South Asia Development Matters South Asia’s Hotspots The Impact of Temperature and Precipitation Changes on Living Standards Muthukumara Mani Sushenjit Bandyopadhyay Shun Chonabayashi Anil Markandya Thomas Mosier South Asia’s Hotspots South Asia’s Hotspots The Impact of Temperature and Precipitation Changes on Living Standards Muthukumara Mani Sushenjit Bandyopadhyay Shun Chonabayashi Anil Markandya Thomas Mosier © 2018 International Bank for Reconstruction and Development / The World Bank 1818 H Street NW, Washington, DC 20433 Telephone: 202-473-1000; Internet: www.worldbank.org Some rights reserved 1 2 3 4 21 20 19 18 This work is a product of the staff of The World Bank with external contributions. The findings, interpre- tations, and conclusions expressed in this work do not necessarily reflect the views of The World Bank, its Board of Executive Directors, or the governments they represent. The World Bank does not guarantee the accuracy of the data included in this work. The boundaries, colors, denominations, and other information shown on any map in this work do not imply any judgment on the part of The World Bank concerning the legal status of any territory or the endorsement or acceptance of such boundaries. Nothing herein shall constitute or be considered to be a limitation upon or waiver of the privileges and immunities of The World Bank, all of which are specifically reserved. Rights and Permissions This work is available under the Creative Commons Attribution 3.0 IGO license (CC BY 3.0 IGO) http:// creativecommons.org/licenses/by/3.0/igo. Under the Creative Commons Attribution license, you are free to copy, distribute, transmit, and adapt this work, including for commercial purposes, under the following conditions: Attribution—Please cite the work as follows: Mani, Muthukumara, Sushenjit Bandyopadhyay, Shun Chonabayashi, Anil Markandya, and Thomas Mosier. 2018. South Asia’s Hotspots: The Impact of Temperature and Precipitation Changes on Living Standards. South Asia Development Matters. Washington, DC: World Bank. doi:10.1596/978-1-4648-1155-5. License: Creative Commons Attribution CC BY 3.0 IGO Translations—If you create a translation of this work, please add the following disclaimer along with the attribution: This translation was not created by The World Bank and should not be considered an official World Bank translation. The World Bank shall not be liable for any content or error in this translation. Adaptations—If you create an adaptation of this work, please add the following disclaimer along with the attribution: This is an adaptation of an original work by The World Bank. Views and opinions expressed in the adaptation are the sole responsibility of the author or authors of the adaptation and are not endorsed by The World Bank. Third-party content—The World Bank does not necessarily own each component of the content contained within the work. The World Bank therefore does not warrant that the use of any third-party-owned individual component or part contained in the work will not infringe on the rights of those third parties. The risk of claims resulting from such infringement rests solely with you. If you wish to reuse a component of the work, it is your responsibility to determine whether permission is needed for that reuse and to obtain permission from the copyright owner. Examples of components can include, but are not limited to, tables, figures, or images. All queries on rights and licenses should be addressed to World Bank Publications, The World Bank Group, 1818 H Street NW, Washington, DC 20433, USA; e-mail: pubrights@worldbank.org. ISBN (paper): 978-1-4648-1155-5 ISBN (electronic): 978-1-4648-1156-2 DOI: 10.1596/978-1-4648-1155-5 Cover photo: © Ismail Ferdous/World Bank. Permission required for reuse. Library of Congress Cataloging-in-Publication Data has been requested. Contents The Book at a Glance. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xi Foreword. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xiii Acknowledgments. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xv About the Authors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xvii Abbreviations. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xix Overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1 A Vulnerable Region. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3 Climate Change and Living Standards. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3 Climate Modeling and Effects. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6 Hotspots. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8 Toward Greater Resilience. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10 References. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11 1 A Vulnerable Region. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13 Progress So Far. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15 A Road Map for Climate-Resilient Development . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16 Note . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18 References. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18 2 Increasingly Hot. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19 Highly Diverse Climate. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19 Unambiguous Historic Temperature Increases. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20 Projecting Future Climate. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22 Selecting Appropriate Climate Models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24 South Asia Continues to Get Hotter . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26 References. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30 3 Climate and Living Standards. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33 Accumulated Knowledge. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35 Analytical Framework. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36 Data Sources. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38 v vi   C o n t e n t s Two Methodological Challenges. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39 Control Variable Selection. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39 Absorbed Climate Effects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40 Temperature Inflection Points . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41 National-Level Empirical Findings . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43 Dealing with Uncertainty . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46 Notes. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48 References. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49 4 Mapping Hotspots . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51 What Is a Hotspot?. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51 The Carbon-Intensive Scenario Leads to More Severe Hotspots . . . . . . . . . . . . . . . . . . . . . . 52 Hotspots Tend to Have Less Infrastructure and Services. . . . . . . . . . . . . . . . . . . . . . . . . . . . 55 The Most Vulnerable Households. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57 Country Hotspots. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58 References. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68 5 Toward Greater Resilience . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71 Money Worth Spending . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72 Reducing Hotspots in Vulnerable Communities and Vulnerable Households . . . . . . . . . . . . 73 Policy Agenda. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73 Note . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78 References. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78 Appendix A  Methodology for Policy Cobenefits . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79 Appendix B  Supplementary Tables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81 Appendix C  Supplementary Maps. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87 Appendix D  Climate Model Selection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91 Appendix E Calculating Gross Domestic Product Based on Shared Socioeconomic Pathways and Hotspots Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 99 Boxes 1.1 Why Do Changes in the Average Weather Matter?. . . . . . . . . . . . . . . . . . . . . . . . . . . . 17 2.1 Understanding Historic and Projected Temperatures for Each Country . . . . . . . . . . . . . 28 3.1 How Climate Change Affects Consumption Expenditures . . . . . . . . . . . . . . . . . . . . . . 34 3.2 The Quadratic Relationship between Climate and Economy . . . . . . . . . . . . . . . . . . . . . 36 3.3 Why the Positive Results for Nepal Are Not an Anomaly. . . . . . . . . . . . . . . . . . . . . . . 45 4.1 Will Mountain and Coastal Areas Benefit from Climate Change?. . . . . . . . . . . . . . . . . 54 4.2 Heat Vulnerability Index for India . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55 4.3 Other Dimensions of Hotspots: Tracking Nonmonetary Effects of Climate Change. . . . 67 Figures O.1 Increases in Temperatures and Changes in Precipitation Patterns Are Linked to Living Standards through a Diverse Set of Pathways. . . . . . . . . . . . . . . . . . . . . . . . . . . . 4 O.2 Temperature and Consumption Have an Inverted U–Shaped Relationship for Countries in South Asia . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5 C o n t e n t s   vii O.3 Annual Temperature Increases Are Projected to Accelerate. . . . . . . . . . . . . . . . . . . . . . . 6 O.4 Monsoon Precipitation Varies Considerably and Projections Are Uncertain. . . . . . . . . . 7 1.1 Some Manifestations of Climate Change. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14 B1.1.1 Increased Average Temperature Causes Increased Likelihood of Extreme Heat Events. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17 2.1 Unambiguous Temperature Trends in South Asia. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21 2.2 An Illustration of Model Selection Criteria. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25 2.3 Historic Trends in Annual Temperature Increases Are Projected to Increase. . . . . . . . . 27 2.4 Monsoon Precipitation Varies Considerably Year to Year, and Projections Are Highly Uncertain . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27 B2.1.1 Annual Temperatures Are Increasing for All Countries, but the Rate of Change Varies. