on initiatives to deal with UFT challenges. It concludes by discussing some future steps, which include a systematic evaluation of the real-life performance of both transportation and land-use initiatives and further research to quantify the potential synergies between transportation and land-use initiatives. Section I deals with freight transport modeling in the urban context. The overview by Tavasszy and de Bok (Chapter 3) presents the key developments in UFT modeling, with a focus on the latest research and innovation directions. It identifies two main lines of development for future research, i.e. (1) increasing sophistication in the description of the behavior of logistics agents, and (2) increased joint use of models by multiple stakeholders as digital twins in a living lab context. The subsequent chapters by Comi and Delle Site (Chapter 4), Sakai et al. (Chapter 5) and Sánchez-Díaz and Castrellon (Chapter 6) focus on more specific issues related to UFT modeling, on models that provide origin-destination (O-D) matrices of goods vehicle flows, agent-based freight microsimulation models, which replicate the interactions among agents with different roles in multiple inter-connected layers, and freight trip generation (FTG) models, respectively. All of them identify limitations of current approaches and future research directions, including the need to (1) address sustainability issues in freight modeling, (2) better describe the heterogeneous scopes of business decisions, consider innovative vehicles and freight and passenger interaction, and (3) improve the process of data gathering and accuracy. Section II starts with an overview of city logistics and UFT operations by Taniguchi et al. (Chapter 7), underlining important factors for operations (e.g., consolidation, technologies, and collaboration), and identifying recent trends that have the potential to improve sustainability. This article paves the way for the three Section II chapters by Björklund and Gammelgaard (Chapter 8), Lebeau et al. (Chapter 9), and Crainic et al. (Chapter 10), that deal respectively with urban freight consolidation and delivery via urban consolidation centers (UCCs), electric cargo bikes for last-mile deliveries, and operations research for planning and managing city logistics systems. The three chapters present the state of the art by analyzing real case studies

6

Figure I.1  Structure of the HCLUF (KN = Key Note; OS = Overview Section; SC = Section Chapter)

Introduction  7

and highlighting their limitations and future research directions. In the case of UCCs, the authors underline how this topic has been poorly addressed from a theoretical (and thus scientific) perspective and identify different research directions that one should follow to promote sustainability goals. The external cost analysis of cargo bikes shows their potential to improve the sustainability of city logistics under certain circumstances, and highlights the need for further research, for example on the development of micro hubs to support cargo bike operations. After a comprehensive review of operations research approaches in Chapter 10 many research perspectives are presented, among which one might recall: (1) modeling multi-tier cases requiring significant research efforts to model time dependencies and synchronization, (2) research on mechanisms for collaboration and information exchange, and (3) research on uncertainty, which has not been adequately addressed so far. Section III focuses on UFT planning, with the overview chapter by Russo and Comi (Chapter 11) describing the process, guiding the definition of a UFT plan. The overview draws on relevant literature and European guidelines and identifies future directions, such as the need to develop a coherent set of scenarios which include the goals expected by society. The section chapters by Dablanc (Chapter 12), Delle Site (Chapter 13), and Haake (Chapter 14) focus on: (1) the use of city and regional spatial planning in urban freight policies, (2) the assessment of innovative city logistics solutions, and (3) planning for the future of UFT. Some case studies are presented, like Paris in the chapter by Dablanc, which helps defining general directions for conceptual guidelines to better integrate logistic facilities into urban spatial planning. Dablanc concludes that policy coordination at different planning levels is of the outmost importance. By revising different methods for ex-ante assessment, Delle Site underlines the absence of the assessment of acceptability, and the need to enrich the analyses by including new distribution schemes involving consolidation and using a life-cycle approach. Haake explores different planning approaches that can be used to address urban freight challenges and points out the need to expand freight-related land-use research, Section IV explores the topic of stakeholder engagement in UFT. The overview chapter by Browne and Goodchild (Chapter 15) examines important issues/challenges, and addresses how cities can more effectively engage with stakeholders. They identify some potential limits of current practices such as the lack of motivation to engage stakeholders and of accountability from a stakeholder perspective, and suggest engaging different stakeholder types (including city residents) and define instruments to demonstrate the effectiveness of stakeholder engagement. The section chapters by Le Pira et al. (Chapter 16) and Quak et al. (Chapter 17) focus on stakeholder engagement practices and methods; in particular, the former focuses on different participatory decision-support tools, and the latter presents the living lab approach. The chapter by Hansson (Chapter 18) discusses the scarcely addressed issue of multi-level governance in UFT. Both Le Pira et al. and Quak et al. identify some caveats of the proposed methods related to data acquisition, modeling, and time efforts in the first case and the need for more case studies and a common understanding and language for living labs in the second one. Hansson acknowledges that questions relating to governance of UFT are attracting more attention in the last years, and identifies the need to address some issues like public accountability, power, and coordination and to account for the complexity of heterogeneous actors involved in networks. Section V is about innovations in UFT, which is addressed in the overview by Jaller et  al. (Chapter 19) by discussing their relevant impact on the UFT system and underlining sustainability issues. The overview identifies future research, policies, and infrastructure

8  Handbook on city logistics and urban freight

requirements for some innovations and suggests in general to investigate evolving consumer behaviors and preferences as well as the equilibrium between online and physical retailing. The three subsequent chapters focus on specific innovative concepts, i.e., hyperconnected city logistics (Crainic et al., Chapter 20), e-commerce (Gatta et al., Chapter 21), and automation in UFT (Paddeu, Chapter 22). Hyperconnected city logistics (HCL) is reframed in a conceptual framework based on a set of fourteen core concepts, leveraged on the innovative Physical Internet concepts. Future research perspectives on this topic are identified, such as the need for advanced HCL models and solution methods, including both optimization and simulation, as well as the need to address the complexity of multi-party decisions and to have more action research and living labs to test HCL concepts in real-life contexts. E-commerce is analyzed by considering its complex relationship with urban freight distribution and citizens’ well-being, and it is suggested that more attention be paid to the development of advanced techniques capable of correctly dealing with the evolving e-commerce context, driven by innovations and stakeholder preferences. New technologies and autonomous vehicles for urban goods distribution are analyzed via a critical review of different technologies and case studies, underlining the importance of understanding the main factors influencing acceptance and adoption of new technologies from an end-consumer’s perspective and assessing their potential to maximize efficiency and safety and to contribute to sustainability. The concluding keynote chapter by McKinnon (Chapter 23) focuses on the environmental impact of logistical activity in urban areas and presents initiatives to reduce it, which, in most cases. are mutually supportive. The chapter identifies and discusses some limitations of the current approach to reduce UFT environmental impact, in particular the need to establish baseline conditions and emission reduction potentials, to evaluate the cost-effectiveness of emission reduction measures, and to account for the potential of digitalization and retail transformation (from shop-based to online retailing). Even if the Handbook is structured into specific thematic sections, all chapters are directly linked with each other. There are 162 cross-references within the Handbook. This means that there are, on average, seven citations of other chapters in each of them, testifying a high degree of connectivity, which is visible in the social network reported in Figure I.2. This has been built using the open-source software Gephi by considering chapters as nodes and crossreferences as links. Some interesting results emerge by performing a simplified social network analysis with some centrality metrics using Gephi (see Figure I.3). The degree centrality index is representative of the number of connections that a node has (in this case a chapter), in particular in-bound connections (in-degree centrality), out-bound connections (out-degree centrality), and overall connections (degree centrality). Browne and Goodchild (Chapter 15) and Le Pira et al. (Chapter 16) have the highest scores in terms of in-degree, meaning that they received many citations by other chapters, while HolguinVeras et  al. (Chapter 2) and Taniguchi et  al. (Chapter 7) are the chapters that made more cross-references to other chapters. Overall, Holguin-Veras et al. (Chapter 2) and Browne and Goodchild (Chapter 15) are the most central chapters from a degree-centrality perspective. These results are consistent with the topic of these chapters, i.e., UFT land use and transport initiatives (Chapter 2) and stakeholder engagement (Chapter 15), which are recalled in almost all chapters. The betweenness centrality index evaluates the importance of a node in relation to its role as a “bridge” between other pairs of nodes. In other words, a central node lies on

Introduction  9

Figure I.2  Social network of chapters In-degree Centrality

Out-Degree Centrality

Degree Centrality

Betweenness Centrality

Figure I.3  Centrality indexes of chapters based on cross-citations the shortest path between many other pairs of nodes, so it acts as an intermediate node, fostering the connections between other nodes. Also, in this case, the keynote chapter by Holguin-Veras et  al. (Chapter 2) confirms its centrality, followed by the chapter by Delle Site (Chapter 13) dealing with assessment of UFT solutions, which is a quite comprehensive topic that stands between many other chapters and has interrelations with all the sections, i.e., with UFT modeling, operations, planning, stakeholder engagement, and innovations.

10  Handbook on city logistics and urban freight

The high level of integration of chapters is an important result of the work done by the editors on the structure of the Handbook. Besides, authors were recommended to refer to other chapters, given that many topics are interrelated. In conclusion, this Handbook aims to provide an up-to-date and comprehensive overview of City Logistics and Urban Freight research in terms of the main issues and theories, and with specific thematic focuses. It is intended to be educational and to reach a wide audience made of researchers, students, and practitioners, and hopefully it has achieved the goals that were originally set when starting the project. Whether the “mission” was accomplished or not will soon be clear, depending both on the number of copies of the Handbook which are sold and the number of citations it will receive. The editors are optimistic by nature and hope for the best. One last hope is that you will enjoy reading this Handbook as much as they enjoyed putting it together. If this is achieved, it would imply that their effort and hard work was not in vain. However, this “long introduction” could not be completed without expressing our deepest recognition and thanks to all the authors who have contributed to HCLUF. Thank you for the patience, time, support, and nice moments you have provided us with.

1. The challenges of freight transport in cities Genevieve Giuliano

INTRODUCTION Large cities – particularly global cities – are the economic engines of the global economy. They are simultaneously financial and knowledge centres, gateways for international merchandise trade, and logistics hubs in the global freight network. As places of large populations and economic activity, cities are also vast local producers and consumers of goods, from automobiles to laptops to food and clothing. Cities are therefore dependent upon an efficient goods movement system. At the same time, goods movement generates serious externalities in the form of air pollution, greenhouse gas (GHG) emissions, congestion, crashes, and noise. Urban freight activity is growing and will continue to grow because of continued globalization and urbanization, rising per capita income, and the growth of e-commerce. For example, US freight tonnage is expected to increase by about 1.2% per year through 2045, an increase of 37% from 2018.1 Cities are increasingly challenged to effectively manage freight demand and reduce its negative impacts. With increasing urban freight flows come increased visibility of freight, conflicts with passenger demand, and public demands to solve freight problems. The twenty-first century has been a period of recognizing freight as an urban problem, research to understand urban freight dynamics, and extensive policy experimentation to mitigate urban freight problems. Nearly a decade ago, Giuliano and colleagues wrote a comprehensive assessment of urban freight research (Giuliano et al., 2013). It presented the major problems of urban freight and inventoried the many policy strategies either implemented or explored to address them. This 2013 assessment provides an appropriate baseline to address urban freight challenges a decade later. This chapter discusses what we have learned about urban freight and how freight problems have changed. It serves as an introduction for many of the following chapters in this Handbook. This chapter is organized as follows. The chapter begins with the current state of knowledge of urban freight challenges. We document the overall increase in freight activity, and then discuss congestion, air pollution, traffic safety, and noise. We conclude that the problems identified a decade ago continue, though sometimes at different magnitudes or forms. In addition, a new problem has emerged: environmental justice, the disproportionate burden low-income and often minority populations experience as a result of freight-related externalities. We then discuss explanations for urban freight problems. We distinguish between last mile activity and freight flows associated with international trade, termed trade node problems. The focus for last mile problems is the rise in e-commerce and its associated impacts on supply chains and delivery patterns. Trade-related freight has not undergone such dramatic changes, but the overall increase in trade generates more congestion and pollution, while new technologies and automation suggest future structural changes. The last section of the chapter presents some concluding observations. 11

12  Handbook on city logistics and urban freight

CURRENT STATE OF KNOWLEDGE We begin this section with what we know about urban freight. Unfortunately, lack of consistent and comparable freight data at the sub-metropolitan level continues to be a constraint on research in this field. In general, as urbanization continues and per capita incomes rise, we can expect consumer demand for physical products to increase, leading to more freight movements. Here we provide a few statistics. Growth in Urban Freight Volumes A source for nationwide truck data in the US is the Federal Highway Administration’s (FHWA) Highway Statistics Series. It provides data on the urban portion of the National Highway System (NHS) by vehicle type. It has some major limitations: the national highway system is a small subset of urban roads, and light trucks used for commercial purposes cannot be separately identified. However, there are no comparable data for urban roads. FHWA data indicate that light-duty vehicles account for about 92% of vehicle miles travelled (VMT) on the urban portion of the NHS. Trucks account for 7% and other vehicles (e.g. motorcycles, buses) account for the remaining 1%. Figure 1.1 shows the growth of vehicle miles travelled on urban highways from 2010 to 2019, the latest year for which data are available. The figure is indexed because the scale of car travel is so much greater than truck travel. Light-duty vehicles represent all vehicles, whether private or commercial, of less than 10,000 pounds gross vehicle weight (GVW). All trucks include vehicles with six tyres or more and at least 10,000 GVW; heavy trucks include combination trucks. Within the truck categories, travel of all trucks increased about as much as light-duty vehicles, and much more than heavy trucks, suggesting relatively more local (last mile) truck travel. More local truck travel is consistent with the rise in e-commerce. Urban Freight Externalities Giuliano et al. (2013) identified the following problems associated with urban freight: air pollution, congestion, safety, parking and circulation, and noise. A decade later, these problems still exist. The spatial distribution of these externalities has led to the recognition of another problem, environmental justice. Air pollution Air pollution is arguably the most critical freight externality. Air pollution is not only a major problem in urban freight, but a growing problem as countries employ more aggressive efforts to reduce greenhouse gas (GHG) emissions. In this section, we discuss both air toxics and GHGs. Trucks account for a significant share of air toxics. Table 1.1 gives emissions data for nitrogen oxides (NOx), volatile organic compounds (VOC), and particulate matter smaller than 10 microns (PM10) or smaller than 2.5 microns (PM2.5) for the US. The transport sector accounts for nearly 60% of emissions of NOx and 22% of VOC, the precursors to ozone. Trucks account for about one-third of the transport share of air pollutants. Due primarily to regulation of engines and fuels, emissions of all air toxics have declined steadily since 2000. However, trade facilities remain ‘hot spots’ due to the intensity of truck and rail traffic.

The challenges of freight transport in cities  13 1.2

1.15

1.1

1.05

1

0.95

0.9

Light Duty Vehicles

All Trucks

Heavy Trucks

2010 2015 2019

Source:   Calculated from FHWA Annual Highway Statistics data

Figure 1.1  Vehicle miles of travel on urban highways, selected years, indexed Table 1.1  US air pollutant emissions, transport and truck shares, 2016 Air pollutant

NOx

VOC

PM10*

PM2.5*

Total (1000 tons)

11,310

16,459

2,440

1,658

Transport share (%)

59.7

22.4

18.9

19.4

Truck share of transport (%)

34.3

N/A

28.6

N/A

Notes:   NOx = nitrogen oxides, VOC = volatile organic compounds, PM10 = particulate matter of 10 microns or less, PM2.5 = particulate matter of 2.5 microns or less* Excludes fugitive dust and wildfires Source:   Environmental Protection Agency (EPA). National Emissions Inventory 2016

The air toxics most damaging to human health are small particulates and ozone, which is a product of NOx and VOC. Small particulate matter (PM2.5) is a well-documented health hazard. Long-term health studies demonstrate that exposure to small particulates increases risk of both mortality and morbidity from asthma, other lung diseases, and cardiovascular disease (e.g. Bose et al., 2015; Di et al., 2017; Madrigano et al., 2013). Recent research has identified a possible link to cognitive impairment as well (Yu et al., 2019). See Chapter 23 (McKinnon) for further discussion. With the threats of global climate change becoming reality, there is increasing policy interest in reducing GHGs from all sectors. Over the past decade, global GHG emissions have grown about 1% annually. High-income regions, including the US, EU-28, Japan, and South Korea have reduced GHG emissions, while fast-growing developing countries have increased GHG emissions. The transport sector accounts for 21% of all global GHG emissions, and

14  Handbook on city logistics and urban freight 1%

2%

11% 2% Road - passenger 9%

45%

Road - freight Aviaon - passenger Aviaon - freight Shipping Rail Other

30%

Source:   Calculated from Our World in Data

Figure 1.2  Global GHG transportation emissions by source, 2018 global transport emissions continue to increase. Within the transport sector, road passenger transport accounts for the largest share, followed by freight road transport (Figure 1.2). GHG emission reductions from the transport sector are difficult to achieve due to the dependence of this sector on high-energy density fuels.2 Heavy vehicles – notably, heavy trucks, locomotives, and aeroplanes – are the most challenging, because of the energy required to move them. As light-duty vehicles become cleaner and switch to alternative fuels, the share of GHGs from freight transport will increase. Policy efforts to reduce transport sector GHG emissions are numerous and include operational efficiency, fuel efficiency, shifts to more energy-efficient modes, and efforts to manage freight demand. For more on this subject, see Chapters 2 (Holguin-Veras et al.) and 23 (McKinnon) in this volume. Congestion There are several commercial providers of urban congestion metrics, notably TomTom and Inrex, that track traffic levels annually for metropolitan areas around the world. These indices show that traffic congestion increased consistently until the occurrence of COVID-19 pandemicrelated business interruptions and stay-at-home orders. Traffic has gradually increased in countries where the pandemic has waned. However, these indices do not separate out truck traffic. One consistent source of truck-related congestion in the US comes from the American Trucking Research Institute (ATRI). It conducts a bottleneck study each year. It uses GPS data from participating trucking firms to estimate travel delay at 300 ‘freight-significant’ highway locations around the US. Locations are ranked on total estimated truck delay. Thus, locations with the same level of speed delay will rank differently based on the volume of trucks. The top 100 bottlenecks are identified each year. We selected the 25 top-ranked bottlenecks in 2009, 2015, and 2019 to compare congestion levels over time (Table 1.2). As measured by daily and peak average speed, congestion at these bottlenecks has increased over time.

The challenges of freight transport in cities  15

Table 1.2   Average daily speed and peak speed, American Transportation Research Institute top 25 bottlenecks Top 25 group averages

2009

2015

2019

Daily average speed (mph)

43

41.24

38.85

Peak average speed (mph)

35.28

32.6

29.71

Source:   Calculated from American Transportation Research Institute Top 100 Truck Bottlenecks Reports 3000

2500 2009

2012

2015

2018

2000

1500

1000

500

0

rural total

rural highways

rural other roads

urban total

urbanhigh ways

urban other roads

Source:   Calculated from Federal Motor Carrier Safety Administration data

Figure 1.3  Large-truck-involved fatal crashes by year, location The rankings are somewhat fluid. Of the top 25 bottlenecks in each of these years, ten appear in all three years, and six appear in two of the three years. The ten that appear in all three years tend to be major freeway-to-freeway interchanges where at least one of the freeways connects ports or industrial zones and intermodal facilities. Some are in downtown areas where overall congestion is high. Safety The main US national sources for information on truck-involved crashes are the National Highway Safety Administration (NHTSA) and the Federal Motor Carrier Safety Administration (FMCSA). Figure 1.3 shows the location of fatal truck-involved crashes in 2009, 2012, 2015,

16  Handbook on city logistics and urban freight

and 2018. Fatal crashes continue to increase in frequency, and crashes in urban areas have increased more than crashes in rural areas. Thus, urban crashes constitute a growing share of the total. Truck-involved crashes have increased most on other roads – arterials and minor roads – suggesting more safety conflicts in urban areas. Pedestrian and bicyclist injuries and fatalities are of particular concern for urban freight. While the number of pedestrians killed in large truck crashes has increased, the share of fatalities has remained stable at 6–7%. National statistics give information on trends, but do not reflect the effects of truck-involved crashes on local communities. We examined truck-involved crashes in a low-income majority-minority area in Los Angeles, Southeast Los Angeles (SELA). The SELA area includes industrial zones and warehouse clusters and is traversed by the major freeway route linking the Los Angeles and Long Beach ports to intermodal facilities in central Los Angeles. Of the Los Angeles region’s approximately one million daily truck trips, about 20% start, end, or go through the SELA area. We used heavy-duty truck crash data from 2015 through 2018 to examine crash patterns. We found that the SELA area has a higher rate of heavy truck crashes on a per square mile basis than Los Angeles County (11.4 vs 2.0 per square mile), and a slightly higher share of fatalities (3.2 vs 2.9%). About 55% of all crashes occurred on local streets. On average, four people were killed and 127 injured per year by trucks on city streets. A small number of crashes involved pedestrians (25 of 407, i.e., 6%) but crashes involving pedestrians made up 38% of the fatal accidents. Of these pedestrian accidents, 42% occurred at legal intersections. Crash risk affects choices of travel mode and route and reduces community quality of life. Parking and circulation Parking and circulation problems have increased as a result of more e-commerce-related activity, as well as local planning efforts to promote transit and non-motorized modes. E-commerce will be discussed in a later section of this chapter and in Chapters 19 (Jaller et al.) and 21 (Gatta et al.). Cities around the world are engaging in transportation planning strategies to restrict motorized vehicles and provide more space for public transport and use of non-motorized modes. The concept of ‘Complete Streets’ exemplifies these efforts. Complete streets reduce space for private vehicles and increase space for bike lanes and sidewalks. The logo from the Massachusetts Department of Transportation (Figure 1.4) is illustrative. It notes that ‘complete streets are for everyone’, but there is no delivery vehicle in the figure. Complete street planning tends to ignore local freight demands; truck parking and loading zones may not be incorporated. The result is conflict among street users. One example is the complete street near the author’s university campus. Figure 1.5 shows a delivery truck parked in the bike lane, forcing bike riders into the general traffic lane. These conflicts also have safety consequences. Conway et al. (2016) examined the impacts of locating bicycle lanes on truck routes in New York City. The study showed that bicyclecommercial vehicle collisions, though rare, occurred disproportionately where bike lanes were located on truck routes. Conway et al. (2018) conducted a comprehensive study of complete streets and freight and found a number of issues. With regard to driving to and from delivery stops, the study found conflicts with other users (e.g. buses), and difficulties manoeuvring narrow intersections, speed bumps, and street access. With regard to parking and loading, there are problems with parking and loading zones, as well as safe paths to buildings and sidewalks. The study recommends a number of solutions. Examples include shared parking and loading areas, right-of-way loading

The challenges of freight transport in cities  17

Complete streets are for everyone

Source:   Massachusetts State Department of Transportation

Figure 1.4  Complete Streets are for everyone stops, bike boxes, leading signal timing for bike traffic, and various technologies to reduce blind spots. Noise The final externality identified in the 2013 report is noise. Noise, loosely defined as an undesirable sound, is part of urban life. Long-term exposure to noise above 75 decibels (dB) can cause hearing loss and may affect both physical and mental health. Road noise accounts for a large share of urban noise, and heavy trucks generate a large proportion of road noise. A busy highway may generate 70 dB, trucks 90 dB, and an aircraft at take-off 120 dB (Rodrigue, 2021). The US Bureau of Transportation Statistics provides a national noise map, with estimates of noise from roads, highways, rail lines, and airports. Airports generate the most noise and rail lines generate more noise than highways. In the US, rail lines carry mostly freight, even in metropolitan areas. Noise maps illustrate both the geographic extent and the level of noise from transport experienced in metropolitan areas.3 With truck traffic growing each year, we can expect that noise pollution will continue to be a problem in urban freight. Environmental justice Our 2013 analysis did not explicitly address environmental justice, though there was growing evidence that the negative impacts of urban freight are not equitably distributed across metropolitan areas. Environmental justice concerns and our understanding of harms generated have increased over the past decade. Residential neighbourhoods located near intermodal facilities, warehouse clusters, or major rail or highway corridors tend to be low-income neighbourhoods and communities of colour. These areas are often pollution hot spots. As noted earlier, air pollution, especially small particulates, has serious negative impacts on human health. There is extensive research documenting higher exposure levels among low-income communities and/or people of colour. Here are just a few examples. Miranda et al. (2011) conducted a national study of the US to estimate exposure to PM2.5 and ozone. They found that non-Hispanic blacks were consistently over-represented in places with the worst air quality. Bravo et al. (2016) used a measure of racial isolation (extent to which minority populations are exposed only to each other) to examine long-term PM2.5 and ozone exposure in the eastern US. They found that racial isolation is associated with higher levels of exposure in cities, suburbs, and rural areas. Yu et al. (2020) conducted a study of commuters in Beijing. Using mobile phone data, commuter locations were tracked throughout the day and local exposure to PM2.5 was estimated. Using housing price as a proxy for income, they found that pollution exposure at home was greater for lower-income households.

18  Handbook on city logistics and urban freight

Source:   Author

Figure 1.5  Truck delivery in the bike lane of a Complete Street A fundamental issue in environmental justice is the relationship between property prices and air pollution or other negative externalities. Air quality is capitalized into land values: property buyers or renters are less willing to pay for locations with excessive air pollution, noise, or other hazards (e.g. Chay & Greenstone, 2005; Huang & Lanz, 2018). Thus, residential areas located near ports or rail terminals have lower rents and are more affordable to lowincome populations. Therefore, it might be argued that low-income households are making a rational choice, trading off the PM2.5 exposure for better housing. A critical question is whether low-income or minority households in fact have a choice. In the US, a history of discriminatory practices in the housing market and the lack of public or subsidized housing suggests no. There is broad evidence that low-income, immigrant, or other minoritized groups experience discrimination in housing or employment that results in disproportionate exposure to environmental hazards (e.g. Toussaint, 2021; Boone & Fragkias, 2013; Hajat et al., 2015). In countries with less discrimination and more supportive housing policies the answer may be different. A second question is whether these spatial patterns simply evolved over time – for example, as ports or intermodal yards expanded, their impact grew, leading to lower rents and more low-income households in the surrounding areas – or whether polluting facilities choose to locate in low-income communities. Warehouse location in the Los Angeles region offers an interesting case study. The environmental justice literature offers three possible explanations for why warehousing and low-income households are co-located (Mohai & Saha, 2007; Mohai et al., 2009). First, warehouse developers prefer places with cheap land and low-wage labour, and these places are often where poor or minority people are concentrated. Second, disadvantaged populations are

The challenges of freight transport in cities  19

less politically empowered, and hence less able to prevent the development of locally undesirable land uses. Third, discrimination in the housing market has constrained housing choices, Yuan (2021) conducted a study of warehouse location changes in the Los Angeles region from 2000 to 2010. Prior research revealed the co-location of warehousing and minority populations (Giuliano & Yuan, 2017). He compared shifts in warehouse location to changes in population. He found that warehousing activity is more likely to be located near minority neighbourhoods, but not necessarily low-income minority neighbourhoods, likely because of land availability. To answer the question of whether minority populations follow warehousing or the reverse, he estimated a causal model. Results showed that warehouses follow minority populations. Part of the solution to environmental justice problems is to mitigate the negative externality. In the case of air pollution, government regulation of vehicle emissions has resulted in major air quality improvements in higher-income countries. Nevertheless, hot spots remain, and many large cities continue to experience unhealthful levels of air pollution. The more difficult part of environmental justice is addressing discrimination that systematically disadvantages people on the basis of income, race, religion, or another factor.

EXPLAINING URBAN FREIGHT PROBLEMS We have now identified the main problems associated with urban freight. The next step is to explain why they exist. To understand urban freight challenges, it is helpful to understand the types of urban freight activity. Metropolitan freight activity is best understood as two main types: freight related to local supply and demand, and freight related to national or international trade. Freight related to local supply and demand is largely a function of population and employment; freight related to global trade depends on a city’s role in the global trade system. Urban Freight Related to Global Trade Large metropolitan areas are the major nodes of the global production network, containing the largest ports, airports, and intermodal facilities. In the United States, 87% of total exports come from metropolitan areas, defined as places with a population of over 50,000. In addition, the top 40 metropolitan areas accounted for about two-thirds of total exports in 2019 (US Census, Foreign Trade Statistics, 2021). Rodrigue (2004) notes that gateway cities are usually located in ‘mega-urban regions’, through which logistics functions are geographically and functionally integrated at the local, regional, and global levels. These regions developed historically as points of trade. With large and concentrated population and economic activity, they generate much of the trade demand and provide the array of expertise for managing global supply chains. Globalization has been facilitated by transportation and communications technology as well as trade liberalization policies (Dicken, 2011). Goods production processes – spatially fragmented but temporally integrated – connect countries and cities into ‘global production networks’, demanding cost-efficient and timely flow of goods (Capineri & Leinbach, 2007). This has resulted in consistent growth in cross-border trade for the past several decades. In the US, total imports more than doubled from 2000 to 2019, whereas total exports grew by 133% (US Census, Foreign Trade Statistics, 2021). The extent to which globalization will continue

20  Handbook on city logistics and urban freight

to grow is uncertain. The 2020 global pandemic revealed the vulnerability of global supply chains that could result in restructuring within smaller geographies. Although it will take some time for pandemic impacts to be understood, it seems reasonable to assume that the level of globalization is less likely to change than regional trade patterns. Urban Freight Related to Last Mile Freight associated with local supply and demand is typically termed ‘last mile’ as it represents the last (or first) link in supply chains. ‘Last mile’ includes delivery or pickup of imports/ exports, intra-metropolitan trade of commodities (local production and consumption), deliveries to or from retailers, etc. Freight related to local supply and demand is increasing for several reasons, including increasing consumer demand and the rise in e-commerce and its impacts on supply chains and distribution patterns. What is driving increased consumer demand? The resident population consumes food, clothing, and shelter, generating the demand for consumer goods and services that now account for nearly 70% of US economy gross domestic product (GDP). As per capita income increases, so does consumption. As households get richer, the demand for higher quality and more diverse commodities increases. Historically, the rise in consumption has been evidenced by the emergence of supermarkets, shopping malls, specialty shopping districts, and, later, big-box retail and ‘lifestyle’ shopping centres. Housing consumption also increases with increased per capita income. In the US, the average size of a new single-family home increased by 51% from 1973 to 2019.4 As dwelling size increases, so does the demand for more household goods. Rodrigue (2021) categorizes consumer-based goods distributions based on supply chain and distribution patterns (Figure 1.6). It can be seen that there is great variety in the frequency of deliveries, delivery locations, parking and loading facilities, and distribution networks. Personal consumption is not the only source of freight demand in cities. Construction activity is a constant in any city. Building materials and equipment are shipped to construction sites, and waste material is carted away. Utility systems must be maintained and repaired, and refuse must be removed from residences and commercial sites. Reverse logistics is the term used for the return of transformed goods as well as returns from online consumption activity. Finally, cities are also nodes of production, and production generates another dimension of freight demand. Manufacturing – from automobiles to computer chips – creates products for export out of the local area. Cities with a large manufacturing presence typically have large associated rail and truck flows. Professional service activities, for example real estate, finance, or medical services, generate a different type of demand: office supplies, computer hardware and software, medical supplies, often in small and frequent deliveries. The geography of city flows is a function of industry mix as well as population characteristics and metropolitan size. Why are last mile problems more prevalent in city cores? Last mile problems tend to be more severe in city cores, where development density is high and street space is limited. To better understand the city core problem, it is helpful to consider how land price drives both location and consumption choices. Density of population, employment, or both is a proxy for urban form. Density reflects the value of land; as land value increases, so does rent, and as rent increases, space is more intensively used. Firms economize on space by

The challenges of freight transport in cities  21

Independent retailing • Small scale, high delivery frequency, small or medium trucks, street as delivery plaorm

Chain retailing • Large scale, truck parking and loading on-site, large trucks, consolidated deliveries, use of third party logiscs services

Food deliveries • Specialized supply chains, cold chain logiscs, oen small suppliers and frequent deliveries

Parcels and home deliveries • Large freight integrators, distribuon center networks, small and medium truck delivery fleets, delivery companies Source:  Adapted from Rodrigue (2021), by permission

Figure 1.6  Types of urban retail distribution allocating less space per worker, as well as minimizing lobby and storage space. In the case of retailing, high land values necessitate more revenue per square foot, which means more rapid turnover of product, narrow aisles and tall shelves, as well as less space devoted to storage, compared with retail activities in lower-density environments. More intensive space utilization implies more delivery trips, all else being equal. Households also economize on space in response to high land values. Dwelling units are smaller and inventory capacity is limited. There is less room to store food or linens. More shopping trips are taken by walking or public transit, constraining the amount of goods that can be carried home. Services conducted inside the home (e.g. clothes washing, food preparation) may be outsourced to cleaners and restaurants. This suggests that, controlling for demographic characteristics, households in high-density areas will shop more frequently, purchase in smaller lots, and consume more out-of-home services. The outcome of high land prices, then, is more frequent goods deliveries to offices, retailers, and service providers. Giuliano et al. (2016) conducted case studies of the Los Angeles and San Francisco metropolitan regions to explore the relationship between freight activity and urban density. They found a significant and positive relationship. Rise of e-commerce and instant deliveries Arguably the most significant change in last mile urban freight is the rapid growth of e-commerce. E-commerce is defined as any type of consumption that takes place via an online platform. The most common type of e-commerce is online shopping: consumers order and pay online, and the goods are delivered to the consumer’s residence. There are variations: the consumer may order online and pick up at a retail establishment, or at a local locker facility. E-commerce includes all types of consumption, from meal deliveries to cat litter or furniture purchases. The emergence of online shopping has transformed where and how goods are produced, distributed, and sold, and how consumers make shopping as well as shopping travel decisions (Mokhtarian, 2004). See Chapters 19 (Jaller et al.) and 21 (Gatta et al.) for a comprehensive discussion of e-commerce.