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28 B3.1.1 Climate Change and Living Standards Are Linked through a Diverse Set of Pathways. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34 B3.2.1 Impacts of Temperature on Productivity Are Well Explained Using a Quadratic Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37 3.1 Temperature and Consumption Have an Inverted U–Shaped Relationship for Countries in South Asia . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42 3.2 Uncertainties of the Predicted Consumption Changes Arise from Differences between Climate Models and Economic Modeling. . . . . . . . . . . . . . . . . . . 48 B4.3.1 Climate Has Diverse Monetary and Nonmonetary Effects on Well-Being. . . . . . . . . . . 67 5.1 Good Development Outcomes Reduce Hotspots. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75 A.1 Effects of Development Outcomes on Hotspots in Sri Lanka under the Carbon-Intensive Scenario by 2050. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80 Maps O.1 Temperatures Have Been Increasing in Much of South Asia. . . . . . . . . . . . . . . . . . . . . . 3 O.2 Annual Average Temperatures Increase by 2050 Relative to 1981–2010. . . . . . . . . . . . . 7 O.3 Severe Hotspots Will Cover a Significant Portion of South Asia by 2050 . . . . . . . . . . . . 9 1.1 South Asia Remains a Region Very Vulnerable to Climate Change and Extreme Events. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14 1.2 South Asia Continues to Be Home to a Large Number of Poor People. . . . . . . . . . . . . 15 2.1 Temperatures Vary Significantly across South Asia. . . . . . . . . . . . . . . . . . . . . . . . . . . . 20 2.2 Average Monsoon Precipitation in South Asia Generally Increases from West to East. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20 2.3 Temperatures Have Been Increasing in Most of South Asia. . . . . . . . . . . . . . . . . . . . . . 20 2.4 No Overall Monsoon Precipitation Trends for Most of South Asia. . . . . . . . . . . . . . . . 22 2.5 Annual Average Temperature Is Projected to Continue Increasing Dramatically under the Climate-Sensitive and Carbon-Intensive Scenarios. . . . . . . . . . 30 4.1 Mild and Moderate Hotspots Are Prevalent Throughout South Asia by 2030 . . . . . . . 53 4.2 Moderate and Severe Hotspots Cover a Significant Portion of South Asia by 2050. . . . 53 B4.2.1 Central India Is the Most Vulnerable to Heat. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55 C.1 Percentage of People in Each Administrative Unit Who Live in Rural Environments. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87 C.2 Average Years of Education of the Head of Household in Each Administrative Unit . . . . 87 C.3 Percentage of People in Each Administrative Unit Who Have Access to Electricity. . . . 88 C.4 Average Travel Time to Market in Hours. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 88 viii   C o n t e n t s C.5 Average Density of Roads. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 88 C.6 Average Population Density per Square Kilometer . . . . . . . . . . . . . . . . . . . . . . . . . . . . 88 C.7 Climate-Sensitive Scenario by 2050: Hotspots Do Not Clearly Overlap with Major Basins. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89 C.8 Carbon-Intensive Scenario by 2050: Hotspots Do Not Clearly Overlap with Major Basins. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89 D.1 Percentage of Annual Precipitation Contained in the Study Seasons. . . . . . . . . . . . . . . 92 D.2 Spatial Density of Station Measurements’ Contribution to the Aphrodite Data Set, 1981 through 2000. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93 D.3 Seasonal and Temporal Consistency of Station Measurements’ Contribution to the Aphrodite Data Set, 1979 through 2005. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95 D.4 Average Monsoon Precipitation, 1981 through 2000. . . . . . . . . . . . . . . . . . . . . . . . . . 95 Tables 1.1 Climate Change Strategies and Action Plans of Countries in South Asia. . . . . . . . . . . . 16 2.1 Results from Climate Model Projections for Two Future Time Frames . . . . . . . . . . . . 26 3.1 Household Surveys in South Asia. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38 3.2 Variables Considered for Household, District, and Geospatial Differences. . . . . . . . . . 39 3.3 Control Variables for Each Country. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40 3.4 Changes in Average Weather Predicted to Have Mostly Negative Effects under Both Scenarios . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43 3.5 Comparison between This Book’s Results and Those of Other Studies in the Same Time Frame. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44 4.1 Hotspot Labels and Definitions. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52 4.2 Millions of South Asians Are Living in Areas Projected to Become Hotspots . . . . . . . . 56 4.3 Locational Characteristics, by Hotspot Category . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56 4.4 Characteristics of the Most Affected Households Compared with National Averages . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57 4.5 Predicted Change in Living Standards and Characteristics of Divisions in Bangladesh . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59 4.6 Predicted Change in Living Standards and Characteristics of the Top 10 District Hotspots in Bangladesh . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59 4.7 Predicted Change in Living Standards and Characteristics of the 10 Most Affected States in India . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60 4.8 Predicted Change in Living Standards and Characteristics of the Top 10 District Hotspots in India. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60 4.9 Predicted Change in Living Standards and Characteristics of Provinces in Sri Lanka under the Carbon-Intensive Scenario in 2050. . . . . . . . . . . . . . . . . . . . . . 62 4.10 Predicted Change in Living Standards and Characteristics of the Top 10 District Hotspots in Sri Lanka under the Carbon-Intensive Scenario in 2050. . . . . . . . 62 4.11 Predicted Change in Living Standards and Characteristics of Provinces in Pakistan under the Carbon-Intensive Scenario in 2050. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63 4.12 Predicted Change in Living Standards and Characteristics of the Top 10 District Hotspots in Pakistan under the Carbon-Intensive Scenario in 2050 . . . . . . . . . . . . . . . 63 4.13 Predicted Change in Living Standards and Characteristics of Regions in Nepal under the Carbon-Intensive Scenario in 2050. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64 4.14 Predicted Change in Living Standards and Characteristics of the Top 10 District Hotspots in Nepal under the Carbon-Intensive Scenario in 2050 . . . . . . . . . . . . . . . . . 64 C o n t e n t s   ix 4.15 Predicted Change in Living Standards and Characteristics of the Top 10 Most Affected Districts in Afghanistan under the Carbon-Intensive Scenario in 2050. . . . . . 66 4.16 Predicted Change in Living Standards and Characteristics of the Top 10 Province Hotspots in Afghanistan under the Carbon-Intensive Scenario in 2050. . . . . . . . . . . . . 66 5.1 Changes in Average Weather Projected under the Carbon-Intensive Scenario Will Disproportionately Impact Severe Hotspots. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72 5.2 Changes in Average Weather Projected under the Carbon-Intensive Scenario Will Reduce Total GDP. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72 5.3 Profile of the Top 10 Percent Resilient Households. . . . . . . . . . . . . . . . . . . . . . . . . . . . 74 B.1 18 Climate Models Assessed . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81 B.2 Regression Results Used for Consumption Predictions. . . . . . . . . . . . . . . . . . . . . . . . . 82 B.3 Changes in Consumption from Base Year 2011 to 2030 and 2050 for Climate-Sensitive and Carbon-Intensive Scenarios from Model Specifications . . . . . . . 83 E.1 Population Projections for Countries with Severe Hotspots . . . . . . . . . . . . . . . . . . . . 100 The Book at a Glance About the Book n Both scenarios show rising temperatures throughout the region in the coming decades, n The World Bank’s regional flagship, with the carbon-intensive scenario leading to South Asia’s Hotspots: The Impact of greater increases. Expected changes in rainfall Temperature and Precipitation Changes on patterns are more complex in both scenarios. Living Standards, brings forth new research on the impact of climate change in South n Changes in average weather are projected Asia by analyzing how changes in average to have overall negative impacts on living temperature and precipitation—referred standards in Bangladesh, India, Pakistan, to as “average weather”—will affect living and Sri Lanka. While negative impacts standards. are sizable under both climate scenarios, they are more severe under the carbon- n The book breaks new ground in intensive scenario. understanding how weather conditions affect living standards by combining n Unlike sea-level rise and extreme weather and analyzing granular temperature and events, changes in average weather will precipitation information and household affect inland areas the most. survey data. n For most countries, changes in average n The book identifies climate “hotspots”— weather will also reduce growth of their areas where changes in average weather gross domestic product (GDP) per capita, are predicted to have a negative impact on compared to what it would be under living standards—in South Asia. present climate conditions. The GDP losses are greater for severe hotspot regions. Main Findings n More than 800 million people—almost n The book analyzes two future climate half of South Asia’s population—currently scenarios — one that is “climate-sensitive,” live in areas that are projected to become in which some collective action is taken to moderate to severe hotspots by 2050 limit greenhouse gas emissions, and one under the carbon-intensive scenario. Most that is “carbon-intensive,” in which no projected hotspots are found to be in action is taken. disadvantaged areas. xi xii  T h e Book at a Glance Recommendations differences in local conditions between hotspots. n Overall, ensuring good development outcomes is the best strategy to build n South Asia’s Hotspots, together with resilience to changes in average weather and existing studies on the impacts of sea- improve hotspots. level rise and extreme events, creates a sound foundation for investing in targeted n The book identifies interventions tailored policies and actions to build climate to each country that could mitigate resilience throughout the region. hotspots. Interventions must account for Foreword C hanges in the Earth’s climate will Hotspots have distinguishing features that have major effects on the people of vary from country to country. A detailed South Asia, which is already one of assessment of their characteristics, and of the most affected regions of the world. A their households’ characteristics, enriches our number of studies have looked at the conse- understanding of how to address climate quences of extreme events—droughts, change. floods, heat waves, and storm surges—as Actions needed to adapt to climate include well as those of sea-level rise in general, and many with which we are already familiar. have found such events to damage the Reducing emissions of greenhouse gases will health and well-being of the population, mitigate the size of the climate impacts, and especially the poor. This book adds to that inclusive economic growth across the subcon- body of knowledge by investigating the tinent will help its population to adapt more effects on living standards of long-term easily. In addition, the potential effects of spe- changes in the climate—of rising average cific actions on the hotspots are discussed. temperatures, but also changes in the pat- These actions vary across countries, and the terns of precipitation. It does so by looking analyses provide further guidance on what at how differences in weather have influ- kinds of policy responses are most likely to be enced living standards across the region in beneficial. recent decades. This book is a major contribution to our The book finds that higher temperatures understanding of how increasing tempera- reduce average living standards in most of tures and changing precipitation patterns South Asia. This finding, combined with the interact with social and economic structures expected changes in climate by 2050, is used at a very granular level across South Asia. to project likely changes in living standards at The book should be of great value to all those a detailed, spatial level. Hotspots are identi- concerned with the development of the region fied where people could be most severely over the coming decades. affected by changes in average temperature Annette Dixon and precipitation. Many of these are in loca- Vice President tions that hitherto have not been seen as par- South Asia Region ticularly vulnerable to climate. World Bank Group xiii Acknowledgments T his book was prepared by a team led by comments and suggestions at ­various stages of Muthukumara Mani (Lead Economist) the book, including, most notably, Marianne under the guidance of Martin Rama Fay, Christophe Pusch, Stephen Hammer, (Chief Economist) and Annette Dixon (Vice Bernice Van Bronkhorst, Chandra Shekar President) of the South Asia Region of the Sinha, Gladys Lopez-Acevedo, Dhushyanth World Bank. The book is a collaborative effort Raju, Fan Zhang, Yue Li, Alex Ferguson, of the Office of the Chief Economist and the Tatiana Nenova, Sanjay Srivastava, Kanta Disaster and Climate Change Unit of the South Kumari Rigaud, and Rafik Hirji. Asia Region. The core team members include The book could not have been produced S u s h e n j i t B a n d y o p a d h y a y, S h u n without the support of Elena Karaban, Chonabayashi, Anil Markandya, and Thomas Yann Doignon, Joe Qian, Jewel McFadden, Mosier. The team was supported in the Rumit Pancholi, Michael Harrup, Jeffrey ­ climate-modeling work by a team of climate Lawrence, Bruno Bonansea and Neelam scientists from the Columbia University’s Chowdhry. Thomas Cohen provided edito- International Research Institute for Climate rial support. and Society, led by Lisa Goddard. The team gratefully acknowledges timely The team greatly benefited from insightful financial support from the Australian comments and guidance from internal peer Department of Foreign Affairs and Trade. In reviewers—Stephane Hallegatte, Emmanuel producing this report, the World Bank Skoufias, Harshadeep Nagaraja Rao, and emphasizes that climate change initiatives Erick Fernandes—and from external peer and projects shall respect the sovereignty of reviewers—Robert Mendelsohn, Rajat the countries involved, and notes that the Kathuria, Saleemul Huq, and Shantanu Mitra. findings and conclusions in the report may The team is also grateful to other colleagues not reflect the views of individual countries from the World Bank for their thoughtful or their acceptance. xv About the Authors M uthukumara Mani is a Lead then he has published on a variety of topics, Economist in the Office of the Chief such as the environmental Kuznets curves, Economist of the World Bank’s overcompliance with water pollution controls, South Asia Region. He primarily works on cli- linkages between poverty and natural resource mate change mitigation and adaptation and management, and autonomous adaptation to water and environmental issues in the region. climate change and its effects on household Prior to joining the region, he led the Bank’s welfare. Sushenjit has worked in the World work on assessing environmental implications Bank in various research capacities. He has an of development policy reforms in the Bank’s MA in economics from Jawaharlal Nehru Environmental and Natural Resources Global University in India and a PhD in economics Practice. His work also has focused on country from the University of Maryland, College Park. environmental assessments, natural resources management, environmental institutions and Shun Chonabayashi is an environmental econ- governance, and trade and climate change omist at the World Bank. He has extensive issues. Prior to joining the Bank in this posi- experience in economic analysis of impacts of tion, he was an economist in the Fiscal Affairs climate change and extreme weather events Department of the International Monetary and has carried out many research projects Fund, where he was responsible for analyzing in both developing and developed countries. environmental implications of macroeconomic His special interests include economic mod- policies and programs and in integrating envi- eling, sustainability, climate change, and the ronmental considerations broadly in the IMF’s poverty-environment nexus. Prior to joining the ­ country programs. Mani has an MA and a World Bank, he was a postgraduate fellow at PhD in economics from the University of the Yale School of Forestry and Environmental Maryland, College Park. Studies. He received a BA from Soka University and an MA from Yale University and is a PhD Sushenjit Bandyopadhyay is an empirical envi- candidate at Cornell University. ronmental economist with a special interest in linkages between climate change and house- Anil Markandya has worked in the field of hold welfare. He studied economics in India resource and environmental economics for and the United States. His first research on more than 35 years. He was a lead author for droughts in India was published in 1992. Since the 3rd and 4th IPCC Assessment Reports on xvii xviii  Ab o u t the Authors Climate Change (which were awarded a share extensive experience in development con- of the Nobel Peace Prize in 2007) as well as texts. At the World Bank, Thomas has for the 5th IPCC Report published in 2014. blended his background with economics to He has worked extensively as an advisor on examine the impacts of climate change on climate change adaptation policies in Europe, people and water resources in South Asia. the United States, and several developing Thomas earned a dual-major PhD in water countries. He was President of the European resources engineering and mechanical engi- Association of Environmental and Resource neering at Oregon State University. His Economics from 2014 to 2015. Anil is a graduate research investigated the linkages Distinguished Ikerbasque Professor at the between climate, cryosphere (snow and Basque Centre for Climate Change in the ice), surface water resources, and hydro- Basque Country, Spain, and a Honorary power in mountain environments. Before Professor of Economics at the University of graduate school, Thomas served as a Peace Bath, UK. He has a PhD from the London Corps volunteer in Kenya, where he taught School of Economics. high school physics and mathematics. He earned his undergraduate degree Thomas Mosier is a research scientist spe- from Reed College, where he majored cializing in climate, energy, and water, with in physics. Abbreviations AGDP agricultural gross domestic product GDP gross domestic product GHG greenhouse gas IIASA International Institute for Applied Systems Analysis IMD Indian Meteorological Department RCP representative concentration pathway RMSE root mean squared error SSP shared socioeconomic pathway xix Overview S outh Asia’s Hotspots: The Impact of the relationship between weather conditions Temperature and Precipitation Changes and living standards are conducted separately on Living Standards is the first book of its for individual countries. The combination of kind to conduct granular spatial analyses of the localized climate projections and household long-term effects of changes in average temper- survey analyses yields a granular picture of ature and precipitation—referred to throughout the expected effects. the book as “average weather”—in one of the The book shows that average temperatures world’s poorest regions. This book builds upon have risen over the past six decades and will accumulated research on climate change by continue to rise. Over the 2050 horizon, it analyzing how trends in average temperatures predicts more warming inland and less warm- and precipitation patterns over the coming ing in coastal areas. Changes in precipitation years will affect living standards. It uses weather patterns have been more mixed, and this data from global climate models to predict diversity will persist in the future. These changes in average weather at the local level. weather changes are expected to result in a The book analyzes these climate data in combi- decrease in living standards in most countries nation with household surveys to explain how in the region, compared with a situation in changes in average weather will affect living which current weather conditions are standards. preserved. Research on the effects of climate change In the coming decades, changes in average has focused mostly on the immediate shocks of weather will have a clearly negative effect on extreme events, such as major storms, droughts, living standards in Bangladesh, India, Pakistan, and floods. Valuable insights have also been and Sri Lanka. Overall, inland areas will be gained on the effects of sea-level rise. This book more severely affected than those near the complements the existing body of knowledge coast. In India and Pakistan, water-stressed by providing granular analyses of projected areas will be more adversely affected compared changes in average weather. It shows how these with the national average. changes in average weather conditions will dif- Many parts of Afghanistan and Nepal are fer across regions. relatively cold at present, so warming will not Furthermore, the book analyzes how have a negative effect on living standards in living standards, measured by per capita con- these countries. In addition, climate change sumption expenditures, will be affected by may increase precipitation in Afghanistan, these changes in average weather. Analyses of which is predicted to have a positive effect. 