22  Handbook on city logistics and urban freight

The rise in e-commerce has been rapid. In the US, the market share of online shopping has increased from about 3.7% in 2008 to 9.5% in 2018 and 13.5% in 2021. The annual rate of growth of online shopping sales has been around 11–13% since 2011, much greater than the rates for total retail sales, at 3–4% (US Census, 2019). The global market share of online shopping is estimated to be nearly 20% in 2021 with a value of nearly $5 trillion. China is by far the largest e-shopping market, accounting for about $2.8 trillion in purchases (Keenan, 2021). Growth of e-shopping accelerated during the COVID-19 pandemic. It remains to be seen whether the growth rate will slow as pandemic restrictions are removed. Two hypotheses have been examined to explain the adoption of online shopping. The first is the diffusion of innovation. Those who are closest to the innovation are the first adopters. These tech-savvy people are more willing to take risks and experiment with the new technology. Once the technology is demonstrated to have benefits, others follow the lead and the technology spreads to the mainstream population. The second hypothesis is based on efficiency. Online shopping gives consumers more choices and more opportunities to obtain product information. Goods from faraway places can be obtained with a click of the mouse, not an hours-long journey (Farag, Schwanen et al., 2007; Anderson et al., 2003). Analyzing what influences consumers’ channel choice is fundamental to understanding the diffusion of online shopping (Maltese et al., 2021). E-commerce is also changing rapidly. The variety of goods available continues to grow, and many new products have emerged, such as ingredients and instructions for home-prepared meals (e.g. Blue Apron) and subscription deliveries of frequently used products. Speed of delivery is also increasing. Large online retailers offer ‘instant deliveries’ (within two hours) in cities, and one-day delivery is now routine in many metropolitan areas. The growth of e-commerce has many impacts. Figure 1.7 shows how e-shopping has affected package delivery from 2004 to 2019. The bars show USPS mail volume; mail volume has declined consistently

Source:   Rodrigue (2021), by permission

Figure 1.7  The growth of e-commerce: Mail carried by USPS and parcels carried by major carriers, US, 2004–2019

The challenges of freight transport in cities  23

since 2007. The lines show parcel deliveries for UPS, USPS, FedEx, and Amazon. Total parcel deliveries have grown from 7.2 to 19.7 billion, a 174% increase. Clearly such shifts in consumer demand have significant impacts on both freight and passenger movements. Dablanc and colleagues were among the first to track the growth of instant deliveries (Dablanc et al., 2017). Their focus was on app-based services that provide on-demand delivery usually within two hours. In contrast to the more conventional one- or two-day deliveries, instant deliveries may be performed by private individuals, independent contractors, or employees. Instant deliveries have led to restructured supply chains and emergence of new business models. Prepared food delivery is one example. ‘Ghost kitchens’, food preparation businesses with no on-site service have emerged to serve only the delivery market. These businesses tend to locate in industrial zones, hence changing geographic patterns of food deliveries. Delivery companies have added specialized food delivery services (e.g. UberEats), and in dense city cores these deliveries may be made by bicycle, scooter, car, or small truck. Market forecasts predict continued expansion and diversification of instant deliveries. Taken together, these trends suggest continued fragmentation of the goods supply chain, as more consumption is individualized and delivered to homes. Fragmentation in turn will generate more truck travel. E-shopping and urban form With the growth of e-shopping comes more deliveries in residential neighbourhoods, more competition for kerb space, and additional externalities. We expect online shopping frequency to be related to urban density. As noted above, shopping is more frequent in dense, urban environments. Under such circumstances, online shopping may be more attractive, especially for bulky or heavy standard products. Table 1.3 gives information on online shopping behaviour from the US 2017 National Household Travel Survey. It includes respondents living in metropolitan areas with neighbourhood population density of at least 1,000 persons per square mile. Respondents were asked about the frequency of their online shopping. Those in the highest-density category have the lowest share of never shopping online and highest share of shopping online more than once per week. They also have the highest average monthly frequency. Note that the marked differences are between the high-density categories and the rest. From a density of 10,000 persons per square mile and below, the rate of online shopping is quite consistent. When we control for demographic characteristics, the effects of density are statistically significant only for the high-density category (Giuliano et al., 2016). Competition and free delivery Impacts of faster deliveries merit further discussion. E-shoppers have options: they can choose the speed of shipping (e.g. standard, two-day, next day, instant) and often the destination (e.g. residence, place of employment, pick-up station or locker, retail establishment). With all else being equal, as the speed of shipment increases, so does the cost of delivery, as products or packages are less likely to be bundled. Thus, the speed of shipment is an inverse proxy for efficiency of the freight delivery. Currently the major e-retailers are competing on free shipping. When shipping is free, even for one- or two-day or instant deliveries, e-shoppers have no incentive to choose cheaper and slower shipping. Free shipping also incentivizes smaller volume purchases and more returns. As long as shipping is free, there is no reason to bundle purchases to save money. As long

24  Handbook on city logistics and urban freight

Table 1.3  Online shopping frequency and population density Sample

Frequency

Residential Density (persons/square mile)

Obs

Share (%)

Never (%)

Less than 1/ More than week (%) 1/week (%)

Average per month

High (>25,000)

2,399

3

36

37

27

3.28

Medium high (10,000–25,000)

8,081

9

40

38

22

2.91

Medium (4,000–10,000)

34,247

37

41

38

21

2.72

Medium-low (2,000–4,000)

28,998

31

40

39

21

2.73

Low (1,000–2,000)

19,989

21

40

40

20

2.66

Source:   2017 National Household Travel Survey

as returns are free, the shopper may purchase three different sizes of shirt, keep the one that fits and return the rest. Both free shipping and free return options are important for the competitiveness of online retailing, so it is not surprising that free shipping is so widely offered. However, free shipping, and especially free one- or two-day or instant shipping is a powerful force of fragmentation of purchases and deliveries, and hence a negative force on the efficiency of freight deliveries. Of course, shipping is not free. To address the high cost of home delivery, carriers have experimented with alternative strategies to replace home delivery services. Delivery lockers and pick-up points are common in some countries in Europe. In the US, Amazon has implemented delivery lockers, and UPS is using ‘Access Points’. The purpose is to generate small clusters of deliveries, thus avoiding the additional costs of delivering to multiple single destinations. However, with free delivery available, there is little attraction for the customer to choose a locker pick-up. Studies show that about 90% of online shoppers request home deliveries in France. The estimate for UPS is 74% (Morganti et al., 2014). Whether online retailers will be able to continue to absorb shipping costs while maintaining a price advantage remains to be seen. Meanwhile, fragmentation of deliveries may be expected to continue. Overall impact of e-shopping The net effect of e-shopping remains uncertain. While delivering multiple packages via truck may be more efficient than several shoppers each making individual trips to purchase the same merchandise, it is doubtful that all individual shopping trips are eliminated. Some e-shoppers may browse in stores, shop both online and in-store, or substitute travel for other purposes. On the freight delivery side, it seems clear that e-shopping leads to more freight shipments. Online shopping requires delivery to a residence or a local, common pickup point. Small-scale deliveries (small packages in small trucks) are less efficient than the large-scale deliveries made to retail establishments. Although e-shopping eliminates at least some largescale deliveries due to loss of in-store business, these losses will be more than offset by the added truck travel generated by small-scale deliveries. Efficiency of freight deliveries is further reduced by the rate of failed local deliveries and the higher rate of returns associated with online shopping. The outcome is increased truck travel. Finally, the trend of faster deliveries

The challenges of freight transport in cities  25

(one- or two-day shipping, or ‘instant delivery’ within two hours) intensifies freight inefficiencies by prioritizing speed over larger loads, all else being equal. Understanding consumers’ preferences between online and offline is essential for estimating their ensuing transport and environmental implications (Marcucci et al., 2021). Strategies to reduce last mile problems Much of the remainder of this Handbook is devoted to solving last mile problems. I present only a brief review here. Last mile problems are often unique and highly localized, and hence typically require customized solutions. Very generally, mitigation strategies may be characterized as voluntary, regulatory, or technological. Often last mile problems do not fall neatly into one jurisdiction such as a municipality, and in some countries local authority is limited by state or federal law. Therefore, voluntary, collaborative industry–government partnerships are established to develop policies or guidelines, such as delivery hours or freight consolidation centres. Incentives may be offered to support off-hours deliveries or the use of low-emission vehicles. In such a context, a participatory planning approach, based on sound behavioural analyses, is needed to identify the most effective and accepted solutions (e.g. Gatta et al., 2019; Le Pira et al., 2017; Marcucci et al., 2017). See Chapters 2 (Holguin-Veras et al.), 9 (Lebeau et al.) and 13 (Delle Site) for more on these strategies. When regulatory authority exists and the political environment is supportive, jurisdictions may impose regulations. Examples include low-emission zones, truck fees, or emissionreduction targets. Technology plays an important role not only in increasing the efficiency of urban freight but also in making it possible to manage existing street and kerb resources as efficiently as possible. For example, smart parking systems can provide information on space availability so that deliveries can be more efficiently planned. On the industry side, efforts to consolidate residential deliveries through lockers and pickup points are aimed at reducing costs and miles travelled. Section II of this Handbook provides many models and examples of last mile strategies. Trade Nodes Global trade generates another layer of freight and logistics demand to serve both local and non-local supply and demand. Whereas domestic freight flow per capita is fairly consistent across cities, international trade flow is concentrated in cities that serve as global trade hubs. Trade-related flows are regional, and therefore have impacts throughout metropolitan areas. Major facilities like ports or warehouse clusters are generators of heavy truck traffic, creating hot spots of congestion, air pollution, noise, and conflicts with passenger traffic. Highways that serve as main linkages with these facilities and the national network have greater concentrations of heavy trucks. Metropolitan areas that serve as major trade nodes have high volumes of both truck and rail traffic. High volumes of rail freight traffic contribute to congestion at rail crossings and conflicts with rail passenger services. In contrast to the rapid increase in e-commerce around the world, global merchandise trade in 2019 was about $0.5 trillion dollars greater than in 2012 (unadjusted dollars). Trade declined by $1.5 trillion in 2020, reflecting the business interruptions associated with the COVID-19 pandemic (see Figure 1.8). Major changes in ocean carrier alliances, the use of ever-larger container ships, and tariff conflicts between the US and China have each played a role in the volatility of global trade. At the national and sub-national levels, changing supply

26  Handbook on city logistics and urban freight 25000000

20000000

15000000

10000000

5000000

0

2012

2013

2014

2015

2016

2017

2018

2019

2020

Source:   WTO Open Data Portal

Figure 1.8  World Merchandise Trades, US $ million, 2012–2020 chains (e.g. migration of manufacturing to Southeast Asia) and competition among ports for large-ship business are leading to changes in ocean shipping flows. Supply chain disruptions during the pandemic may further affect supply chains as global producers seek more flexible and resilient product chains, which could mean geographically smaller and less dispersed production and distribution chains. The lack of dramatic growth in international trade suggests that trade node problems are much the same today as they were a decade ago. Congestion Last mile freight activity particularly impacts local roads, whereas trade-related goods movement impacts highways and road systems near major freight facilities. Last mile freight activity includes more light- and medium-duty trucks, whereas trade-related freight moves on rail and heavy-duty trucks. Port, airports, intermodal facilities, and warehouse clusters are attractors of heavy-duty trucks, and highways linking these facilities typically have higher-thanaverage shares of truck traffic. Heavy-duty trucks (HDTs) have a disproportionate impact on congestion due to both performance of the vehicles and avoidance behaviour of other drivers (Moridpour et al., 2015; Kong et al., 2016). Although it is well understood that the performance characteristics and avoidance behaviour of other drivers cause HDTs to have a disproportionate impact on traffic, the extent to which truck traffic contributes to congestion and therefore delay of other vehicles is less well understood (Simoni et al., 2020). Giuliano et al. (2018) developed the concept of ‘freight impact area’, defined as segments of highways with high congestion and a high volume of

The challenges of freight transport in cities  27

trucks. The measure of impact is total peak hour delay for all vehicles. The measure was applied to the Los Angeles and San Francisco regions in California. Table 1.4 gives results for the top 15 freight impact areas in the Los Angeles region. Together the top 15 freight impact areas account for 53,615 hours of vehicle delay each afternoon peak period per day. Over the course of a year, the delay would be about 14 million vehicle hours – substantial by any standard. We cannot distinguish what share of the delay is caused by trucks. However, we can say that congested locations with high proportions of trucks are associated with significant delays for both passenger and freight vehicles. How trade-related freight affects metropolitan areas also depends on mode shares. In the US, rail carries over 40% of all freight ton-miles; in the EU-28, the share is around 17%. The vast majority of rail infrastructure in the US is owned by the freight rail carriers; urban passenger rail services typically lease access from the freight carriers to operate passenger service. This arrangement leads to conflicts between passenger and freight demands. Passenger service does not generate profits for the freight carrier and at the same time constrains freight operations. Road is the dominant mode for freight in the EU-28. Metropolitan and intercity rail passenger services in the EU-28 are far more developed than in the US, and rail ownership structures differ across countries, with the legacy of each nation having its own rail system helping to explain the dominance of road transport in the EU. Figure 1.9 shows mode shares for 2010, 2015, and 2018 for the US and 2010, 2015 and 2019 for the EU-28, respectively. Mode shares have remained relatively stable, but the trend is in the wrong direction from an energy efficiency perspective, as in both regions the road share is increasing. Warehousing and distribution facilities A second aspect of trade-related impacts is the growth and evolution of warehouse and distribution facilities. Warehousing and distribution have been influenced by continued globalization, competition and restructuring of wholesaling and retailing, and technology. Rodrigue (2021) describes the shift from push to pull distribution systems, where products are delivered on an as-needed, just-in-time basis to minimize inventory holdings. The objective is velocity – move the item from production to sale as quickly as possible. These distribution systems have been complicated by e-shopping and short delivery times. Distribution centres must be close to the population served to provide short delivery times. At the same time, warehouse Table 1.4  Descriptive statistics of key variables of the top 15 freight impact areas on the National Highway System in the Los Angeles region Variable

Average

Minimum

Maximum

Length (miles)

2.24

0.11

9.95

Average total volume per direction, PM Peak

43,241

32,529

61,762

Average truck volume per direction, PM Peak

3,252

2,594

4,062

Average share of trucks (%)

8.00

4.38

10.85

Congestion speed (miles/hour)

18.61

8.10

29.93

Total peak hour all-vehicle delay (vehicle×hour)

3,574.3

344.0

11,741.6

Total peak hour freight delay (truck×hour)

285.0

20.4

1,103.7

Source:   Giuliano et al. (2018)

28  Handbook on city logistics and urban freight

(a)

US Freight Mode Shares, ton-miles

100% 90%

12.5%

11.6%

11.5%

42.1%

41.1%

40.5%

45.0%

47.0%

47.6%

0.3% 2010

0%3 2015

0.4% 2018

80% 70% 60% 50% 40% 30% 20% 10% 0%

Air

(b) 100.0%

Truck

Railroad

Domesc water

EU-28 Freight Mode Shares, ton-km 6.9%

6.5%

5.6%

75.7%

75.3%

77.4%

17.4%

18.2%

17.0%

2015

2019

90.0% 80.0% 70.0% 60.0% 50.0% 40.0% 30.0% 20.0% 10.0% 0.0%

2010 Rail

Road

Inland water

Source:   US Bureau of Transportation Statistics, Eurostat

Figure 1.9  Freight mode shares, US and EU-28

The challenges of freight transport in cities  29

automation has increased scale economies: the upfront expenses of automation require largescale operation to recoup costs. The demand for large-scale facilities translates to demand for cheap land, which is more available at the periphery of metropolitan areas – far away from instant delivery customers. The result is an emerging network of distribution. Import products are shipped to large warehouses which serve as hubs for consolidating shipments to other domestic markets and distributing product to fulfilment centres located closer to the metropolitan area population. These in turn serve as hubs for local distribution. Until the rapid rise of e-commerce, the dominant trend in warehouse location was decentralization, driven by scale economies and the search for low-priced land. Decentralization has been documented in the US (e.g. Atlanta, Los Angeles, and Chicago (Dablanc & Ross, 2012; Dablanc et al., 2014; Goodchild & Dubie, 2016), Sweden, the UK, and Japan (Heitz et al., 2018; Allen et al., 2012; Sakai et al., 2015). Kang (2020) conducted a more extensive study of the 48 largest US metropolitan areas using data from 2003 to 2013. He found that warehouse location with respect to the Central Business District (CBD) increased by 1.06 miles over that time period. Location shifts were greater for larger warehouses and for the largest metropolitan areas, where land prices are higher. Centralizing effects are demonstrated by rent and vacancy rates of in-city space. According to CBRE data, rental rates have increased faster than the long-term average (about 2% per year) every year since 2012, and for the past five years have increased by 5% annually (SCDigest Editorial Staff, 2020). Rents for smaller in-city warehouses have increased much more, reflecting the growing demand for spaces that make same-day deliveries possible. Centralizing effects are also seen in the conversion of urban spaces to fulfilment and distribution functions. In the US, vacant department store, industrial, or office space are being adapted for warehousing (Panepinto, 2018), and some retailers are repurposing their space to include local fulfilment operations (Sparkman, 2019). In Paris, joint public/private ventures have resulted in mixed use redevelopment that includes logistics spaces; see Chapter 12 (Dablanc). To conclude, warehousing trends are affecting all parts of metropolitan areas. Solving trade node problems Giuliano (2020) identifies a long list of potential near-term mitigation strategies and evaluates them with respect to cost, effectiveness in reducing congestion, co-benefits, technical difficulty, and implementation feasibility. Strategies fall into four categories: infrastructure, efficiency improvements, emission reduction, and policy tools. Strategies identified as being most effective are summarized in Table 1.5. Giuliano (2020) does not include Table 1.5  Effective strategies to reduce trade node problems Category

Strategy

Infrastructure

On-dock rail, railroad grade separations

Efficiency improvements

Integrated freight information systems, cargo matching, smart truck parking, terminal appointment systems, freight advanced traffic management systems

Emission reduction

Low emission standards, more energy-efficient modes

Policy tools

Port cargo and gate tolls, terminal turn time limits for trucks

30  Handbook on city logistics and urban freight

some obvious possibilities, for example truck mileage taxes, truck-only highway lanes, or zero-emission heavy-duty trucks due to high costs and/or limited political feasibility. We may expect that political feasibility will change as climate change impacts become more serious. In the longer term, the solution space will change. The beginnings of heavy-truck automation are being demonstrated in long-haul travel and will eventually be tested in urban environments. New technologies of production – 3D printing – will likely restructure global production systems, and possibly dramatically shrink supply chain geographies. Section V of this Handbook discusses new technology and innovations.

CONCLUSIONS This chapter has provided an overview of the many challenges facing urban freight. Congestion, air pollution, traffic safety, and noise are challenges that have existed for decades. Continued globalization, urbanization, per capita income growth, and changing consumption patterns assure that goods movement will continue to increase. I have also identified environmental justice as a growing challenge, one that will require not only regulation, but also less race, ethnic, or religious discrimination in society. I noted that big changes in freight are in motion. I conclude with some thoughts on the future. First, the transportation revolution taking place will affect last mile transport. Platform-based services such as ride-hailing, car sharing, and bike sharing are ubiquitous in large cities. Platform-based services are now penetrating goods movement with instant delivery services and cargo-matching services. Robots are used widely in warehousing and drones are in various stages of application for local delivery. Delivery robots and automated vehicles are delivering food and other items. Clearly last mile freight activity is changing fast, and our efforts to address congestion and safety will have to account for this more complex delivery system. At the same time, technology offers new opportunities to achieve more efficient and sustainable last mile transport. Second, this Handbook is being written at a time of great uncertainty with regard to international trade. Tariff conflicts between the US and China, the US renegotiation of the Canada-US-Mexico trade agreement, Britain’s withdrawal from the EU, and the impacts of the pandemic remain to be sorted out. Some reshoring of key products or industries may occur as reliability becomes more valuable. Thus, the future growth and geographic pattern of international trade is difficult to predict. At the same time, we expect more aggressive climate change mitigation efforts, and these will add pressure to parts of the supply chain that have been largely exempt (e.g. ocean carriers, US railroads). California provides an example of what other states and countries may do to decarbonize the freight sector. California policies include a broad carbon cap and trade system, specific targets for achieving a zero-emissions truck fleet, regulations to reduce carbon in existing fuels, regulations requiring ocean carriers to run on electricity while in port, and regulations being planned to achieve zero-emission cargo handling equipment at ports.5 In addition, California is funding research and development on battery and hydrogen-fuel-cellpowered locomotives. Efforts to reduce GHG emissions throughout the economy will surely accelerate change in the freight sector.

The challenges of freight transport in cities  31

NOTES 1. Calculated from Freight Analysis Framework data, available at https://data​ .bts​ .gov​ /stories​ /s​ / Moving​-Goods​-in​-the​-United​-States​/ bcyt​-rqmu. 2. Energy density is measured as the amount of energy available per volume of fuel. 3. The US National Noise Map is available at https://data​.bts​.gov​/stories​/s​/National​-Transportation​Noise​-Map​/ri89​-bhxh/. Examples are not shown here as they are difficult to interpret without colour. 4. Source: US Census data, https://www​.census​.gov​/construction​/chars​/xls​/squarefeet​_cust​.xls. 5. See Purdon et al. (2021), for details on California policy.

REFERENCES Allen, J., Browne, M., & Cherrett, T. (2012). Investigating relationships between road freight transport, facility location, logistics management and urban form. Journal of Transport Geography, 24, 45–57. Anderson, W. P., Chatterjee, L., & Lakshmanan, T. R. (2003 September). E-commerce, transportation, and economic geography. Growth and Change, 34(4), 415–432. Boone, C., & Fragkias, M. (2013). Urbanization and sustainability: Linking urban ecology, environmental justice and global environmental change. New York: Springer. https://doi​.org​/10​.1007​ /978​-94​- 007​-5666​-3. Bose, S., Hansel, N. N., Tonorezos, E. S., Williams, D. L., Bilderback, A., Breysse, P. N., … McCormack, M. C. (2015). Indoor particulate matter associated with systemic inflammation in COPD. Journal of Environmental Protection, 6(5), 566–572. https://doi​.org​/10​.4236​/jep​.2015​.65051. Bravo, M., Anthopolosa, R., Bell, M. L., & Mirandaac, M. L. (2016). Racial isolation and exposure to airborne particulate matter and ozone in understudied US populations: Environmental justice applications of downscaled numerical model output. Environment International, 92–93, 247–255. Capineri, C., & Leinbach, T. R. (2007). Globalized freight transport: Conclusions and future research. In T. Leinbach & C. Capineri (Eds.), Globalized freight transport: Intermodality, e-commerce, logistics and sustainability (p. 259). Cheltenham, UK: Edward Elgar Publishing Ltd. Chay, K., & Greenstone, M. (2005). Does air quality matter? Evidence from the housing market. Journal of Political Economy, 113(2). https://doi​.org​/10​.1086​/427462. Conway, A., Tavernier, N., Leal-Tavares, V., Gharamani, N., Chauvet, L., Chiu, M., & Yeap, X. B. (2016). Freight in a bicycle-friendly city: Exploratory analysis with New York City open data. Journal of the Transportation Research Board, 2547(1), 91–101. https://doi​.org​/10​.3141​/2547​-13. Conway, A., Williamson, J., Gjorgjievska, M., Chem, Q., Prasad, L., & Xing, C. (2018). Complete streets considerations for freight and emergency vehicle operations. New York State Energy Research and Development Authority. Retrieved from https://www​.metrans​.org​/assets​/upload​/complete​%20streets​ %20freight​%20booklet​_v6​_ hires​- 0​.pdf. Dablanc, L., & Ross, C. (2012). Atlanta: A mega logistics center in the Piedmont Atlantic megaregion (PAM). Journal of Transport Geography, 24, 432–442. Dablanc, L., Ogilvie, S., & Goodchild, A. (2014). Logistics sprawl: Differential warehousing development patterns in Los Angeles, California, and Seattle, Washington. Transportation Research Record: Journal of the Transportation Research Board, 2410(1), 105–112. Dablanc, L., Morganti, E., Arvidsson, N., Woxenius, J., Browne, M., & Saidi, N.  (2017). The rise of on-demand ‘instant deliveries’ in European cities.  Supply Chain Forum: An International Journal, 18(4), 203–217. https://doi​.org​/10​.1080​/16258312​.2017​.1375375. Di, Q., Wang, Y., Zanobetti, A., Wang, Y., Koutrakis, P., Choirat, C., … Schwartz, J. D. (2017). Air pollution and mortality in the medicare population. New England Journal of Medicine, 376(26), 2513–2522. https://doi​.org​/10​.1056​/ NEJMoa1702747. Dicken, P. (2011). Global shift: Mapping the contours of the world economy (6th ed.). New York: Guilford Press.

32  Handbook on city logistics and urban freight

Farag, S., Schwanen, T., Dijst, M., & Faber, J. (2007, February). Shopping online and/or in-store? A structural equation model of the relationships between e-shopping and in-store shopping. Transportation Research Part A: Policy and Practice, 41(2), 125–141. Gatta, V., Marcucci, E., Delle Site, P., Le Pira, M., & Carrocci, C. S. (2019). Planning with stakeholders: Analysing alternative off-hour delivery solutions via an interactive multi-criteria approach. Research in Transportation Economics, 73, 53–62. Giuliano, G. (2020). Urban logistics: The regional dimension. Chapter 3. In M. Browne, S. Behrends, J. Woxenius, G. Giuliano & J. Holguin-Veras (Eds.), Urban logistics: Management, policy and innovation in a rapidly changing environment (pp. 52–81). London: Kogan Page Ltd. Giuliano, G., O’Brien, T., Dablanc, L., & Holliday, K. (2013). Synthesis of freight research in urban transportation planning. NCFRP Report 23. Washington DC: National Cooperative Freight Research Program (refereed). Giuliano, G., & Yuan, Q. (2017). Location of warehouses and environmental justice. Report 16-1.1g. Los Angeles: MetroFreight Center of Excellence, University of Southern California. Retrieved from www​.metrans​.org​/sites​/default ​/files​/research​-project ​/ MF​%201​.1g ​_ Location​%20of​%20warehouses​ %20and​%20environmental​%20justice​_ Final​%20Report​_021618​.pdf. Giuliano, G., Kang, S., Yuan, Q., & Shin, E.-J. (2016). Understanding freight flows in cities: The role of density. Presented at the Volve Research and Education Foundation Conference on Urban Freight, Gothenburg, Sweden. Giuliano, G., Showalter, C., Yuan, Q., & Zhang, R. (2018). Managing the impacts of freight in California. National Center for Sustainable Transportation Research Report, University of California, Davis. Retrieved from https://escholarship​.org​/uc​/item​/6614p4js. Goodchild, A., & Dubie, M. (2016). Logistics sprawl in Chicago, Illinois. Presented at 14th World Conference on Transport Research, Shanghai, China. Hajat, A., Hsia, C., & O’Neill, M. S. (2015). Socioeconomic disparities and air pollution exposure: A global review. Current Environmental Health Reports, 2(4), 440–450. https://doi​.org​/10​.1007​/s40572​ -015​- 0069​-5. Heitz, A., Dablanc, L., Olsson, J., Sanchez-Diaz, I., & Woxenius, J. (2018). Spatial patterns of logistics facilities in Gothernburg, Sweden. Journal of Transport Geography, 83, article 102191. https://doi​ .org​/10​.1016​/j​.jtrangeo​.2018​.03​.005. Huang, X., & Lanz, B. (2018). The value of air quality in Chinese cities: Evidence from labor and property market outcomes. Environmental and Resource Economics, 71(4), 849–874. https://doi​.org​ /10​.1007​/s10640​- 017​- 0186-8 Kang, S. (2020). Why do warehouses decentralized more in certain metropolitan areas? Journal of Transport Geography, 88, 102330. https://doi​.org​/10​.1016​/j​.jtrangeo​.2018​.10​.005. Keenan, M. (2021). Global ecommerce explained: Stats and trends to watch in 2021. Industry Insights and Trends, May 13, 2021. Retrieved from www​.shopify​.com​/enterprise​/global​-ecommerce​-statistics. Kong, D., Guo, X., Yang, B., & Wu, D. (2016). Analyzing the impact of trucks on traffic flow based on an improved cellular automation model. Discrete Dynamics in Nature and Society, 2016, 14. https:// doi​.org​/10​.1155​/2016​/1236846. Le Pira, M., Marcucci, E., Gatta, V., Inturri, G., Ignaccolo, M., & Pluchino, A. (2017). Integrating discrete choice models and agent-based models for ex-ante evaluation of stakeholder policy acceptability in urban freight transport. Research in Transportation Economics, 64, 13–25. Madrigano, J., Kloog, I., Goldberg, R., Coull, B. A., Mittleman, M. A., & Schwartz, J. (2013). Longterm exposure to PM2.5 and incidence of acute myocardial infarction. Environmental Health Perspectives, 121(2), 192–196. https://doi​.org​/10​.1289​/ehp​.1205284. Maltese, I., Le Pira, M., Marcucci, E., Gatta, V., & Evangelinos, C. (2021). Grocery or @grocery: A stated preference investigation in Rome and Milan. Research in Transportation Economics, 87, 101096. Marcucci, E., Gatta, V., Marciani, M., & Cossu, P. (2017). Measuring the effects of an urban freight policy package defined via a collaborative governance model. Research in Transportation Economics, 65, 3–9. Marcucci, E., Gatta, V., Le Pira, M., Chao, T., & Li, S. (2021). Bricks or clicks? Consumer channel choice and its transport and environmental implications for the grocery market in Norway. CITIES, 110, article 103106. https://doi​.org​/10​.1016​/j​.cities​.2020​.103046.

The challenges of freight transport in cities  33

Miranda, M. L., Edwards, S. E., Keating, M. H., & Paul, C. J. (2011). Making the environmental justice grade: The relative burden of air pollution exposure in the United States. International Journal of Environmental Research and Public Health, 8(6), 1755–1771. https://doi​.org​/10​.3390​/ijerph8061755. Mohai, P., & Saha, R. (2007). Racial inequality in the distribution of hazardous waste: A national-level reassessment. Social Problems, 54(3), 343–370. Mohai, P., Pellow, D., & Timmons Roberts, J. (2009). Environmental justice. Annual Review of Environment and Resources, 34(1), 405–430. Mokhtarian, P. L. (2004). A conceptual analysis of the transportation impacts of B2C e-commerce. Transportation, 31(3), 257–284. https://doi​.org​/10​.1023​/ B​:PORT​.0000025428​.64128​.d3. Morganti, E. S., Seidel, C., Blanquart, L., Dablanc, B., & Lenz, B. (2014). The impact of e-commerce on final deliveries: Alternative parcel delivery services in France and Germany. Transportation Research Procedia, 4, 178–190. https://doi​.org​/10​.1016​/j​.trpro​.2014​.11​.014. Moridpour, S., Mazloumi, E., & Mesbah, M. (2015). Impact of heavy vehicle on surrounding traffic characteristics. Journal of Advanced Transportation, 49(4), 5325–5552. https://doi​.org​/10​.1002​/atr​ .1286. Panepinto, J. A. Sr. (2018). E-commerce industry's last mile needs create new demand for old warehouse space. National Real Estate Investor, October 4, 2018. Retrieved November 12, 2018 from: www​ .nreionline​ .com​ / industrial​ /e​ - commerce​ -industry​ - s​ -last​ -mile​ - needs​ - create​ - new​ - demand​ - old​ -warehouse​-space. Purdon, M., Giuliano, G., Wircover, J., Murphy, C., Ziaja, S., Kasier, C., … Seguin, J. (2021). Climate and transportation policy sequencing in California and Quebec. Review of Policy Research, 38(5), 596–630. Rodrigue, J.-P. (2004). Freight, Gateways and mega-urban regions: The logistical integration of the Bostwash Corridor. Tijdschrift voor Economische en Sociale Geografie, 95(2), 147–161. Rodrigue, J.-P. (2021). The geography of transport systems (5th ed.). London: Routledge. Retrieved from https://transportgeography​.org/​?page​_id​=1698. Sakai, T., Kawamura, K., & Hyodo, T. (2015). Locational dynamics of logistics facilities: Evidence from Tokyo. Journal of Transport Geography, 46, 10–19. SCDigest Editorial Staff. (2020). Supply chain news: Market for US warehouse space will remain tight in 2020, but shippers getting back some leverage. Supply Chain Digest, January 28, 2020. Retrieved from www​.scdigest​.com ​/ONTARGET​/20 ​- 01​-28​_TRENDS​_US​_WAREHOIUSE ​_ MARKETS​.PHP​ ?cid​=16289​&ctype​=content. Simoni, M. D., Marcucci, E., Gatta, V., & Claudel, C. G. (2020). Potential last-mile impacts of crowdshipping services: A simulation-based evaluation. Transportation, 47(4), 1933–1954. Sparkman, D. (2019). Retail-to-warehouse conversions are real. Material Handling and Logistics News. Retrieved from www​.mhlnews​.com​/warehousing​/article​/22055462​/retailtowarehouse​-conversions​ -are​-real​?utm​_ source​=linkedin​.com​&utm​_ medium​=social​&utm​_ campaign​=www​.mhlnews​.com​ -warehousing​-reta​&utm​_content​=28651315. Toussaint, M. (2021). Environmental racism and the quest for justice. META from the EEB (news channel of European Environmental Bureau), January 12, 2021. Retrieved from https://meta​.eeb​.org​ /2021​/01​/12​/quete​-de​-justice​-contre​-le​-racisme​-environnemental/. Yu, X., Zheng, L., Jiang, W., & Zhang, D. (2020). Exposure to air pollution and cognitive impairment risk: A meta-analysis of longitudinal cohort studies with dose-response analysis. Journal of Global Health, 10(1), 010417. https://doi​.org​/10​.7189​/jogh​.10​.010417. Yuan, Q. (2021). Location of warehouses and environmental justice. Journal of Planning Education and Research, 41(3), 282–293.

DATA SOURCES American Transportation Research Institute, Top 100 Truck Bottlenecks. Retrieved from https:// truckingresearch​.org​/atri​-research ​/ bottlenecks​-congestion​-infrastructure​-funding/. Eurostat data browser. Retrieved from https://ec​.europa​.eu​/eurostat​/databrowser​/view​/tran​_ hv​_frmod​/ default​/table​?lang​=en

34  Handbook on city logistics and urban freight

Federal Motor Carrier Safety Administration, Large Truck and Bus Crash Facts. Retrieved from www​ .fmcsa​.dot​.gov​/safety​/data​-and​-statistics​/ large​-truck​-and​-bus​-crash​-facts. Our World in Data. Retrieved from https://ourworldindata​.org​/co2​-emissions​-from​-transport. US Census Characteristics of New Housing. Retrieved from: www​.census​.gov​/construction​/chars. US Census, Foreign Trade Statistics data portal. Retrieved from www​.census​.gov​/foreign​-trade​/data​/ index​.html. US Department of Transportation, Bureau of Transportation Statistics, National Transportation Statistics. Retrieved from www​.bts​.gov​/topics​/national​-transportation​-statistics. US Department of Transportation, Bureau of Transportation Statistics, National Noise Map. Retrieved from https://data​.bts​.gov​/stories​/s​/ National​-Transportation​-Noise​-Map​/ri89​-bhxh/. US Department of Transportation, Bureau of Transportation Statistics, Moving goods in the United States. Retrieved from https://data​.bts​.gov​/stories​/s​/ Moving​-Goods​-in​-the​-United​-States​/ bcyt​-rqmu. US Department of Transportation, Federal Highway Administration, National Household Travel Survey 2017. Retrieved from https://nhts​.ornl​.gov/. US Department of Transportation, National Transportation Noise Map, March 2021. Retrieved from https://maps​.dot​.gov​/ BTS​/ Nat​iona​lTra​nspo​r tat​ionN​oiseMap/. US Environmental Protection Agency, National Emissions Inventory data portal. Retrieved from www​ .epa​.gov​/air​-emissions​-inventories​/national​-emissions​-inventory​-nei. World Trade Organization open data portal, WTO STATS. Retrieved from data​.wto​.org​​/en.

2. Integrated transportation and land-use program to improve metropolitan freight system performance José Holguín-Veras, Carlos Rivera-González, Benjamin Caron, Julia Coutinho Amaral, and Abdelrahman Ismael

INTRODUCTION Metropolitan areas are efficient markets, concentrating economic interactions, employment opportunities, and the trade of services and goods. These economic interactions require the movement of goods from one place to another, creating freight flows between the various stages of supply chains. Although large portions of supplies come from rural areas, metropolitan areas play an outsized role in manufacturing (about 80% of manufactured products in the US are made in metropolitan areas). Metropolitan areas concentrate the demand, and it is where the majority of the world’s population live, and where 68% of the world population is projected to live by 2050 (United Nations Department of Economic and Social Affairs, 2018). For example, about 80% of the cargo transportation in the United States (US) has either its origin or destination in the top 100 metropolitan areas. In spite of being essential for economic activities in the cities, the transportation of goods from suppliers—possibly passing through distributors outside of metropolitan areas, and then finally reaching receivers at the urban core—creates large amounts of negative externalities, including congestion, pollution, noise, and infrastructure damage, among others. The dichotomy between the benefits of transportation to the economy and the externalities it produces poses a dilemma; restricting freight activity to reduce externalities hampers the economy, causing huge economic losses. Social pressure can (and usually does) lead to bans and restrictions on freight traffic, creating detrimental effects on the economy, and potentially contributing more to the externalities that they were trying to address in the first place. Yet, research shows that the social objective of minimizing externalities is aligned with the objectives of the private sector of increasing supply chain efficiency and minimizing cost (Quak & de Koster, 2006; Holguín-Veras et al., 2011). This alignment of objectives is not coincidental; it reflects the deep interconnections between supply chains, the economy, and the externalities produced by freight activity. In this context, the best way to foster such efficiency of freight activity is to implement transportation and land-use initiatives that preserve and enhance the multiple benefits associated with the production and consumption of supplies, while, at the same time, addressing the associated externalities. Such nuanced policies are needed because of the pervasiveness of freight activity. The reality is that freight activity does not occur only at the large cargo facilities commonly associated with heavy freight traffic (e.g., ports, rail terminals, distribution centers). In fact, the majority 35

36  Handbook on city logistics and urban freight

of all freight activity takes place in the core of urban areas, where vans and small/mid-sized trucks deliver goods to establishments and households. To understand how pervasive freight activity is, it is useful to classify the various economic activities performed in metropolitan areas into two groups according to their freight demand needs: (i) freight-intensive sectors (FIS), and (ii) service-intensive sectors (SIS). FIS includes sectors that require the production and consumption of freight for its activities (e.g., manufacturing, restaurants, retail), while SIS includes sectors that are focused on services rather than physical goods (e.g., educational and financial institutions). These transactions among businesses are typically referred to as business-to-business (B2B) transactions, while those involving deliveries to consumers are labeled business-tocustomer (B2C). The importance of the latter stems from the rise of e-commerce and the externalities it generates. While B2B deliveries tend to be clustered in commercial areas, B2C deliveries are spread across the metropolitan area according to population densities. This scattered demand of smaller packages to residential areas, along with the increasing demand for expedited deliveries, impedes carriers from consolidating deliveries and using larger delivery vehicles, which leads to a sharp rise in the number of small trucks and vans in urban areas, increasing congestion and pollution even more. The e-commerce growth trend, discussed in Chapter 1 by Giuliano, Chapter 19 by Jaller et al., and Chapter 21 by Gatta et al., suggest that these externalities will be intensified in the near future. Worldwide sales of retail e-commerce reached a value of $3.53 trillion in 2020, and are expected to account for 23% of the global total retail sales by 2023 (Statista, 2020). In the US, for example, it is estimated that B2C deliveries account for 55% of all deliveries (Holguín-Veras et al., 2020c). Adding urgency to the implementation of proactive freight management efforts are the potential large increases in freight traffic associated with the “on-demand” economy, i.e., deliveries expected to be made within an hour or less after the order was placed. Ironically, given their economic importance and success, most metropolitan areas are not designed to support efficient supply chain activity. The sad reality is that both transportation and land-use policy, more often than not, either ignore the unique needs of freight activity or tend to over-rely on regulatory approaches that often produce unintended effects. There is a major need to deploy enlightened transportation and land-use initiatives that work together to simultaneously address the private objective of enhancing freight efficiency and the social objective of minimizing externalities. Of great interest in this chapter is the potential role of freightefficient land uses (FELU) initiatives, which are defined as land-use policies that minimize the external and private costs related to the production, transportation, and consumption of goods. Throughout this chapter, the term “initiative” is used to refer to policies, programs, and projects. Transportation initiatives are actions that could impact the supply side and/or the demand side of transportation. They are subdivided into seven categories: (1) infrastructure management, (2) parking/loading area management, (3) vehicle-related strategies, (4) traffic management, (5) pricing, incentives, and taxation, (6) logistical management, and (7) freight demand management. FELU initiatives are subdivided into six categories: (1) long-term planning, (2) zoning, (3) site/building design, (4) facility/infrastructure management, (5) parking/loading area management, and (6) pricing, incentives, and taxation. The inclusion of FELU initiatives is justified by the necessity of considering land-use planning to foster harmony among the social and economic activities that compete for space in urban areas. The fundamental tenet of this document is that the main goal of transportation and FELU management and planning is to maximize the benefits associated with the production and

Integrated transportation and land-use program  37

consumption of goods, while mitigating the externalities generated by the resulting freight traffic. To achieve this goal, transportation and FELU initiatives must: (1) Gain insight into the effects of trends in technology and the economy, particularly novel freight technologies and e-commerce, that are fostering changes in consumer behavior, transportation systems, and land use; and (2) Exploit the synergies that could be derived from the joint use of transportation and FELU initiatives. To understand why it is important to exploit the synergies between transportation and FELU initiatives, it suffices to highlight the interconnections between transportation activity and land-use patterns, which are the two sides of the proverbial coin. Essentially, land-use patterns influence transportation and, conversely, transportation accessibility influences land-use patterns. Of great importance to this chapter are the effects of land-use patterns on supply chains. Land-use patterns that lead to compact supply chains—with supply chain stages that are relatively in proximity to each other—significantly reduce the total amount of freight vehicle kilometers traveled (VKT). In contrast, dispersed supply chains—with stages that are relatively far from each other—translate into increasing amounts of VKT for the same level of economic activity. Increasing VKT is a problem for both the private sector which experiences higher transportation costs and for the communities that are impacted by the externalities produced by the freight vehicles. Because of the two-way interactions between transportation and land use, single-sided approaches that influence only one side are bound to be ineffective in the long run. The best outcomes will be obtained if transportation and land-use initiatives are used in coordination for maximum effectiveness. In this way, FELU initiatives would foster more compact supply chains while transportation initiatives aimed at reducing the externalities that impact local communities that live near large traffic generators such as urban distribution centers, would make it easier for these facilities to locate closer to the areas with high concentrations of deliveries. Another important reason to proactively implement coordinated transportation and FELU initiatives is the long-lasting effects of land-use policy and planning which can last for decades and even centuries. It is key to consider the longterm implications of land-use policy on freight transportation activity and the associated externalities. This chapter is divided into three additional sections. The second section provides an overview of transportation and land-use management and planning. The third section elaborates on possible transportation and FELU initiatives that can improve the efficiency of supply chains. Finally, the fourth section presents concluding remarks.