1 2  SOUTH AS I A ’ S HOTS P OTS These predicted positive effects do not groups. The hotspots analysis contained account for the projected negative effects of herein can serve as a development blue- natural disasters and extreme events, to which print by providing region-specific insights these countries are highly vulnerable, accord- on the effects of these changes and ways to ing to other studies. adapt. Scenarios representing atmospheric emis- The analyses in the full book complement sions of greenhouse gases (GHGs) and their a body of well-documented work on emer- associated atmospheric concentrations are gency response and disaster preparedness, referred to as representative concentration with a view to informing long-term develop- pathways (RCPs). The trends and impacts ment planning to build climate change resil- just described will occur under both a ience. The findings can help governments, aid c limate-sensitive scenario (RCP 4.5) and ­ agencies, and others involved in development a carbon-intensive scenario (RCP 8.5). In the efforts expand beyond policies to tackle natu- former, some collective global action is under- ral disasters and vulnerability of coastal areas. taken to reduce the GHG emissions that are a At the regional level, the book shows the major cause of climate change. In the latter certainty of adverse long-term effects in South scenario, the assumption is that there is no Asia under all climate change scenarios. global action. Adverse effects on living stan- Smaller effects under the climate-sensitive sce- dards in South Asia are greater under the nario emphasize the need for nations to work carbon-intensive scenario. together to reduce GHG emissions, as called The current research adds to the under- for by the Paris Agreement of 2015. The link standing of the effects of climate change between climate effects and living standards, through a more granular approach that yields especially among the poorest populations, predictions at the district level. The book provides an economic argument for stronger identifies “hotspots”— districts where rising mitigation efforts. average temperatures and changing precipita- At the local level, the hotspot analyses tion patterns will have a notable negative provide guidance for decision makers in effect on living standards. Almost half of South Asia on where to focus investments South Asia’s population now lives in areas that increase resilience to the effects of that are projected to become moderate to changes in average weather. Investing now severe hotspots under the ­ carbon-intensive in building resilience will equip populations scenario. in South Asia that are particularly vulnera- The book uses granular information from ble to climate change with the needed tools the South Asia Spatial Database to examine and resources to break the downward spiral the characteristics of the hotspots and of the of poverty and inequality, helping them households that are located in them. The become drivers of growth and sustainable analyses reveal that hotspots tend to be more development. For example, prioritizing disadvantaged districts, even before the effects investments in climate resilience based on of changes in average weather are felt. needs identified by the book’s hotspots Hotspots are characterized by low household modeling can get resources to where they consumption, poor road connectivity, limited will be most needed in coming decades. The access to markets, and other development research discusses how specific actions— challenges. such as moving people out of agriculture, This level of granularity provides new increasing educational attainment, and pro- awareness of how effects will differ from viding access to e ­ lectricity—could ease the country to country and from district to dis- decline in living standards caused by trict throughout the region. Such granular- changes in average weather. The research ity increases the ability of decision makers also points out that the actions with the to focus resilience-building efforts on the greatest potential to make a difference vary most vulnerable locations and population across countries and locations. O v e r v i e w   3 A Vulnerable Region MAP O.1  Temperatures Have Been Increasing in Much of South Asia South Asia is recognized as being very vulnera- ble to climate change. The region’s varied geog- raphy combines with regional circulation patterns to create a diverse climate. The glaci- ated northern parts—which include the Himalayas, Karakoram, and Hindu Kush mountains—have annual average temperatures at or below freezing, whereas much of the Indian subcontinent averages 25°C to 30°C (77°F to 86°F). Both the hot and cold extremes are challenging for human well-being, and cli- mate change heightens these challenges. Increasing average temperatures and changes in seasonal rainfall patterns are already having an effect on agriculture across South Asia. Low-lying Bangladesh and the Maldives are increasingly vulnerable to flood- ing and cyclones in the Indian Ocean. The sci- entific literature suggests that such events will grow in intensity over the coming decades. Sources: Mani et al. 2018; data from Harris et al. 2014. Note: Changes are based on trend analysis between 1950 and 2010. Dhaka, Karachi, Kolkata, and Mumbai— urban areas that are home to more than standards. In this analysis, household con- 50 million ­ people—face a substantial risk of sumption expenditures are used as a proxy flood-related damage over the next century. for living standards. Average annual temperatures throughout Rising average temperatures can affect living many parts of South Asia have increased sig- standards through diverse pathways, such as nificantly in recent decades, but unevenly agricultural and labor productivity, health, (map O.1). Western Afghanistan and south- migration, and other factors that affect eco- western Pakistan have experienced the largest nomic growth and poverty reduction increases, with annual average temperatures (figure O.1). They can dampen agricultural ­ rising by 1.0°C to 3.0°C (1.8°F to 5.4°F) from productivity, leading to a decline in living stan- 1950 to 2010. Southeastern India, western dards for agriculture-dependent households. Sri Lanka, northern Pakistan, and eastern A warmer climate can also increase the Nepal have all experienced increases of 1.0°C propagation of vector-borne and other infec- to 1.5°C (1.8°F to 2.7°F) over the same period. tious diseases, resulting in lost productivity and The precise magnitude of the estimated tem- income. At the same time, a warmer climate perature changes varies across locations, but can increase productivity in historically colder the direction of the changes is unambiguous. regions, such as mountainous areas. Days of extreme heat are generally corre- lated with lower worker productivity, espe- Climate Change and Living cially in areas that are already warm. A Standards changing climate can force people out of their Climate change includes rising temperatures, traditional professional domains, resulting in changing precipitation patterns, and intensi- individuals not earning as much income. fying extreme events, such as storms and Previous research on climate change in droughts. All these have profound repercus- South Asia and associated policy prescriptions sions for societies, from sudden economic dis- has focused on disaster-resilient infrastructure ruptions to a longer-term decline in living and emergency responses, such as building 4  SOUTH AS I A ’ S HOTS P OTS FIGURE O.1  Increases in Temperatures and Changes in survey data for Pakistan are designed to rep- Precipitation Patterns Are Linked to Living Standards resent provincial conditions, whereas survey through a Diverse Set of Pathways data for India can show district conditions. The book focuses on the impact of changes Health in average weather on living standards. Such changes in averages can be projected with greater confidence than changes in extreme Agriculture Increases Living events. Although extreme events cause major in temperature standards disruptions to consumption, they generally and changes in (consumption precipitation Migration expenditures) are of relatively short duration, and con- sumption bounces back after relief and reha- bilitation efforts have been undertaken. In Productivity contrast, the effects of long-term changes in climate, such as average temperatures and Source: Mani et al. 2018. precipitation patterns, are recurring and will require adaptation to overcome. The book uses household consumption cyclone shelters and coastal embankments. expenditures as a metric that expresses the There has also been a focus on strengthening monetary dimensions of living standards early warning systems in areas that are highly because it is objectively quantifiable. It is well vulnerable to flooding, storm surges, and sea- understood that nonmonetary dimensions of level rise. The benefit from these investments is well-being matter as well. However, the focus to reduce the economic shocks associated with on per capita consumption expenditures extreme weather events. makes the analyses in this book consistent However, little effort has been made to with the literature on