OVERVIEW OF TRANSPORTATION AND LAND-USE MANAGEMENT AND PLANNING This section provides an overview of the freight management process necessary to achieve the objectives outlined in the previous section. As part of the discussion, the basic principles of the suggested process are outlined together with the major components of freight transportation and land-use planning.

38  Handbook on city logistics and urban freight

Improving the efficiency of the freight system requires integrating the needs and impacts of freight into the process of planning and maintaining urban areas. The following seven principles offer guidance for planning and designing more efficient freight systems. These principles are meant to guide practitioners and policymakers alike across a variety of fields on a path to better incorporate freight into current planning and management. They support the goal of maximizing the economic benefits of freight in urban areas while minimizing its negative impacts on other roadway users, the environment, and local communities. The principles in this section build on principles from the NCHRP 08-111 Guidebook “Effective DecisionMaking Methods for Freight-Efficient Land Use” (Holguín-Veras et al., 2020c), and have been expanded to include guidance for both transportation and land use. The principles are: ●









Minimize the private and external costs of supply chains and their various stages: Making a concerted effort to minimize the cost of freight for both the private sector and the local community is crucial for making efficient decisions. In particular, it is important to consider the external costs, which are often difficult to measure and neglected in many business decisions. Understanding these costs means that decision makers must have a working knowledge of the impacts of freight on all stakeholders and across a broad geographic region. This principle underpins all aspects of freight transportation, from locating sites of production and storage facilities to managing the road network and the vehicles using it. Mitigate, or eliminate, the externalities at supply chain nodes and large traffic generators (LTGs): The freight traffic at facilities that handle large amounts of freight—manufacturing sites, warehouses, and distribution centers—can cause major disturbances to transportation systems and local communities. Efforts to plan and manage these types of facilities can substantially reduce the externalities produced. By incorporating novel infrastructure, operational schemes, and/or demand management strategies, these facilities can reduce the negative impacts experienced by the supply chain and the surrounding community. Reduce the distance traveled at supply chain stages, up- and downstream: Supply chains can be lengthy and complex, involving the transfer of freight at numerous facilities including suppliers, manufacturers, distribution centers, and retail stores. The further that a vehicle travels, the more pollution, fuel costs, and vehicle and infrastructure deterioration it produces. By supporting the development of more compact supply chains, urban planners can improve the efficiency of supply chains by reducing the freight vehicle miles traveled. Land-use decisions and transportation policies can guide freight stakeholders to densify logistics facilities that are part of the same supply chain, reserve strategic areas for key logistics facilities, and improve roads and other transportation infrastructure that connects these facilities. Recognize and account for local conditions: Urban planners must recognize the importance of freight to local economies and acknowledge their own role in supporting, or opposing, freight efficiency. By focusing on the local context and the existing conditions, practitioners can more precisely and accurately identify ongoing issues regarding freight activity and their potential causes. Doing so minimizes the risk of implementing projects that cause unintended negative consequences. Understand the behavior of participants in supply chains: It is imperative to understand how the various participants in supply chains—shippers, carriers, receivers, and

Integrated transportation and land-use program  39





others—will react to public-sector decisions, as doing so minimizes the risk of unintended consequences. For example, if faced with restrictions on large trucks, freight carriers will likely rely on a larger number of smaller vehicles, which could have the unintended effect of increasing congestion. Of particular importance is to consider the role played by receivers and households, as they create much of the demand for freight transportation. These issues will be further detailed in Chapter 16 by Le Pira et al. Proactively engage all stakeholders in transportation and land-use decisions: It is important to include freight stakeholders in planning and managing transportation and land use, as they often provide valuable ideas, advice, and feedback. In addition, stakeholders will have specialized knowledge of what the most pertinent freight issues are and potential solutions to address these issues. Engaging with stakeholders during the planning and policymaking processes will allow for a better understanding of how a particular decision will impact the freight community and reduce the risk of unintended negative consequences. Stakeholders include those impacted by the production and transportation of freight, such as residents and local businesses. A comprehensive list of urban freight stakeholders is presented in the overview chapter on stakeholder engagement (Chapter 15 by Browne and Goodchild). Employ a rational process for evaluation and selection of initiatives: During the planning process, there will come a time at which planners must select the preferred project or program alternatives. By this stage, a specific transportation challenge has been identified, along with a set of potential initiatives on how to solve it. It is critical for planners to have a thorough appreciation of each alternative’s potential benefits, impacts, challenges, costs, and time for implementation. This allows the public sector to consider the potential economic, political, and time constraints they could face while implementing the alternative. In essence, it is important to ensure that the benefits attributable to each initiative selected for implementation are materially larger than the associated costs.

The remainder of this section outlines the general process that transportation and land-use practitioners and policymakers should follow to manage and plan urban freight systems. At the risk of oversimplifying, they do this by setting goals and objectives, developing a set of potential ways to achieve these goals, and evaluating and selecting the alternative with the best-expected outcome. The steps outlined below are not meant to be prescriptive, but instead offer a general process applicable to a wide range of conditions. The decision-making process starts with the identification of a freight issue and ends with implementing a solution and looking back to study the effectiveness of the solution in addressing the issue. Definition of Goals and Objectives to Be Pursued Once the freight issue to be addressed and the context surrounding the issue have been identified, the next step toward solving the issue is to set goals and objectives. For more details on freight likely issues to be addressed in cities, see Chapter 1 by Giuliano. The goals should represent values that are shared among all stakeholders so that everyone is invested in finding a solution. In addition, when all stakeholders are included in the process, it minimizes the risk of making decisions that cause unintended consequences. In most cases, goals and objectives express the desire to achieve higher levels of performance and/or mitigation or solution of issues that impact many stakeholders.

40  Handbook on city logistics and urban freight

Within the process of setting goals and objectives, it is important to recognize the role and responsibilities of every stakeholder in achieving these criteria. Regarding freight issues, transportation practitioners should ensure that the needs and impacts of freight vehicles are sufficiently integrated into the planning process. Similarly, land-use practitioners should strive to develop FELUs to better incorporate freight facilities into urban land-use planning and policymaking. The Guidebook for NCHRP 08-111 “Effective Decision-Making Methods for Freight-Efficient Land Use” defines FELUs as follows: Freight-Efficient Land Uses are the land-use patterns that minimize the social costs (private plus external costs) associated with both the supply chains and the economic activities that consume and produce goods, at all stages of production and consumption, including reverse and waste logistics.

This definition highlights the importance of considering both private and external costs in considering the impacts of freight on society. The private cost is the cost incurred by the private sector, such as costs for the construction and operation of facilities, whereas the external cost is the impact of freight facilities on the broader community, such as pollution by vehicles and factories that damages public health. The social costs are the sum of the private and external costs. Especially in the case of freight facilities, which often have large external costs, the social cost can be substantial. Because of this, land-use planning and business decisions need to consider the total social cost, along with the private-sector cost, to avoid excessive but preventable negative impacts on local communities. Understanding Local Conditions and the Root Causes of Key Issues Understanding the local conditions and root causes of the issues to be addressed is one of the most important steps: ignoring the local context or identifying the wrong sources of issues could steer the entire effort down the wrong path. A careful study of the local context will likely lead practitioners to the key sources of the freight issue they are addressing. Efforts to understand local conditions typically include both technical analyses, such as traffic counts, and qualitative studies, such as discussions with stakeholders. Involving freight stakeholders in this step is especially important because, although freight vehicles are the most visible signs of freight activity, the sources of the problem may lie elsewhere. Often, receivers— such as retail stores and households that order goods online—play an important role in both producing and solving freight issues, because they are the ones that create the freight traffic. Discussions with stakeholders, who are often more familiar with certain aspects of the freight system than practitioners who make freight decisions, may uncover useful knowledge about sources of freight issues and potential solutions. While studying the local conditions, it is important to gain insights into the composition of the local economy and to understand how pertinent the transportation of freight is to the economic activity of a region, as well as its negative impacts on other stakeholders. A lack of understanding of the role of freight transportation on the local economy could result in harsh restrictions on freight vehicles, negative unintended consequences, and harm to the local economy. In addition, understanding the spatial patterns of the establishments involved in supply chains as producers, transporters, or consumers of the supplies is important to gain insight into patterns of freight traffic as well as potential problem areas. Estimating the amount of freight that is transported within an area each day can also help practitioners gain an

Integrated transportation and land-use program  41

understanding of the magnitude of the needs of the economy. Chapter 5 of the Guidebook for NCHRP 08-111 “Effective Decision-Making Methods for Freight-Efficient Land Use” details how to use population, number of establishments, or employment as indicators to estimate freight flows and traffic. More detailed models estimate freight volumes using the number of establishments along with the industry sector and the employment size of each (Holguín-Veras et al., 2017). Practitioners can easily apply these models in their regions using the Freight and Service Activity Trip Generation Software (FASTGS) developed at Rensselaer Polytechnic Institute in the USA. Definition of Key Performance Measures In general, the goal of urban freight planning and management is to maintain and improve the efficient movement of goods. Although there is no perfectly efficient freight system, the goal is to achieve a more efficient freight system. There are many ways to measure efficiency, and the term often refers to the simultaneous combination of many metrics. One broad view of an efficient freight system is one that minimizes the negative impacts of freight—including pollution, noise, congestion, and safety concerns—while maximizing the benefits of freight activity and the economic competitiveness of an urban area. Nonetheless, it is important to have quantitative performance measures to know the degree to which the goals and objectives have been achieved and the effectiveness of the implemented solution. For example, decision makers may consider both travel-based metrics, such as vehicle miles traveled or the total emissions created by freight vehicles, and quality-of-life and livability metrics, such as the impact of congestion and the number and severity of accidents involving freight vehicles. For a more detailed discussion, refer to Barber & Grobar (2001) and Gunasekarana et al. (2004). Role of Land-Use Management and Planning A central way that communities manage freight and the externalities produced by freight activity is through land-use plans and policies. Therefore, it is critical to design land uses that are conducive to supporting the efficiency of freight. Land-use planners should pursue a FELU plan to maximize the benefits of freight while minimizing the negative externalities from freight. The overall process of creating a FELU plan can be divided into three critical steps: first, understanding the local conditions of the city or region; second, identification of priorities and opportunities (based on meticulous analysis of local conditions); and third, selection of FELU initiatives that best suit the local context. Throughout these steps, it is critical to have continuous stakeholder engagement, as insights from both the public and private sectors are necessary to develop a robust and flexible FELU plan capable of shaping the landuse ordinance and transportation of the region. In order for a FELU plan to be effective, it is critical that it be built on an understanding of the local conditions of the geographic area. A solid assessment of land-use patterns, the geographic distribution of economic activities, and supply chain patterns in the area is required to understand the current context holistically. Furthermore, it is key to consider the importance of the FIS of the economy. In addition, the identification of large traffic generators, defined as infrastructures that either generate or attract a large number of freight vehicles, is critical to pinpoint which buildings and facilities generate or attract a significant amount of freight vehicles. Partnering up with communities and the private sector is key to understanding the

42  Handbook on city logistics and urban freight

problems they currently face. Furthermore, the private sector is a powerful ally for gaining qualitative knowledge about the supply chain activities. Overall, gaining solid knowledge about local conditions is indispensable, given that it works as the foundation for the next two steps. Once a solid understanding of the local conditions is obtained, the next step is to identify the priorities for the FELU plan. The fundamental questions for the identification of freight priorities are: What are the most pressing freight problems that impact both local communities and the public sector? And what land-use planning and regulations can help to address the problems identified? It is key that during the identification of priorities the public sector is aware of all the potential opportunities that FELU plans introduce to traditional planning, such as land-use reservation for urban logistical spaces. It is also crucial that the public sector is well informed about the economic and technological trends that could impact their communities in several ways, such as the rise of e-commerce deliveries. The last step to achieve a successful FELU plan is to use a context-based criterion to select a set of initiatives that match the objectives. Similar to how “one-solution-fits-all” never works, policy planners should consider the unique challenges and opportunities their geographic area has, and which set of FELU initiatives (complementary land-use and transportation initiatives) would be the most beneficial to them. In essence, each city should take ownership of both the objectives of the FELU plan and the steps—in the form of FELU initiatives—that they will follow to achieve their plan. Lastly, it is critical to reiterate how getting the feedback and guidance of communities, the private sector, and academia is key to achieving a robust and flexible FELU plan. Preliminary Identification of Potential Initiatives The goal of this step is to select the initiatives that have the greatest potential to solve a particular freight issue with as few negative impacts as possible. In this sense, “initiatives” refer to projects, programs, regulations, policies, or other mechanisms that can help practitioners and planners achieve the desired objective. By the end of this step, the search for the best solution is narrowed down to a more manageable set of alternatives. During this process, practitioners should consider the nature of the problem, the geographic scope, the problem source, the financial and practical constraints, and stakeholder concerns, among other factors. To help guide the selection of potential initiatives, our research team created the “Initiative Selector for Fostering Freight System Performance and Freight-Efficient Land Use”, a webpage that allows practitioners to input a scenario and provides them with suggestions about potential initiatives that could be solutions to the specified problem (Holguín-Veras et al., 2021). Each initiative is accompanied by a detailed description, an assessment of the expected cost and risk of unintended effects, and examples of situations where the initiative has previously been applied. The Initiative Selector can be found at https://cite​.rpi​.edu​/iselector/. This tool is not meant to replace the traditional design process, but rather to help practitioners identify alternatives they might not otherwise consider. The Initiative Selector is designed to be used in many contexts since it provides both land-use initiatives developed as part of NCHRP Project 08-111 (Holguín-Veras et al., 2020c) and complementary transportation initiatives that were updated from NCFRP Report 33 (Holguín-Veras et al., 2015). As part of the process of selecting potential initiatives, it is useful for practitioners and researchers to examine previous experiences to gain insight into the effectiveness of an initiative for addressing a particular issue. In addition to the Initiative Selector, several articles

Integrated transportation and land-use program  43

and reports have been developed to consolidate knowledge on best practices and document examples of the implementation of some initiatives. In a series of two papers, Holguín-Veras et  al. (2020a), Holguín-Veras et  al. (2020b) review a multitude of public-sector initiatives, evaluate the performance of each, and use a survey to investigate practitioners’ experiences with transportation initiatives and their perceptions of the positive and negative impacts of each. SUGAR (2011) identifies and describes dozens of examples of best practices regarding urban logistics that can be implemented by the public sector. BESTUFS (2007) guides both the public and private sectors on good practices regarding urban freight. Formulation and Performance Analysis of Solution Alternatives Once the preliminary identification of potential initiatives is done, the next step is to study each in detail. Although the study of these initiatives should be led by the public sector, the participation of key stakeholders is critical to developing a thorough appreciation of their potential benefits and impacts. For estimating the performance of the potential initiatives, it is necessary to be thorough, but pragmatic. Specifically, efforts related to data collection and modeling must be proportionate to the scope and potential impact of each initiative. In essence, data collection efforts and modeling exercises should be reserved for the most critical and impactful projects. For those initiatives that require data collection efforts, it is key to obtain information about each possible alternative in the initiative, including cost, time, and effort for implementation, and potential risks and benefits. The public sector would lead a process to have, in detail, a clear idea of what each initiative involves. For the modeling process, it is critical to not get carried away by the technocratic side. Robust models are always necessary to represent reality in an appropriate way, however, it is also critical to have insight from key stakeholders such as the private sector and the community. In-depth interviews with the stakeholders provide a different perspective and, ultimately, help planners achieve a holistic view of the potential initiatives. The expected outcome of this analysis is a robust identification of the potential initiatives highlighting their challenges and opportunities, in addition to a preliminary conclusion regarding the merit of each initiative. For a review of traditional and promising modeling approaches, the reader can refer to the overview chapter (Chapter 3) by Tavasszy and de Bok and Chapter 4 by Comi and Delle Site, Chapter 5 by Sakai et al., and Chapter 6 by Sánchez-Díaz and Castrellon. Evaluation and Selection of Preferred Alternatives The process of evaluating and selecting alternatives requires judgment on how each potential initiative would meet the goals and objectives mentioned earlier in this chapter. Ex-ante evaluating stakeholder policy acceptability supports policy makers in taking well-thought-out decisions (Le Pira et al., 2017). Although there are several evaluation and selection techniques, they can be categorized into two main groups. The first group contains economic techniques that aim to translate any potential impacts into monetary values. Even non-market impacts, such as environment-related impacts, are quantified into monetary values using valuation techniques (Bateman et al., 2002). Once all potential impacts produced during the project’s economic lifespan are translated into monetary values, they are categorized as either benefits or costs. Lastly, economic indicators such as Net Present Value, Internal Rate of Return, and Benefit/Cost Ratio are produced.

44  Handbook on city logistics and urban freight

The second group of evaluation techniques is multi-criteria techniques, which do not translate impacts into monetary values (see Chapter 13 by Delle Site). For this evaluation method, customarily a matrix that associates the performance of each alternative with the objectives of the project is conducted. If quantitative analysis is available, these values could be placed in the corresponding cells. Contrastingly, if qualitative analysis is available, an indicator of the impact should be placed in the corresponding cell, i.e., low, medium, or high. It is critical that, regardless of the method selected, the stakeholders are involved in the selection criteria of initiatives. Stakeholder agreement on the evaluation of alternative initiatives is key to assuring their long-term support and to advancing with their implementation. Usually, there is no clear best alternative because of the trade-offs involved. In those cases, it is critical to get input from all of the stakeholders about the relative importance of different decision criteria, as this will assist the selection process. Creation of Action Plan Once there is a consensus with respect to which initiatives are going to be implemented, the next step is to create an action plan that identifies the recommended policies, programs, processes, and initiatives to be implemented. This plan should be presented as a draft to all of the stakeholders, advisory groups, and agency leadership for their review and comments. It is key to collect information about the reactions and concerns to the draft to address any issues. For initiatives that require a detailed engineering design, additional data must be collected to support the design. Lastly, in some cases, pilot tests could be planned to gain insight into the potential benefits, costs, challenges, and opportunities of the proposed initiative. The output of this action plan is the specific actions needed to implement the initiatives selected. Pilot Testing and Implementation In cases where there is no consensus about the anticipated performance of a potentially transforming initiative, conducting well-designed pilot tests could be an excellent course of action. First, it demonstrates to the stakeholders that the public sector is taking all necessary precautions before implementing an initiative that could potentially impact them. Second, it allows for quantifying the real-life impacts of the initiative where a small group of stakeholders experience the initiative and can give feedback. Third, it reassures stakeholders that only successful pilot-tested initiatives will be implemented in the long term. For pilot testing to be meaningful, it must be designed carefully. Poorly designed pilot tests could create scenarios where inadequate initiatives appear to be successful or adequate initiatives appear unsuccessful. Follow-Up, Reassessment, and Modification Planning and management are best seen as continuous processes given how the needs and issues of cities are in constant flux. As a result, the public sector should periodically revisit and reassess the objectives and goals they want to achieve. A rigorous evaluation of recommended and implemented initiatives is vital to assess what worked and what needs to be adjusted to achieve the goals and objectives. It is important to monitor the programs and initiatives to ensure that they work as expected.

Integrated transportation and land-use program  45

Another benefit of following up and reassessing is to reassure the stakeholders that the public sector is being as meticulous as possible in ensuring the success of the suggested initiatives. Furthermore, by building trust with stakeholders, it creates the perfect scenario for future synergetic freight and land-use planning in conjunction with the private sector and local communities. This participative approach, if properly conducted, can lead to the implementation of both impactful and bold initiatives that will have long-lasting effects in the city. Stakeholder Engagement To ensure the success of transportation and land-use management and planning, it is key to engage stakeholders in all stages of the decision-making process. In essence, such engagement with the private sector and communities is what enriches the understanding of the freight problems and provides invaluable insight into how to solve them. Usually, policy decisions related to transportation and land-use management can result in unintended problems (Jones et  al., 2009). However, with healthy interaction between stakeholders, these unintended impacts can be reduced to a minimum. It should not come as a surprise that the private sector and the community play a key role in understanding the local conditions of urban freight. In addition, private sector participation in pilot tests is critical to determining whether a potential freight initiative is successful or not.

OVERVIEW OF TRANSPORTATION AND LAND-USE INITIATIVES The objective of this section is to provide the reader with an overview of different groups of initiatives that can be implemented to enhance supply chain efficiency and support FELU planning. The groups range from supply-oriented to demand-oriented for both transportation and FELU initiatives. Within each group, there is a further subdivision considering how initiatives within a group can be heterogeneous. It is important to point out how there are some initiatives, named joint initiatives, that require collaboration between transportation and land-use agencies in order to be implemented. Table 2.1 shows the categorization of initiatives (Holguín-Veras et al., 2020c). Transportation Initiatives This subsection provides a brief description of the groups of transportation initiatives, all of which aim to enhance the efficiency of supply chains while minimizing externalities. Although implementing a single initiative can produce positive results, these initiatives should not be perceived as mutually exclusive solutions, rather they are complementary in some cases and standalone in others. Ultimately, combinations of initiatives can be explored to produce synergy among them and maximize potential benefits, e.g., last-mile delivery practices with on-street parking and loading. For further information about each initiative and worldwide implementation examples, the reader can refer to Holguín-Veras et al. (2015, 2020a, 2020b, 2020c).

46  Handbook on city logistics and urban freight

Table 2.1  Transportation and FELU initiatives Transportation initiatives

FELU initiatives

Group

Subgroups

Group

Subgroups

Facilities/infrastructure management*

Major improvements

Long-term planning

Planning tools

Parking/loading area management*

On-street parking and loading

Minor improvements

Strategies Zoning

Off-street parking and loading Vehicle-related strategies

Technologies and programs

Traffic management

Access and vehiclerelated restrictions

Regulatory controls Discretionary approaches

Site/building design

Site design building design

Traffic control and lane management

Facilitiy/infrastructure management*

Major improvements

Pricing

Parking/loading area management*

Off-street parking and loading

Incentives

Pricing, incentives, and taxation

Pricing

Time access restrictions

Pricing, incentives, and taxation

Taxation Logistical management

Intelligent transportation systems

Incentives Taxation

Last-mile delivery practices Freight demand management Demand management Note:   * denotes groups which contain joint initiatives

Facilities and infrastructure management This group of initiatives enhances freight mobility through investments in infrastructure and facility improvements. Such investments are necessary due to increases in freight traffic volume and truck sizes time, which ultimately leads to outdated and inadequate infrastructure. Within this group of initiatives there are two subgroups. The first one, “Major Improvements”, as the name suggests, considers major infrastructure projects, whereas the second group, “Minor Improvements”, deals with smaller projects that are more economical and faster to implement. Major improvements This subgroup focuses on large-scale improvements that require a significant amount of time, cost, and planning efforts. Nonetheless, these major-infrastructure projects shape the development of the region or urban area. Building new infrastructure, e.g., ring roads, or upgrading the existing one to accommodate freight traffic, e.g., widening the turning radii of roads for large trucks, are examples of this subgroup.

Integrated transportation and land-use program  47

Minor improvements The second subgroup of initiatives considers minor infrastructure improvements that do not require as many resources. This subgroup focuses on localized interventions that aim to fix or improve a particular local problem. These infrastructure changes can be on the roadway, e.g., implementing acceleration and deceleration lanes to allow smooth merging of trucks on roadways. These changes can also target sidewalks, e.g., adding ramps for handcarts and forklifts to facilitate the loading and unloading process, hence reducing parking time of delivery trucks and increasing parking capacity. Parking and loading area management Usually, city centers have parking problems due to limited parking supply and high demand for parking spaces. The heart of the issue is the competition for parking spaces between different types of vehicles, i.e., private cars, public transit, and freight vehicles. Given the operational constraints freight vehicles face, e.g., time window constraints, they tend to double-park or occupy sidewalks, resulting in the necessity for a group of initiatives dedicated to the management of parking and loading areas. Within this group of initiatives there are two subgroups. The first, onstreet parking and loading, targets improving on-street parking conditions, whereas the second subgroup, off-street parking and loading, is concerned with off-street parking spaces. On-street parking and loading This subgroup of initiatives focuses on the better management of available curb space in cities. Customarily, curb spaces can be used for either commercial, transit, or private cars. By redistributing the inventory of available curb spaces and increasing the amount of dedicated freight vehicle parking, negative externalities, e.g., congestion or emissions, can be alleviated. Furthermore, by introducing a vehicle parking reservation system, drivers would reserve curb space in advance, reducing possible cruising or double parking, and ultimately improving local traffic conditions. Although most initiatives in this subgroup address parking regulations, they do not necessarily target increasing on-street parking capacity, which is the case of peak-hour clearways, which prioritize road capacity by prohibiting curb-side parking or stopping during specified times. The success of the implementation of these initiatives depends on local demand conditions and whether parking or road capacity should be prioritized to enhance efficiency. Off-street parking and loading The second subgroup of initiatives addresses facilities that have dedicated spaces for freight vehicles, e.g., loading docks or staging areas, that are off-street. By enhancing building codes or upgrading the requirements needed for facilities that attract large amounts of cargo, e.g., shopping malls or large stores in Central Business Districts (CBDs), parking-associated traffic near those facilities would be alleviated. Another possibility is to implement time-sharing parking initiatives that would allow different road users to share off-street parking spaces during specific times of the day. Off-street parking is strongly related to building codes; by updating building codes to include clear design guidelines and specifications for dedicated freight areas, developers will design and build facilities with appropriate parking spaces and loading areas that meet the needs of the freight vehicles that serve the building. Vehicle-related strategies This group of initiatives exploits the available vehicle technologies and practices to enhance supply chain efficiency, while minimizing negative externalities of noise and emissions; refer to Chapter 23 by McKinnon, which stresses the importance of switching to freight vehicles

48  Handbook on city logistics and urban freight

with cleaner, lower carbon energy, and Chapter 9 by Lebeau et al. New technologies are incentivized by regulations such as emission standards for freight vehicles or low-noise delivery programs. It is known that adopting vehicles with new technologies might require large initial investments from carriers, so there is a synergy opportunity between the initiatives in this subgroup and the initiatives in the pricing, incentives, and taxation group. Furthermore, the adoption of new technologies requires the public sector to invest in adequate infrastructure for such new technologies, e.g., charging stations for electric vehicles. Lastly, there is an opportunity for synergy with initiatives of the freight demand management group, e.g., low-noise delivery programs while doing off-hour deliveries, since it addresses one of the major concerns of the local population (noise caused by urban deliveries) while doing deliveries in the night. Traffic management This group of initiatives focuses on traffic control measures and restrictions to improve traffic conditions by reducing congestion, emissions, and conflicts between freight vehicles and other network users. These initiatives can be grouped into three subgroups: (1) access and vehiclerelated restrictions, (2) time access restrictions and traffic control, and (3) lane management. As discussed amply in Holguín-Veras et al. (2015), many of the initiatives discussed here have a significant risk of causing negative, unintended effects. These risks could be mitigated by means of consultations with the key participants in supply chains, such as freight carriers, shippers, and receivers. Access and vehicle-related restrictions This subgroup focuses on initiatives that use restrictions to control the access of trucks to certain areas or specific roads. The restrictions can be based on vehicle size, weight, payload, or even type of engine. Usually, the goal of the restrictions is to protect the structural integrity of roads, bridges, and tunnels from overweight and/or oversized trucks, but they can also be applied to improve livability by constraining certain engine specifications and emissions. However, it is important to keep in mind that restrictions might generate unintended consequences. For instance, restricting the traffic of large vehicles promotes the substitution of one large vehicle for many smaller vehicles to meet the equivalent demand, increasing the number of vehicles on the road and increasing congestion even more. Also, most of the initiatives in this subgroup promote the diversion of truck traffic to other routes that will end up receiving larger traffic volumes, potentially causing congestion and damage on those routes. Therefore, it is important to carefully consider the implementation of any type of restriction, so as not to transfer the issue from one location to another, or even exacerbate the issues further. Another aspect that has to be taken into consideration with initiatives that target sustainability and livability by reducing emissions through constraining engine specifications is the high investments they would require from carriers to renew their fleets. Instead of promoting restrictions, the public sector can incentivize the process of fleet renewal by implementing initiatives that provide preferential parking spots for environmentally friendly vehicles in urban areas where parking is usually a major issue. Finally, it is also possible to restrict truck traffic by mandating predesignated routes for trucks, or by offering information about the geometry and structure of routes that can accommodate each truck type. This type of initiative protects the integrity of infrastructures, reduces congestion, emissions, and dangers around sensitive facilities like schools, and grants more protection to pedestrians and cyclists. The designated truck routes must have good enough infrastructure

Integrated transportation and land-use program  49

to receive the additional traffic flow, and preferably must be in locations where local communities will not be affected by the externalities generated by the traffic of freight vehicles. Time access restrictions This subgroup of initiatives utilizes time restrictions and bans to limit deliveries to specific times to reduce congestion and emissions. These restrictions differ in terms of specified times, which can be in the daytime or in the nighttime. Usually, these initiatives are not well received by establishments because they might require businesses to stay open after their regular working hours to receive goods. However, they are welcomed by citizens as they reduce congestion and increase the livability of urban areas. Nevertheless, since the receivers have the power to decide the time of deliveries, restricting the traffic of large freight vehicles during a certain period of day might force carriers to deliver the goods in smaller vehicles to bypass the restriction and meet the receivers’ requests. With the reduced capacity of smaller vehicles, carriers are also forced to use more vehicles to meet the demand of receivers. This phenomenon motivates a shift from fewer larger delivery vehicles to more numerous smaller delivery vehicles, increasing the total number of freight vehicles traveling in the city and worsening congestion. Traffic control and lane management These strategies promote the efficient use of available road capacity by studying the traffic conditions and by promoting the separation of different types of vehicles. This subgroup also aims to improve the safety and mobility of all road users by segregating vehicle types where possible. This separation can be achieved via multiple initiatives that dedicate certain lanes for a specific type of vehicle, or by assigning periods of time through the day for the traffic of a certain type of vehicle. The separation of different types of vehicles can improve safety and mobility. Another initiative that serves the purpose of efficient allocation of road capacity is enhanced traffic impact analysis (enhanced TIA). This initiative widens the scope of traditional traffic impact studies by integrating measures to reduce externalities of freight traffic of new developments within the surrounding areas. Enhanced TIA integrates many of the initiatives proposed in this chapter, e.g., off-hour delivery programs, time slotting of pickup and deliveries, among others. Enhanced TIA can be also integrated with some of the FELU initiatives (discussed later in this chapter), e.g., conditional use requirements. The goal of enhanced TIA is to require the implementation of such initiatives when designing new developments which allows early assessment and mitigation of freight traffic impacts on local communities. However, this requirement would translate into an increase in the time needed for planning and designing such new developments. Pricing, incentives, and taxation The common aspect of this group of initiatives is that they rely on incentives, both monetary and non-monetary, to foster freight efficiency. The strategy is to use incentives to promote desired behaviors and achieve public goals such as revenue generation, promotion of novel technologies, among others. Pricing The main objective of pricing initiatives is to foster a more rational use of the existing capacity of infrastructure and the reduction of freight externalities. The two initiatives in this group are road pricing and parking pricing. In theory, incrementing the transportation cost at a certain

50  Handbook on city logistics and urban freight

location or period will promote a reduction of traffic. However, research shows that carriers have limited power regarding the time of the deliveries (Holguín-Veras et al., 2006), which is controlled by the receivers, making road pricing an ineffective strategy by which to foster carriers’ behavior changes. Still, road pricing can be implemented to generate revenue that can be used to foster other initiatives, making it a good alternative to promote freight efficiency when combined with other initiatives. However, parking pricing can foster behavior changes when implemented by itself. It can be used to promote proper allocation of curb space, improve traffic conditions by reducing parking dwell times (Taniguchi et al., 2012), and also to generate revenue. However, for parking pricing to work as intended, it must be accompanied by the provision of a suitable number of spaces for use by freight vehicles. Incentives Monetary and non-monetary incentives are given by programs that aim to foster sustainable practices. The combination of incentives and penalties is likely to have an impact on the behavior of freight agents. For example, the public sector can provide incentives for carriers that use environmentally friendly vehicles, and charge penalties to carriers that use vehicles out of the desired standards. The initiatives in this group are recognition programs, certification programs, and operational incentives for electric/low emission vehicles. Recognition and certification programs are voluntary programs in which businesses can be publicly acknowledged for their good practices. Operational incentives for electric/low emission vehicles provide incentives, such as preferential access to restricted areas or lanes to foster the use of environmentally friendly vehicles (BESTUFS, 2007). For example, in Norway the use of “clean vehicles” for last-mile deliveries gives access to priority lanes. Taxation Taxation is a common strategy to raise revenues and incentivize behavior changes for the public interest. The public sector can regulate taxes of goods and activities according to the desired behavior. An example is tax incentives for the purchase of electric vehicles or for companies that use environmentally friendly equipment (City Ports, 2005; US Environmental Protection Agency, 2013). Logistical management This group of initiatives targets reducing negative externalities by introducing new technologies or promoting changes in last-mile delivery practices. These types of initiatives can play a major role in increasing the energy efficiency of urban freight movement as pointed out in Chapter 23 by McKinnon. Intelligent transportation systems (ITS) ITS can greatly help the freight sector in many ways: it can provide real-time information, guide trucks through alternative routes, and warn drivers when their trucks exceed height restrictions through vertical height detection systems. Real-time information systems can provide trucks with real-time information on congestion and delays, accidents, weather, parking, toll facilities, speed restrictions, weighing stations, and waiting times, among others. Another example of the practical application of ITS is dynamic routing, which uses the realtime information provided by the aforementioned systems to guide trucks to safer, faster, and regulatory-compliant routes to ensure more efficient operations.

Integrated transportation and land-use program  51

Last-mile delivery practices The last-mile portion of deliveries is often one of the most expensive parts of the supply chain due to congestion and the restrictive operational environment, hence the need for precise strategies that specifically target last-mile deliveries. One example among these strategies is time-slotting of pickups and deliveries at LTGs. Since LTGs usually have limited loading and unloading areas, coordinating delivery and pickup times tend to relieve some of the negative impacts of last-mile deliveries. Thus, staggering the arrival of delivery trucks throughout the day can reduce congestion and double parking at LTG locations. Other initiatives include driver training programs and anti-idling programs. The former aims to change driver behavior and improve delivery efficiency by training drivers on how to save fuel and reduce noise and emissions throughout the delivery process. The latter seeks to use technology, regulations, and incentives to reduce the pollution caused by truck idling. Lastly, the pickups/deliveries to alternate locations initiative tries to consolidate e-commerce deliveries by delivering goods to delivery lockers and other local freight collection points, allowing customers to pick up their goods themselves. However, these locations should be positioned in a way that minimizes customers’ deviations from their daily routines, or else it may increase overall traffic of passenger cars trying to retrieve their goods. Demand management This group of initiatives targets the modification of freight demand patterns—time, frequency, mode, or destination of deliveries—as a method of redistribution of freight traffic, which ultimately reduces negative externalities. By shifting the time of deliveries to off-peak hours or to the off-hours (7:00 pm to 6:00 am), when there is no traffic congestion, delivery trucks can be more efficient, and there will be a reduction in peak hour congestion and emissions. As an example, a voluntary off-hour delivery program is one of the most important and successful initiatives of this group. It proved to be very effective in London for the 2012 Summer Olympics and also in New York City (Holguín-Veras et al., 2015, 2020c). It is important to emphasize the voluntary nature of the program because a peak-hour delivery restriction leads to unintended consequences of increasing the number of smaller delivery vehicles, as mentioned in the last-mile delivery practices subgroup. This program targets receivers, who are the key decision makers of delivery times, by giving them financial incentives to request offhours deliveries, resulting in a win-win situation for both carriers and receivers. Another method of changing freight demand patterns is by promoting receiver-led delivery consolidation programs with the objective of reducing the frequency of deliveries. This requires the consolidation of deliveries at one of the suppliers’ facilities at the request of the receiver, and then the supplier does the final delivery. Hence, the number of delivery trips is reduced, saving time for receivers, and increasing the productivity of suppliers while at the same time saving transportation costs for suppliers. The mode selected to do the deliveries also impacts the efficiency of the supply chain and the negative externalities produced, hence the need to encourage mode shift programs, which aim to switch to alternative delivery modes e.g., cargo bikes to reduce congestion and pollution. Small-scale mode changes have been successful through the use of “cargocycles” in France, and the use of Cargotram in Switzerland (SUGAR, 2011). Lastly, demand patterns can also change in terms of the destination of deliveries. For instance, if the demand is located closer to the supply, freight trips will be shorter, more efficient, and will generate less externalities. Since the spatial distribution of demand is closely

52  Handbook on city logistics and urban freight

related to land-use patterns, the reader can refer to the description of FELU initiatives (below), in particular the subgroups of long-term planning and zoning for an expanded discussion and insights on how the destination of deliveries can affect the efficiency of supply chains. Freight Efficient Land-Use Initiatives The purpose of this subsection is to describe the groups of initiatives that can support FELU design and planning. As discussed in Chapter 23 by McKinnon, these types of initiatives can play a major role in improving air quality in urban areas. The groups, discussed below, are based on the characterization of the supply chain starting from supply-related to demandrelated land-use initiatives. Long-term planning This group consists of land-use initiatives that provide a path for city planners to move toward FELU. To start, it is important to develop and implement a FELU program. Essentially, these two steps will establish the goals and initiatives to strive for the vision of FELU by creating a roadmap for implementing initiatives in the FELU plan. Within the set of long-term planning strategies, several initiatives will help support FELU. This set can be categorized into three subgroups. The first subgroup aims towards the densification of logistical activities close to urban cores and promotes logistical mixed use, which has the purpose of allowing complementary land uses to be closer to each other. For example, by having distribution centers in urban areas, the distance covered by trucks to do the last mile would be reduced. The second subgroup aims to preserve existing logistic areas and reserve new locations that could be strategic in the future for the supply chain. The intent of these strategies is to avoid systematic inefficiencies that arise when relocating logistical areas to the outskirts of the region, such as increasing the vehicle miles traveled and increasing the externalities produced by freight vehicles. The third subgroup considers the co-location of auxiliary facilities near major gateways aiming to reduce the physical distance of key logistical facilities, e.g., airports, with supporting facilities, e.g., maintenance yards, in order to have a compact development of these critical logistical facilities. Zoning This group of initiatives uses zoning tools to foster FELUs because zoning is a powerful tool that can impact both freight vehicle operations and land-usage mandates. Therefore, zoning must be used carefully given that it might bring unintended negative impacts. Currently, there are two main groups of zoning initiatives: (1) regulatory controls and (2) discretionary tools. Regulatory controls The regulatory control subgroup consists of initiatives that aim to use existing institutions like planning boards and zoning regulations to promote FELUs. Currently, there are three types of zoning that can be used for this purpose: (1) overlay zoning, (2) form-based zoning, and (3) hybrid zoning. Overlay zoning adds a new set of rules to a pre-existing zone. As an example, the city of Baltimore (Maryland) used overlay zoning for the Maritime Industrial Zoning Overlay District (MIZOD) to reserve areas for maritime industries (Envision Freight, 2010).

Integrated transportation and land-use program  53

Form-based zoning regulates building form instead of lot usage to force developers to achieve a desired urban form. Nonetheless, by incorporating space requirements and logistical needs in the form-based zoning, a design that supports FELU can be achieved. As an example, the neighborhoods of Gridley, Allin, and Prickett in Bloomington (Illinois) specified logistical requirements, such as parking and loading areas (City of Bloomington, 2007). Hybrid zoning combines Euclidean zoning—traditional zoning that focuses on separating uses for land—with other types of zoning. Hybrid zoning can mandate developers to have infrastructure or open spaces dedicated to logistical activities. As an example, in Albany (New York), following a hybrid zoning approach, the city required general industrial and light industrial sites to account for logistical needs (City of Albany, 2017). Discretionary tools Discretionary approaches consider initiatives that can either restrict or promote the construction of infrastructure to meet the specific needs of a particular area. For example, creating a specialpurpose district might preserve (or prevent) real estate developments, or revitalize a specific area of the city. This is helpful to achieve more compact supply chains because land-use planners can use these approaches to either restrict the construction of real estate or build supportive infrastructure that facilitates the transportation of goods. For the special case of suburban communities with freight-intensive facilities in their vicinities, e.g., communities besides maritime ports, by enhancing the subdivision regulations, more freight-efficient site designs can be achieved. Site/building design This group consists of initiatives that change the design of sites and buildings to mitigate, or even eliminate, the negative externalities produced by freight activity. It is important to consider how the design of sites and buildings can result in the production of excessive negative externalities and inefficient freight operations. For example, older buildings, the design of which pre-dates modern freight vehicles, may not have been built to handle the types of freight vehicles in use today. Site This subgroup aims to enhance the site layout to minimize the negative externalities of freight activity. These efforts involve identifying how the site design can be improved to mitigate freight issues, which should involve discussions with freight stakeholders and the local community. However, upgrading existing sites to comply with these standards may increase development costs. Such initiatives include foster context-sensitive planning and design, conditional use requirements to foster FELUs, and requiring provision of buffers. Fostering context-sensitive decisions ensures that developers consider the impact that site designs have on all stakeholders. Implementing conditional use requirements is a way to require that developers incorporate freight improvements appropriately into the design of sites. Requiring buffer areas—such as buffer zones, zoning setbacks, or planting strips that physically separate land uses that are deemed incompatible with each other—increases the quality of life for local communities that experience excessive noise and safety issues from nearby freight facilities. Building Another way to foster FELUs is to focus on the design of the buildings within the site. Implementing design requirements and providing recommendations will lead to buildings

54  Handbook on city logistics and urban freight

that are designed to handle freight in a more efficient way. As with site design, these initiatives may carry a high financial cost for owners who need to upgrade existing buildings. One such initiative is to require provisions for logistics areas within buildings. This is especially important given current freight trends: due to the increasing popularity of online shopping, many packages arrive at offices and residential buildings. Without designated areas for logistics to receive, store, and distribute these packages, the risk of theft increases, freight follows an inefficient path through the building, and freight disrupts visitors and tenants. Many of these initiatives to improve building design have evolved out of emerging needs in the era of internet deliveries, highlighting the need for planning agencies to periodically review and update building codes and design guidelines to make sure that buildings are designed to meet the current freight needs. Pricing, incentives, and taxation This group of initiatives explores innovative ways to develop and fund programs that promote freight efficiency and mitigate the negative externalities of freight activity. These efforts have often been led by the public sector but require private-sector participation and involvement as well. This group is unique in that these initiatives do not require costly construction and may offer short- or medium-term results. Pricing Pricing initiatives charge developers a fee to offset the burden on the community of the construction of a new facility. Use of impact fees and proffers is a common tactic for local governments to raise funds. When these programs are set up to charge developers based on the magnitude of externalities they produce, they can encourage design and management practices that promote freight efficiency. However, implementing development fees may impede economic growth and encourage developers to look elsewhere to regions without these fees. Incentives These initiatives aim to set up programs that encourage developers to invest in freight-efficient infrastructure and operations. The programs within this group are intended to be voluntary, meaning that developers choose whether or not to participate. These programs could be designed to encourage specific aspects of planning, such as a site/building design that includes off-street parking and logistics areas, or specific aspects of management, such as a facility that utilizes off-hour deliveries. Developers who participate would receive rewards, which could involve tax benefits, density bonuses, public-sector support, or public recognition. An advantage of these programs is that they encourage the private sector to absorb the bulk of the cost of investing in freight facilities; however, this investment may cause unintended increases in congestion and noise. Taxation Taxation initiatives use the tax system to foster freight efficiency. An example of this is using tax increment financing, which designates a portion of the taxes collected to be spent on improvements within the same district. With tax increment financing, the public sector can complete community projects without increasing the total tax owed, which may encourage private-sector development and investment in the targeted district. However, these programs may take away from development outside of the district (Merriman, 2018).

Integrated transportation and land-use program  55

Joint Initiatives As explained earlier in this chapter, there is an inherent relationship between land use and transportation planning and policy, hence there is a set of joint initiatives. In essence, joint initiatives require the efforts of both land-use and transportation agencies. These initiatives are within the facilities/infrastructure management and parking/loading areas management groups. Facilities and infrastructure management Major improvements The joint initiatives in this subcategory focus on major improvements that require the efforts of transportation agencies as well as the approval of land-use agencies to change the land use of an area, for example, the implementation of a new logistical facility such as an intermodal terminal or an urban consolidation center. Parking and loading area management Off-street parking and loading Sometimes it is necessary to engage land-use agencies to promote off-street parking. This happens when the current building codes do not meet adequate standards for freight vehicles. Commercial centers and large stores often have limited space for trucks to maneuver and park. Outdated buildings that cannot accommodate modern large trucks are one example of such limitations, which lead delivery trucks to occupy the curb parking area that is already limited and competitive in commercial areas. By changing building codes to provide adequate off-street parking, the competition for on-street parking will decrease. Additionally, regulations can be made to provide long-term truck parking and staging areas, when needed. Taken together, this chapter provides a comprehensive view of the wide range of public sector initiatives—both transportation and land use—that could be considered to foster the beneficial effects associated with the production and consumption of supplies and reduce the negative impacts of freight traffic. Notwithstanding the work done, there are a number of areas where further research is needed. To start, there is a major need for a systematic evaluation of the real-life performance of both transportation and land-use initiatives, to ensure transportation and land-use planners are aware of previous experiences. It is also important to conduct further research to quantify the potential synergies between transportation and land-use initiatives. Addressing these research gaps could go a long way toward effective urban freight management and planning

CONCLUDING REMARKS This chapter provides a succinct overview of an integrated transportation and land-use urban freight management and planning process. In addition, it presents a comprehensive description of several groups of initiatives that can be used to enhance supply chain efficiency while minimizing externalities. The foundations of the chapter are seven principles, presented in the second section, that guide practitioners in designing more efficient freight systems while accounting for the interplay between supply chain efficiency and land use. In essence, at the risk of oversimplifying, the principles aim to make supply chains more compact, to engage

56  Handbook on city logistics and urban freight

and understand the behavior of all stakeholders in transportation and land-use decisions, and ultimately to maximize social welfare while having an efficient and dynamic supply chain. It should be noted that managing and planning urban freight systems is a tremendous challenge, considering all of the stakeholders involved (e.g., local communities, carriers, etc.) and the conflicting interests they could have. A general process of transportation and land-use management and planning is outlined to provide an overview of how transportation and land-use decisions should be made. The process has eleven stages, discussed in the section section, starting with defining goals and identifying a freight issue and ending with implementing a solution and looking back to study the effectiveness of the solution at addressing the identified issue. With these eleven stages, stakeholder engagement is critical to enriching the understanding of freight issues, gaining insights into possible solutions, and ultimately reducing the possibility of unintended impacts from transportation policies. In addition, whenever there is no agreement about the viability of a potential transforming initiative, conducting well-designed pilot tests is the recommended course of action. The chapter also provides a brief characterization of a wide spectrum of initiatives that can promote supply chain efficiency and minimize the externalities generated by freight transportation. Transportation and land-use planning organizations can find great benefit from this categorization, considering how it works as a state-of-the-art inventory for freight-efficient design. The categorization includes seven groups of transportation initiatives and six groups of FELU initiatives. Given the intrinsic relationship between transportation and land use, there are groups of joint initiatives that require collaboration between transportation and land-use agencies to be implemented in a way that maximizes their impact. Furthermore, the implementation of initiatives within groups and between transportation and land use should not be understood as mutually exclusive, but rather as complementary in some cases, and standalone in others. The groups of initiatives were organized from supply-oriented to demandoriented. For instance, the first group of transportation initiatives focuses on the physical infrastructure of the network, e.g., construction of ring roads. Conversely, the last group of transportation initiatives focuses on demand patterns that generate freight trips, e.g., off-hour deliveries. Details about the initiatives are publicly available at https://cite​.rpi​.edu​/iselector/. With this electronic database, users can input filters to specify the issues they are addressing and the local context surrounding the issues, and the software will output initiatives that could be effective solutions along with a one-page summary including a qualitative assessment of performance and examples of implementation. As discussed throughout the chapter, the incorporation of freight needs into transportation and land-use planning is not a linear process; instead, it involves the coordination of several stakeholders with sometimes conflicting interests, and it requires tremendous effort from public agencies. However, confronting this challenge head-on comes from the necessity of addressing the interplay between supply chain efficiency, land-use ordinances, and freight activity. It is necessary to use all the tools available to address this challenge. Central to this quest is the explicit consideration of the insight provided in Chapter 1 by Giuliano and Chapter 23 by McKinnon. The first provides a comprehensive review of the challenges encountered by freight transportation in cities, in particular the issues that must be addressed, such as air pollution, noise, congestion, parking, and the more recently discussed environmental justice. The second discusses the sources of urban freight emissions and how to reduce them. This chapter take those issues and provides a set of tools and initiatives that aim to

Integrated transportation and land-use program  57

make supply chains more compact, involve all stakeholders along the supply chain to come to a consensus, minimize negative impacts on local communities, and ultimately maximize welfare.

REFERENCES Barber, D., & Grobar, L. (2001). Implementing a statewide goods movement strategy and performance measurement of goods movement in California. Long Beach: California State University, METRANS. Bateman, I. J., Carson, R. T., Day, B., Hanemann, W. M., Hanley, N., Hett, T., … Pearce, D. W. (2002). Economic valuation with stated preference techniques: A manual. Northampton: Edward Elgar. BESTUFS. (2007). Good practice guide on urban freight transport. BESTUFS. London: University of Westminster, 84. City of Albany. (2017). City of Albany unified sustainable development ordinance. Clarion Associates, LLC, 1–327. City of Bloomington. (2007). Section 44.6-26 Gridley, Allin, & Prickett (GAP) form-based code, 1–43. City Ports. (2005). City ports project interim report. M. Panebianco & M. Zanarini. Bologna, European Project City Ports Interreg Programme III B CADSES. Interim Report: 209. Envision Freight. (2010). Case study: Baltimore maritime industrial zoning overlay district. Gunasekarana, A., Patelb, C., & McGaugheyc, R. E. (2004). A framework for supply chain performance measurement. International Journal of Production Economics, 87(3), 333–347. Holguín-Veras, J., Wang, Q., Xu, N., Ozbay, K., Cetin, M., & Polimeni, J. (2006). Impacts of time of day pricing on the behavior of freight carriers in a congested urban area: Implications to road pricing. Transportation Research, Part A: Policy and Practice, 40(9), 744–766. Holguín-Veras, J., Xu, N., De Jong, G., & Maurer, H. (2011). An experimental economics investigation of shipper-carrier interactions on the choice of mode and shipment size in freight transport. Networks and Spatial Economics, 11(3), 509–532. Holguín-Veras, J., Amaya-Leal, J., Wojtowicz, J., Jaller, M., González-Calderón, C., Sánchez-Díaz, I., … Browne, M. (2015). Interactive planning guide: Improving freight system performance in metropolitan areas. National Cooperative Freight Research Program. Washington, DC: Transportation Research Board, 1–212. Holguín-Veras, J., Lawson, C., Wang, C., Jaller, M., González-Calderón, C., Campbell, S., … Ramirez-Rios, D. (2017). Using commodity flow survey and other microdata to estimate the generation of freight, freight trip generation, and service trips: Guidebook. National Cooperative Highway Research Program / National Cooperative Freight Research Program. Washington, DC: Transportation Research Board. Holguín-Veras, J., Leal, J. A., Sánchez-Diaz, I., Browne, M., & Wojtowicz, J. (2020a). State of the art and practice of urban freight management part I: Infrastructure, vehicle-related, and traffic operations. Transportation Research, Part A: Policy and Practice, 137, 360–382. Holguín-Veras, J., Leal, J. A., Sanchez-Diaz, I., Browne, M., & Wojtowicz, J. (2020b). State of the art and practice of urban freight management part II: Financial approaches, logistics, and demand management. Transportation Research Part A: Policy and Practice, 137, 383–410. Holguín-Veras, J., Wang, C., Ng, J., Ramírez-Ríos, D., Wojtowicz, J., Calderón-Quevedo, O., … Haake, D. (2020c). Planning freight-efficient land uses: Methodology, strategies, and tools: Guidebook. National Cooperative Highway Research Program. Holguín: Transportation Research Board Holguín-Veras, J., Ramirez-Rios, D., & Wojtowicz, J. (2021). Initiative selector for fostering freight system performance, energy efficiency, and freight-efficient land use. Retrieved October 1, 2021, from https://cite​.rpi​.edu​/iselector/. Jones, E., Chatterjee, A., & Marsili, R. (2009). A collaborative plan for curbside freight delivery in Washington D.C. ITE Journal, 79(5), 22–25. Le Pira, M., Marcucci, E., Gatta, V., Inturri, G., Ignaccolo, M., & Pluchino, A. (2017). Integrating discrete choice models and agent-based models for ex-ante evaluation of stakeholder policy acceptability in urban freight transport. Research in Transportation Economics, 64, 13–25.

58  Handbook on city logistics and urban freight

Merriman, D. (2018). Improving tax increment financing (TIF) for economic development. Policy Focus Report, Lincoln Institute of Land Policy. Quak, H., & de Koster, M. B. M. (2006). The impacts of time access restrictions and vehicle weight restrictions on food retailers and the environment. European Journal of Transport and Infrastructure Research, 6(2), 131–150. Statista. (2020, October 26). E-commerce worldwide – Statistics & facts. Retrieved February 15, 2021, from https://www​.statista​.com​/topics​/871​/online​-shopping/. SUGAR. (2011). City logistics best practices: A handbook for authorities. Bologna, Italy: SUGAR, 276. Taniguchi, E., James, J., Barber, R., Imanishi, Y., & Debauche, W. (2012). Public sector governance of urban freight transport. PIARC World Road Association. US Environmental Protection Agency. (2013). Smartway transport partnership. Retrieved June 25, 2013, from www​.epa​.gov​/smartway​/partnership​/index​.htm. United Nations Department of Economic and Social Affairs. (2018, May 16). 68% of the world population projected to live in urban areas by 2050, says UN. Retrieved February 15, 2021, from https://www​. un​.org​/development​/desa​/en​/news​/population​/2018​-revision​-of​-world​-urbanization​-prospects​.html.

SECTION I MODELLING AND SIMULATION

3. Overview of urban freight transport modelling Lóri Tavasszy and Michiel de Bok

INTRODUCTION The urban freight transport system has many unique characteristics and concerns that make it stand out from national or global freight transport systems. Cities are places of mass consumption, where logistic processes serve the urban fabric as a place of destination for retail goods and construction materials. Urban freight involves also first-mile transport for manufactured goods, waste and consumer-to-consumer (C2C) shipments. Last-mile transport usually ends at the consumers’ homes or the place where they pick up the goods for personal use; at an intermediate stage, the freight will pass through warehouses, cross-docking centres or fulfilment centres. It is here where product flows are broken down into individual consumption units and mass individualization of society becomes most visible. Also, as distances are usually short, the modes of transport exclude the typical long-distance transport by air, sea and rail and include active modes such as walking and cycling. Commodity mixes include mostly consumer goods traffic, service traffic and construction materials, whereas the share of industrial goods is low. Freight delivery traffic mixes intensively with other purposes of transport inside residential, business and recreation areas. There is fierce competition for urban space, extending well beyond that needed for transportation alone and including parking, waiting, (un)loading, manoeuvering, registration, storage, handling, etc. The high population densities introduce additional risks to the safety and health of citizens. Each of the stakeholders involved in these activities is powerful, in different ways. As the urban consumer is at the end of the chain, their demands are followed by the entire supply chain. Retailers and service providers are parts of big conglomerates and, with their decisions, can make a large impact on city liveability. This concatenation of stakeholder interests and powers makes the urban freight transport system a challenging one to describe in models. The engagement of stakeholders is a separate line of research: Browne and Goodchild provide an overview in Chapter 15 of this Handbook. Urban freight models generally require integrated modelling approaches that combine both optimization and simulation, such as dynamic flow simulation and multi-agent systems (Taniguchi et  al., 2003) The main function of urban freight transport models is to provide comprehensive information to all stakeholders about the current and expected performance of the system under different future social, economic and technological scenarios: the models are used to assess the effects of these scenarios (Crainic et al., 2009; Comi et al., 2014; Holguín‐Veras et al., 2018). At their core, these models traditionally focused on the intensity of transport services within and around the urban area, to support long-term investments and policies. Nowadays, the models provide numbers needed for the design of infrastructural provisions, new services, business models and even governance arrangements between stakeholders. More and more models are also becoming a basis to support experimentation with socioeconomic and technological innovations (like crowd-shipping platforms) in multistakeholder, value-driven innovation processes. Models are also becoming a component of 60

Overview of urban freight transport modelling  61

urban traffic control towers for real-time flow management. Finally, we see that the notion of cities as places of mostly on-demand, last-mile movements is slowly becoming obsolete. Consumer-to-consumer trade is growing and the circular economy will amplify return flows. The above developments imply that the field of urban freight transport modelling is dynamic and innovating fast. Literature on modelling urban freight processes is overwhelmingly normative in nature and focuses on optimization of decisions. Even for applications in city logistics, much modelling that relates to company-level decision making uses operational-research-based optimization as the main tool. Crainic et al. provide an updated overview of operations research for planning and managing city logistics systems in Chapter 10 of this Handbook. In contrast, descriptive approaches, such as discrete choice modelling, aim to provide a picture of representative decision behaviour for a previously defined population of firms (e.g. all firms in a city). Both model types have existed next to each other and are sometimes used within one modelling framework (Tavasszy and De Jong, 2013). Our focus in this section is on descriptive and predictive models as they are statistically validated as a (sufficiently) truthful representation of urban freight processes – which is not necessarily the case with optimization models. The aim of the current chapter is to provide an overview of the key recent developments in urban freight modelling, with a focus on the latest research and innovation directions. In that sense, it introduces the subsequent chapters (Chapters 4, 5 and 6) in this section. These present three examples of the most recent modelling developments in urban freight demand. Each example deals with one of the particular levels of urban freight demand models shown in Table 3.1: production and attraction of freight trips, the simulation of freight patterns, and vehicle flow. Sánchez-Díaz and Castrellon (Chapter 6) provide a contemporary overview of freight demand generation models in the urban context, predicting freight and/or vehicle flow using establishment data. Their chapter also discusses the linkage to microsimulation agentbased models. Comi and Delle Site (Chapter 4) present a forecasting model for restocking activity and tour planning for the retail sector in the urban context, representing logistics functions explicitly. Sakai et al. (Chapter 5) present a state-of-the-art multiagent simulation model for city logistics. Logistic agent behaviour is simulated across different dimensions of decision making, with a focus on the choice of carrier and vehicle activity Table 3.1  Framework of logistics structures and models Logistics sub-structure

Model ype

Production structures or intersectoral trade

Production and consumption functions (CGE) models, Input/output models, Freight (trip) generation

Spatial supply structures (firm level) or trade relations (aggregate)

Supplier choice models (firm level) Gravity model (aggregate)

Distribution structures or channels

Distribution structure models Shipment size and frequency models

Transport structures

Carrier (type) choice, Mode and/or route choices Vehicle type choice Detailed routing and scheduling models

62  Handbook on city logistics and urban freight

The remainder of this chapter is organized as follows. The second section presents a conceptual framework for freight modelling to help position the work on freight modelling presented in this Handbook. The third section gives a short overview of different generations of urban freight models, and how they have evolved from the conventional four-step and operations research models into the multistakeholder, multiobjective simulation models in use today. The fourth section summarizes the key new challenges in urban freight transport, elaborates on the resulting changes in user environment and model requirements, and discusses how these requirements are being translated in a new generation of urban freight transport models. Finally, we introduce the subsequent chapters of this section on urban freight modelling.

CONCEPTUAL MODELS OF URBAN FREIGHT Transport services serve a demand that is derived from various supply chain activities in between consumption and production, including distribution, packaging, inventory management, etc. In order to predict transport flows in a way that they relate to the real-world economy, this background needs to be understood. Models do not need to model all details, but merely effectively summarize real-world activities using aggregate structures, and reproduce the results of logistics decision making across different functional layers in the supply chain. Real-world decision-making structures in logistics are complex. Riopel et al. (2005) mapped and categorized logistic decisions in relation to transport. Figure 3.1 shows in a stylized way how these decisions are interconnected. The importance of their study lies in that it provides a comprehensive conceptual model rooted in supply chain management, which mathematical modelling efforts can use as a starting point. Next to eight transport-focused decision problems, they also identified 40 contextual decisions that directly or indirectly influence these eight transport decisions which vary according to specific market characteristics. The transport-related decisions include: ● ● ● ● ● ● ● ●

Transport modes. Types of carriers (own account or for hire, specialization, etc.). Carriers. Degree of consolidation (e.g. hybrid channels or combined). Transport fleet mix. Assignment of customers to vehicles. Vehicle routing and scheduling. Vehicle load plans.

The relevant contextual decisions are the ones that shape the demand for transport, span all areas of supply chain management, and can be found in the following areas: ● ●



Strategic level decisions, e.g. setting customer service decisions. Tactical level decisions, including physical facility network design (e.g. number and location of distribution centres) and communication and information network design. Operational level decisions, including demand forecasting, materials handling, procurement and supply management, production, product packaging, inventory management, order processing and warehousing.

Overview of urban freight transport modelling  63

Source:   Riopel et al. (2005)

Figure 3.1  Network of logistics decisions surrounding the transport function (see Appendix 3.1 for reference numbers) A full simulation of all these decisions to replicate all the individual steps that firms in the real world go through would be a daunting task. It is also probably unnecessary, as the main challenge in modelling is to identify the smallest subset or aggregate of decisions which provides a sufficient representation of reality. As a result, decision-support models that integrate this entire spectrum of decisions do not exist and probably never will. Models always focus on subsets of the logistics systems, on specific relations or on an aggregate of multiple decisions. They will thus summarize or ignore many details in this system, resulting in varying degrees of sophistication and external validity. Nevertheless, a complete framework remains an important point of reference. A relatively new concern in modelling are the dynamics of decision making (Tavasszy, 2020). In policy and innovation circles, response times in the transport system are increasingly recognized as becoming more important. Consider climate change, for example: most policies include a target year for achieving a desired effect (e.g. 55% emission reduction by 2030). The dynamics of decision making inside the system will determine how quickly the system will respond and whether climate mitigation targets can be met in time. Simulation models allow these dynamics to be included, but unfortunately there is still little empirical

64  Handbook on city logistics and urban freight

knowledge about the responsiveness of parts of the urban freight system. Strategic decisions will only be taken a limited number of times during the lifetime of a company, whereas the tactical investment decisions have a very long turnaround (years to decades) due to their capital intensity. Operational decisions may reveal stronger dynamics, but some decisions – like modes of transport and product packaging – may have a very long review frequency. The aggregation of many decisions may also lead to long system response times, even if many decisions are taken relatively quickly. When the total response time of the system on a policy is important, and key decisions are taken slowly, insight into dynamics is of critical importance to assess the feasibility and usefulness of policies. Ultimately, researchers and practitioners have developed a practice of thinking about aggregate structures by partitioning the entire network of logistics decisions into larger blocks, that are relatively easy to model. Of course, the drawback of this is that information about detailed decisions is lost, that response mechanisms are not modelled completely or truthfully and that they may even provide misleading information. But, given the unsurmountable constraints of data availability and with the help of rigorous statistical validation, the method of aggregation has allowed science to progress. As a basic conceptual model of the freight system, we find representations at different levels of aggregation. A crude one is the division into four key structures (Rodrigue, 2020): production structures, distribution structures, supply structures and transport structures. Herein, one faintly recognizes the traditional, passenger-transport-oriented approach of the four-step model – where distribution structures are irrelevant and transport structures are modelled with mode and route choice. From this starting point in the 1970s, increasingly sophisticated conceptual frameworks have evolved for freight transport. In a stepwise fashion, more and more logistical detail was added to the four-step framework for the purpose of improving predictive freight flows models – this included the explicit consideration of distribution structures, the inclusion of an agent-based view and the rooting of framework in multi-stakeholder ontologies. The SMILE model (Tavasszy et al., 1998) recognized the difference between spatial flow structures for trade and those for transport. The former concern the inter-regional trade relations bridging producers and consumers (P/C relations (see De Jong & Ben Akiva, 2007)) and the latter concern the places where freight transport assignments mark their origin and destination (O/D relations (ibid.)). The two structures are bridged by distribution networks, where warehouses also act as origin or destination of freight movements. If one only wants to consider mode-specific O/D relations, intermodal transport networks also play a role, where transhipment centres will act as origin or destination. The SMILE model considered both interactions, also included a model of trade networks and was empirically implemented for the Netherlands. An urban freight model that appeared at the same time, which also modelled distribution centres explicitly, was GoodTRIP (Boerkamps et al., 2000). This model had the same logic as part of SMILE but its empirical implementation was limited. Later, Roorda et  al. (2010) presented the FREMIS model architecture that takes an explicit agent-based view, meaning that decisions of consumers and firms act as a starting point. The considered decisions differ from those of Riopel et al. (2005) in the detailed decisions and aggregates these into three types of artificial “contracts”: commodity contracts, business contracts and logistic contracts. Prototype applications implementing parts of this framework were built for the Greater Toronto Area. Anand et al. (2014) considered dynamic interaction agent decisions explicitly in an agent-based model implementation for the city of Rotterdam, the Netherlands.

Overview of urban freight transport modelling  65

Although the number of decisions considered is limited (sourcing, inventory, shipment size and routing), the conceptual framework of the model had a unique feature in the sense that it was built on a generic ontology for city logistics, based on linguistic processing and knowledge mapping of verbal and written text material. As we will discuss later, the latest generation of urban freight models, like Mass-GT (de Bok et al., 2021), SimMobility Freight (Sakai et al., 2020) and POLARIS (Stinson et al., 2020), have adopted the same range of decision-making problems. In other words, they have mainly moved forward in the empirical implementation of this same architecture, by using microsimulation and new data sources. Research on predictive transport models has still not addressed several decisions from the Riopel framework, including, in the transport spheres, (1) type of carrier, (2) degree of consolidation of flows (internal or external), (3) fleet composition or (4) vehicle load plans. In the wider set of operational decisions, this gap is even larger and includes (5) material handling, (6) product packaging, (7) inventory management and (8) warehousing. Tactical and strategic decisions are rarely modelled at firm level, if at all, with the exception of the actual transport and distribution networks. New research work could address the necessity and feasibility of including these decisions in operational modelling frameworks. In summary, Table 3.1 displays the prevailing, commonly used structuring of logistics decisions, with the most popular quantitative models used to portray these structures (partially or completely) in the right-hand column. For each structure and model type, several studies have appeared, many in an urban context, producing empirically validated models for different regions in the world, from urban to global level (see Tavasszy et al., 2020 for further elaboration on these cases). The modelling methodologies applied have evolved gradually from zone-based models in the 1970s to agentbased approaches nowadays, where the latest models simulate the actions of individual firms. We note that these are not yet the individual decision makers, who would be consumers or responsible managers, but actions of abstract firms which assume specific sequence and speed of decisions as output from the firm as a whole. In the next section, we explore this evolution further and review the development of urban freight modelling approaches through time.

EVOLUTION OF URBAN FREIGHT MODELLING METHODOLOGIES Over the past decades, urban freight transport models have improved in their representation of logistics agent behaviour and spatial resolution and have become more accurate at addressing the complexities of today. We refer readers interested in general reviews of freight transport models to the most recent overview of de Jong, de Bok and Thoen (2021). We can discern three generations of models as they have evolved from the first iterations in the 1980s: aggregate, disaggregate and microsimulation models. They can be best characterized by the level of detail of their inputs and outputs, where disaggregate data refers to firm or shipment level, and aggregate data to the level of traffic analysis zones or geographical regions (Figure 3.2). The first generation of models was built on aggregate data, largely by analogy to the fourstep models in passenger transport, with aggregate choice models or zone-level empirical models like the direct demand models and gravity models. The second generation used disaggregate data and started to represent logistics process like trip chaining and physical distribution. The models were still applied at the zonal level. The third generation of models has taken the step to simulate actions at the individual agent level, which allows for an explicit model

66  Handbook on city logistics and urban freight

Note:   A = aggregate; D = disaggregate

Figure 3.2  Three generations of freight models Table 3.2  Evolution of urban freight model systems Generation

Agent level of underlying theory

Spatial resolution of outputs

Time resolution of decisions

Time resolution of output

Type of application

1 (1970–)

Aggregate

Zone

Year-on-year

Public policy

2 (2000–)

Disaggregate

Yearly performance

3 (2010–)

Zone + Firm

Event-based

4 (Future)

Place + Agent

Week/month

Public + private co-innovation

of connected logistics decisions but makes empirical validation more challenging. The focus of all these models has still been to produce information about yearly flows of goods. A next generation of models needs the disaggregate/disaggregate form as a basis to go in more depth in terms of showing operational details and shorter-cycled changes in the system. Table 3.2 summarizes these characteristics. We describe these generations in more detail below. First Generation: Aggregate Approaches The first series of freight transport models are founded on classic theories of economic activity and transport costs, formulated for aggregate agents: an average firm that represents all firms in one region (or, more generally, spatial unit of analysis or zone). Despite the obvious risk of aggregation bias, the zone based approach is an effective and proven method to simulate how a population of firms behave on aggregate. As the mathematical form is light and broadly known, operational models are quickly estimated and easy to validate and interpret. Contemporary models still carry the fundamental step-wise DNA of these models: transport costs affect decisions as to where to source products, how to choose the efficient mode of transport or how to route shipments. The scope of these models can be continental, national or regional. Data to develop these models (aggregate national or regional trade or transport

Overview of urban freight transport modelling  67

statistics) have been around for decades, and sometimes are even available in time series. Most freight policy scenarios worldwide have been based on this type of model. Practical cases are numerous and include most, if not all, large-scale freight models, like the European Transtools I and II (Hansen & Rich, 2011) and the current national freight model BasGoed for the Netherlands (de Jong et  al., 2011). Furthermore, the national freight model SMILE (Tavasszy et al., 1998) can be classified here. Although it used firm and shipment data to simulate freight movements, the estimated behavioural models were aggregate in nature. For the reasons sketched above, this generation of models still forms the cornerstone of freight policy analysis. For that reason, it is important to continue scientific work to improve their validity. Recent streams of research work attempt to tackle the following challenges: ●







Extending functional properties of models to model the effect of changes in distribution channels or intermodal transport chains (see, e.g. de Bok et al., 2018). Understanding and extending limits of validity. When aggregate models hide complex underlying behavioural patterns that are relevant for policy, these errors need to be understood and may give rise to model improvements (see e.g. Holguín-Veras et al., 2011). Representing heterogeneity of populations. Instead of assuming a non-existing average firm or commodity in the system, segmenting models or refining assumptions about underlying distributions can improve validity (see e.g. Marcucci & Gatta, 2014; Piendl et al., 2019). Modelling response sensitivity or response dynamics in the system. System dynamics models, time series regression models or econometric modes can provide new policyrelevant insights (see e.g. Ferrari (2014) or Davydenko et al., 2021).