poverty and inequality. understand the diverse effects of changes in There is a wide range of model formula- average weather. These effects could be sub- tions that could potentially be used to estimate stantial, given the implications of weather con- the relationship between weather and living ditions for agricultural productivity, health, standards. Similar to previous studies, this migration, and other factors. Addressing this research uses a reduced-form model. Reduced- knowledge gap is important. Increasing evi- form models do not make assumptions about dence shows that changing temperatures and the channels through which external factors seasonal precipitation patterns have already such as weather affect living standards, and altered the growing seasons of regions in cannot provide a causal analysis. Instead, these Bangladesh, India, and Pakistan, and have models seek to capture the aggregate relation- resulted in serious health and productivity ship between external factors and ­ outcomes— damage (Burke, Hsiang, and Miguel 2015). which, in this case, are changes in average Less understood are the economic implications weather and living standards. of these long-term changes for households and The book confirms that there is an communities. optimal temperature range that is correlated The book adds to the accumulated knowl- with higher consumption expenditures rela- edge on climate change in South Asia through tive to locations where temperatures a combination of spatially granular weather are either hotter or colder (­figure O.2). The data and statistical household analyses. The overall relationship is similar between coun- weather data are derived from predictions tries, but the optimal temperature differs. from global climate models that are especially This indicates that there may be some relevant for South Asia. The household sur- ability for countries to adapt to long-term veys are designed to be representative of con- changes in temperature. Nationally, temper- ditions at different levels of administrative atures in Bangladesh, India, Pakistan, aggregation, varying by country. For example, and Sri Lanka are already above their O v e r v i e w   5 FIGURE O.2  Temperature and Consumption Have an Inverted U–Shaped Relationship for Countries in South Asia a. Afghanistan b. Bangladesh 0 0 –1 –5 Consumption change (%) Consumption change (%) –2 –10 –3 Average temperature Average temperature –15 –4 –20 –5 –25 0 5 10 15 20 25 22 23 24 25 26 27 28 Annual average temperature (°C) Annual average temperature (°C) c. India d. Nepal 0 0 –10 Consumption change (%) Consumption change (%) –5 –20 Average temperature Average temperature –30 –10 –40 –50 –15 0 5 10 15 20 25 30 0 5 10 15 20 25 Annual average temperature (°C) Annual average temperature (°C) e. Pakistan f. Sri Lanka 0 0 –5 Consumption change (%) Consumption change (%) –10 –5 –15 Average temperature Average temperature –20 –10 –25 –30 –15 –35 5 10 15 20 25 30 24 25 26 27 28 29 Annual average temperature (°C) Annual average temperature (°C) Source: Mani et al. 2018. Note: Blue-shaded region indicates 90 percent confidence interval. 6  SOUTH AS I A ’ S HOTS P OTS optimal values. This means that at the number means greater overall emissions and national level, any further increase in aver- atmospheric concentrations—and therefore age temperature will have a negative effect the potential for more severe climate change. on consumption expenditures. Temperatures The 2015 Paris Agreement on climate in Nepal are still less than the inflection change sets a target of limiting average global point, meaning that increases in tempera- temperature increases to 2°C (3.6°F) relative tures are predicted to have positive effects to preindustrial conditions. RCP 4.5 repre- on consumption. Nationally, Afghanistan is sents a future in which some collective action close to its optimal temperature; however, is taken to limit GHG emissions, with global consumption expenditures are less sensi- annual average temperatures increasing 2.4°C tive to temperature in Afghanistan than (4.3°F) by 2100. Therefore, the book labels in the other countries analyzed. RCP 4.5 as a “climate-sensitive” development scenario. RCP 8.5 is closer to a scenario in which no actions are taken to reduce emis- Climate Modeling and Effects sions, and global annual average tempera- The primary driver of climate change is tures increase 4.3°C (7.5°F) by 2100. The GHG emissions, with human-caused emis- book labels RCP 8.5 as a “carbon-intensive” sions as the major contributor. Projecting development scenario. future climatic changes requires creating a Global climate models are the primary scenario that projects the amount, timing, tool for projecting how a given RCP scenario and type of future GHG emissions by will affect the Earth’s climate. Climate mod- human activities. els are designed to approximate fundamental The international community has devel- laws of physics, modeling interactions oped multiple scenarios to account for uncer- between the atmosphere, land, and oceans. tainty about the path the world will take. This research considers 18 global climate Scenarios representing atmospheric emissions models covered by the Climate Model of GHGs and their associated atmospheric Intercomparison Project (CMIP5), and concentrations are referred to as RCPs. This assesses their performance in reproducing book uses ­climate projections corresponding historic weather patterns observed in South to RCPs 4.5 and 8.5. With RCPs, a higher Asia. On the basis of this performance crite- rion, 11 models are selected that perform best. The research uses these 11 climate mod- els to project long-term changes in average temperature and precipitation throughout FIGURE O.3  Annual Temperature Increases Are Projected to Accelerate South Asia. The average prediction by these climate 25.5 Historic time series models is that annual average temperatures in 25.0 Historic average (1981–2010) South Asia will increase 1.6°C (2.9°F) by 24.5 Climate-sensitive (RCP 4.5) 2050 under the ­ climate-sensitive scenario, Carbon-intensive (RCP 8.5) and 2.2°C (3.9°F) under the carbon-intensive Temperature (°C) 24.0 Historic trend (0.14˚C/decade) scenario. These increases are relative to 1981– 23.5 2010 conditions (figure O.3). 23.0 Projected changes in precipitation are 22.5 highly uncertain, in part because they are 22.0 heavily dependent on cloud microphysics, 21.5 which are difficult to represent in current 21.0 global climate models. The average climate 1960 1970 1980 1990 2000 2010 2020 2030 2040 2050 model prediction is that average monsoon Sources: Mani et al. 2018; data from Harris et al. (2014) and 11 climate models. precipitation will increase 3.9 percent under Note: RCP = representative concentration pathway. the climate-sensitive scenario and 6.4 percent O v e r v i e w   7 MAP O.2  Annual Average Temperatures Increase by 2050 Relative to 1981–2010 Source: Mani et al. 2018. Note: Changes are for 2036 through 2065 relative to averages for 1981 through 2010. under the carbon-­intensive scenario by 2050 Nepal, and high-elevation areas of India. For (figure O.4). example, people in the mountain regions rely If average precipitation increases, some extensively on streamflow from snow and gla- areas that have historically experienced low ciers. Warming will affect the timing and avail- rainfall could benefit. It is also likely that ability of water resources, which could have extreme precipitation events will become profound effects. In addition, mountain regions more common, especially because of the may be less resilient to natural disasters. large simultaneous temperature increases. Extreme precipitation events would cause FIGURE O.4  Monsoon Precipitation Varies Considerably and an increase in damage and economic dis- Projections Are Uncertain ruption, whereas decreasing precipitation Historic time series 210 would result in less overall water availabil- Historic average (1981–2010) ity in South Asia, which would reduce agri- 200 Climate-sensitive (RCP 4.5) Precipitation (mm/month) cultural yields and water security in some 190 Carbon-intensive (RCP 8.5) areas (map O.2, panels a and b). 180 The book shows that failure to reduce 170 GHG emissions and take measures to build 160 climate change resilience will lead to dimin- ished economic performance in most South 150 Asian countries. At the same time, changes in 140 average weather may have some benefits for 130 Afghanistan, Nepal, and high-elevation areas of India because of their cold climates. 1960 1970 1980 1990 2000 2010 2020 2030 2040 2050 However, not all effects of increasing tem- Sources: Mani et al. 2018; data from Harris et al. (2014) and 11 climate models. peratures will be positive in Afghanistan, Note: RCP = representative concentration pathway. 8  SOUTH AS I A ’ S HOTS P OTS Changes in average weather are predicted This diversity in the way that living stan- to reduce living standards in Bangladesh, dards react to changing weather conditions India, Pakistan, and Sri Lanka, relative to can be interpreted as implicitly capturing the what they would have been with the same cli- effects of differences in institutional settings, mate as today. By 2050, under the carbon- economic structures, and policy frameworks intensive scenario the declines are projected across countries. The diversity may also to be 6.7 percent for Bangladesh, 2.8 percent reflect differing degrees of adaptive capability for India, 2.9 percent for Pakistan, and by households and communities to weather 7.0 percent for Sri Lanka. conditions. For countries with severe hotspots— Hotspots are labeled mild when projected Bangladesh, India, and Sri Lanka—the consumption spending declines by less than negative impacts are predicted to be even ­ 4 percent, moderate for declines of 4 percent greater. Translated into gross domestic to 8 percent, and severe for declines exceed- product (GDP) per capita, changes in aver- ing 8 percent. age weather are predicted to reduce income Hotspots are primarily predicted to occur in severe hotspots by 14.4 percent in in Bangladesh, India, Pakistan, and Bangladesh, 9.8 percent in India, and Sri Lanka. In these countries, projected 10.0 percent in Sri Lanka by 2050 under changes in average weather are expected to the carbon-intensive scenario compared to result in an overall decrease in per capita the climate of today. consumption expenditures. The analyses do Climate effects are smaller