Second Generation: Modelling Logistics Behaviour with Firm- or Shipment-Level Data With a second-generation model, we distinguish approaches that have more detail concerning logistic behaviour, including lower-level decisions, decision-maker preferences and interactions between decisions. In these models, several layers of logistics network modelling were added to the freight demand models. The focus of these models is on simulation of firm behaviour at the zonal level, taking into consideration constraints such as location and availability to transfer goods between modes in multimodal transport chains or multitier distribution structures. The methodologies applied in these models represent advances in understanding of logistics processes, new techniques in discrete choice modelling and increased use of detailed data. Theoretically, connections are made between flow models and random utility-based discrete choice models, transforming the formulation from physical flow models using aggregate agents (spatial zones) to choice models for disaggregate agents (decision makers or firms). The attractiveness of these models is their behavioural validity at the firm or shipment level. In addition to national or regional statistics, these models especially build on micro-level (firm- or shipment-level) information from commodity flow surveys, freight trip diaries or establishment surveys. Due to their high costs, large-scale shipper surveys or commodity flow surveys are only available in a handful of countries (US, France, Sweden, Norway, Japan). This has resulted in a limited number of applications. As with the first-generation models, the application of second-generation models has remained mainly at a zonal level, combining aggregate and disaggregate approaches to link

68  Handbook on city logistics and urban freight

zonal flow models to firm-level behavioural models. Where surveys include stated preference interviews, special care must be taken to calibrate models to real-world flows, which often involves tuning a model to aggregate statistics. Representatives of this generation of models include the ADA models in Scandinavia (De Jong & Ben Akiva, 2007), TransTools III (Jensen et al., 2019), TriMode (Williams et al., 2017) and the Strategic Freight Model for Flanders (Grebe et al., 2016). This stream of modelling is uniquely suited to explore decision making in the logistics sector. In principle, every one of the 48 logistics decisions lends itself to studying the alternatives in their unique context, the decision processes, objectives and preferences of actors. As decisions can be recorded at the individual level, this is a fruitful area of study. Although traditionally (as in the abovementioned models) the emphasis has been on multimodal routing choices, in recent times, the emphasis has been on understanding departure-time choice, ordering behaviour (see e.g. Chapter 4 by Comi and Delle Site), firm-level freight trip generation (see Chapter 6 by Sánchez-Díaz and Castrellon ) and distribution choices (see Chapter 5 by Sakai et al.). Third Generation: Microscopic Simulation The second generation of freight models showed that valid disaggregate choice models are feasible. From here, the next step was to explicitly represent the agents and their decisions instead of aggregating the results to zonal level. In this third generation, we use the microscopic simulation approaches. The scope of these models is large-scale simulation of freight demand for all agents in a study area. Models simulate how all firms behave individually, taking into account the preferences and constraints of these agents explicitly. Typically these models take a step down in aggregation level, not just for their estimation, but also for their application: individual actors are modelled explicitly and their behaviour is included in the output of the model. Also, dynamics portrayed become explicit, adding detail to the typical yearly flows of the previous generations by using an event-based approach. In addition, these models are also shipment based: the logistic decision making explicitly represents the units of transport in order to better model the decision behaviour around consolidation of goods transport. This stream of models implied a deeper representation of logistic behaviour across networks and freight service layers, and the simulation of vehicle patterns and routing. Theories typically used include discrete choice models and Monte Carlo simulation or microsimulation of network usage. Dynamic agent-based models also fall within this category: they are a novel approach incorporating learning and emerging behaviour into the simulation. The combination of discrete-choice models, based on stated preference data, and agent-based models within an integrated modelling framework can be fruitfully adopted to ex-ante assess stakeholders’ policy acceptability accounting for heterogeneity and interaction effects (Le Pira et al., 2017). The recent microsimulation implementations of freight transport demand models often take advantage of increasing automation in data collection, providing more detailed, extensive or dense transport surveys. Types of data collection include logistic data on freight demand (establishment survey) or freight transport (trip travel diaries). Often data sets are extended with other combinations of automated data collection (GPS, roadside camera registration) or available geographic information about the location of activities (Yang et  al., 2022; Mohammed et al., 2023). The microsimulation approaches operate with more spatial

Overview of urban freight transport modelling  69

detail, and are designed to fit better with the urban context in the transport domain. The models are used to explore city logistics developments or evaluate policies at this level. Recent examples of operational microsimulation models include ULLTRA-SIM (Sakai et al., 2019), Simmobility Freight (Sakai et al., 2020), MASS-GT (De Bok et al., 2021), POLARIS (Stinson et al., 2020) and MATSim (Bean & Joubert, 2021). Fourth Generation: Living Labs and Digital Twins as a New Context All the above models are designed to support policy making, rather than private strategy building or management. This implies that the models are part of very slow and long decision cycles: infrastructural and regulatory policies that take years to decide, to implement and to take effect. In future generations of models, we expect the main paradigm change to lie in this area: a change of focus towards support to much shorter decision cycles. We see two stages of shortening decision cycles. The first involves decision making by multiple stakeholders around collaborative innovations in city logistics (see also Chapter 16 by Le Pira et al.). Through experimentation in living labs, public and private partners co-create changes in the urban freight landscape. The second stage builds on new opportunities in sensing and information processing, allowing urban management to include city logistics. This operationallevel decision making in smart cities relies on quick estimates of expected effects, currently built around urban management dashboards, but with a full digital twin for the city as ultimate vision for the future. The heart of a digital twin is a model of the city, including its behaviour, for example Lim et al. (2019) or Marcucci et al. (2020). Below, we further develop our ideas for this next generation of urban freight models.

CHANGING REQUIREMENTS AND NEW DIRECTIONS FOR MODELLING As introduced above, the urban freight context is changing in many directions for various reasons. Consumption is slowly becoming mass individualized, implying smaller and more frequent shipments with high service levels. Consumers have become important generators of freight movements due to e-commerce returns, consumer-to-consumer shipments (e.g. preowned products) and return of used products and materials. In response, stakeholders are extending their business by collaboration and changing business models, working on relatively complicated innovations (e.g. crowdshipping, city hubs, collaborative schemes, the physical internet). This creates a strong evolution in business models, such as with firms deciding to take on roles which fall outside their original boundaries. For example, carriers may take on forwarding services and e-commerce sales platforms may incorporate physical delivery, crowdshipping or financial services. Figure 3.3 shows how different agents in a city can take on multiple, competing as well as complementary roles. In this specific case, final delivery is taken over from the courier by the urban consolidation centre operator while the courier assumes the role of logistics service provider and network coordinator. Models can also be used to explore new solutions, such as real-time bay reservation and monitoring (Comi et al., 2018), which require the use of telematic tools. The increasing complexity of logistics business models affects all private and public stakeholders, not just in relation to markets of logistics services but also in transport equipment, information technology

70  Handbook on city logistics and urban freight

Source:   Zenezini et al. (2018)

Figure 3.3  Blurring of business model boundaries in collaborative city logistics and real estate. Many new questions arise concerning the role of public authorities to regulate service markets, and the position of private companies in collaborative networks (Zenezini et  al., 2018). In contrast to earlier government-induced city logistics measures, which are sometimes taken on without regard for sustainable business models and therefore fail (Allen et  al., 2007) stakeholders now engage in collaborative processes to identify shared values, break down Big-Bang innovations into smaller steps and create consensus on the feasibility of innovations before investing. The close involvement of all stakeholders proves to be critical for achieving a sustainable state of the new system. This living lab approach is slowly becoming standard practice in multistakeholder city logistics innovation (Quak et al., 2016; Gatta et al., 2017; Fredriksson et al., 2021). A more detailed contemporary overview of this approach is provided in Chapter 17 by Quak et al. For simulation models to be relevant to facilitate the evolution of these new concepts and business models in urban freight transport, these models need to represent these stakeholders and the diversity in urban freight demand. Living Labs and Digital Twins: Requirements Living labs and digital twins, characterized by behavioural and simulation models, play an important role in supporting participated planning processes, where reactions to structural change and policy measure implementations are investigated (Marcucci et al., 2020). To facilitate the living lab movement, cities are becoming increasingly smart, shortening their own decision cycles to experiment with temporary changes in regulations and new approaches to

Overview of urban freight transport modelling  71

urban traffic management. Urban freight models can support this development. The model requirements differ considerably from those in policy evaluations, however: ●







Experimentation cycles in city logistics living labs are shorter than the conventional policy-making cycles (months instead of years); models will have to be set up quickly for cities and provide answers within days or weeks. As industry wants to understand effects on operations, the models need to include a description of agent- (firm- or establishment-) level impacts of innovations, including shorter (operations) and longer-term (market position) effects. In order to simulate logistics processes consistently, the models need to have a representation of freight shipments. Conventional vehicle-based models fall short in adequately simulating impacts of consolidation or cooperation between stakeholders on shipment patterns. Models are used as predictive dashboards towards a larger stakeholder community and will need to be comprehensive in terms of the relevant agents and impacts accounted for. Also, they will require acceptance by all involved.

A subsequent development which requires even more detail and speed in models involves urban management. Cities are leaning more and more towards the use of real-time information about the state of their systems to optimize the use of urban space by access control and pricing, and to mitigate negative impacts with traffic control. These operational, control room functions require models that can provide even shorter-term forecasts of user behaviour at a lower, agent-specific level of detail. Here, urban freight models become part of a new context of the smart city movement and are developing towards digital twins of cities (Farsi et al., 2020). The fast digitalization of the sector, with rapidly increasing data availability, supports this change (Bukrinskaya & Dyukova, 2019). Numerous sensors allow immediate tracking of freight shipments, freight traffic and its impacts. New information that is becoming widely available in digital form includes historical records and streaming data concerning logistics services (planned and executed tours and trips, service times, stops, etc.); cargo (e.g. digital bill of lading, cargo appearance, etc.); vehicles (position and driving conditions, driver behaviour, etc.), traffic (intensity, safety, compliance, etc.) and the environment (pollution, weather, etc.). As the amount of data available is abundant and operational in nature, it is useful for dynamically adaptive decision making. Most of these data, however, provide partial information and it remains a challenge to smartly combine and link different sources of data, whether they are static statistics or dynamic operational data. Eventually, it is also conceivable that urban management decisions are automated on the basis of these predictions in a modelbased, predictive control cycle. As these decisions in urban management are short cycled, this application requires even faster analysis which is accurate for a wide variety of situations. Here, artificial intelligence will be used more and more to tune models and their predictions to observed reality, and modelling will become increasingly data driven. Research Directions Short-term research directions deal with a successful application of these new sources of information into simulation models for urban freight transport to support strategic policy making.

72  Handbook on city logistics and urban freight

The increasing data availability and changing use of urban freight models call for guidelines to develop empirical descriptive urban freight models. The third generation of microsimulation models simulates individual agents and interdependent logistic decision making: new data allow the development of these models, but require a comprehensive architecture and smart intelligent procedures to combine and link different sources of data. The complexity of these models requires a smart development procedure that differs from the conventional theory-driven approach. Similar to software development or complex product design, a minimum viable product (MVP) principle can be adopted to follow an evolutionary approach. In this approach, the urban freight model starts with a simple descriptive and data-driven baseline model with as little choice modelling as possible, and, in a stepwise process, complexity is added. These additional steps can include the implementation of a choice model for tour formation, or delivery time modelling, or the further segmentation of logistics agents. This approach was also adopted in the development of the MASS-GT model (De Bok & Tavasszy, 2018). Each intermediate version of the model allows learning about the model design and cases in city logistics. Subsequently, as a longer-term research and development direction, models will have to evolve by gaining experience in innovation processes and by functioning within an urban management context. This evolution will include the following capabilities: ●









Applicability within multistakeholder living labs around innovation themes such as (1) horizontal and vertical integration of services as well as (2) changes in fiscal arrangements from government and regulations for use of urban space. Adoption of multistakeholder frameworks for modelling including dynamic business models and a linkage to performance measurement using industry data. Process-wise, these models will need to be embedded in city logistics living labs, both technologically and socially. Ability to function in a fast-paced, model-based predictive control cycle, allowing sensing of performance of cities, prediction of expected future states, calculation of optimal control measures and actuation of measures for urban management. Ability to optimize across different objectives of different actors, suggesting or prescribing a promising course of action. This will require a merger with optimization-focused models of freight distribution (see, for example, Rezaei et al. (2020) and Kim et al. (2021) for two examples in a physical internet context).

The above directions of research will together shape the development of the fourth generation of urban freight models.

CONCLUSIONS In this chapter, we have summarized the past and current developments in urban freight modelling. We have also presented a vision for future development. Over the past decades, models have undergone a development which has entailed increasing use of disaggregate data and a recognition of the main logistics decisions by which firms respond to changes in the environment, policy induced or otherwise. The latest generation of models is characterized by the use of microsimulation aiming to reproduce tactical and operational logistics processes. The user environment of models is predominantly one of long-term

Overview of urban freight transport modelling  73

policy making, where models inform policy makers about possible long-term futures of the city under different policy scenarios. Nowadays we are seeing more collaborative policy making coupled, where public and private decision makers create new decision-making arenas and innovation processes. In this changing context, we have sketched an evolutionary path of freight modelling which revolves around two lines of development: ●



An increasing sophistication in the description of behaviour of logistics agents, by explicit modelling of decisions at the individual firm level and at the level of supply chains. This requires the modelling of more decisions than before based on disaggregate data related to logistics operations, and creates increased external validity of models. From a policy perspective, the increased joint use of models by multiple stakeholders as digital twins in a living lab context. This requires closing of the sensing-actuation loop by direct linkages to streaming data about logistics processes, and produces new model outputs which related directly to decision processes of stakeholders. Models are increasingly data driven.

The above two developments converge in the notion of models as digital twins of urban freight systems. They are built on the principles of event- or agent-based simulation, work at a detailed spatial resolution and are integrated in an urban management control cycle. They are transparent towards both public and private stakeholders and create a strong sense of face validity with these actors. They allow a fast processing of large volumes of incoming data and can suggest promising courses of action to managers. This specification of decision support for urban freight systems brings many challenges – it will require social scientists, engineers and computer scientists, to work more closely together than ever before.

REFERENCES Allen, J., Thorne, G., & Browne, M. (2007). BESTUFS good practice guide on urban freight transport. Brussels: European Commission Framework Research Program. Anand, N., van Duin, R., & Tavasszy, L. (2014). Ontology-based multi-agent system for urban freight transportation. International Journal of Urban Sciences, 18(2), 133–153. Bean, W. L., & Joubert, J. W. (2021). An agent-based implementation of freight receiver and carrier collaboration with cost sharing. Transportation Research Interdisciplinary Perspectives, 11, 100416. Boerkamps, J. H., van Binsbergen, A. J., & Bovy, P. H. (2000). Modeling behavioral aspects of urban freight movement in supply chains. Transportation Research Record, 1725(1), 17–25. Bukrinskaya, E. M., & Dyukova, O. M. (2019). Digital transformation of city logistics infrastructure. In International Conference on Digital Technologies in Logistics and Infrastructure (ICDTLI 2019). Atlantis Press. Comi, A., Donnelly, R., & Russo, F. (2014). Urban freight models. In L. Tavasszy & G. de Jong (Eds.), Modelling freight transport (pp. 163–200). Amsterdam: Elsevier. Comi, A., Schiraldi, M. M., & Buttarazzi, B. (2018). Smart urban freight transport: Tools for planning and optimising delivery operations. Simulation Modelling Practice and Theory, 88, 48–61. Crainic, T. G., Ricciardi, N., & Storchi, G. (2009). Models for evaluating and planning city logistics systems. Transportation Science, 43(4), 432–454. Davydenko, I. Y., Tavasszy, L. A., & Quak, H. (2021). On the cost elasticity of inter-regional distribution structures: A case study for the Netherlands.  European Journal of Transport and Infrastructure Research, 2021, 3–4165.

74  Handbook on city logistics and urban freight

de Bok, M., & Tavasszy, L. (2018). An empirical agent-based simulation system for urban goods transport (MASS-GT). Procedia Computer Science, 130, 126–133. de Bok, M., de Jong, G., Tavasszy, L., Van Meijeren, J., Davydenko, I., Benjamins, M., … Van den Berg, M. (2018). A multimodal transport chain choice model for container transport. Transportation Research Procedia, 31, 99–107. de Bok, M., Tavasszy, L., Kourounioti, I., Thoen, S., Eggers, L., Nielsen, V. M., & Streng, J. (2021). Simulation of the impacts of a zero-emission zone on freight delivery patterns in Rotterdam. Transportation Research Record, 2675(10), 776–785. de Jong, G., & Ben Akiva, M. (2007). A micro-simulation model of shipment size and transport chain choice. Transportation Research: Part B, Methodological, 41(9), 950–965. de Jong, G. C., Burgess, A., Tavasszy, L.,Versteegh, R., de Bok, M., & Schmorak, N. (2011). Distribution and modal split models for freight transport in the Netherlands. European Transport Conference, Glasgow. de Jong, G., de Bok, M., & Thoen, S. (2021). Seven fat years or seven lean years for freight transport modelling? Developments since 2013. Journal of Transport Economics and Policy (JTEP), 55(2), 124–140. Farsi, M., Daneshkhah, A., Hosseinian-Far, A., & Jahankhani, H. (Eds.). (2020). Digital twin technologies and smart cities. Berlin: Springer. Ferrari, P. (2014). The dynamics of modal split for freight transport. Transportation Research: Part E: Logistics and Transportation Review, 70, 163–176. Fredriksson, A., Janné, M., Nolz, P., de Chennevière, P. D. R., van Lier, T., & Macharis, C. (2021). Creating stakeholder awareness in construction logistics by means of the MAMCA. City and Environment Interactions, 11, 100067. Gatta, V., Marcucci, E., & Le Pira, M. (2017). Smart urban freight planning process: Integrating desk, living lab and modelling approaches in decision-making. European Transport Research Review, 9(3), 32. Grebe, S., de Jong, G., de Bok, M., van Houwe, P., & Borremans, D. (2016). The strategic flemish freight model at the intersection of policy issues and the available data. European Transport Conference, Barcelona. Hansen, C. O., & Rich, J. (2011). Trans-tools. In S. Helmreich & H. Keller (Eds.), FREIGHTVISIONsustainable European freight transport 2050 (pp. 43–62). Berlin, Heidelberg: Springer. Holguín-Veras, J., Jaller, M., Destro, L., Ban, X., Lawson, C., & Levinson, H. S. (2011). Freight generation, freight trip generation, and perils of using constant trip rates. Transportation Research Record, 2224(1), 68–81. Holguín‐Veras, J., Campbell, S., González‐Calderón, C. A., Ramírez‐Ríos, D., Kalahasthi, L., Aros‐Vera, F., … Sanchez‐Diaz, I. (2018). Importance and potential applications of freight and service activity models. In E. Taniguchi & R. G. Thompson (Eds.), City logistics 1: New opportunities and challenges (pp. 45–63). London/New York: Wiley. Jensen, A. F., Thorhauge, M., de Jong, G., Rich, J., Dekker, T., Johnson, D., … Nielsen, O. A. (2019). A disaggregate freight transport chain choice model for Europe. Transportation Research: Part E, 121, 43–62. Kim, N., Montreuil, B., Klibi, W., & Kholgade, N. (2021). Hyperconnected urban fulfillment and delivery. Transportation Research: Part E, 145, 102. Le Pira, M., Marcucci, E., Gatta, V., Inturri, G., Ignaccolo, M., & Pluchino, A. (2017). Integrating discrete choice models and agent-based models for ex-ante evaluation of stakeholder policy acceptability in urban freight transport. Research in Transportation Economics, 64, 13–25. Lim, K. Y. H., Zheng, P., & Chen, C. H. (2019). A state-of-the-art survey of Digital Twin: Techniques, engineering product lifecycle management and business innovation perspectives. Journal of Intelligent Manufacturing, 31, 1313–1337. Marcucci, E., & Gatta, V. (2014). Behavioral modeling of urban freight transport. In J. Gonzalez-Feliu, F. Semet, & J. L. Routhier (Eds.), Sustainable urban logistics: Concepts, methods and information systems (pp. 227–243). Berlin, Heidelberg: Springer. Marcucci, E., Gatta, V., Le Pira, M., Hansson, L., & Bråthen, S. (2020). Digital twins: A critical discussion on their potential for supporting policy-making and planning in urban logistics. Sustainability, 12(24), 10623.

Overview of urban freight transport modelling  75

Mohammed, R., Nadi, A., Tavasszy, L. & de Bok, M. (2023). A data fusion approach to identify distribution chain segments in freight shipment databases. Transportation Research Records [in press]. Piendl, R., Matteis, T., & Liedtke, G. (2019). A machine learning approach for the operationalization of latent classes in a discrete shipment size choice model.  Transportation Research: Part E,  121, 149–161. Quak, H., Lindholm, M., Tavasszy, L., & Browne, M. (2016). From freight partnerships to city logistics living labs–Giving meaning to the elusive concept of living labs. Transportation Research Procedia, 12, 461–473. Rezaei, J., Pourmohammadzia, N., Dimitropoulos, C., Tavasszy, L., & Duinkerken, M. (2020). Co-procurement: Making the most of collaborative procurement. International Journal of Production Research, 58(15), 4529–4540. Riopel, D., Langevin, A., & Campbell, J. F. (2005). The network of logistics decisions. In A. Langevin, & D. Riopel (Eds.), Logistics systems: Design and optimization (pp. 1–38). Boston, MA: Springer. Rodrigue, J. P. (2020). The geography of transport systems. Routledge. Roorda, M. J., Cavalcante, R., R., McCabe, S., & Kwan, H. (2010). A conceptual framework for agentbased modelling of logistics services. Transportation Research: Part E: Logistics and Transportation Review, 46(1), 18–31. Sakai, T., Kawamura, K., & Hyodo, T. (2019). Evaluation of the spatial pattern of logistics facilities using urban logistics land-use and traffic simulator. Journal of Transport Geography, 74, 145–160. https://doi​.org​/10​.1016​/j​.jtrangeo​.2018​.10​.011. Sakai, T., Alho, A. R., Bhavathrathan, B. K., Dalla Chiara, G., Gopalakrishnan, R., Jing, P., … & Ben-Akiva, M. (2020). SimMobility freight: An agent-based urban freight simulator for evaluating logistics solutions. Transportation Research: Part E: Logistics and Transportation Review, 141. Stinson, M., Auld, J., & Mohammadian, A. K. (2020). A large-scale, agent-based simulation of metropolitan freight movements with passenger and freight market interactions. Procedia Computer Science, 170, 771–778. Taniguchi, E., Thompson, R. G., & Yamada, T. (2003). Predicting the effects of city logistics schemes. Transport Reviews, 23(4), 489–515. Tavasszy, L. A. (2020). Predicting the effects of logistics innovations on freight systems: Directions for research. Transport Policy, 86, A1–A6. Tavasszy, L., & De Jong, G. (2013). Modelling freight transport. Amsterdam: Elsevier. Tavasszy, L., van der Vlist, M., & Ruijgrok, C. (1998). Scenario-wise analysis of transport and logistics systems with a SMILE. In Selected Proceedings of the 8th World Conference on Transport Research Antwerp, July 12–16, 1998. Tavasszy, L., de Bok, M., Alimoradi, Z., & Rezaei, J. (2020). Logistics decisions in descriptive freight transportation models: A review. Journal of Supply Chain Management Science, 1(3–4), 74–86. Williams, I., TRT Trasporti e Territorio, Garratt, M., Wright, C., & MDS Transmodal. (2017). Trimode freight & logistics model of Europe. Paper presented at the European transport conference. Retrieved March 2, 2023, from www​.trt​.it​/wp​/wp​-content​/uploads​/2018​/05​/3​-TRIMODE​-Freight​ -Logistics​-model​.pdf. Yang, Y., Jia, B., Yan, X.Y., Li, J., Yang, Z., & Gao, Z. (2022). Identifying intercity freight trip ends of heavy trucks from GPS data. Transportation Research Part E: Logistics and Transportation Review, 157. Zenezini, G., van Duin, J. H. R., Tavasszy, L., & De Marco, A. (2018). Stakeholders’ roles for business modeling in a city logistics ecosystem: Towards a conceptual model. In E. Taniguchi & R. G. Thompson (Eds.), City Logistics 2: Modeling and Planning Initiatives (pp. 39–58). London/New York: Wiley.

76  Handbook on city logistics and urban freight

APPENDIX 3.1: LOGISTICS DECISIONS (RIOPEL & LANGEVIN, 2005) Strategic Planning Level 1. Definition of customer service 2. Customer service objectives 3. Degree of vertical integration and outsourcing Physical Facility (PF) Network 4. PF network strategy 5. PF network design Communication and Information (C&I) Network 6. C&I network strategy Inventory Management 7. C&I network design Demand Forecasting 8. Forecasts of demand magnitude, timing and locations 9. Inventory management strategy 10. Relative importance of inventory 11. Control methods 12. Desired inventory level 13. Safety stock Production 14. Product routing 15. Facilities layout 16. Master production schedule 17. Production scheduling Procurement and Supply Management 18. Procurement type 19. Specifications of goods procured 20. Suppliers 21. Order intervals and quantities 22. Quality control Transport 23. Transport modes 24. Types of carriers 25. Carriers 26. Degree of consolidation 27. Transport fleet mix 28. Assignment of customers to vehicles 29. Vehicle routing and scheduling 30. Vehicle load plans

Overview of urban freight transport modelling  77

Product Packaging 31. Level of protection needed 32. Information to be provided with the product 33. Information media 34. Type of packaging 35. Packaging design Material Handling 36. Unit loads 37. Types of material handling equipment 38. Material handling fleet mix 39. Material handling fleet control Warehousing 40. Warehousing mission and functions 41. Warehouse layout 42. Stock location 43. Receiving/shipping dock design 44. Safety systems Order Processing 45. Order entry procedures 46. Order transmission means 47. Order picking procedures 48. Order follow-up procedures

4. Estimating and forecasting urban freight origin–destination flows Antonio Comi and Paolo Delle Site

INTRODUCTION This chapter deals with the methods that can be used to estimate the origin–destination (O–D) matrices of urban goods vehicle flows. This is a key input, within the classical what-if approach to transport planning (Cascetta, 2009; Nuzzolo & Comi, 2013). The interest of the policy maker is in predicting the impacts on traffic flows, on environment and on the safety of socioeconomic development scenarios and transport supply design scenarios. The latter include regulation-, pricing-, infrastructure- and ITS- (Intelligent Transport Systems) based policies. Urban goods mobility includes different segments: traditional distribution flows to shops and food-and-drink outlets, and e-commerce flows (including restocking, deliveries to endconsumers’ pick-up points, reverse logistics), road works and construction sites, waste collection as well as maintenance services. Modelling of goods transport demand in urban areas is a relatively new field of research. The reader can refer to Chapter 3 by Tavasszy and de Bok in this Handbook for an overview of the evolution of this research sector. Different approaches have been proposed. The approaches can be classified on the basis of the unit of reference: the vehicle trip (trip-based models), the weight of the goods (commodity-based models) or the delivery of the goods (delivery-based models). Trip-based models are able to represent the different urban goods mobility segments. Commodity-based models and delivery-based models are able to represent the distribution flows to shops and to food-and-drink outlets and e-commerce flows. Commodity-based and delivery-based models are able to simulate the interactions between the economic and spatial characteristics of demand and the supply of logistics infrastructure and services better than trip-based models. It is important to underline that these models are typically based on revealed preference data. However, especially when used for planning purposes, they should also consider stated preference data so as to estimate stakeholders’ reactions to policy changes (e.g., Gatta & Marcucci, 2014; Le Pira et al., 2017; Comi, 2020; Marcucci et al., 2021). This chapter is organized as follows. Trip-based, commodity-based and delivery-based models are reviewed in, respectively, the second, third and fourth sections. The fifth section deals with an integrated framework able to model the purchasing, restocking and delivering stages. This section includes application examples. The sixth section discusses directions for future research while conclusions are drawn in the seventh section.

TRUCK-BASED MODELS FOR THE ESTIMATION OF O–D VEHICLE FLOWS The formulation of truck-based models considers the trips of goods vehicles. To obtain the spatial distribution, the simplest approach is to consider gravity models (Spielberg & Smith, 1981). 78

Estimating and forecasting urban freight origin–destination flows  79

Trips between two traffic zones are in direct proportion to the trips produced in the origin zone and the trips attracted by the destination zone, and in inverse proportion to the generalized travel cost between the two zones. These models can represent round trips only, i.e., trips from origin to destination and back to origin. They fail to represent tours where the intermediate stops are in different traffic zones. A tour is an ordered sequence of zones. The first approach that considers tours is that of the WIVER model, which was later included in the VISEVA software (Ambrosini et al., 2008; Friedrich et al., 2003; Lohse, 2001; Müller & Schneider, 2009). The O–D matrix of trips is estimated and tours are not identified. Therefore, tours are modelled only implicitly. The need for estimating the number of trips per tour appears particularly critical in forecasting, because this parameter depends, in principle, on the future travel costs. A second approach is based on entropy (Wang & Holguín-Veras, 2008). A macro-state defines the number of tour flows, i.e., the number of journeys following the given sequence of zones for each tour. A micro-state defines the tour followed by a specific vehicle. According to Wilson’s proposal derived from statistical mechanics (Wilson, 1970), entropy is a measure of probability of the macro-states of the system. The maximum entropy, i.e., the most probable macro-state, is the one that is associated with the highest number of micro-states. The following example clarifies the situation. For the sake of simplicity, this example uses trips instead of tours. Assume we have vehicle A and vehicle B. They both start their trip from the depot (node 0) and have to visit consignees 1 and 2. There are three macro-states: 1. The first macro-state has one trip from 0 to 1 and one trip from 0 to 2. Two micro-states are associated with this macro-state: either vehicle A goes to 1 and vehicle B to 2, or vehicle A goes to 2 and vehicle B goes to 1. 2. The second macro-state has two trips from 0 to 1 and zero trips from 0 to 2. Only one micro-state is associated with it: both vehicles go to 1. 3. The third macro-state has two trips from 0 to 2 and zero from 0 to 1. Only one micro-state is associated with it: both vehicles go to 2. The maximum entropy macro-state, i.e., the one with the highest number of micro-states and, therefore, the most probable is the first macro-state. The approach identifies the tours and is able to compute the most probable vehicle flow for each tour. Then, O–D flows, i.e., O–D trips, are obtained based on the tour-O–D pair incidence matrix, providing identification of tours by O–D pair. The number of all possible tours in a real network is clearly enormous. In the proposal by Wang and Holguín-Veras (2008), tours are identified based on a survey of a sample of transport companies. Only these tours are considered in the model. Tour productions, tour costs and the total cost on the network are also based on the survey. The authors provide an application to Denver, and find that the mean absolute percentage error between simulated tour flows and observed tour flows is in the range 6–7%. An entropy model with additional constraints on traffic counts is considered in GonzalesCalderon and Holguín-Veras (2019). This model is applied to a toy network only. The approach shows limitations in the simulation of both current and future scenarios. Additional tours that are not considered in the survey can take place in the network but are not modelled even in the current scenario. In forecasting, the model should be re-estimated on the basis of the changes in trip productions and in generalized costs. In principle, the set

80  Handbook on city logistics and urban freight

of tours should be regarded as endogenous, because in the future this would be changed as a consequence of the new generalized costs. This would require a tour-choice model which is left for future research.

COMMODITY-BASED MODELS FOR THE ESTIMATION OF O–D FREIGHT FLOWS Commodity-based models for estimating O–D freight flows in urban areas consider the movement of goods as the reference unit. Such a class of models allows accurate simulation of the mechanism underlying the generation of the freight transport demand. In fact, although the production of such a demand depends on freight types, given that the same vehicle can be used for their transport, such a characteristic cannot be represented by truck-based models. Commodity-based models are hence more specific for assessing strategic actions, such as land use or zoning decisions at the local level, that could determine the location of the origin or destination of goods. Traditionally, such a class of models focuses on shop restocking and delivering, i.e., vehicle flows from warehouse/distribution centres to trade or service locations, such as shops, food-and-drink outlets and service activities (Ogden, 1992; Anand et al., 2012; Comi et al., 2014; Anand et al., 2015). Only recently has there been a shift towards an integrated approach (Russo & Comi, 2010, 2012; Gonzalez-Feliu & Peris-Pla, 2019; Comi, 2020; Russo & Comi, 2020) to consider the flows deriving from in-store and online shopping. The two approaches are explored in the following sub-sections. Commodity O–D Flows Although the final output of such a class of models remains the estimation of vehicle O–D matrices, they focus on commodities required by end-consumers (either private or business) to satisfy their needs. Such models are usually developed by a sequence of models aggregated into two levels: ●



Commodity level, i.e., estimation of quantity O–D flows; at this level, models entail calculation of: ● attraction flows related to the freight required for satisfying end-consumer demand; ● acquisition, which allows attracted freight flows to be spatialized and the O–D matrices to be estimated; Vehicle level, which allows quantity flows to be converted into vehicle flows; at this level, the models concern the determination of: ● restocking trip-chains in terms of quantity delivered at each stop, zone and vehicle needed for restocking; ● time used as well as the journey chosen for restocking sales outlets.

With respect to the main scope of this section, the commodity level is described below and a further simplified step is introduced for obtaining the O–D vehicle flows. For further details on vehicle level, as well as on conversion of commodity flows into vehicle ones, readers should refer to Russo and Comi (2010), Comi et al. (2014), Nuzzolo and Comi (2014), Comi et al. (2021), as well as the following sections of this chapter.

Estimating and forecasting urban freight origin–destination flows  81

Freight type (s) and time period play a key role in estimating commodity flows; however, for simplicity of notations, they will be taken as understood, unless otherwise stated. Therefore, store let Qod be the average quantity flow of freight type s attracted between zone o and zone d within the identified study area in a given time period (e.g., day, time slice, week), which can be estimated as follows: store Qod = Qdstore × p éëo /d ùû (4.1)

where ● ● ●

store Qod is the average quantity flow of freight attracted by zone d and coming from zone o; store Qd is the average freight quantity attracted by zone d obtained by an attraction model; p éëo / d ùû is the probability that freight attracted by zone d comes from zone o (e.g., warehouse location zone); it represents the acquisition share obtained by a discrete choice acquisition model.

In addition, in order to take into consideration that freight with destination in a traffic zone d can be related to e-shopping (e.g., at-home delivery), the total quantity of freight attracted to zone d belonging to the study area, QTd , can be determined as follows: QTd = Qdstore + Qdonline (4.2)



where Qdonline is the quantity of freight bought online by users that is required to be delivered in zone d. The attraction flow is modelled through a model that allows us to obtain the average flow of freight that arrives in each zone of the study area in order to satisfy end-consumer demand. In general, each end-consumer can purchase the goods required in different shops or, in the case of some freight types, he/she can buy or consume them in commercial concerns such as cafés and restaurants. The attraction model is usually a regressive-type model in which the average daily quantity of freight attracted by zone d, Qdstore , is estimated as follows: Qdstore =



å

P p =1

b p × AD p,d +

å

N ASA j =1

b j × ASA j ,d (4.3)

where ●





AD p,d is the number of retail employees of type p (e.g., retail employees at foodstuff shops) in zone d; ASAj,d is the j-th land-use variable of zone d (e.g., it could be a dummy variable equal to 1 if the proportion of retail employees to inhabitants in zone d is higher than 35%); βp and βj are parameters to be calibrated.

Further modelling approaches for estimating quantities with destination in establishments located in zone d can be found in Chapter 6 (Sánchez-Díaz & Castrellon) of this Handbook. To simulate the origin of freight flow for each attraction zone within the study area, the acquisition model is used. It simulates the choice of an origin among possible alternatives to

82  Handbook on city logistics and urban freight

get the freight to be sold. The share of freight attracted by zone d coming from zone o (e.g., places where production places/firms, distribution centres and warehouses are located) is usually obtained through random utility models as follows:

å b p [o / d ] = å exp æçè å exp æç è



PO

po =1

× AI po,o +

po

N

PO

o ' =1

po =1

å b × C ö÷ø + å b ×C U

u

u =1

u , od

U

b po × AI p ,o '

u =1

u

ö u ,o ' d ÷ ø

(4.4)

where ● ●

● ●

p[o/d] is the probability that the freight attracted by zone d comes from zone o; AIpo,o is the generic po attribute measuring the commodity-flow production of zone o (e.g., the number of warehouse employees of zone o); Cu,od is the generic component u of travel cost (e.g., travel distance) between o and d; βpo and βu are parameters to be calibrated.

Traditionally, random utility models in a gravitational form are specified and calibrated (Comi, 2020). Assuming to consider AIo, the number of warehouse employees of zone o, as a measure of production power of zone o, and Cod the travel distance between the O–D pair od, the probability p[o/d] is calculated as follows:

( ) å ( AI ) × (C )

p [ o / d ] = ( AI o ) × Cod b1

b1

N

o¢=1

o'

b1

o¢d

b1

(4.5)

Once the O–D flows are obtained, for performing ex-ante assessment of freight transport system, it could be useful to characterize them in terms of distribution channels (Ghiani et al., 2004). In fact, freight flow is usually from producers to retailer and HORECA (HOtel, CAfé, Restaurant) activities (and thus to end-consumers) and the number of intermediaries between sender and receiver can influence the composition and characteristics of freight (quantity) flows. According to Russo and Comi (2010), let c be the distribution channel identified with a generic path from production/international trade to residence/consumption zone, and let Y be the set of inside or outside zones (Y = W or Z) with respect to the urban study area (Figure 4.1). As proposed by Russo and Comi (2010), the distribution channels can be analyzed by means of an aggregation into two classes depending on who the decision maker is: ●



hyper-channel r (cr); it includes all distribution channels in which the decision maker chooses how, where and when to go (or to send their transport employees) to bring the goods for restocking d (pull hyper-channel); hyper-channel l (cl); it includes the other distribution channels in which the decision maker cannot be considered the receiver and in which many different decision makers can be considered; in this case, the receiver suffers the choice of other subjects involved, who choose how and from where the freight must be delivered and can give some addressing regarding the delivery time (push hyper-channel).

Referring to the O–D freight quantity flows for restocking, let QTod [cY] be the quantity of freight moved between the macro-area Y (inside or outside the study area where the zone o belongs) and zone d using the distribution channel c. This quantity can be calculated as follows:

Estimating and forecasting urban freight origin–destination flows  83 Zone o

Zone d (Zone o)

Zone w (Zone d) (Zone o) (Zone z)

Zone z (Zone w) (Zone d) (Zone o)

Retailing

Residence/ Consumption zone

Wholesaling/ Consolidation point

Wholesaling/ Consolidation point

Producing/ International trade Goods flows controlled by end-consumer (e.g. family)

Freight flows controlled by retailer

Freight flows controlled by producer

Source:   Russo & Comi, 2010

Figure 4.1  Graph of distribution channels QTod [cY ] = QTod × p[cY /d ] (4.6)



where p[cY/d] is the probability that the freight flows to a shop in zone d, uses the hyperchannel c and comes from macro-zone Y; it can be obtained by a hyper-channel and restocking area choice model. Then, the probability, p[cY/d], of choosing the hyper-channel c and the restocking area Y can be expressed as: p[cY / d ûù = p éë c / ds] × p éëY / cd ùû (4.7)

where ●



p[c/d] is the probability of choosing hyper-channel c for restocking activities located in zone d; p[Y/cd] is the probability of choosing the restocking area Y, having chosen the hyperchannel c for restocking activities located in zone d.

Once the quantity of goods bought/sold in zone d is delivered using the hyper-channel c from macro-zone Y, QTod[cY], has been estimated by the above models, the restocking quantity flows characterized for vehicle type v, QVod, departing from zone o, can be obtained as follows: QVod éë vcY ùû = QTod ëécY ùû × p éë v / d ùû (4.8)

where: ●

QVod[vcY] is the freight quantity bought/sold in zone d transported on O–D pair od by vehicle type v;

84  Handbook on city logistics and urban freight ●

p[v/d] is the probability that the freight on O–D pair od is transported by vehicle type v, obtained by a vehicle choice model.

Finally, assuming that each vehicle only delivers to activities located in the same traffic zone, the number of vehicles of type v required for restocking the retail outlets of the study area can be estimated as follows: VCod éë v ùû =



å

c = cr ,cl

å

Y =W , Z

QVod éë vcY ùû q éë vcYd ùû (4.9)

where: ● ●

VCod[v] is the total average number of freight vehicles of type v moving on O–D pair od; q[vcYd] is the total average freight quantity delivered to activities in zone d by vehicle type v from restocking area Y through hyper-channel c.

The Integrated Commodity O–D Flows Acquisition and restocking flows (i.e., connection between zones where the retailer takes the freight and zones where she sells them) are mainly generated/produced to satisfy end-consumer demand. Therefore, logistics models have to take into account end-consumer choices. Shopping has only recently been analyzed as a component of goods mobility. It has also been found that changes in shopping attitudes (e.g., buying online) or actions impacting on purchasing behaviour of end consumers (e.g., location of shopping zones, transport mode used for shopping) can also affect restocking and delivery (Barone et al., 2014; Wiese et al., 2015; Comi & Nuzzolo, 2016; Kunytska et al., 2021). The integration of purchasing activity into commodity-based models consists of adding a further modelling (i.e., purchasing) level which allows us to simulate end-consumer behaviour for shopping (both in-store/shop and online) through the simulation of purchase decision making and hence of trips performed to reach the store/shop. At this stage, the quantity bought by end consumers in order to satisfy their needs is estimated, and the freight flows attracted by each traffic zone are identified. In particular, based on empirical evidence (Comi, 2020, 2021), the quantity of goods purchased by end consumers is strictly related to socio-economic characteristics of the end consumer (i) as well as by the type of shop (k) and transport mode (m) used. Therefore, such dimensions should be pointed out in modelling development. Following that proposed by Russo and Comi (2010) and developed/uploaded in Comi (2020), the quantities required by each internal zone to satisfy end-consumer needs can be obtained as follows:

å QI [k ] = å å QI [k ] (4.10) =å å å å å D [ km ] × p [ dim / mk ] × dim

QI dstore =

K

k =1

store d

K

EC

k =1

i =1

K

EC

N

M

DIM

k =1

i =1

o =1

m =1

dim =1

i od

i , store d

i

where ●

QI dstore éë k ùû is the quantity of goods bought/sold in retail outlet k in zone d within the study

area;

Estimating and forecasting urban freight origin–destination flows  85 ●

● ●



QI di,store éë k ùû is the quantity of goods bought/sold in retail outlet k in zone d given by the

demand of end-consumers belonging to category i (among all EC categories indentified) and living/working in a zone within the study area; dim is the dimension of purchases, e.g., expressed in kg; pi[dim/mk] is the probability that a trip concludes with a purchase of dimension dim conditional upon undertaking a trip to retail outlet k for a purchase using transport mode m among all possible M; i Dod éë km ùû is the average number of (e.g., weekly) shopping trips with origin in zone o undertaken by the end-consumer belonging to category i for purchasing goods in retail outlet type k (i.e., small shop, supermarket, hypermarket) located in zone d by using transport mode m.

It should be noted that the goods quantity required in zone d also depends on the end consumers who purchase there and live/work in a zone outside the study area (QEdstore éë k ùû ). This rate can be determined, for example, through direct estimates based on surveys carried out at the border of the study area. In the general, this quantity can be further characterised for endconsumer category i. i Referring to Eq. 4.10, the average number of shopping trips, Dod éë km ùû , can be obtained by simulating the production of purchases as follows: i Dod éë km ùû =



ACQoi,store i × p éë dk /o ùû × pi éë m /dko ùû shacqoi,store (4.11)

= Doi × pi éë dk /o ùû × pi éë m /dko ùû where ●









ACQoi,store is the average number of (e.g., weekly) in-store purchases of goods made by the end consumer belonging to category i and living in zone o; shacqoi,store is the average number of purchases of goods made by end consumers belonging to category i on each shopping trip, obtained by a trip generation model; pi[dk/o] is the probability that users of category i, undertaking a trip from o, travel to destination zone d for purchasing goods at retail outlet (shop) type k (e.g., small shop, supermarket, hyper-market), obtained by a shop type and location model; pi[m/dko] is the probability that users, travelling between o and d for purchasing goods in shop type k, use transport mode m, obtained by a mode choice model; Doi is the (e.g., weekly) average number of trips undertaken by end consumers belonging to category i for purchasing goods with origin in zone o.

Then, the calculation of purchases (Eq. 4.11) made in-store or online (disaggregated by goods type) can be estimated according to the following sequence of models:

ACQoi,h = ni éëo ùû × moi,h = ni éëo ùû ×



å

NY y =1

y × pi éë yh / o ùû (4.12)

where ●

ACQoi,h is the number of purchases (e.g., weekly) of goods made by end consumers belonging to category i (e.g., students, employees) and living in zone o, through the shopping mode h (e.g., in store or on line);

86  Handbook on city logistics and urban freight ● ●



ni[o] is the number of end consumers belonging to category i and resident in zone o; moi,h is the average number of purchases (e.g., weekly) of goods, made using the shopping mode h by end consumers belonging to category i and living in zone o; pi[y h / o] is the probability to make y (among all the possible NY) purchases of goods by end-consumers belonging to category i and resident in zone o using the shopping mode h; it can be obtained by a purchase choice model.

Furthermore, moving from the number of e-purchases, and introducing a purchase-choice model for online purchases as done earlier, the quantity, Qdonline , required by the end-consumer buying online at zone d can be estimated (see Eq. 4.2).

DELIVERY-BASED MODELS FOR THE ESTIMATION OF O–D FREIGHT FLOWS The delivery (movement)-based models focus on pick-ups and deliveries. The use of delivery (movement) as a reference unit allows the provision of a direct link between generators and transport operators, through the use of the same reference unit. The models proposed within the delivery-based approach consist of a sequence of statistic-descriptive models. Similarly to a commodity-based model, let NDod ëé sk ûù be the average total delivery flow of goods type s transported between zone o and zone d for restocking retail businesses (of type k; e.g., small shop, supermarket, shopping centre) or home delivery in a given time period (e.g., week, time slice, day). For simplicity of notation, as introduced in the earlier section, the freight type (s) and the time period will be taken as understood unless otherwise stated. Thus, the average delivery flow, NDod ëé k ûù , can be estimated as follows: NDod éë k ùû = NDd éë k ùû × p éëo / dk ùû (4.13)

where ●



NDd éë k ùû is the average number of deliveries required in zone d in outlet type k for satisfying selling demand obtained through an attraction delivery model; p éëo / dk ùû is the probability that deliveries of goods attracted by zone d come from zone o (e.g., warehouse or courier location zone); it represents the acquisition share obtained, for example, by a discrete choice acquisition delivery model.

The attraction delivery model allows us to obtain the average flow of freight that arrives in each zone of the study area in order to satisfy economic activity demand (e.g., retailing demand). Such a model is a regressive-type model in which the average daily deliveries attracted by zone d, NDd, is estimated as follows:

NDd =

å

N p =1

b p × AD p,d +

å

N ASA j =1

b j × ASA j ,d (4.14)

Estimating and forecasting urban freight origin–destination flows  87

where ● ●



AD p,d is the number of retail employees of type p in zone d; ASAj,d is the j-th land-use variable of zone d (i.e., it could be a dummy variable equal to 1 if the proportion of retail employees to inhabitants in zone d is higher than 35%); βp and βj are are parameters to be calibrated.

Subsequently, spatialization is provided through an acquisition delivery model as is used in commodity-based ones. A first example of this class of model was developed in France and implemented in FRETURB software by LET (Gonzalez-Feliu et al., 2012; Toilier et al., 2018). Based on this French approach, CityGoods was developed by Gentile and Vigo (2006). The prototype model was tested on several cities in Emilia-Romagna, a region in Northeast Italy. The objective was to build a demand-generation model to estimate the yearly number of operations generated by each zone. Once O–D delivery flows are estimated, further stages point out the transport used as well as the vehicle used. In fact, freight can be transported and hence each establishment can be restocked by different transport services according to which transport service is used. The possible transport service types (NR) are (Nuzzolo & Comi, 2014): ● ● ● ●

retailer on own account; retailer by third party (i.e., transport company or courier that offers small-sized shipments); wholesaler on own account; wholesaler by third party.

Then, the average delivery O–D flow carried out by transport service type r on pair od, NDod éërk ùû , can be determined as follows: NDod éërk ùû = NDod éë k ùû × p éër / ok ùû (4.15)



where p éër / odk ùû is the probability that a retail outlet k is restocked by transport service type r obtained by a transport service-type model. The involved decision maker manages the restocking delivery process. Examples of such models developed for simulating the choice made by retailers or HORECA managers were provided by Nuzzolo and Comi (2014) and Russo and Comi (2016). Similarly to commodity-based models, the deliveries can be characterized by type of vehicle v, and, according to the assumptions introduced for Eq. 4.9, by the number of vehicles of type v required for restocking the retail outlets of the study area can be estimated as follows: VCod éë v ùû =



å å NR

NK

r =1

k =1

NDod éërk ùû × p éë v / rkd ùû nd éë vrk ùû (4.16)

where: ● ●

VCod[v] is the total average number of freight vehicles of type v moving on O–D pair od; p[v/rkd] is the probability to deliver using vehicle type v;

88  Handbook on city logistics and urban freight ●

nd[vrk] is the total average number of deliveries to activities of type k (among the NK ones available) in zone d by vehicle type v, using transport service r (among the NR ones available).

INTEGRATED MODELLING FRAMEWORK FOR THE ESTIMATION OF O–D FREIGHT FLOWS In this section, an integrated modelling framework is recalled as illustrated in Figure 4.2 (Nuzzolo & Comi, 2014). Methods and models used for goods transport in urban areas have shifted towards an integrated approach to goods restocking, e-delivering and shopping/purchasing mobility, also pushed by the increase of e-commerce. According to the urban goods movements reported in Figure 4.2, the modelling framework consists of three main modelling stages as detailed below: purchasing, restocking and delivering (Figure 4.3). The modelling system is a multi-step model and considers a disaggregated approach for each decisional level. Each proposed model is framed within the random utility discrete choice setting (Ben-Akiva & Lerman, 1985). Each decision maker is considered to be a rational decision maker who maximizes utility relative to her/his choices. In particular, the considered decision makers are: ● ●

● ●

end consumers: they decide how much and where to buy; retailers: they sell to end consumers and decide where to buy goods to restock their stores and the type of transport service to use (own account or third-party); wholesalers/distributors: they decide the type of transport service to use; carriers/couriers: they provide transport services.

Figure 4.2  Modelling urban O–D flows

Estimating and forecasting urban freight origin–destination flows  89

The Purchasing Stage Purchasing allows us to simulate end-consumer behaviour for shopping (both in-store/shop and online) through the simulation of purchase decision making and hence of trips performed to reach store/shop. At this stage, the quantity bought by end consumers in order to satisfy their needs is estimated, and the freight flows attracted by each traffic zone are identified. Therefore, according to the general modelling framework plotted in Figure 4.3, the output of this stage are the attracted quantities, which can be obtained as described in the second section. In particular, moving from Eq. 4.2, let QTd [ sk ] be the total quantity of goods attracted by zone d for selling in store and bought on line by end consumers who asked to be delivered here, it is equal to: QTd éë sk ùû = Qdstore éë sk ùû + Qdonline éë sk ùû =



= QI dstore éë sk ùû + QEdstore éë sk ùû + Qdonline éë sk ùû

(4.17)

PURCHASING Trip

shop type and location

shopping trip O-D matrices

produced trips

mode choice retail activities

trip generation

mode O-D matrices

quantity purchase model

produced in-store purchases

purchase generation

produced on-line purchases

RESTOCKING - Quantity

DELIVERING tour Freight Vehicle O-D matrices

attracted quantities

store acquisition

delivery tour model sub-system

inhabitants, visitors

warehouses and distribution activities

(at shop)

delivery O-D matrices for transport and courier services

store delivering model sub-system

e-purchase delivering model sub-system

quantity O-D matrices

(deliveries at shop)

PURCHASING Purchase

Figure 4.3  Mixed/integrated modelling framework

(deliveries at home)

model

data

90  Handbook on city logistics and urban freight

where ●







QTd éë sk ùû is the total quantity of goods of type s sold/required in retail or food-and-drink outlets of type k in zone d; QI dstore ëé sk ûù is the quantity of goods of type s bought/sold in retail outlet k in zone d given by the demand of end consumers living/working in a zone within the study area; QEdstore ëé sk ûù is the quantity of goods of type s bought/sold in retail outlet k in zone d given by the demand of end consumers living/working in a zone outside the study area; Qdonline ëé sk ûù is the quantity of goods of type s bought in e-retail given by the demand of end consumers living/working in a zone within the study area.

Subsequently, the average total quantity flow of goods type s transported between zone o and zone d for restocking retail businesses or home delivery in a given time period (e.g., week, time slice, day), QTod ëé sk ûù , can be estimated as follows: QTod éë sk ùû = QTd éë sk ùû × p éë o / ds ùû (4.18)

where ●



(

)

QTd éë sk ùû = QI dstore éë sk ùû + QEdstore éë sk ùû + Qdonline éë sk ùû , Eq. 4.17, is the average freight quantity required in zone d for satisfying end-consumer demand obtained from the models described in the earlier sub-sections; p éëo / ds ùû is the probability that goods of type s attracted by zone d come from zone o (e.g., warehouse or courier location zone); it represents the acquisition share obtained, for example, by a discrete choice acquisition model.

The Restocking Stage Restocking, given the quantity attracted by each traffic zone (due to in-store/shop and online purchases), allows us to estimate the restocking quantity O–D matrices characterized by freight types: NDod éërs ùû =



QTod éë sk ùû × p éër / sd ùû (4.19) q éërs ùû

where ●

● ● ●

NDod éërs ùû is the number of deliveries of freight type s carried out by transport service type r on O–D pair od; QTod éë sk ùû is the average freight quantity flow on O–D pair od; q ëérs ûù is the average freight quantity delivered with service type r (shipment size model); p éër / sd ùû is the share or the probability of having deliveries by service type r (transport service model).

The change, from quantity/delivery O–D matrices to vehicle O–D matrices, is not direct, particularly in urban areas where freight vehicles undertake complex routing patterns involving

Estimating and forecasting urban freight origin–destination flows  91

trip chains (tours). Each restocker jointly chooses the number and the location of deliveries for each tour and hence defines their tours, trying to reduce the related costs (e.g., using routing algorithms). The Delivering Stage The last stage of the recalled modelling framework is the delivering one. It points out the generation of tours performed for delivering goods to end consumers and to stores/shops. Following Nuzzolo and Comi (2014) and Comi et al. (2021), the freight vehicle O–D matrices can be obtained from the delivery O–D matrices through a delivery tour model, which uses a two-step procedure: ●



from delivery O–D matrices, by computation of the number of delivery tours departing from each zone within the study area (tour generation sub-model); definition of freight vehicle O–D matrices from delivery tours (delivery location sub-model).

Given a study area divided into homogeneous sub-areas/zones (zoning) and the relative delivery O–D matrix (average deliveries departing from warehouse zone o and transferred to zone d), let NDod be its generic element. The freight vehicle O–D matrices, which satisfy the earlier matrix, can then be estimated by using an aggregate multi-step delivery tour model. The total number of tours To[vntrs] departing (generated) from zone o at time t made by vehicle type v with n stops/deliveries for performing deliveries from warehousing located in zone o can be determined as follows (tour generation sub-model): To éë vntrs ùû = To éërs ùû × p éë t / ros ùû × p éë n / tros ùû × p éë v / ntros ùû

(4.20) NDo éërs ùû = To éërs ùû = no éërs ùû

å ND éërs ùû å n × p éën / tros ùû N

o =1 NN

od

n =1

where ●

● ● ●





To[rs] is the number of tours departing (generated) from zone o using transport service r for transport freight type s; NDo[rs] is the average number of deliveries performed departing from origin zone o. p[t/ros] is the share or probability of undertaking tours at time t; p[n/to] is the share or probability of undertaking tours with n stops/deliveries (trip chain order model); p[v/ntros] is the share or probability of undertaking tours by vehicle type v (e.g., light goods vehicles, heavy goods vehicles); no ëérs ûù is the average number of deliveries performed by NN tours departing from zone o.

Finally, the number of vehicles VCdi d j from the origin zone of tour o (e.g., location of depots) on (di-dj) pair can be estimated as follows:

92  Handbook on city logistics and urban freight

VCdi d j éë vntros ùû =

å

NN g =1

VCd g+1d g éë vntros ùû i j

= To éë vntrs ùû ×

å

NN g =1

(4.21)

p éë d gj +1 / dig vntros ùû

The above probabilities p[t], p[n], p éë d gj +1 / dig ùû can be obtained by statistic-descriptive or probabilistic-behavioural models. As shown in Comi et al. (2021), calibration of such models could require establishment-/driver-based surveys for obtaining tour-related data (e.g., stop purpose, commodity and quantities handled) which are resource (time and money) consuming. They require the driver to stop the vehicle and answer some questions posed by the surveyor. Nowadays, thanks to the opportunity offered by telematics, AVM (Automated Vehicle Monitoring) data allow the vehicle’s movements to be tracked passively and statisticdescriptive models to be obtained (Comi et al., 2021; Comi & Polimeni, 2021). Examples of Applications Such a modelling framework was validated through comparison between revealed and observed flows both in terms of goods sold at shops and vehicle flows in some road sections (Nuzzolo & Comi, 2014). It was used for performing the assessment of: ● ● ● ●

measures for pollutant emission reduction; land-use governance scenarios; effects on freight restocking under a future scenario when demographic changes occur; effects on parking demand.

As shown in Filippi et al. (2010), the modelling framework was used for assessing the effects of some measures designed to reduce pollutant emissions in the inner area of Rome: vehicle fleet renovation and introduction of an urban consolidation centre. The finding was that an urban consolidation centre can be more effective in reducing environmental externalities than measures based on vehicle fleet renewal. These issues are further discussed in Chapter 8 by Björklund and Gammelgaard and Chapter 9 by Lebeau et al. in this Handbook. As regards land-use governance scenarios, the model was implemented in the city of Padua in northern Italy (Nuzzolo et al., 2014). Padua is characterized by a concentric structure with three types of urban space according to the distance from the centre: the central area, where the density of end consumers and small retailers is usually higher; the first ring, with medium end-consumer density and the presence of warehouses; and the second ring, where end-consumer density is low and large shopping malls and freight distribution facilities are located. The strategy of clustering warehouses, distribution centres and large retail outlets in the first ring can have impacts in terms of reducing both freight distribution and shopping trip distances. Indeed, with respect to the other land-use scenarios, this solution entails a reduction in freight distribution vehicle-kilometres-travelled and a small increase in the number of carbased shopping trips, which is offset by a considerable reduction in the number of car shopping trip vehicle-kilometres travelled. In addition, following Filippi et al. (2010), COPERT (COmputer Programme to calculate Emissions from Road Transport; Eggleston et  al., 2000) methodology was adapted for the

Estimating and forecasting urban freight origin–destination flows  93

urban and metropolitan contexts and the external effects in terms of pollutant emissions were then estimated. We observed that the scenario of an expansion of large retail outlets in the second ring is the least favourable: although the vehicle-kilometres of commercial vehicles for freight distribution decreased by 3.9%, the purchase trips by car increased by 2.9% and the total equivalent vehicle-kilometres increased by 2.8%. In a scenario with an increase in warehousing activities and distribution centre density in the first ring, there is a more sizable decrease in freight distribution vehicle-kilometres in the urban area (5.0%), although the variation in car shopping trips remains constant. The scenario with a clustering strategy of freight distribution and retail activities in the first urban ring would be more favourable, with a decline of 7.4% in freight distribution vehicle-kilometres and a 12.8% decline in shopping trips–kilometres by car. These variations correspond to a 12.7% decrease in total equivalent vehicle-kilometres. In this last scenario, we also observe a considerable reduction in heavy goods vehicle-kilometres (5.7%). Land-use planning for more sustainable urban freight is further discussed in Chapter 12 by Dablanc in this Handbook. The modelling framework was also used to assess the effects on freight restocking under a future scenario when demographic changes occur in a medium-sized urban area (Nuzzolo & Comi, 2014; Nuzzolo et al., 2014). The results indicate that the effects of demographic changes on shop restocking flows can be significant. For example, the shifting of middle-aged adults into a later age group could result in an increase in shopping trips to nearby shops, mainly to small- and medium-sized retail outlets. This could lead to a consequent increase in carkilometres, with subsequent environmental, social and economic impacts. Furthermore, an expectation of an increase in e-shopping could reduce consumer trips. The models were implemented to analyse parking demand. Comi and Polimeni (2021) evaluated the delivery tours and linked such tours with requests for parking for performing deliveries. The delivery traffic within the northeastern region of Veneto in Italy has been studied. The most notable findings are that tours consume a large portion of the total vehicle-hours for delivering and the share of stop time varies with the low values of round trips (i.e., 5%). We can see that, as the number of stops increases, the time spent for at-customer operations decreases, with an incidence for longer tours of about 30%. Furthermore, tours increase with population. Such a trend is understandable. In fact, in areas with high population density, the shops tend to be smaller, with few spaces for storage. This produces a larger number of tours and stops, given that smaller deliveries are performed and during the same tour more than one customer is served. In addition, in order to support the analyses of city logistics scenarios, a decision support system (DSS) called CLASS (City Logistics Analysis and Simulation support System) was developed (Comi & Rosati, 2013). It allows analysis and simulation of the effects of city logistics measures and the effects of exogenous scenario changes, such as land-use and socioeconomic characteristics, and the estimation of traffic-related impacts (e.g., traffic pollutant emissions and road accidents).

RESEARCH DIRECTIONS As shown throughout this chapter, methods and models for estimating and forecasting urban freight O–D flows moved to link end-consumers’ needs with freight operators’ activities. This gives new opportunities for city planners and technicians in using more effective tools for

94  Handbook on city logistics and urban freight

assessing new scenarios. In fact, they allow considering how changes in end-consumers’ shopping behaviour could impact on freight flows. At the same time, it opens up new opportunities for further development. The incoming requests from city planners engaged in developing a more sustainable urban mobility through the promotion of active mobility (Nigro et al., 2022; Comi et al., 2022) should be highlighted, as shown by literature surveys on end consumers’ choice of transport mode when they travel for shopping (Comi, 2021), or the actions of shifting people from motorized transport to active mobility could impact on end-consumers’ way of buying (e.g., quantity bought, type of retailand-food outlets). In addition, the creation of more accessible and liveable areas can attract shoppers with new attitudes and requests, and impact on restocking and e-delivering operations. Modellers should also be pushed towards the integration of urban freight flow methods and models with the land-use ones, as done in passenger mobility methodologies (e.g., LUTI – Land Use Transport Interaction – models; Russo & Musolino, 2012; Hounwanou et al., 2018). On the other hand, in combination with methodological innovations, it is expected that the advances in modelling framework are followed by innovation in model parameter estimation, as well as in modelling framework calibration and validation. The advanced modelling framework allows modellers to capture the behaviours of different actors involved in the process, and to be open towards new needs in terms of data. The consolidated methodologies should be extended in order to include increasing amount of data coming from the network as well as from the tracking and tracing of end-consumers travelling for shopping (or buying online) and trucks driving in urban areas (Russo & Comi, 2021). Finally, the current modelling framework refers to distribution flows, mainly food-anddrink outlets and e-delivering, whereas the other commercial flows are increasing their consistency in the urban areas. Therefore, the modelling framework could be extended to include additional commercial flows: reverse logistics, road works and construction sites, waste collection, and service (maintenance) trips, which could have a strong impact on urban space and road use. For example, the duration of the activity is one of the key differences between freight and service operations. Whereas, in the freight case, most deliveries and pick-ups may take as little as a few minutes, service activities typically last much longer.

CONCLUSIONS The development of DSSs is useful when appraising the impacts of urban goods transport measures and policies. The methods and models that have been reviewed in this chapter provide essential inputs to the evaluation stage, which is dealt with in depth in Chapter 13, ‘Assessment of innovative city logistics solutions’, by Delle Site in this Handbook. As such, the present chapter contributes to the development of of economically, environmentally and socially sustainable mobility solutions. The main components of urban goods mobility were identified. The pros and cons of current demand modelling frameworks that have been developed for supporting city logistics decision making were pointed out, and future research challenges identified. For example, the focus of existing frameworks is on the distribution activities to shops and households, but the other commercial flows need to be modelled within the current frameworks.

Estimating and forecasting urban freight origin–destination flows  95

The opportunities offered by telematics can be further exploited. Automated vehicle monitoring and floating car data (AVM/FCD – Floating Car Data), including GPS (Global Positioning System) for freight vehicles are increasingly available due to the deployment of telematics in companies for both fleet management and insurance purposes. They provide the means to investigate vehicle movements. The existing frameworks consider short-term behavioural responses to measures and policies. The frameworks can be extended to model transport/land-use interactions, with particular regard to longer-term location decisions related to urban freight centres and large shopping centres.

REFERENCES Ambrosini, C., Meimbresse, B., Routhier, J. L., & Sonntag, H. (2008). Urban freight modelling: A review. In E. Taniguchi & R. G. Thompson (Eds.), Innovations for city logistics (pp. 197–212). Hauppauge, NY: Nova Science Publishing. Anand, N., Quak, H., Van Duin, R., & Tavasszy, L. (2012). City logistics modelling efforts: Trends and gaps – A review. Procedia – Social and Behavioural Sciences, 39, 101–115. Anand, N., van Duin, R., Quak, H., & Tavasszy, L. (2015). Relevance of city logistics modelling efforts: A review. Transport Reviews, 35(6), 701–719. Barone, V., Crocco, F., & Mongelli, D. W. E. (2014). Models of choice between shopping and E-shopping. Materials engineering and mechanical Automation, 442, 607–616. Ben-Akiva, M., & Lerman, S. R. (1985). Discrete choice analysis: Theory and application to travel demand. Cambridge, MA: MIT Press. Cascetta, E. (2009). Transportation systems analysis: Models and applications. New York: Springer. Comi, A. (2020). A modelling framework to forecast urban goods flows. Research in Transportation Economics, 80, 100827. https://doi​.org​/10​.1016​/j​.retrec​.2020​.100827. Comi, A. (2021). Shopping and transport modes. In R. Vickerman (Ed.), International encyclopaedia of transportation, volume 5 (pp. 98–105). United Kingdom: Elsevier Ltd. https://doi​.org​/10​.1016​/ B978​0​- 08​-102671​-7​.10412​-9. Comi, A., & Nuzzolo, A. (2016). Exploring the relationships between e-shopping attitudes and urban freight transport. Transportation Research Procedia, 12, 399–412. https://doi​.org​/10​.1016​/j​.trpro​. 2016​.02​.075. Comi, A., & Polimeni, A. (2021). Forecasting delivery pattern through floating car data: Empirical evidence. Future Transportation, 1(3), 707–719. https://doi​.org​/10​.3390​/futuretransp1030038 Comi, A., & Rosati, L. (2013). CLASS: A city logistics analysis and simulation support system. Procedia – Social and Behavioral Sciences, 87, 321–337. https://doi​.org​/10​.1016​/j​.sbspro​.2013​.10​.613 Comi, A., Donnelly, R., & Russo, F. (2014). Chapter 8. Urban freight models. In L. Tavasszy & J. De Jong (Eds.), Modelling freight transport (pp. 163–200). Amsterdam: Elsevier. https://doi​.org​/10​.1016​/ B978​- 0​-12​- 410400​- 6​.00008-2 Comi, A., Nuzzolo, A., & Polimeni, A. (2021). Aggregate delivery tour modelling through AVM data: Experimental evidence for light goods vehicles. Transportation Letters, 13(3), 201–208, https://doi​. org​/10​.1080​/19427867​.2020​.1868178 Comi, A., Polimeni, A., & Nuzzolo, A. (2022). An innovative methodology for micro-mobility network planning. Transportation Research Procedia, 60, 20–27. https://doi​.org​/10​.1016​/j​.trpro​.2021​.12​.004 Eggleston, S., Gorißen, N., Hassel, D., Hickman, A. J., Joumard, R., Rijkeboer, R., … Zierock, K. H. (2000). COPERT III: Computer programme to calculate emissions from road transport. Brussels: European Environment Agency. Filippi, F., Nuzzolo, A., Comi, A., & Delle Site, P. (2010). Ex-ante assessment of urban freight transport policies. Procedia – Social and Behavioral Sciences, 2(3), 6332–6342.

96  Handbook on city logistics and urban freight

Friedrich, M., Haupt, T., & Nökel, K. (2003). Freight modelling. Data issues, survey methods, demand and network models. In Proceedings of the 10th international conference on travel behaviour research, Lucerne. Gatta, V., & Marcucci, E. (2014). Urban freight transport policy changes: Improving decision makers’ awareness via an agent-specific approach. Transport Policy, 36, 248–252. Gentile, G., & Vigo, D. (2006). A demand model for freight movements based on a tree classification of the eco-nomic activities applied to city logistic. City Good, Presentation in the 2nd roundtable, BESTUFS workshop TFH, Wildau. Ghiani, G., Laporte, G., & Musmanno, R. (2004). Introduction to logistics systems planning and control. Hoboken, NJ: Wiley, John & Sons, Inc. Gonzales-Calderon, C. A., & Holguín-Veras, J. (2019). Entropy-based freight tour synthesis and the role of traffic count sampling. Transportation Research: Part E: Logistics and Transportation Review, 121, 63–83. Gonzalez-Feliu, J., Ambrosini, C., Pluvinet, P., Toilier, L., & Routhier, J. L. (2012). A simulation framework for evaluating the impacts of urban goods transport in terms of road occupancy. Journal of Computational Science, 3(4), 206–215. Gonzalez-Feliu, J., & Peris-Pla, C. (2019). Impacts of retailing land use on both retailing deliveries and shopping trips: Modelling framework and decision support system. In IFAC-PapersOnLine 51-11. Amsterdam: Elsevier Ltd., 606–611. Hounwanou, S., Comi, A., Gonzalez-Feliu, J., & Gondran, N. (2018). From city center to urban periphery: Retail-store movement and shopping trip behaviours. Indications from Saint-Etienne. Transportation Research Procedia, 30, 363–372. https://doi​.org​/10​.1016​/j​.trpro​.2018​.09​.039 Kunytska, O., Comi, A., Danchuk, V., Vakulenko, K., & Yanishevskyi, S. (2021). Optimizing last mile delivering through the analysis of shoppers’ behaviour. In G. Ierpiński and E. Macioszek (Eds.), Decision Support Methods in Modern Transportation Systems and Networks. Lecture Notes in Networks and Systems, vol 208 (pp. 129–147). Cham: Springer. https://doi​.org​/10​.1007​/978​-3​- 030​71771​-1_9 Le Pira, M., Marcucci, E., Gatta, V., Inturri, G., Ignaccolo, M., & Pluchino, A. (2017). Integrating discrete choice models and agent-based models for ex-ante evaluation of stakeholder policy acceptability in urban freight transport. Research in Transportation Economics, 64, 13–25. Lohse, D. (2001). Verkehrsnachfragemodellierung mit n-linearen Gleichungssystemen. Proceedings of the 1. AMUS conference, Schriftenreihe Stadt Region Land, Heft 71, Institut für Stadtbauwesen RWTH Aachen. Marcucci, E., Gatta, V., Le Pira, M., Chao, T., & Li, S. (2021). Bricks or clicks? Consumer channel choice and its transport and environmental implications for the grocery market in Norway. Cities, 110, 103046. https://doi​.org​/10​.1016​/j​.cities​.2020​.103046 Müller, S., & Schneider, S. (2009). A methodology for deploying VISEVA-W/VISUM for large area goods transport modelling. Working Paper, Institute of Transport Research, Berlin-Adlershof, https://elib​.dlr​.de​/66323/ Nigro, M., Castiglione, M., Colasanti, F. M., De Vincentis, R., Valenti, G., Liberto, C., & Comi, A. (2022). Exploiting floating car data to derive the shifting potential to electric micromobility. Transportation Research Part A, 157, 78–93. https://doi​.org​/10​.1016​/j​.tra​.2022​.01​.008 Nuzzolo, A., & Comi, A. (2013). Tactical and operational city logistics: Freight vehicle flow modelling. In M. Ben-Akiva, H. Meersman & E. Van de Voorde (Eds.), Freight Transport Modelling (pp. 433– 451). Emerald Group Publishing Limited. https://doi​.org​/110​.1108​/9781781902868​- 021 Nuzzolo, A., & Comi, A. (2014). Urban freight demand forecasting: A mixed quantity/delivery/vehiclebased model. Transportation Research: Part E: Logistics and Transportation Review, 65, 84–98. https://doi​.org​/10​.1016​/j​.tre​.2013​.12​.014 Nuzzolo, A., Comi, A., & Papa, E. (2014). Simulating the effects of shopping attitudes on urban goods distribution. Procedia – Social and Behavioral Sciences, 111, 370–379. https://doi​.org​/10​.1016​/j​ .sbspro​.2014​.01​.070 Ogden, K. W. (1992). Urban goods movement. Hants: Ashgate. Russo, F., & Comi, A. (2010). A modelling system to simulate goods movements at an urban scale. Transportation, 37(6), 987–1009. https://doi​.org​/10​.1007​/s11116​- 010​-9276-y

Estimating and forecasting urban freight origin–destination flows  97

Russo, F., & Comi, A. (2012). The simulation of shopping trips at urban scale: Attraction macro-model. Procedia – Social and Behavioral Sciences, 39, 387–399. https://doi​.org​/10​.1016​/j​.sbspro​.2012​.03​.116 Russo, F., & Comi, A. (2016). Restocking in touristic and cbd areas: Deterministic and stochastic behaviour in the decision-making process. Transportation Research Procedia, 12, 53–65. https://doi​. org​/10​.1016​/j​.trpro​.2016​.02​.047 Russo, F., & Comi, A. (2020). Behavioural simulation of urban goods transport and logistics: The integrated choices of end consumers. Transportation Research Procedia, 46, 165–172. https://doi​ .org​/10​.1016​/j​.trpro​.2020​.03​.177 Russo, F., & Comi, A. (2021). Sustainable urban delivery: The learning process of path costs enhanced by information and communication technologies. Sustainability, 23, 13. https://doi​.org​/10​.3390​/ su132313103 Russo, F., & Musolino, G. (2012). A unifying modelling framework to simulate the Spatial Economic Transport Interaction process at urban and national scales. Journal of Transport Geography, 24, 189–197. Spielberg, F., & Smith, S. A. (1981). Service and supply trips at federal institutions in Washington, DC area. Transportation Research Record, 834, 15–20. Toilier, F., Gardrat, M., Routhier, J. L., & Bonnafous, A. (2018). Freight transport modelling in urban areas: The French case of the FRETURB model. Case Studies on Transport Policy, 6,(4), 753–764. Wang, Q., & Holguín-Veras, J. (2008). Tour-based entropy maximization formulations of urban commercial vehicle movements. Proceedings of the Association for European Transport (AET) Conference, Leeuwenhorst Conference Centre, The Netherlands, October 6–8, 2008. Wiese, A., Zielke, S., & Toporowski, W. (2015). Shopping travel behaviour. International Journal of Retail and Distribution Management, 43(4/5), 469–484. Wilson, A. G. (1970). Entropy in urban and regional modelling. London: Pion Limited.