under the not include Bhutan and the Maldives because climate-sensitive scenario. This finding high- ­ adequate climate projection data are not lights the importance of taking actions to available. reduce GHG emissions, and provides an addi- In general, hotspots tend to be less densely tional economic justification for continuing to populated and have poorer infrastructure, work toward meeting the targets established such as fewer roads, which hinder their inte- under the Paris Agreement. gration with the broader society. Inland areas are predicted to be more affected by projected changes in average weather than coastal areas Hotspots and mountainous regions (map O.3, panels a South Asian megacities—such as Chennai, and b). However, there are also hotspots in Dhaka, Karachi, Kolkata, and Mumbai—are some areas where precipitation is expected to often said to be climate hotspots because they increase, such as southeast Bangladesh, parts are vulnerable to extreme events and sea-level of the Kashmir Valley, and the southern tip rise, including coastal flooding and storm of India. surges. In this book, however, hotspots Identifying hotspots is not as simple as are defined as areas where changes in average finding the regions where changes in average weather will adversely affect living weather are projected to be the largest. Even standards. if climate were to change by similar magni- Hotspots are the result of two interrelated tudes in two locations, the response would factors: (a) the magnitude of predicted depend on the historic relationship between changes in average weather at the local level; weather and living standards at the locations. and (b) the relationship between weather and For example, if two countries are identical living standards in that location. The magni- except that one relies more heavily on agricul- tude of predicted changes in average weather ture, the magnitude of impact attributable to is estimated using global climate models. weather during the growing seasons can be The relationship between weather and living expected to be larger in the agriculture-heavy standards is estimated using country-specific country, all other factors held equal. The household surveys and is therefore different same principle applies to other pathways by across countries. which weather impacts living standards O v e r v i e w   9 MAP O.3  Severe Hotspots Will Cover a Significant Portion of South Asia by 2050 Source: Mani et al. 2018. as well. For example, Ramanathapuram mitigation efforts to minimize the effects of District in India and Jaffna District in Sri climate change, such as reducing GHG emis- Lanka are ­ separated by about 100 kilometers sions, can positively affect living standards and have relatively similar weather. However, throughout the region. Ramanathapuram does not emerge as a Because climate impacts vary from region hotspot in the analysis, whereas Jaffna to region, the hotspots provide a blueprint for emerges as a moderate to severe hotspot. prioritizing investments and actions to build Under the carbon-intensive scenario, doz- resilience. A look at where changes in average ens of inland hotspots in the center of South weather are predicted to impact living stan- Asia would shift from moderate in 2030 to dards in individual countries reveals the diver- severe by 2050. Coastal areas do not gener- sity of findings from the analyses. ally experience this additional deterioration in In Bangladesh, Chittagong Division is the living standards. However, they could be neg- most vulnerable to changes in average atively affected by other consequences of cli- weather, followed by Barisal and Dhaka mate change, such as sea-level rise and a likely divisions. Chittagong is relatively more increase in storms and other extreme events. developed in terms of infrastructure com- Overall, more than half the region will be a pared with the national average, and is also hotspot by 2050 under the carbon-intensive characterized by fewer households engaged scenario, with 45 percent of the present popu- in agriculture. However, the area includes lation of South Asia—800 million people— hill tracts, which are vulnerable to changes living in areas projected to become moderate in average weather. Over the years, the or severe hotspots. Under the climate-sensi- Chittagong hill tracts have experienced out- tive scenario, the number of people affected breaks of vector-borne diseases and defores- would be 375 million, or 21 percent of the tation that have resulted in major landslides population. This finding demonstrates that and the destruction of property. 10  SOUTH AS I A ’ S HOTS P OTS In India, inland states in the central, Toward Greater Resilience northern, and northwestern regions emerge as the most vulnerable to changes in aver- At the highest level, an agenda for building age weather. Chhattisgarh and Madhya resilience includes sustaining economic Pradesh—which are predicted to have a growth and ensuring shared prosperity. ­ l iving standards decline of more than Development is indeed the best adaptation 9 percent—are the top two hotspot states, strategy, since it is associated with improved followed by Rajasthan, Uttar Pradesh, and infrastructure, market-oriented reforms, Maharashtra. Chhattisgarh and Madhya enhanced human capabilities, and a stronger Pradesh are also low-income states, home institutional capacity to respond to the to large tribal populations. Changes in increasing threat of natural disasters. But the average weather could therefore have agenda must also include creating an incen- important implications for poverty tive framework for private action, committing reduction. public resources to mitigation and adapta- In Sri Lanka, the Northern and North tion, and prioritizing spending. Western provinces emerge as the top two The full book identifies and highlights cli- hotspots, followed by the much less densely mate hotspots where communities and house- populated North Central Province. The holds are likely to be particularly vulnerable Northern Province is home to large numbers to changes in average weather. An important of poor and displaced people, so the effects finding of the book is that focusing location- of changes in average weather will add to specific resilience-building efforts on the most these ­ c hallenges. The North Western vulnerable areas and population groups can Province, in turn, is one of the driest regions reduce hotspots. of Sri Lanka. The public sector can help build resilience The highly urbanized and densely popu- among these communities through actions lated Western Province, which includes that support adaptation, such as helping Colombo, is also predicted to experience a develop drought-resistant crops and provid- living standards decline of 7.5 percent by ing weather forecasts and climate risk assess- 2050, compared with a situation without ments. In addition, the public sector can changes in average weather. This is a substan- establish a policy framework for adaptation tial drop, with potentially large implications that creates incentives for private action, for the country, given that the province con- including (a) regulatory and insurance tributes more than 40 percent of Sri Lanka’s instruments that convey the correct incen- GDP. tives for adaptation; (b) pricing and other In Pakistan, Sindh Province emerges as policies that encourage the efficient use of the most vulnerable hotspot, followed by energy, water, agriculture, and other natural Punjab. Sindh has the second-largest econ- resources; and (c) facilitating market access omy in the country. Its GDP per capita is 35 and providing fiscal incentives for research percent above the national average, and and development to exploit existing technol- contributes around 30 percent of Pakistan’s ogies or develop new ones in the energy, GDP. The province’s highly diversified water-supply, agricultural, forestry, and live- economy ranges from heavy industry and stock sectors. finance in and around Karachi to a substan- No single set of interventions will work in all tial agricultural base along the Indus River. hotspots. For example, inland areas in India Punjab, which is the most densely popu- emerge as severe hotspots, whereas in Sri Lanka, lated province, has the largest economy in the postconflict northern coastal areas are most Pakistan. It contributes 53 percent of the vulnerable. The household characteristics of country’s GDP and is known for its relative these areas also differ from one another, so inter- prosperity. ventions must be tailored to the specific context. O v e r v i e w   11 Understanding these diverse effects is critical to In the future, economic growth and struc- help countries design appropriate policies for tural changes will cause people to migrate to building long-term resilience in communities cities, leaving behind their agricultural and and households. other climate-sensitive practices in rural areas. The book investigates specific investment Although this could potentially make more of and policy options that countries could con- the population climate-resilient, urban migra- sider to attenuate or offset the the negative tion also will create new climate impacts. impacts of projected changes in average Urban populations will face a number of weather. health risks exacerbated by events such as For Bangladesh, the analysis suggests heat waves and flooding. that enhancing opportunities in the nonag- Another challenge is to ensure that resil- ricultural sector could potentially reduce ience strategies and actions are inclusive, to the effect of changes in average weather on avoid inequality in growth and opportunity. living standards. A 15 percent increase in The projected emergence of many moderate nonagricultural employment would attenu- and severe hotspots under the carbon-inten- ate the effect of weather changes from –6.7 sive scenario shows the need for resilience percent to –1.4 percent. Similarly, a 30 policies to target impoverished populations percent increase in the share of nonagricul- and highly vulnerable regions. tural employment would not only reduce It is worth noting that, although fraught the negative effect of changes in average with risks, changes in average weather pres- weather but would also result in increased ent opportunities for households, communi- living standards. ties, and nations. Decisions about adaptation In India, the analyses discuss three options: strategies, developing skills, and engaging increasing educational attainment, reducing with the communities will determine the qual- water stress, and expanding the nonagricul- ity of life of the next generation and beyond. tural sector. The analyses predict that increas- ing the average educational attainment by 1.5 References years would reduce the magnitude of decline in living standards from –2.8 percent to –2.4 Harris, I. P. D. J., P. D. Jones, T. J. Osborn, and D. percent. Reducing water stress by 30 percent, H. Lister. 2014. “Updated High-Resolution Grids of Monthly Climatic Observations: The and increasing employment in nonagricul- CRU TS3.10 Dataset.” International Journal of tural sectors by the same percentage, would Climatology 34 (3): 623–42. yield similar benefits. IPCC (Intergovernmental Panel on Climate In Pakistan, the analyses reveal that Change). 2013. Climate Change 2013: The expanding electricity access by 30 percent Physical Science Basis—Contribution of above current levels would reduce the living Working Group I to the Fifth Assessment standards burden from –2.9 percent to –2.5 Report of the Intergovernmental Panel on percent. Climate Change . Cambridge: Cambridge In Sri Lanka, increasing the share of the University Press. nonagricultural sector by 30 percent relative Mani, M., S. Bandyopadhyay, S. Chonabayashi, to current levels would change the sign of the A. Markandya, and T. Mosier. 2018. South Asia’s Hotspots: Impact of Temperature living standards impact from –7.0 percent to and Precipitation Changes on Living Standards. 0.1 percent. Reducing travel time to markets South Asia Development Matters. Washington, and increasing average educational attain- DC: World Bank. ment would also ease negative impacts on Taylor, K. E., R. J. Stouffer, and G. A. Meehl. living standards. If implemented together, 2012. “An Overview of CMIP5 and the such interventions would likely yield signifi- Experiment Design.” Bulletin of the American cantly positive climate cobenefits. Meteorological Society 93 (4): 485–98. A Vulnerable Region 1 C limate change is already a pressing The symptoms of climate change are multi- issue for South Asia. Temperatures faceted, including sea-level rise, shifts in aver- have been rising across the region, age temperature and precipitation patterns, and are projected to continue increasing for and increasing frequency of extreme events the next several decades under all plausible such as storms and droughts. These climatic climate scenarios (IPCC 2014). Precipitation changes have profound effects on societies, response to global emissions is more difficult such as greater frequency of flooding events, to estimate. There is evidence, though, that more year-to-year variability in agriculture historic precipitation patterns are changing, productivity, a greater demand for water and that these changes will become stronger (which may be more difficult to meet), and and less predictable. increased instances of heat-related medical South Asia is recognized as being highly problems. Furthermore, these and other cli- vulnerable to climate variability and change mate change impacts will cause economic dis- (map 1.1). Increasing average temperatures ruption in South Asia, with the effects and changing seasonal rainfall patterns are continuing to grow over time (figure 1.1). already affecting agriculture across the region. These have been well articulated in various Low-lying Bangladesh and the Maldives are on IPCC reports and country studies. the global front line of countries at risk for sea- The effect of extreme events and sea-level level rise—a result of glacier melt induced by rise is clear and well documented in the climate change—and increasing vulnerability region (Bronkhorst 2012; Dasgupta and oth- to flooding and cyclones in the Indian Ocean. ers 2015; Hallegatte and others 2017; Sanghi Major cities such as Dhaka, Karachi, Kolkata, and others 2010). Quantification of hazard, and Mumbai—which are home to more than exposure and vulnerability (which together 50 million people and growing—face the define “climate change risk”) is now a robust greatest risk of flood-related damage over the science; the economic consequences are exten- next century. In addition, extreme temperature sively studied, but many uncertainties remain. events such as the 2015 heat wave that killed Hallegatte and others (2017) move beyond more than 3,500 people also threaten the asset and production losses to focus on how region. There are many such areas and regions natural disasters affect people’s well-being. By in South Asia that are extremely vulnerable to examining well-being instead of asset losses, climate change impacts. their book provides a deeper view of natural 13 14   SOUTH ASIA’S HOTSPOTS MAP 1.1  South Asia Remains a Region Very Vulnerable to Climate Change and Extreme Events Source: Maplecroft Climate Vulnerability Index 2017. Note: This index evaluates 42 social, economic, and environmental factors to assess national vulnerabilities across three core areas. These include exposure to climate-related natural disasters and sea-level rise; human sensitivity in terms of population patterns, development, natural resources, agricultural dependency, and conflicts; and assessing future vulnerability by considering the adaptive capacity of a country’s government and infrastructure to combat climate change. FIGURE 1.1  Some Manifestations of Climate Change disasters that takes better account of people’s vulnerability. With this lens, they find that poor people are significantly more impacted Changes in average by natural disasters than nonpoor people. Impacts on agriculture productivity, temperature health, labor productivity, and migration Much of the focus related to climate and precipitation change in the region has been on emergency response, including the building of cyclone Changes in shelters and coastal embankments as well characteristics Loss of property, assets, and human life as strengthening early warning systems in of extreme areas highly vulnerable to flooding, storm events surges, and sea-level rise. This is an important course of action because the effects of extreme Sea-level Coastal erosion and ooding, asset events are significant from an economic rise damage, and habitat loss perspective. These investments help increase immediate political capital and have medium- to long-term benefits associated with reducing A V u l n e r a b l e R e g i o n    15 the adverse economic impact of future MAP 1.2  South Asia Continues to Be Home to a extreme events. Large Number of Poor People This book looks at the impact of long-term changes in average temperature and precipita- tion. “Long-term” changes in the average spe- cifically refers to changes in the mean of 30 consecutive years of weather for a given parameter. The 30-year mean can be calcu- lated for a specific season or to represent annual conditions. Throughout the book, “average weather” is used to refer to long- term changes in average seasonal temperature and precipitation. There has been little effort to understand the spatial heterogeneity of changes in average weather—specifically, increases in tempera- ture and changes in precipitation patterns— and their implications for agricultural productivity, health, and migration. This is an important knowledge gap since increasing evi- dence suggests that these gradual changes are already disrupting the growing season for Source: Household survey data (see table 3.1). areas in Bangladesh, India, and Pakistan, and Note: PPP = purchasing power parity. are causing serious health damage and pro- ductivity losses (Burke, Hsiang, and Miguel 2015). The economic implications of these subnational levels (table 1.1). Furthermore, changes in average weather for households all countries in South Asia have pledged to and communities–and their possible contribute to global emissions reductions thresholds or inflection points–are even less under the Paris Agreement through their sub- understood. This book aims to estimate these mitted intended national contributions.1 effects, focusing on changes in average Going forward, adapting to long-term cli- weather and its impacts on household living matic shifts such as increasing temperatures standards across the region. and changing seasonal precipitation patterns will involve a portfolio of actions—from improving infrastructure to introducing mar- Progress So Far ket reforms and building household and insti- The South Asia region has recently witnessed tutional capacity. Given that such actions favorable economic growth and is gearing up will incur a cost, there will inevitably be to capitalize on opportunities provided by trade-offs; therefore, governments must pri- urbanization, economic diversification, and oritize efforts. In addition to internal a young population. At the same time, the resources, both international public and pri- region is also home to one-third of the world’s vate funds and resources will be needed to poor (map 1.2). build resilience. Decisions must be made in Countries in the region now better under- some cases with incomplete information, and stand that adapting to climate change and all countries will face the dilemma of either building resilience are essential courses of not taking early action—with the risk of action to sustain the benefits of their growing incurring very high future costs—or acting economies. They recognize climate change as early on—when the pressure on public and a national priority and have been formulating private resources is intense—and eventually strategies and action plans at the national and realizing the actions were redundant. 