5. Evaluating city logistics solutions with agent-based microsimulation Takanori Sakai, Peiyu Jing, André Romano Alho, Ravi Seshadri, and Moshe Ben-Akiva

INTRODUCTION Urban distribution systems have been evolving continuously over the past decades. This evolution has come with advances in technologies and innovation in logistics practices (e.g., delivery consolidation, parcel lockers, cargo cycles, and crowd-shipping). Furthermore, dynamics in supply chains and the growth in online shopping have been contributing to the transformation of urban distribution systems. Moreover, the COVID-19 pandemic has triggered significant changes in both goods flow in cities and shopping behavior of individuals. The need to achieve greater efficiency and sustainability is growing, as is the need to evaluate city logistics policies and innovative solutions. Agent-level behavior and the interactions between different agents are key elements in a comprehensive evaluation of policies and solutions. For example, a solution such as the adoption of a new mode for last-mile deliveries might affect the decisions and interactions of individual consumers/receivers, shippers, and carriers. It could also have impacts at the system level – on passenger and freight flows – and network performance which, in turn, more broadly influences individuals and businesses. The interaction between passenger and freight flows has been getting more relevant in the past few decades. For example, these interactions are critical for analyzing online shopping, parcel lockers, and on-demand vehicles serving both passengers and freight. Agent-based microsimulation models aim to take into account these complex interactions between agents and the transportation network, and thus are suitable to support stakeholders’ decision making (Le Pira et al., 2017). First, in this chapter, we characterize the class of agent-based freight simulation models, highlighting the differences from other model classes as already anticipated in the overview Chapter 3 by Tavasszy and De Bok. Next, we describe the framework of SimMobility, an agent-based simulation laboratory, with a focus on its urban freight model – SimMobility Freight – to illustrate an agent-based urban freight microsimulation model and its application. We detail how the simulator can consider agents’ interactions in various dimensions. Following this, we present a case study on the impacts of congestion pricing policies for a prototypical city in the US. Then, we discuss the limitations and future research needs in urban freight microsimulation. Lastly, we provide a brief conclusion at the end of this chapter.

AGENT-BASED MICROSIMULATION MODELS Microsimulation models are, in general, advantageous in terms of the range of applications and the analytical granularity while entailing greater computational burden and data 98

Evaluating city logistics solutions  99

requirements for calibration and implementation. The fundamental difference from the models described in the earlier chapters is that the model replicates the decision making of individual agents, considering the environment which each agent faces. Such a setting allows the model to be sensitive to various factors. Both individuals/households and companies/business establishments can be considered as agents in the models. These agents play roles as shippers/ producers, receivers/consumers, carriers, and/or drivers. Synthetic agents are the key input of a microsimulation model and are generated often based on publicly available statistics, while land-use and transportation network information are required like other classes such as freight trip generation models and origin–destination (O–D) flow models (see Chapter 4 by Comi and Delle Site and Chapter 6 by Sánchez-Díaz and Castrellon). One of the earliest applications of an agent-based framework in freight models is the GoodTrip model (Boerkamps et al., 2000). This framework has been further elaborated in more recent developments such as Roorda et al. (2010) and Cavalcante and Roorda (2013), MASS-GT (de Bok & Tavasszy, 2018), Polaris (Stinson et al., 2020), and SimMobility Freight (Sakai et al., 2020). Since early 2000, improvements have been made in various aspects including the commodity coverage (e.g., from only food retail supply chain to all, or almost all, types of goods flows), the coverage of decisions (e.g., carrier selection and parking location choice), and modeling and calibration techniques. Similar to urban freight O–D flow models, covered in Chapter 4 by Comi and Delle Site, e-commerce freight flows are also getting into the scope of agent-based microsimulations (Stinson et al., 2020; Sakai et al., 2022), adding to business-to-business (B2B) flows. The agent-based freight model is often characterized by a multi-layer structure and modeling techniques considering the logic of logistics decision making: Multi-layer structure. A model often includes multiple layers. Typically, each of supply chain (output: commodity flows), logistics/transportation planning (output: transportation operation plans), network usage (output: network demand and performance), and network management (output: supply of system capacity and service) is simulated in different but inter-connected layers. Sometimes, they are framed as markets. For example, the GoodTrip model has layers called commodity market, transport service market, traffic service market, and infrastructure market which correspond to the above four layers. It should be noted that some agent-based models focus only on a specific layer without considering the interactions with other layers. For example, a tour-based microsimulation system proposed by Hunt and Stefan (2007) does not explicitly consider commodity flows/shipments. In the multi-layer setting, the information flow is not necessarily one-directional. The layers can transfer information for not only demand (shipments, tours, and trips) but also network performance and logistics costs obtained from network traffic simulation. Logic of logistics decision making. To replicate the decision making of agents, the adoption of the logic of logistics decision making is essential. In the simulation, each agent makes decisions as a rational agent, minimizing the cost and/or maximizing the profit. Thus, the techniques on logistics and operations research, such as Economic Order Quantity (EOQ) model or vehicle loading and routing model, are used for freight simulation (although a fully fledged Vehicle Routing Problem often cannot be incorporated in city-scale freight simulation due to the computational burden). The consideration of logistics elements allows the simulation for connecting commodity flows and vehicle tours; the consistency between the two is maintained properly only by replicating the planning by stakeholders, which entails a disaggregate, agent-based framework.

100  Handbook on city logistics and urban freight

A comprehensive framework that covers major layers can generate the complete picture of shipments, vehicle operation schedules (set of tours), and network traffic and performance, maintaining their consistency. An analyst can identify vehicles which go through congested links in the network as well as shipments loaded in those vehicles with information about commodity types, size, shippers, and receivers.

SIMMOBILITY’S URBAN FREIGHT SIMULATOR This section introduces an agent-based urban freight simulator in SimMobility – SimMobility Freight. Among many proposed, agent-based, urban freight simulators which are conceptually or partially developed, SimMobility Freight is fully developed and calibrated for evaluating real-world logistics solutions. This section does not cover the details of the model specification. The detailed model specifications of business-to-business freight and e-commerce demand simulation models are available, respectively, in Sakai et al. (2020) and Sakai et al. (2022). Furthermore, the latest updates in the pre-day logistics model are available in Jing (2021). SimMobility Platform SimMobility is a demand-and-supply urban transportation simulation platform. The simulation in SimMobility is fully disaggregated; decisions are simulated and recorded at the agent level. The first generation of SimMobility focused on the simulation of passenger-centric mobilities (Adnan et al., 2016). It has been extended to simulate urban freight transportation and deliveries, inclusive of e-commerce (Sakai et  al., 2020; Sakai et  al., 2022). The latest version of the platform is comprehensive, jointly simulating passenger, business-to-business (B2B) and e-commerce freight flows. SimMobility simulates agents’ decisions in three temporal dimensions – long-term, midterm, and short-term (Figure 5.1). The long-term model simulates decisions, which typically involve time periods longer than a day (e.g., residential locations, business locations, and shipper contracts). The mid-term model simulates activities and deliveries at the day level (e.g., individuals’ activities and freight deliveries). Finally, the short-term model is a micro-simulator of transportation operations with high temporal resolution. The simulation time-step is typically a fraction of a second. In general, the platform applies a dynamic plan-action, and activity-based framework, which allows for evaluating both short- and long-lasting effects of disruptions and unexpected events on network performance. Furthermore, SimMobility is capable of simulating the decisions of supply agents, such as public transit and on-demand service operators (Oh et al., 2020b; Oh et al., 2020a; Alho et al., 2021). The framework of SimMobility allows for evaluating impacts of new modes and services, traffic management, last-mile solutions, post-pandemic scenarios, disruptions, land use, and infrastructure for both passengers and freight agents. SimMobility Freight SimMobility Freight is a part of SimMobility, consisting of models which simulate the decisions of freight agents, including the individuals engaged in online shopping. The major components of SimMobility Freight are shown in Figure 5.2. The long-term model includes

Evaluating city logistics solutions  101

Source:   Authors

Figure 5.1   SimMobility structure

Source:   Authors

Figure 5.2   Freight models

102  Handbook on city logistics and urban freight

establishment and fleet ownership synthesis, overnight parking selection, and shipment (or commodity flow) generation, covering both B2B and e-commerce shipments. The mid-term model handles the simulations of “pre-day” logistics planning and “within-day” vehicle operations and the mesoscopic traffic simulation. The “pre-day” simulator generates the planned schedule of freight vehicles on a given day. The “within-day” simulator functions jointly with the mesoscopic traffic simulator, replicating the demand–supply interactions in a transportation network. The shipment simulator in the long-term model The two distinct simulators in the long-term model – one is for the B2B shipments and the other is for e-commerce shipments (Figure 5.3) – are both summarized below. B2B shipments The main input of the B2B shipment simulation is the establishment population with details of location, employment size, floor area, function type (office, factory, store and restaurant, logistics facilities, others), and industry type (e.g., chemical manufacturing), as well as the network information (travel time, distance, and transportation cost between zones). The freight generation model simulates for each establishment, by commodity type, the annual production (total weight of outbound shipments per year) and consumption (total weight of inbound shipments per year), as well as outbound and inbound shipments. With the shipper contract

Source:   Authors

Figure 5.3  Shipment models (left: B2B; right: E-commerce)

Evaluating city logistics solutions  103

model, the total consumption of each establishment is allocated to contracts, and a shipper (supplier) is assigned to each of these contracts. The output – shipper contracts – defines the commodity flows: origin, destination, commodity type, and total size (weight). Lastly, the model for shipment size and frequency determines the shipment size for each contract, outputting B2B shipments. For an average day simulation, shipments are randomly selected based on the shipment frequency. E-commerce shipments The e-commerce simulation’s main input is the household population as well as delivery option characteristics (delivery fee, speed, and time slot). “Delivery option” includes both home delivery and pickup options for which an individual needs to travel to a pick-up point. A household-based demand model simulates e-commerce adoption, monthly e-commerce expenditure, value for each e-commerce order, and delivery option choice for each item category (groceries, household goods, and others (e.g., clothing, books, and electric appliances)). Each home delivery order is converted to packages. Following on, a distribution facility is assigned for fulfilling each package delivery, which generates e-commerce shipments. The delivery demand for a distribution facility is, in turn, used to determine the inbound shipments (i.e., consumption) of the facility. Predicted delivery orders and pickups are used as inputs for the passenger activity simulation. This allows SimMobility to capture the substitution and/or complementary effects of e-commerce on passenger travel. Like B2B shipments, e-commerce shipments to be fulfilled on an average day are randomly selected based on the frequency. The mid-term model The main inputs of the mid-term model (Figure 5.4) are the combined B2B and e-commerce shipments, vehicle fleets, and the network information. Pre-day logistics planning: The pre-day logistics planning model first simulates the carrier selection process for each shipper. This selection process considers the fleet size of carriers, the distances between shipper and carrier establishments, and type of shipments handled (Courier, Express, and Parcel (CEP) services or not). A shipper might select more than one carrier, in case the total size of shipments is large, and a single carrier cannot handle it. Once shipments are assigned to carriers, the vehicle operations planning model assigns shipments, one by one, to vehicles and schedules pickups and deliveries. In the simulation, multiple alternatives of “vehicle operations plans” are generated and the best alternative with respect to the total cost is chosen. The output is vehicle tours, which consist of at least one pickup and one delivery trip. Within-day vehicle operations: The models in within-day simulator are used when the vehicle tours are assigned to the network. The route choice model simulates the driver’s route for each trip considering the network performance at the time of trip. Delivery/pickup parking choice model, on the other hand, simulates selection of parking type (loading/ unloading bay, public car park, or on-street parking) considering the availability and the congestion in parking lots. The supply model – a mesoscopic traffic simulator – simulates traffic in the network and compute the network performance. There is a feedback mechanism – Day-to-Day Learning – allowing the use of network performance in the following

104  Handbook on city logistics and urban freight

Source:   Authors

Figure 5.4  The mid-term model simulation of pre-day logistics planning (and even the long-term simulator), affecting shipments and vehicle operations plans. Applications SimMobility Freight has been used for evaluating a series of urban freight policies and solutions, including: ● ● ● ● ●

Overnight freight vehicle parking policy (Gopalakrishnan et al., 2020) Freight consolidation centers (Mepparambath et al., 2021) Night/Off-peak deliveries (Sakai et al., 2020) Freight-on-Demand (Alho et al., 2021) Congestion pricing (Jing, 2021)

In the next section, we present the impact analysis of congestion pricing policies for demonstrating the capability of the simulator.

Evaluating city logistics solutions  105

DEMONSTRATION: EVALUATION OF CONGESTION PRICING IMPACTS Overview Congestion pricing can reshape travel demand by impacting a series of pre-day and withinday decisions. This type of initiative is described in Chapter 2 by Holguin-Veras et al. in this Handbook. For passengers, activities, and trips, travel mode, destination, time of day, and route choices are influenced by pricing. For freight, congestion pricing reshapes shipments and goods vehicle flows through the impacts on shipment size and frequency, vehicle operations planning, and route choice. Furthermore, the impacts of congestion pricing are usually not distributed evenly across population segments. Congestion pricing directly increases travel costs of the agents who travel within or enter the toll zone at specific time periods. The increase in cost motivates them to modify their transportation-related decisions and, in turn, changes the network conditions. The study area in this demonstration is a synthetic prototype city classified as AutoInnovative and detailed in Oke et al. (2019). This type of city is featured by modernization and industrialization, high auto-dependency, and high transit mode share, as well as high metro and population density. It mainly represents North American cities (e.g., Boston, Washington D.C., Chicago, Toronto). The prototype city is synthetized based on population, land use, demand, and supply characteristics of the Greater Boston Area in the U.S. Major statistics of the city are summarized below. ● ● ● ● ●

Zoning: 164 municipalities, 2,727 traffic analysis zones (TAZs) Area: 7.32 thousand km2 Population: 1.74 million households, 4.60 million residents Business establishments: 130 thousand Vehicles: 2.47 million passenger vehicles, 0.378 million goods vehicles

First, the base (do-nothing) scenario is simulated to identify the temporal and spatial patterns of congestion and quantify proper toll rates based on principles of marginal cost pricing (Lentzakis, Seshadri, Akkinepally, Vu, & Ben-Akiva, 2020). This base scenario represents an average day with the average level of passenger and freight travel demand. Furthermore, three congestion pricing scenarios are simulated. The common features of the tested pricing schemes are as follows: ●



Vehicle-type-specific: Different toll rates are applied in accordance with the passengercar-unit. Based on the maximum laden weight, goods vehicles are classified as light goods vehicles (LGV) (no more than 3.5 ton), heavy goods vehicles (HGV) (3.5–16 ton), and very heavy goods vehicles (VHGV) (more than 16 ton), for which the PCU is 1.5, 2, and 2.5, respectively. Time-period-specific: Targeting to mitigate congestion during the observed traffic peak periods, tolls are applied during the AM peak (8:00-10:00) and the PM peak (16:0019:00). Tolls are also applied 30 minutes before and after the peak periods with lower rates, as ramp up/down periods to avoid abrupt changes in demand.

106  Handbook on city logistics and urban freight ●

Location-specific: Tolls are applied for a specific zone of the city which is the shaded zone in Figure 5.5. The toll zone is designed to cover most of the highly congested TAZs and has a high population density. It spans over 791 km2 (10.8% of total) and has 1.85 million residents (40.2% of total).

The three scenarios are different with respect to how tolls are charged. The first scenario is distance-based tolling. The toll charge is linear to the distance traveled within the toll zone and peak periods with an upper limit for each vehicle. The second scenario is cordon-based tolling, in which vehicles are charged to enter the toll zone during peak periods through radial road links. The third scenario is area-based tolling, wherein any vehicle crossing through or traveling within the toll area is charged a flat toll per day for unlimited travel. The toll rates for different vehicle types and time to enter the zone are shown in Table 5.1. Model Sensitivities to Tolling Policies Figure 5.6 illustrates how the model system in SimMobility Freight considers the impacts of tolling policies. In this analysis, the commodity flows between shippers and receivers are fixed while shipment size and frequency are sensitive to changes in total logistics costs and e-commerce demand is sensitive to delivery fees. At the pre-day level, carriers take into account freight transportation costs, including distance-dependent, time-dependent, and toll cost, in vehicle operations planning, aiming to minimize the total cost. At the within-day level, drivers consider toll rate in their route choice for each trip. The use of the feedback loop allows the agents to receive updated information about the transportation network for their decision-making and update decisions. The simulations are repeated until the mode-specific travel time, waiting time, and costs reach convergence, and thus, the simulation represents an average day under equilibrium. Results and Discussions In this section, we compare the results of the base scenario (Base) and the following three scenarios in terms of shipment size, freight vehicle load factor, freight transportation cost, vehicle kilometers traveled (passenger and freight), and social welfare. ● ● ●

Distance-based pricing scenario (Distance) Cordon-based pricing scenario (Cordon) Area-based pricing scenario (Area)

For comparability, some results are aggregated based on the departure time and OD pairs. External trips have both origin and destination outside of the toll zone, Internal trips have both origin and destination within the toll zone, and Connection trips have either origin or destination within the toll zone. In the following tables, the percentages in parenthesis show the changes from Base. Shipment size: Table 5.2 shows the average shipment size (weight). In congestion pricing scenarios, the shippers increase the shipment size to varying degrees. This, in turn, leads to less shipment frequency, contributing to the reduction in goods vehicle trips. In Distance and

Evaluating city logistics solutions  107

Source:   Authors

Figure 5.5  The prototype city and toll zone Notes: Links: freeways; Polygons: traffic analysis zones; Shaded area: toll zone

108  Handbook on city logistics and urban freight

Table 5.1  Toll rates of congestion pricing policies Passenger vehicle

LGV

HGV

VHGV

Distance-based

Per unit distance toll rate (USD/km)

0.14–0.32

0.20–0.48

0.27–0.64

0.34–0.80

Cordon-based

Toll charge (USD/entry)

2.10–5.80

3.15–8.70

4.20–11.60

5.25–14.50

Area-based

Toll charge (USD/day)

2.65

4.00

5.50

6.60

Note:   Maximum toll per vehicle-day in distance-based tolling is 10 USD (passenger vehicle), 15 USD (LGV), 20 USD (HGV) and 15 USD (VHGV)

Figure 5.6  Model mechanism for evaluating tolling impacts on freight Area, the sizes of internal shipments increase more than those of connection shipments, while, in Cordon, it is the opposite. Freight vehicle load factor: Table 5.3 shows the average load factor. Note that tour type is defined based on the OD of the first trip. Only Distance shows significant improvement in load factor of internal and connection tours (2–6%). The table also shows the result for a subset of large carriers with large fleets and numbers of shipments to carry (>100 shipments). For them, the improvement in load factor is much more significant compared to other carriers in pricing scenarios (in Distance, 6–12% for internal tours and 4–7% for connection tours). However, for internal tours in Cordon and Area, the load factors decrease since these tours have lower

Evaluating city logistics solutions  109

Table 5.2  Average shipment size (unit: kg) Type of shipment

Internal shipments

Connection shipments

Base

Distance

Cordon

Area

Base

Distance

Cordon

Area

All

149.3

153.9 (+3.1%)

151.0 (+1.1%)

154.2 (+3.3%)

154.3

156.4 (+1.4%)

156.3 (+1.3%)

158.8 (+2.9%)

Non-e-commerce

287.2

293.3 (+2.1%)

290.8 (+1.2%)

295.5 (+2.9%)

312.5

318.3 (+1.9%)

317.8 (+1.7%)

320.4 (+2.5%)

E-commerce

14.1

14.7 (+4.3%)

14.5 (+2.9%)

14.8 (+4.7%)

15.6

15.9 (+1.9%)

15.9 (+1.9%)

16.1 (+3.2%)

Table 5.3  Freight vehicle average load factor Internal tours

Connection tours

Base

Distance

Cordon

Area

Base

Distance

Cordon

Area

 LGV

51.2%

52.7% (+2.8%)

51.0% (–0.4%)

51.1% (–0.2%)

53.0%

54.0% (+1.9%)

53.5% (+1.0%)

53.2% (+0.3%)

 HGV

49.3%

51.7% (+4.9%)

49.0% (–0.6%)

49.0% (–0.6%)

48.7%

49.7% (+2.0%)

49.4% (+1.4%)

49.0% (+0.7%)

 VHGV

31.6%

33.4% (+5.9%)

31.5% (–0.3%)

31.1% (–1.6%)

35.8%

36.7% (+2.5%)

36.6% (+2.1%)

36.1% (+0.9%)

 LGV

71.7%

75.9% (+5.8%)

70.9% (–1.2%)

71.3% (–0.6%)

75.3%

78.7% (+4.4%)

77.3% (+2.7%)

76.2% (+1.1%)

 HGV

64.9%

69.4% (+6.8%)

63.9% (–1.6%)

64.4% (–0.9%)

65.1%

69.2% (+6.4%)

67.4% (+3.6%)

67.0% (+3.0%)

 VHGV

46.3%

51.7% (+11.7%)

45.4% (–1.8%)

46.0% (–0.6%)

54.3%

58.2% (+7.2%)

57.0% (+5.1%)

56.0% (+3.3%)

All carriers

Large carriers

costs (no toll collected) but bigger travel time savings. Cordon and Area improve load factors of connection tours with smaller magnitudes than Distance. Freight transportation costs: Three components of freight transportation costs are considered: toll, time-dependent (driver wages), and distance-dependent (fuel, tire, vehicle operations costs) component. Table 5.4 shows the per-ton-km freight transport cost wherein all trips (tolled or not) are considered. For all classifications, the time component decreases under congestion pricing mainly due to less congestion and higher speed. The distance component is not significantly affected. For internal trips, in Distance and Area, the per-ton-km transportation cost increases but, in Cordon, slightly decreases. As for connection trips, the per-ton-km cost increases under all congestion pricing scenarios. Vehicle kilometers traveled (VKT) (passenger and freight): Table 5.5 shows the total VKT of passenger cars, public transit, and goods vehicles classified. With congestion pricing,

110  Handbook on city logistics and urban freight

Table 5.4  Freight transport costs (unit: USD per ton-km)

Internal trips (n=0.34 mil.)

Connection trips (n=0.19 mil.)

External Trips (n=0.47 mil.)

Base

Distance

Cordon

Area

Total cost

8.88

9.51 (+7.1%)

8.77 (–1.3%)

9.55 (+7.5%)

Toll component

0.00

0.91

0.03

0.87

Time component

6.23

5.98 (–4.2%)

6.10 (–2.1%)

6.04 (–3.0%)

Distance component

2.65

2.62 (–1.1%)

2.64 (–0.4%)

2.64 (–0.4%)

Total cost

8.79

9.09 (+1.0%)

8.88 (+1.0%)

9.15 (+4.1%)

Toll component

0.00

0.56

0.17

0.51

Time component

6.16

5.93 (–3.9%)

6.06 (–1.7%)

5.98 (–2.9%)

Distance component

2.63

2.60 (–1.1%)

2.65 (+0.7%)

2.66 (+1.0%)

Total cost

8.63

8.67 (+0.5%)

8.65 (+0.2%)

8.59 (+0.5%)

Toll component

0.00

0.07

0.02

0.00

Time component

5.99

5.94 (–0.8%)

5.96 (–0.5%)

5.94 (–0.8%)

Distance component

2.64

2.66 (+0.8%)

2.67 (+1.1%)

2.65 (+0.4%)

Table 5.5  Vehicle kilometers traveled (in million, all day)

Internal Trips

Connection Trips

Cars

Base

Distance

Cordon

Area

27.03

24.33 (–10.0%)

27.43 (+1.5%)

25.68 (–5.0%)

Public transit

0.90

1.00 (+11.8%)

0.89 (–0.9%)

0.95 (+6.1%)

LGV

1.46

1.33 (–9.0%)

1.45 (–0.9%)

1.37 (–6.5%)

HGV

0.16

0.14 (–11.2%)

0.15 (–0.4%)

0.14 (–7.9%)

VHGV

0.08

0.07 (–10.6%)

0.08 (–2.1%)

0.08 (–5.5%)

Cars

50.43

47.80 (–5.2%)

50.10 (–0.7%)

49.11 (–2.6%)

Public transit

1.64

1.73 (+5.2%)

1.69 (+3.0%)

1.70 (+3.6%)

LGV

2.49

2.36 (–5.5%)

2.40 (–3.8%)

2.42 (–3.0%)

HGV

0.40

0.37 (–6.0%)

0.38 (–3.9%)

0.39 (–3.0%)

VHGV

0.45

0.42 (–6.9%)

0.44 (–2.6%)

0.43 (–2.9%)

External

Cars

78.01

78.29 (+0.4%)

78.19 (+0.2%)

78.23 (+0.3%)

Trips

Public transit

2.39

2.39 (+0.0%)

2.39 (–0.0%)

2.39 (+0.1%)

LGV

3.25

3.24 (–0.1%)

3.24 (–0.2%)

3.24 (–0.2%)

HGV

0.65

0.65 (+0.3%)

0.65 (+0.2%)

0.65 (+0.2%)

VHGV

0.74

0.76 (+2.6%)

0.75 (+0.9%)

0.75 (+1.7%)

the total number of car and goods trips decreases for the entire city. Among three types of goods vehicles, the VKT decrease of LGV is lower in relative magnitude compared with HGV and VHGV. Note that this city has a very large share of LGV trips (91%) whereas HGV trips (6%) and VHGV trips (3%) only consist of a small share, so the goods vehicle VKT reduction is greatly attributed to the reduction in VKT by LGV.

Evaluating city logistics solutions  111

Table 5.6  Change in social welfare compared with Base in an average weekday Category

Item (mil. USD/weekday)

Distance

Cordon

Area

Change in producer surplus

Toll earning

+5.50

+1.14

+5.27

Public transport revenue

+0.0545

+0.0124

+0.0586

Fuel tax revenue

–0.0864

–0.0152

–0.0872

Passengers

–3.44

–1.01

–3.73

Freight shippers/carriers

–0.167

–0.0529

–0.228

Change in consumer surplus

Table 5.7  Change in consumer welfare compared to Base in an average weekday Numerator

Unit

Distance

Cordon

Area

Change in daily passenger surplus

USD/passenger

–0.836

–0.246

–0.926

USD/ trip

–0.237

–0.208

–0.257

USD/person-km

–0.183

–0.160

–0.201

USD/shipper

–3.85

–1.22

–5.26

USD/shipment

–0.115

–0.0878

–0.134

USD/veh-tour

–0.486

–0.351

–0.537

USD/veh-km

–0.257

–0.185

–0.289

Change in daily freight shipper/carrier surplus

Social welfare: Table 5.6 summarizes the change in total social welfare relative to Base. Congestion pricing reduces the use of private vehicles and encourages public transit ridership, thus improving public transportation revenue and reducing fuel tax revenue. The magnitude of the change of these two revenues is much smaller than that of toll earning, indicating that the producer agency profits from congestion pricing. On the consumer side, all pricing scenarios reduce the passenger surplus and freight shipper/carrier surplus, mainly as a result of increased travel costs. Although they benefit from reduced travel time, it cannot compensate for the increased travel cost. Table 5.7 shows the change in consumer welfare of both passenger and freight per individual, trip/tour, or VKT on an average weekday. Overall, all three congestion pricing schemes improve total social welfare, logistics operations efficiency, and reduce private car and goods vehicle VKT compared against Base. Although travel-time-dependent cost decreases, the surplus of shippers still decreases as the magnitude of toll costs is larger. Distance outperforms Cordon and Area. It is the most effective in improving logistics efficiency and total social welfare as well as reducing peak-period congestion and VKT. Furthermore, it is more efficient since by design, it properly internalizes congestion costs and penalizes the heavy users for the congestion they cause.

LIMITATIONS AND FUTURE RESEARCH DIRECTIONS IN AGENTBASED MICROSIMULATION We discussed and demonstrated the capability of an agent-based microsimulation model. The simulator, for example, SimMobility, can be used for analyzing a variety of policies and

112  Handbook on city logistics and urban freight

solutions; however, there are still several key interactions and sensitivities to be enhanced and/ or incorporated in the future. Below, we discuss the key subjects where existing limitations hamper evaluating a set of novel trends, policies, and solutions in the class of agent-based urban freight microsimulation models. Logistics network: The supply side simulation is not suitable for some purposes, for example, to answer questions such as “what happens if some fulfillment centers of some vendors cannot be used any more due to a disruption?” Specifically, the logistics network is typically not well represented in the urban-level simulators and that is also the case for most microsimulators. In SimMobility Freight, the definition of shipment used is “a quantity of goods shipped together from one location to another without transshipment”, which means the inbound and outbound shipments are not explicitly connected. Thus, fulfillment and sorting facilities, where items are put together and turned to a package, and cross-docking facilities, where packages are transshipped, are not explicitly distinguished. To consider transshipments under the scope of agent decisions properly, the simulator should have high-level agents, which coordinate the decisions of individual “establishment” agents. These agents should be able to decide locations of the facilities under their management and delivery channels in addition to productions, consumptions, and shipment sizes. In the real world, the roles of, and the relationship between, different companies and establishments are intricate. Thus, the key challenges lie in the development of the theoretical framework which can describe the heterogeneous scopes of business decisions as well as the preparation of the data to inform the model. Delivery mode choice: The increase in parcel deliveries, technological innovations, and the growing concerns about environmental impacts, motivate the deployment of unconventional delivery modes such as crowd-shipping, cargo bikes, delivery robots, and drones (see Chapter 19 by Jaller et al., Chapter 21 by Gatta et al., and Chapter 22 by Paddeu et al. in this Handbook). For example, SimMobility Freight has been used to evaluate the potential impacts of the deployment of novel delivery modes, assuming the adoption of the modes. These modes are suitable for some types of deliveries but not for others (e.g., cargo bikes are typically used in high-density urban centers). To predict the city-wide impacts of the penetration of novel delivery modes, the model needs to consider how each carrier plans deliveries using not only conventional goods vehicles but the combination of alternative modes. This indicates room for improvement in models that connect shipments to freight mode choices. Passenger-freight activity interactions: The interactions between passenger and freight activities are more relevant as increasingly more last-mile deliveries are destined directly to individual consumers, parcel lockers, or other pickup points. The relationship between online and on-site shopping is not straightforward and is still being investigated by many researchers. However, the significant increase in online shopping in many cities around the world during the 2020 COVID-19 pandemic, which decreases shopping trips, highlights the importance of simulating the substitution/complementarity effects between online and on-site shopping. Such substitution is also connected to the above discussion on logistics networks since a delivery channel for home deliveries is different from that for retail stores. Furthermore, the recent trend on the supply side also makes this subject important. Crowd-shipping is becoming more prevalent and traditionally passenger-centric modes of transportation, such as taxis, are used for deliveries of parcels and groceries. The model needs to make the situations of passenger activities and freight activities consistently, considering the sharing of types of fleets. An initial demonstration of this connected SimMobility Freight with Mobility-On-Demand, as Freight-On-Demand (Alho et al., 2021).

Evaluating city logistics solutions  113

CONCLUSION In this section, we described the class of agent-based urban freight microsimulation models, detailing SimMobility Freight as an example of this class. A key advantage of microsimulations is the capability of analyzing a diverse set of policies and solutions at a high resolution, which is critically important to evaluate novel policies and solutions for sustainable urban freight transportation. For example, SimMobility can identify carriers, vehicles, shipments, shippers, and receivers which are directly and indirectly affected by a congestion pricing policy imposed in specific areas and duration (Jing, 2021). Also, the simulator’s capability has been demonstrated in past and ongoing studies focusing on policies and solutions, which are not limited to be freight centric (Gopalakrishnan et al., 2019; Alho et al., 2021; Sakai et al., 2020a; Mepparambath, Cheah, Zegras, Alho, & Sakai, 2021). Nevertheless, there are identified gaps to be addressed in the agent-based microsimulation models to make them available for tackling emerging trends and policy issues.

REFERENCES Adnan, M., Pereira, F. C., Azevedo, C. M. L., Basak, K., Lovric, M., Feliu, S. R., … Ben-Akiva, M. (2016). SimMobility: A multi-scale integrated agent-based simulation platform. In 95th Annual Meeting of the Transportation Research Board, Washington, DC. Alho, A. R., Sakai, T., Oh, S., Cheng, C., Seshadri, R., Chong, W. H., … Ben-Akiva, M. (2021). A simulation-based evaluation of a cargo-hitching service for e-commerce using mobility-on-demand vehicles. Future Transportation, 1(3), 639–656. Boerkamps, J., van Binsbergen, A., & Bovy, P. (2000). Modeling behavioral aspects of urban freight movement in supply chains. Transportation Research Record: Journal of the Transportation Research Board, 1725(1), 17–25. Cavalcante, R., & Roorda, M. J. (2013). Freight market interactions simulation (FREMIS): An agentbased modelling framework. Procedia Computer Science, 19, 867–873. de Bok, M., & Tavasszy, L. (2018). An empirical agent-based simulation system for urban goods transport (MASS-GT). Procedia Computer Science, 130, 126–133. Gopalakrishnan, R., Alho, A. R., Sakai, T., Hara, Y., Cheah, L., & Ben-Akiva, M. (2020). Assessing overnight parking infrastructure policies for commercial vehicles in cities using agent-based simulation. Sustainability, 12(7), 2673. Hunt, J. D., & Stefan, K. J. (2007). Tour-based microsimulation of urban commercial movements. Transportation Research: Part B, Methodological, 41(9), 981–1013. Jing, P. (2021). Design and evaluation of urban congestion pricing policies with microsimulation of passenger and freight (Doctoral Dissertation). Massachusetts Institute of Technology. Lentzakis, A. F., Seshadri, R., Akkinepally, A., Vu, V. A., & Ben-Akiva, M. (2020). Hierarchical density-based clustering methods for tolling zone definition and their impact on distance-based toll optimization. Transportation Research Part C: Emerging Technologies, 118, 102685. Le Pira, M., Marcucci, E., Gatta, V., Inturri, G., Ignaccolo, M., & Pluchino, A. (2017). Integrating discrete choice models and agent-based models for ex-ante evaluation of stakeholder policy acceptability in urban freight transport. Research in Transportation Economics, 64, 13–25. Mepparambath, R. M., Cheah, L., Zegras, P. C., Alho, A. R., & Sakai, T. (2021). Evaluating the impact of an urban consolidation centre and off-hour deliveries on freight flows to a retail district using agent-based simulation. 100th Annual Meeting of the Transportation Research Board, Washington, DC. (No. TRBAM-21-00992). Oh, S., Seshadri, R., Azevedo, C. L., Kumar, N., Basak, K., & Ben-Akiva, M. (2020a). Assessing the impacts of automated mobility-on-demand through agent-based simulation: A study of Singapore. Transportation Research, Part A: Policy and Practice, 138, 367–388.

114  Handbook on city logistics and urban freight

Oh, S., Seshadri, R., Le, D. T., Zegras, P. C., & Ben-Akiva, M. E. (2020b). Evaluating automated demand responsive transit using microsimulation. IEEE Access, 8, 82551–82561. Oke, J. B., Aboutaleb, Y. M., Akkinepally, A., Azevedo, C. L., Han, Y., Zegras, P. C., … Ben-Akiva, M. E. (2019). A novel global urban typology framework for sustainable mobility futures. Environmental Research Letters, 14(9), 095006. Roorda, M. J., Cavalcante, R., McCabe, S., & Kwan, H. (2010). A conceptual framework for agent-based modelling of logistics services. Transportation Research: Part E: Logistics and Transportation Review, 46(1), 18–31. Sakai, T., Alho, A. R., Bhavathrathan, B. K., Dalla Chiara, G., Gopalakrishnan, R., Jing, P., … BenAkiva, M. (2020). SimMobility freight: An agent-based urban freight simulator for evaluating logistics solutions. Transportation Research: Part E: Logistics and Transportation Review, 141, 102017. Sakai, T., Hara, Y., Seshadri, R., Alho, A., Hasnine, M. S., Jing, P., … Ben-Akiva, M. (2022). Householdbased E-commerce demand modeling for an agent-based urban transportation simulation platform.  Transportation Planning and Technology, 45(2), 179–201. Stinson, M., Auld, J., & Mohammadian, A. K. (2020). A large-scale, agent-based simulation of metropolitan freight movements with passenger and freight market interactions. Procedia Computer Science, 170, 771–778.

6. Freight trip generation models: using establishment data to understand the origin of urban freight traffic Ivan Sánchez-Díaz and Juan Pablo Castrellon

INTRODUCTION Cities are complex and diverse systems where most current human interactions take place, involving cultural, social, economic and environmental aspects. Understanding city dynamics is a recurrent starting point for researchers and practitioners with the aim of building better cities in which to live, that generally relies on the quantification of patterns of how people, infrastructure and economic activity are organized and interrelated (Youn et al., 2016). Urban freight transport is an important activity that allows decision makers to get a feel for the economic development of cities, regions and countries (Wang et al., 2020). Understanding freight generation as a consequence of economic activities is fundamental for supporting demand-driven infrastructure planning and effective freight policy design. Although several stakeholders have recognized the importance of this analysis from a quantitative perspective, there is a lack of generally accepted methodologies to estimate urban freight patterns under efficiency, accuracy and replicability criteria (Errampalli et al., 2021; Janjevic et al., 2019). Most of the sources used to implement local authorities’ initiatives are based on perceptions or case-based data, which are crucial for understanding and assessing pilot-scale trials but which become insufficient for large-scale implementation. Ironically, in the upsurge of the fourth industrial revolution, one of the main challenges for urban freight management continues to be the lack of suitable data and quantitative analytic techniques to obtain insights from them, the so-called Big-No-Data paradox (Gonzalez-Feliu, 2019). Local authorities in charge of urban freight management initiatives often rely on certain companies sharing their data or on traffic counts collected on-demand because there is no clear data collection and analysis strategy (NCFRP Report 25, 2013). Urban freight transport studies imply collecting, analyzing and predicting data about freight flows, fleet capacities and frequency of deliveries, among other features of establishments’ dynamics inside the city. These estimations are essential for: (i) improving the urban system knowledge, (ii) facilitating the formulation of suitable initiatives and (iii) enhancing the public sector’s decision-making processes. Capturing and using these data for decision purposes are challenging tasks that have attracted the attention of researchers and private and public stakeholders and resulted in the development of several techniques and technological adoptions. From the transport engineering discipline, the conventional four-step transport modelling system has been a widely used tool to cope with this challenge by estimating and forecasting travel demand at different geographical levels based on statistical and econometric techniques. These steps refer to trip generation, trip distribution, modal split and trip assignment. This chapter deploys theoretical and practical implications for the first step through the Freight Trip 115

116  Handbook on city logistics and urban freight

Generation (FTG) modelling approach. Chapter 4 by Comi and Delle Site of this Handbook proposes an integrated framework for the second step. Details on the third and fourth steps are covered by Ortúzar and Willumsen (2011). FTG modelling has given a feasible framework by which to estimate the total number of freight vehicles arriving at establishments (Ogden, 1992). It is a tool for characterizing zones in an urban setting by the frequency of deliveries. FTG considers the type of industrial, commercial or service activity, size of the establishment and the number of employees, among other business features (Ortúzar & Willumsen, 2011). Establishments’ data connect freight traffic analysis and infrastructure planning to economic activities. By estimating FTG, it is possible to assess traffic impacts of different activities, to extrapolate results based on statistical principles and to aggregate urban freight systems at distinct levels of analysis. This procedure is useful for dimensioning new transport infrastructure, accommodating demand to existing facilities (e.g., freight parking) and considering the environment with which freight traffic interacts (e.g., pedestrians, passenger cars, air pollution) (Alho & de Abreu e Silva, 2017). Inferential statistical models are the base for linking establishments’ activities and attributes to freight generation patterns. Explanatory variables, such as employment, site area, land use and commodity type, among others, can be used to estimate freight demand (weight, vehicle trips) through several modelling techniques available in the FTG literature. Academics and practitioners need to have clear foundations about the implications of each modelling methodology to achieve accurate estimations on trip generation based on case study objectives and proposed use. According to Holguín-Veras et al. (2013a), FTG accuracy strongly depends on methodological decisions such as data collection quality and representativeness. Accuracy also depends on the selected classification system to group establishments in aggregated analysis, explanatory variables choice, statistical validity of the techniques and appropriateness of aggregation procedures. This chapter gives an overview of each of these decisions based on recent contributions to the body of knowledge and links those theoretical implications to practice and contexts particularities. The FTG application field is vast and varied. It covers from logistics facilities design and applications at the building level to applications at zone, city, regional or even national level. This chapter uses the city of Stockholm as a case study. It also establishes linkages among FTG models and other modelling techniques e.g., demand models, addressed in Chapter 4 by Comi and Delle Site, agent-based models, and microsimulation, discussed in detail in Chapter 5 by Sakai et al. of this Handbook. Through the content of this chapter, authors provide a detailed explanation of theoretical and practical considerations about FTG models. The sections are organized as follows: the second section deploys FTG research taxonomy, the third section shows a case study for the city of Stockholm, the fourth section mentions current limitations and future directions and the fifth section concludes the chapter.

CURRENT STATE OF KNOWLEDGE Ogden (1992) defined urban freight demand modelling as the intention of urban planners and modellers to analyze how needs, desires and fashions of an urban community can impact freight movements. In this intention, there are delivery-, vehicle- and commodity-based models

Freight trip generation models  117

(Comi et al., 2012). The latter two categories focused attention on traditional transport problems related to the estimation of the quantity of trucks trips and the magnitude of goods movements, respectively. Delivery-based models capture the logic of logistics operations related to the tour nature of freight pick-ups and deliveries. Within the delivery-based category, FTG models allow estimation of freight traffic based on the analysis of supplier–receiver logistics decisions regarding quantities, frequencies, locations and vehicle sizes. For a deeper understanding of the advantages and limitations of each modelling category, readers can refer to Chapter 4 by Comi and Delle Site in this Handbook. Holguín-Veras et al. (2012) defined the set of factors that should be considered in developing FTG methods and models, which are determinants of the accuracy and relevance of the models for the studied context. Among the factors to be considered, they included dependent and independent variables, levels of aggregation and geography, estimation techniques, data collection techniques, and model validation and transferability. Several approaches to FTG modelling have shaped this knowledge field from theoretical foundations to practical implementations. Figure 6.1 shows a taxonomy of FTG that will be the guide to analyze the implications and scope that research contributions have made in specific applications, purposes, economic sectors and contexts. Dependent and Independent Variables Dependent variables are the FTG models’ outcome. They give the values needed for planning and policy decision purposes since they allow traffic impact quantification. The number of vehicle trips during a specific period is the most common outcome in FTG models. It represents logistics decisions from establishments in terms of freight transport needs according to the business ordering policy. The common practice is to consider this output as a continuous

Figure 6.1  FTG research taxonomy

118  Handbook on city logistics and urban freight

variable that is estimated from linear relationships with its parameters i.e., independent variables (Sánchez-Díaz, 2020). Freight weight and value are other possible outcomes that are more often produced by freight generation (FG) models, especially those with a regional focus rather than urban communities. Outcomes from FTG models feed distribution models explained in Chapter 4 by Comi and Delle Site as freight attraction parameters related to the freight required to satisfy endconsumer demand. For instance, FTG models provide estimates for the parameter ND.d ëé k ûù of delivery-based models that refers to the average number of deliveries required in zone d in outlet type k to satisfy demand. In terms of the independent variables, one of the main concerns of researchers when building FTG models is the choice of the significant variables that better explain freight demand in the study areas. Willumsen et al. (2006) listed some of the most important factors that influence freight movements, i.e., locational factors, the range of products needed and produced, characteristics and nature of the products, size of the firms, geographical factors, seasonal variations and pricing factors. Several contributions have validated the importance and significance of each of these factors with the use of statistical, econometric and machine-learning techniques. In terms of economic factors, Holguín-Veras et al. (2011) analyzed the performance of linear models based on employment as the sole independent variable to explain truck trips made per day and deliveries received per day. The empirical data taken from the study area – New York – showed low significance of the employment factor alone. However, they found that the combined effect of employment and industry sector grouping by specific categories performed better. Alho and de Abreu e Silva (2017) expanded the analysis by adding establishment area into the estimation of weekly trips demand. Though they found these factors were statistically significant, they realized that the magnitude of the effect depended on the model type used for linking explanatory to the decision variable. Recent evidence has shown that employment-based models outperform area-based ones in terms of freight movement prediction accuracy (Pani et al., 2020). Employment, establishment area and industry sector are the most common factors used in the literature, due to data availability. Nonetheless, other approaches suggested including more factors that are also significant in predicting and explaining freight generation. Commodity type, number of customers, total sales per year, type of facility are some of the factors proposed in Bastida and Holguín-Veras (2009). Pani et al. (2018) suggested the use of multiplevariable FTG models rather than single-variable ones, subject to data availability. They added the analysis of ‘year in business’ as a factor for explaining freight activities to the existing studies. Regarding geographical factors, Sánchez-Díaz et al. (2016) pointed out the need for locational approaches for FTG models and proposed the inclusion of land market value, district classification and geographic location as explanatory land-use variables. This paper also included network characteristic variables, such as distance to truck routes, distance to the primary network, minimum and mean distance to a large traffic generator and width of the street. Results showed that this mixed approach, including economic and geographical factors, outperforms previous models which considered economic factors alone. Readers can refer to Chapter 12 by Dablanc in this Handbook for further reflections on the role of geography and land use, not only with respect to FTG models, but in general in urban freight planning. As a final remark regarding explanatory variables for FTG estimations, Sánchez-Díaz (2017) concluded that the key premises for models’ development are the size of an establishment

Freight trip generation models  119

(measured in area or employment), type of commercial sectors and the intermediary role of the establishment within the supply chain which may affect both the FG and the ordering policy. In any case, data availability and fieldwork efforts will determine factor inclusion and accurate analysis to explain freight movements in urban contexts. Classification Systems and Their Aggregation Procedures FTG models employ aggregated analyses that allow understanding of extended freight behaviour beyond the zone where establishments provided primary information. These aggregated analyses assume homogeneous characteristics within the groups or categories selected for the establishments. One of the ways for aggregating according to either economic activity or landuse characteristics is the use of national/regional classification standards that can vary among countries, regions or even cities (Sánchez-Díaz et al., 2015a). Choices regarding classification systems should depend on information availability at public sources about explanatory variables chosen for the FTG models. Holguín-Veras et al. (2013a) concluded that economic classification systems are significantly better than land-use-based classification systems for Freight Generation (FG) and FTG modelling. A global economic-based classification system is defined by the United Nations as the International Standard Industrial Classification of All Economic Activities (ISIC). FTG models in Pani & Sahu (2019) used this system. In Europe, most codes come from the Classification of Economic Activities in the European Community (NACE) (Müller et  al., 2015). Some studies from France use the French Classification of Activities (NAF) (GonzalezFeliu & Peris-Pla, 2017), whereas Swedish studies use the Standard Industrial Classification (SNI) (Sánchez-Díaz, 2017). FTG research in the US is based mainly on the North American Industry Classification System (NAICS) (Holguín-Veras et  al., 2012). In Asia, researchers have used the National Industrial Classification (NIC) (Pani et al., 2018) and Korean Standard Statistical Classification (KSIC) for Indian and Korean cases, respectively. Several levels of disaggregation can be observed, starting from the main economic activity sectors and then defining different hierarchies of subcategories. Gonzalez-Feliu and Sánchez-Díaz (2019) and Pani and Sahu (2019) analyzed the implications of FTG estimations using different category frameworks and aggregation levels on the classification systems. Those works agreed that categorization has little impact on the model quality if the functional form (i.e., the generation pattern) is well identified and studied. For accuracy purposes, researchers recommend having enough disaggregation to capture behaviours of establishments with compatible logistics decisions. In general terms, NAICS- or ISICbased classification systems are widely used and provide high levels of confidence. Pani and Sahu (2019) reflected on the fact that adopting such classifications, based on a global reference, is crucial for improving the comparability of freight studies conducted around the world. Gonzalez-Feliu and Sánchez-Díaz (2019) evaluated classification-systems-based and functional-forms-based aggregation procedures’ impact on FTG model quality. For the former, the disaggregation impact on FTG analysis distinguished eight categories at the first stage, then each of those was divided into one or more categories to conform to a second stage of 27 categories and then a final stage with 43 categories by using the activity of the firm as a classification criterion. Changing the segmentation criteria to a functional form based on employment, the initial categories were split into 17, 72 and 105, respectively. Gonzalez-Feliu and Sánchez-Díaz

120  Handbook on city logistics and urban freight

(2019) concluded that more detailed categories do not always result in more accurate FTG estimates, in addition to the fact that detailed disaggregation of the classification codes implies that stronger efforts in data collection are needed to ensure representativeness. Data analysis also showed that functional forms can reduce the need for high levels of disaggregation as it is demanded in classification-systems-based procedures. Pani and Sahu (2019) confirmed this conclusion and found that shipment size (tons/shipment) was an effective variable by which to conduct segmentation. Geographical aggregation procedures One of the attributes that has made FTG models become popular is their disaggregated nature, with a focus on establishment/firm study levels. This advantage allows researchers to reach detailed conclusions about logistics decisions by avoiding the biases of aggregated models. Nonetheless, planning decisions need aggregated data to consolidate general infrastructure demands based on forecasting transport demand, to test transport policy measures or predict impacts on traffic (Ben-Akiva & de Jong, 2008). General conclusions about freight demand behaviour in a delimited, studied urban area imply calculations from the FTG disaggregated estimations. The procedure of providing average estimates over a group or category of establishments from their particular features is the aggregation process (Holguín-Veras et al., 2013b). Efficiency in data collection in this process could lead to savings in a project’s budget so research should pay special attention to developing optimal and at the same time reliable designs. According to the scale of FTG estimations, aggregation procedures have different scope and practical implications. Gonzalez-Feliu and Sánchez-Díaz (2019) defined FTG models into three categories, according to their scale: microscopic level for estimations at a single establishment; mesoscopic for a neighbourhood or a street; and macroscopic for an urban area or a city. Holguín-Veras et al. (2011) proposed three spatial aggregation procedures to generate FTG estimations from microscopic studies, based on employment as the explanatory variable for FTG models. Results can also apply to other variables if the model structure remains the same. Ducret and Gonzalez-Feliu (2016) compared these aggregation procedures for an FTG model in France. The authors concluded that there was no statistically significant difference among the various aggregation procedures for modelling purposes so choices can be made indistinctly and based on information availability or practical needs. However, instead of constant values for freight generation, they proposed a representation of its statistical distribution specially for highly aggregated categorization. In mesoscopic and macroscopic studies, aggregation procedures consist of the definition of groups of establishments that share relevant features within a geographic area, e.g., economic activity, employment, geographic area. FTG models look to generate insights into homogeneous groups of classes. Estimation Techniques Research on FTG models has come up with several estimation techniques to explain freight trips from the abovementioned explanatory variables. Fischer et al. (2001) listed three major methods for estimating truck trips: trip rates, linear regression models using the least-squaresregression analysis technique, and commodity-flow models. Holguín-Veras et  al. (2013b) also included additional recently developed approaches, i.e., time series, input–output, spatial

Freight trip generation models  121

regression and cross-classification methods, multiple classification analysis (MCA) and neural networks. Alho and de Abreu e Silva (2017) and Günay et al. (2016) compared the performance and accuracy of the generalized linear model (GZLM), the ordinal logit model and the cross-classification model (the MCA and partition method). They concluded that the crossclassification method is recommended for practical purposes. Nonetheless, Pani et al. (2018) found that most of the FTG studies use the ordinary least square (OLS) regression approach due to its advantage in explaining the relation between freight activity and causal variables. Trip rates The most elementary method by which to estimate FTG is trip rates. The method consists of calculating the average FTG for a specific type of establishment (ITE, 2004). More elaborate rates can be calculated by dividing the total FTG in a study zone by the number of employees, by the total commercial area or by the number of establishments. Holguín-Veras et al. (2011) highlighted the risk of using constant trip rates as they do not consider logistics decisions of receivers, mode choice of shippers and the indivisibility of trips. In essence, an increase in business size does not necessarily imply an increase in FTG, and, as shown by Holguín-Veras et al. (2012), FTG can be independent of the number of employees for some commercial sectors. This finding highlighted the need to study functional forms and estimate functions using statistical principles. Cross-classification analysis An alternative method to trip rate models is the so called cross-classification or category analysis. Its estimations are based on a function of establishments’ stratified features that explains freight trips. The main method’s assumption is the homogeneous behaviour of trip estimations within each stratum or category. Data reliability per category, as discussed in the section titled ‘Estimation Techniques and Levels of Aggregation’, is key to the quality of this method. According to Stopher and McDonald (1983), the FTG estimate results from the product of the number of individuals in each category by the trip generation rate calculation taken from primary data collection processes. Ortúzar and Willumsen (2011) suggested a careful process for choosing the categories in a way that frequency distributions of explanatory variables have the minimum standard deviation at each category. MCA is one of the most widely used methods of this kind. As an extension of the Analysis of Variance, ANOVA, it predicts average freight trips according to a certain combination of data attributes (explanatory variables in a form of integer dummy variables), significant for the response variable. Statistical goodness-of-fit measures, i.e., F-statistic, correlation ratio, coefficient of determination (r2), provide quality indicators of the proposed models. Alho and de Abreu e Silva (2017) described regression trees/partition methods as alternative approaches to MCA that make partitions according to independent variables’ values and their relationship to the dependent variable. Linear regression models This procedure explains the number of trips generated using a linear relationship and, as independent variables, employment, business area, economic activity, etc. The regression is expressed as follows:

Y j = q 0 + q1 X1 j + q 2 X 2 j +¼+ q k X kj + e j . (6.1)

122  Handbook on city logistics and urban freight

where Yj represents the estimated number of trips generated at the establishment, zone or economic sector j; Xij is the value for the i-th explanatory variable; θi is part of the k model regressors of the straight line fitted to the data; θ0 is the intercept of that line, and εj is the error term. Sánchez-Díaz et  al. (2016) proposed two particular cases of linear regression models to incorporate spatial effects on the explanatory variables. Spatial lag models add a regressor in the form of a spatial lag variable and a spatial error model which incorporates the spatial effect as part of the error term to correct for the spatial autocorrelation bias. Modellers should pay special attention to avoid methodological problems when using linear regression models as explained in Ortúzar & Willumsen (2011): multicollinearity (i.e., linear relation among explanatory variables); the number of regressors included that may lead to an overfitted model (Akaike information criterion (AIC) helps to evaluate this problem); the calculation of r2 under the inclusion of more regressors; and the non-linearity effect of the independent variables on the output (exponential, logarithmic or potential relations). For the final one, Ortúzar and Willumsen (2011) suggest procedures such as variable transformations (e.g., logarithmic) and the use of dummy variables. Non-linear regression models Another type of regression model is a non-linear model in which the independent variable is modelled as a non-linear function of the explanatory variables. Sánchez-Diaz et  al. (2016) and Sánchez-Díaz (2018) showed that FTG models can be modelled using linear-logarithmic (lin-log) or logarithmic-logarithmic models. Sánchez-Díaz et  al. (2016) showed that, when the explanatory variables are the logarithmic transformation of employment, lin-log models are better than linear models for several sectors (i.e., construction, manufacturing, wholesale, retail and food services). These results show that establishments with more employees attract more freight trips, but freight trip attraction (FTA) increases at a diminishing marginal rate. Sánchez-Díaz (2018) studied the specific case of accommodation and food sector establishments and found that non-linear models were the best method by which to model FTG. In this study, the log-log model was the best model specification, showing that FTA, freight trip production (FTP) and FTG are better estimated as a concave function of employment or area. Regression models with discrete outcome As explained by Alho and de Abreu e Silva (2017), GZLMs represent the more general form of the OLS, where the latter assumes normality for the probability distribution of the response variable and binary or continuous explanatory variables. GZLMs can have different forms. One form that has been used widely to model FTG is regression models with discrete outcome. Discrete models assume multinomial distribution for the response variable, with the link function as cumulative (e.g., logit/probit), and continuous/ordinal/categorical explanatory variables. Modelling FTG as a discrete outcome instead of a continuous variable can be more accurate. Discrete choice models use a function of covariates to explain the odds of FTG taking a discrete value. However, the wide range of FTG (e.g., from once-monthly trips for a shoe retailer to six trips per day for a restaurant) means the amount of data necessary for these models is substantial and the cost of data collection can become impractical for most cities.

Freight trip generation models  123

Count models Another type of discrete outcome model that deserves its own category is count models. The most common ones are Poisson and negative binomial models. These models consider FTG as a non-negative integer and parametrize the Poisson or negative binomial distribution to add covariates that make the outcome data more likely. While Poisson assumes that the mean and variance of the outcome variable are equal, the negative binomial relaxes this assumption and is better to model FTG data, which is usually an overdispersed variable. Zero-inflated negative binomial (ZINB) models are used when count data are overdispersed and there are several zeros. This kind of model appears to be the most appropriate to represent heterogeneity within establishments of the same category but with variable freight generation behaviour. It handles the challenge of the significant presence of zeros in the responses (Washington et al., 2020). This zero-behaviour could be explained by the fact that establishments do not generate any freight delivery/shipment, or that the frequency is so low that respondents answered zero because they cannot recall the last delivery, e.g., offices as used in Sánchez-Díaz (2020). The ZINB regression model is formulated as: 1





é 1 ù a ê ú yi = 0 with probability pi + (1 - pi ) ê a ú (6.2) 1 ê + li ú ëa û é ææ 1 êGçç èa yi = y with probability (1 - pi ) ê è ê ê êë

ö 1a yù ö ÷ + y ÷ ui (1 - ui ) ú ø ø ú ú æ1ö G ç ÷ y! ú a è ø úû

1 a

, y = 1, 2, 3¼ (6.3)

where Y = (y1, y2, …, yn) are independent events, ui = (1/α)/[(1/α) + λi]. If α is close to zero, then a Poisson model is a better fit. The parameters of the model are estimated using maximum likelihood procedures. Two-step FTG models FTG can also be modelled as a combination of linear and discrete choice models that consider the error terms to be correlated. Sánchez-Díaz et  al. (2016) highlighted the importance of modelling FTG by differentiating between FTA and FTP of freight trips since logistics implications change for establishments with one or both behaviours. Jaller et al. (2015) called those firms that produce and attract trips “intermediaries” and proposed a two-step FTG model. The first step identifies the type of establishment (intermediary or not) using a binary variable δ that is 1 if it produces freight trips and 0 otherwise.

FTP = d ( bC ) (6.4)

where β represents the vector of estimable parameters and X the vector of explanatory variables. The probability that an establishment has a positive FTP can be calculated from the following expression:

124  Handbook on city logistics and urban freight

P=



eUi (6.5) 1 + eUi

where Ui is the utility function for establishment i used in the binary logit/probit model, derived from random utility theory as follows:

Ui = a + bi C i + f FTPi + e i (6.6)

where ϕ is the vector of estimable parameters, α is the intercept and εi corresponds to a random disturbance that, in the case of logit models, is assumed to follow a Gumble distribution. FTPi is the second step of the model formulation that expresses the number of trips produced by an establishment as a continuous function:

FTPi = q Z i + hi (6.7)

where θ is the vector of estimable parameters, Zi is the vector of explanatory variables for the establishment i and ηi is a continuous random disturbance. Table 6.1 presents a summary of the estimation techniques addressed in this section. Data Collection Techniques The quality and representativity of FTG models strongly depend on the establishment’s data collection techniques to feed explanatory variables’ vectors and to validate results. Commodity flow surveys, establishment surveys, direct observation and traffic counts are traditional techniques reported in the FTG research literature (Holguín-Veras et al., 2016a). Nonetheless, with the emergence of sensors, GPS datalogs, cameras, app-based applications and data analytics capabilities, new opportunities are now available to obtain more data from urban freight trip generators (Puente-Mejia et al., 2020). For instance, Cheah et al. (2021) conducted an FTG study at a building level in Singapore, using survey apps as a data collection tool. Choices regarding data collection processes go beyond budget and context constraints but they should also consider study disaggregation levels and the number of explanatory variables needed (Gonzalez-Feliu & Sánchez-Díaz, 2019). Holguín-Veras and Jaller (2014) affirmed that establishment surveys, coupled with secondary data collection, are the most reliable instruments to provide data quality and availability. They also pointed out the need for considering modelling requirements to adapt data collection techniques to models’ structure, algorithms’ assumptions, analytics, calibration and forecast. Sánchez-Díaz (2017) enumerated the advantages and pitfalls of the main data sources for FTG models and local traffic analysis. The summary is shown in Table 6.2. SánchezDíaz (2017) confirmed the relevance of establishment-based surveys since they provided key insights from shippers and receivers for planning and public policy development. Model Validation and Transferability One of the main questions after conducting FTG studies is the possibility of applying estimated rates for establishments located in other zones, cities or even countries. At the intra-national level, Holguín-Veras et al. (2013b) applied two methods to assess FTG model transferability.

Freight trip generation models  125

Table 6.1  Summary of estimation techniques Estimation technique

Description

Advantages

Disadvantages

Trip rates

FTG divided by number of establishments

• Simplicity • Work with small samples

• No statistical assessment • Neglect role of ordering and inventory strategy

Cross classification analysis

FTG rates for different sectors of establishments

• Allows non-linear variations • Simplicity

• Assumes homogeneous FTG for each stratum

Linear regression

Linear relationship with economic variables

• Multiple explanatory variables • Flexible

• Risk of multicollinearity • Risk of overfit • Neglects nonlinear effects

Nonlinear regression

Nonlinear relationships (e.g., logarithmic or quadratic functions) with explanatory variables

• Captures that FTG increases at a diminishing marginal rate of business size

• Challenging spatial aggregation • Sensitive to sample design

Regression models with discrete outcome

FTG modelled as non-negative integer

• Nature of dependent variable in line with indivisibility of trips • Identify establishments that do not generate any freight (e.g., intermediaries, some offices)

• Require large amounts of data • More complex models

Two-step models (e.g., continuous-discrete)

Model two variables in an integrated framework

• Model discrete and continuous outcomes • Identify establishments that do not generate any freight (e.g., intermediaries, some offices)

• Model complexity

One model estimated FTG rates for establishments with an already known rate for comparison purposes, whereas the other consisted of an econometric analysis of estimations collected from same-industry establishments located in different US cities (i.e., furniture stores and grocery stores). International comparisons have been reported in Alho et al. (2015) and Ibeas et al. (2012). Their research compared retail FTG rates between cities in different countries using statistical measures of the estimation error (root-mean-square error, RMSE). Conclusions drawn from the abovementioned studies suggest that transferability is valid for the specific industry sectors that were part of their research, but no generalization to all

126  Handbook on city logistics and urban freight

Table 6.2  Data sources for FTG and local traffic analysis: advantages and disadvantages Data source

Main advantages

Main disadvantages

Traffic counts

• Simple to measure and estimate

• Freight trips not linked to establishments • Little insight into causality; role of logistics disregarded • Limited use for policy analysis

Secondary sources: aggregated data

• Limited data collection efforts • Low cost

• Freight trips not linked to establishments • Little insight into causality; role of logistics disregarded • Limited use for policy analysis • Disaggregation techniques assume employment proportionality • Trips within zones are not considered • Often limited to heavy commercial vehicles (>3 tons)

Transport operators' data

• High detail about shipment size

• Estimates per establishment are partial (not all operators interviewed) • Little insight into causality; role of logistics disregarded • Freight trips not linked to establishments • Limited use for policy analysis

Establishment-based surveys

• Estimates at establishment level • Connects FTG to establishment attributes • Allow forecasting and policy analysis • Allow aggregation at any geographical level

• Higher cost • Require cooperation from commercial establishments • Require modelling efforts • Limited knowledge about routes and vehicles • Do not capture through traffic

Source:  Sánchez-Díaz (2017)

industry sectors can be derived from this result. In essence, Holguín-Veras et  al. (2013b) showed that there is evidence that (i) FTG models estimated from data in one city can be applied to other cities in the same country to obtain acceptable estimations, and that (ii) applying the right functional form can deliver better estimates than including location (e.g., State) as explicit variables. Pani et al. (2020) confirmed some of these findings and called for future research on the effects of locational effects on FTG to reach definitive conclusions about FTG transferability for various industry sectors. Applications FTG models have been the origin of high-impact applications for urban freight management. Worldwide reports show the impact of FTG on parking management (Gardrat & Serouge, 2016; Jaller et al., 2013; Kalahasthi et al., 2022), policy design, e.g., off-hour deliveries (Holguín-Veras et  al., 2016b; Ukkusuri et  al., 2016) and receiver-led consolidation

Freight trip generation models  127

(Holguín-Veras & Sánchez-Díaz, 2016c). They have also provided foundations for facility design (ITE, 2020; Ogden, 1992) and macroscopic models for city planning and infrastructure (McLeod & Curtis, 2020; Sánchez-Díaz et al., 2015b).

CASE STUDY: FTG ANALYSIS FOR STOCKHOLM This section illustrates how to apply the techniques discussed via a real-life case study. The city of Stockholm serves as the case study for presenting FTG models’ scope in the commercial sectors of accommodation, food services, retail, manufacturing, wholesale and offices. This research was reported in Sánchez-Díaz (2018, 2020). Data Collection Techniques The city of Stockholm, delimited by the congestion zone, is home to 11,279 establishments. For this study, a sample of 354 establishments was selected using a random draw and considering budget constraints. The data collection process had three phases that included internet surveys and mail-in mail-out surveys, computer-assisted telephone interviews and in-person interviews to ensure a high response rate and respondent suitability. The questionnaires included enquiries about the type of commercial activity and business, number of employees, area, the number of weekly deliveries and shipments, the amount of cargo attracted, the number of carriers, the number of vehicles used for deliveries, receiving facility and the time of day for the deliveries (SánchezDíaz, 2018). Statistical nonparametric tests served to assess the statistical significance of the relationships among the variables. The latter were classified by their nature, i.e., continuous, discrete and categorial, and by attribute. Primary information was complemented by official information from the Swedish Office of Statistics (SCB) that included data about employment, postcode and commercial sector. Classification Systems The data were segmented in seven different sectors using the SNI system, i.e., retail nonperishable, retail perishable, accommodation, food services, wholesale, manufacturing and offices. Table 6.3 shows FTG patterns for the eight categories that include non-freight-intensive (non-FIS) and freight-intensive sectors (FIS) as defined in Holguín-Veras et al. (2016b). In general, descriptive information shows how offices have high employment per square meter and low FTG per establishment. Dependent and Independent Variables The study defined FTG as the sum of FTA, FTP, combined freight trips (i.e., both a shipment and delivery take place), and service trips (i.e., a service, such as replenishment, takes place). For sectors with a more intensive freight activity, the dependent variable estimated was the continuous number of weekly freight trips. For sectors with less frequent behaviour of freight trips, the response variable was modelled as the non-negative discrete number of trips

128

0.7

62

1 if yes, 0 if no

e-commerce

8.5

0.7

473.6

77

m2

Area

1.7 26.5

85

84

Interval freight

Employee no.

5.0

FG

85

57

1 if yes, 0 if no

Trips/week

Intermediary

Interm. FTP

7.9 4.6

Employment

85

85

Trips/week

Deliveries/week

FTG

FTA

162%

155%

190%

n.a.

148%

n.a.

146%

0.5

15.0

0.5

1.0

0.0

0.0

1.0

77.0

3,780.0

320.0

5.0

35.5

1.0

41.5

23

23

23

23

21

23

23

818.7

7.8

2.7

17.5

0.9

10.7

26.6

Mean

SNI: Manufacturing

230.7

6.2

4.0

3.9

0.3

7.2

Obs.

40

43

42

14

43

43

43

326%

180%

n.a.

159%

n.a.

123%

139%

CV

193%

109%

n.a.

60%

n.a.

81%

81%

10.0

1.0

1.0

1.0

0.0

0.5

0.5

Min

10.0

1.0

1.0

1.0

0.0

1.0

1.0

13,000.0

56.5

6.0

100.0

1.0

57.0

127.5

Max

2,500.0

35.5

6.0

8.0

1.0

20.0

26.0

SNI: Food services (restaurants and cafés) Obs. Mean CV Min Max

Obs.

Max

1.0

887.0

30.0

6.0

30.0

1.0

35.0

40.0

Max

Unit

Min

0.0

20.0

1.0

1.0

0.1

0.0

0.3

0.25

Min

Variable

CV

n.a.

92%

118%

n.a.

135%

n.a.

98%

92%

CV

SNI: Offices Mean

0.4

221.1

55

m2

Area

3.0 4.6

61

63

Interval freight

Employee no.

4.1

FG

63

44

1 if yes, 0 if not

Trips/week

Intermediary

Interm. FTP

6.9

9.8

Mean

Employment

63

63

Trips/week

Deliveries/week

FTG

Obs.

FTA

SNI: Retail non-perishable Variable Unit

Table 6.3  Descriptive statistics of freight generation by establishments categories in Stockholm

129

0.6

17.0 483.4

28

28

26

Interval freight

Employee no.

m2

FG

Employment

Area

1.0

Source:  Sánchez-Díaz (2018, 2020) Notes:  CV – Coefficient of Variation. n.a. - not applicable

3.6

15.3

28

27

1 if yes, 0 if no

Trips/week

Intermediary

18.3

33.1

Mean

Interm. FTP

28

28

Trips/week

Deliveries/week

FTG

FTA

Obs.

Unit

Variable

SNI: Wholesale

479.0

55

m2

Area

4.4 6.1

59

65

Interval freight

Employee no.

12.6

FG

65

38

1 if yes, 0 if no

Trips/week

Intermediary

Interm. FTP

21.8

Employment

65

Deliveries/week

FTA

Mean 29.2

CV

197%

216%

n.a.

155%

n.a.

108%

123%

CV

165%

137%

n.a.

150%

n.a.

177%

155%

Min

34.0

2.0

1.0

1.0

0.0

0.1

1.0

Min

18.0

1.0

2.0

0.3

0.0

1.0

1.0

Max

5,000.0

165.0

6.0

120.0

1.0

75.0

195.0

Max

3,500.0

41.0

6.0

102.0

1.0

250.0

250.0

37

46

46

37

46

46

46

3,730.2

24.1

4.3

5.0

0.8

14.4

18.4

Mean

153%

136%

n.a.

104%

n.a.

121%

109%

CV

Obs.

Obs.

65

Unit

Trips/week

Variable

FTG

SNI: Accommodation

SNI: Retail perishable Min

250.0

1.0

1.0

0.5

0.0

0.5

1.0

Max

27,000.0

130.0

6.0

22.5

1.0

90.0

102.0

130  Handbook on city logistics and urban freight

generated per week. Employment was the independent variable used in the regression models. Table 6.3 displays the descriptive statistics. As expected, FIS present high FTG with wholesale as the most prominent sector in the demand for freight. An explanation for this behaviour can be given by its nature of an intermediary node in supply chain freight flow. All the establishments in this sector declared this intermediary operation. For the non-FIS, the FTG of the accommodation establishment was more than twice that of the food service or office establishments. The latter represents only 7.9 deliveries per week with a dispersion of 162%, which is relatively high compared to results from FIS. Surprisingly, the share of intermediaries of the office establishments sector is 70% which can be explained by their active documents’ deliveries/shipments. Offices also have the highest mean number of employees (26.5), with up to 320 employees reported. Among the delivery items reported by the respondents, roller cages and packages are the most common delivery items. In terms of weight, 43% of establishments received 50 kg per week at most, mostly offices and retail non-perishable, 30% between 51 kg and 500 kg, and 29% of more than 500 kg. Regarding the type of vehicles that provided freight transport services, cars were used for 10% of the trips, light vehicles like vans and pick-ups for 35% of the trips, medium-duty trucks for 33%, heavy trucks represented 7% of the deliveries and other vehicles, such as cargo bikes and electric small vehicles, made up 9% of the trips. Most of the deliveries (66.1%) took place at the establishment’s entrance. Loading docks are common only for operations at retail perishable and accommodation establishments (42.9% and 38.2% of the operations in these sectors, respectively). Estimation Techniques and Levels of Aggregation Linear regression models estimated weekly freight trips for FIS based on the explanatory variables of employment and area. Employment and area were incorporated in separate models to avoid multicollinearity. A ZINB model was selected to estimate the offices’ FTG. Table 6.4 shows the results for the two models; t-statistics are displayed in parentheses under each parameter, with only variables which were significant at the 10% level of confidence being retained.​ Since the number of employees, the area and the commercial activity are the main factors with which to estimate FTG, it was possible to use official data from SCB to extend the analysis beyond the sample of the fieldwork. As discussed in Sánchez-Díaz (2018), this involves the challenge of mismatches between headquarters address and freight operation facilities. It also implies taking the risk of having outdated information. However, by applying aggregation techniques properly and with some data wrangling, it was possible to produce FTG estimates for the entire city. Table 6.5 presents the estimates for the municipality of Stockholm. The FTG estimation of trips per week is 164,955. Wholesale represents the sector with the highest proportion (32.2%), followed by retail perishable (18.1%), retail nonperishable (14.3%), food services (12.5%), offices (12.1%), manufacturing (8.3%) and accommodation (2.5%). These models estimate FTG, which does not necessarily mean that it represents the actual number of freight vehicles entering the area. One single vehicle could fulfil several deliveries/ shipments and the number of stops per vehicle varies across sectors. Of the almost 30,000 freight trips generated daily in Stockholm (City of Stockholm, 2015), 85% come from freightintensive sectors and 15% from offices. Freight-intensive sectors contribute the most to FTG in

131

65

29

Retail perishable

Wholesale

(2.56)

0.52 (5.61)

24.31

(3.18)

(5.23)

4.54

0.29

(4.80)

(2.78)

11.48

1.53

(2.30)

(1.82)

14.73

0.42

(6.06)

(1.84)

7.88

0.25

(5.22)

(1.70)

(7.31)

6.95

0.03

Emp.

3.94

Const. (2.93)

28.20

Adv.

0.22

0.75

0.22

0.34

0.06

0.06

0.46

r2

36.85

26.76

17.967

30.88

8.81

6.75

9.17

RMSE

26

55

37

23

55

40

77

(2.75)

24.80

(2.84)

10.66

(3.53)

6.60

(3.03)

19.51

(3.61)

6.01

(6.14)

3.80

Const.

Area models Obs.

(4.73)

1.58

(4.12)

4.53

(6.02)

0.32

(15.21)

0.87

(2.10)

1.64

(1.89)

0.91

(1.83)

0.22

Area (2.43)

26.90

Adv.

0.13

0.55

0.72

0.39

0.13

0.18

0.42

r2

40.05

32.97

11.53

29.52

8.77

9.81

9.49

RMSE

Notes:   Const. denotes the intercept of the model. Emp. denotes the parameter for the number of employees. The unit for area is 100 m2. Numbers in parentheses represent t values – only those values where P