16  SOUTH AS I A ’ S HOTS P OTS TABLE 1.1  Climate Change Strategies and Action Plans of Countries in South Asia Country National Strategy, Policy, or Action Plan Year Afghanistan Intended Nationally Determined Contribution 2015 Bangladesh Bangladesh Climate Change Strategy and Action Plan 2009 Intended Nationally Determined Contribution 2015 Bhutan National Adaptation Plan 2015 Intended Nationally Determined Contribution 2015 India National Action Plan on Climate Change 2009 Intended Nationally Determined Contribution 2015 Maldives, The Strategic National Action Plan for Disaster Risk Reduction and Climate Change Adaptation 2010 Intended Nationally Determined Contribution 2015 Nepal National Framework on Local Adaptation Plans for Action 2011 Intended Nationally Determined Contribution 2015 Pakistan National Climate Change Policy 2012 Intended Nationally Determined Contribution 2015 Sri Lanka National Adaptation Plan for Climate Change Impacts in Sri Lanka 2015 Intended Nationally Determined Contribution 2015 The best approach is for countries to gather The book uses granular spatial information the existing information about the most likely from the World Bank’s South Asia Spatial causes of the problem and assess the pros and Database and national household surveys to cons—costs and benefits—of alternative understand the characteristics of hotspots at actions. There is, therefore, a pressing need to the household and district levels. The infor- provide decision makers with the economic mation will be useful from a national perspec- rationale for investing in resilience to changes tive (for example, when designing a social in average weather as part of their adaptation welfare program) as well as a local one (for strategies. In addition, governments require example, determining which investments information on the types of interventions that would be most needed in each community, will build resilience and the locations where accounting for local socioeconomic character- the investments are most needed. istics and climate-related risks). Detailed analyses are carried out for Afghanistan, A Road Map for Climate-Resilient Bangladesh, India, Nepal, Pakistan, and Development Sri Lanka—the South Asian countries for which the necessary household survey and This book identifies climate hotspots, defined climate data are available. as locations that will be adversely affected by The objective is to investigate the spatial changes in average weather. The book does not patterns of historic and projected changes focus on short-term manifestations of climate in average weather across South Asia and change, such as extreme events, or slow onset their effects on living standards. To this events, such as glacier melt or sea-level rise. end, the book attempts to answer three However, the analysis does capture some long- specific questions related to changes in term effects of changes in extreme temperature average weather: and precipitation (box 1.1). Overall, the analy- sis builds on the existing well-documented •  What changes in average temperature and work on emergency response and disaster pre- precipitation will occur in different loca- paredness with a view to inform long-term tions across South Asia? development planning, public sector programs, •  How will these changes affect living and public and private sector projects. standards? A V u l n e r a b l e R e g i o n    17 BOX 1.1  Why Do Changes in the Average Weather Matter? The term climate change refers to changes in the FIGURE B1.1.1  Increased Average Temperature frequency or magnitude of weather. Climate Causes Increased Likelihood of Extreme Heat Events change therefore encapsulates a wide variety of PDF (28°C) phenomena, including changes in average tem- 34.6°C 37.6°C 4.5 PDF (31°C) perature and precipitation and changes in the fre- 95th percentile (28°C) 4.0 95th percentile (31°C) quency or severity of extreme events (such as 3.5 tropical storms or heat waves). Often, much of Probability (%) 3.0 the attention related to climate change is on extreme events and sea-level rise, which are more 2.5 immediately visible through the profound effects 2.0 these events have on communities. From a long- 1.5 term perspective, both changes in the average 1.0 weather and extreme events matter. 0.5 In many cases, changes in extreme events can 0 be explained through changes in average 15 20 25 30 35 40 weather. As a practical example, over the past Temperature (°C) 30 years there has been an increasing number of Source: World Bank calculations. heat-related deaths in South Asia and around Note: The blue-shaded region represents the probability that the 95th percentile will be exceeded for the normal distribution having a mean of 28°C, and the the world. The increase has been driven by more red-shaded region represents the additional probability of exceedance for this frequent, longer, and more intense heatwaves magnitude event when the normal distribution is shifted such that the mean is 31°C. The standard deviation used to produce these probability distribution during the summer. While heatwaves are an functions is 4°C. PDF = probability distribution function. extreme event, their changes are explained well through analyzing shifts in the average distribu- Any rise in the average temperature could tion of temperature (McKinnon and others thus potentially lead to a rise in the number of 2016). Similarly, the analyses in this book, which days that are extremely hot. This increase in focus on shifts in average weather, are able to heat has repercussions for a myriad of sectors, capture some changes in extreme events. including health, farming, and energy systems. The concept that changing averages can cap- More extreme heat raises the risk of heat-related ture changes in extreme weather is demon- illnesses, such as heat exhaustion, and allows strated visually in figure B1.1.1. In this figure, it insects to move into new areas, potentially is supposed that the average temperature of a increasing the spread of vector-borne diseases. location is 28°C (represented by the blue solid It could also stress crops accustomed to a milder line) and that climate change shifts the mean climate and worsen drought conditions. In temperature +3°C (represented by the red solid addition, extreme heat is associated with air line). Assuming the shape of the underlying stagnation, which could trap pollutants and probability distribution remains constant, this worsen respiratory illnesses such as asthma. shift would increase the likelihood that temper- Similarly, shifting the average of the precipita- ature exceeds 34.6°C (the assumed 95th percen- distribution would mean a greater likelihood tion ­ tile of the 28°C distribution) by 13 percent. This of no precipitation or extreme precipitation, increased likelihood is represented by the red- corresponding to an increasing likelihood of shaded area. droughts or flooding, respectively. 18  SOUTH AS I A ’ S HOTS P OTS •  What are the characteristics of the places significant agreement to address climate and people most affected by these change since the issue first emerged as a changes? major political priority decades ago. Countries committed to keep global tempera- Understanding answers to these important tures from rising more than 2°C by 2100, questions will help countries and communi- with an ideal target of keeping temperature ties build resilience to changes in average rise less than 1.5°C. weather in the region through the following: •  Designing interventions across locations References that address the challenges posed by Bronkhorst, V. B. 2012. “Disaster Risk higher temperatures and uncertain long- Management in South Asia: A Regional term precipitation patterns Overview.” Working Paper 76302, World •  Helping local communities design social Bank, Washington, DC. protection programs that can build cli- Dasgupta, S., A. Zaman, S. Roy, M. Huq, mate resilience given the current spatial S. Jahan, and A. Nishat. 2015. Urban Flooding of Greater Dhaka in a Changing distribution of poverty through a location- Climate: Building Local Resilience to Disaster specific focus Risk. Directions in Development. Washington, •  Highlighting the costs, trade-offs, and DC: World Bank. opportunities for countries to build cli- Hallegatte, S., A. Vogt-Schilb, M. Bangalore, and mate resilience while realizing their J. Rozenberg. 2017. Unbreakable: Building the growth potentials Resilience of the Poor in the Face of Natural Disasters. Washington, DC: World Bank. The rest of this book is structured as IPCC (Intergovernmental Panel on Climate follows: chapter 2 provides an overview of Change). 2014. “Climate Change 2014: historic climate trends in the region, as well as Synthesis Report; Contribution of Working future projections, using extensive and sophis- Groups I, II and III to the Fifth Assessment ticated climate modeling; chapter 3 provides Report of the Intergovernmental Panel an analytical framework linking changes in on Climate Change.” IPCC, Geneva. average weather and living standards by Maplecroft Climate Vulnerability Index 2017. estimating their effects on household con- http://www.maplecroft.com/about/news/ccvi​ sumption; chapter 4 identifies future climate- .html. induced hotspots at the national and local McKinnon, K. A., A. Rhines, M. P. Tingley, and P. Huybers. 2016. “The Changing Shape of levels; and chapter 5 provides policy Northern Hemisphere Summer Temperature recommendations. Distributions.”  Journal of Geophysical Research: Atmospheres 121 (15): 8849–68. Sanghi, A., S. Ramachandran, A. de la Fuente, Note M. Tonizzo, S. Sahin, and B. Adam. 2010. 1. At the Paris climate conference (Conference Natural Hazards, Unnatural Disasters : The of Parties [COP21]) in December of 2015, Economics of Effective Prevention. 195 countries reached the world’s most Washington, DC: World Bank. Increasingly Hot 2 T emperatures have been rising in most climate that is already uncomfortably hot much parts of the globe, and South Asia is of the year. Both the hot and cold extremes are no exception (IPCC 2013). These tem- challenging for human well-being. perature increases are going to continue, with Precipitation patterns are similarly diverse, some variation based on location and the with portions of the region receiving as little level of global collective action taken to limit as 100 mm of average annual precipitation and greenhouse gas (GHG) emissions. As outlined others receiving nearly 5,000 mm (map 2.2). in chapter 1, these changes will have varied The South Asian monsoon typically occurs consequences. Preparing for climate change during the months of June through September impacts is critical and requires understanding and is the most important climatic feature in projections of changes over the horizons use- terms of effect on the region’s people. During ful to planners. For this purpose, this chapter the premonsoon season, temperatures are outlines the diverse historic climatic condi- typically the highest of any point during the tions across South Asia and develops esti- year. The onset of monsoon rains quickly mates of future changes based on an ensemble reduces temperatures to more comfortable of the most suitable climate models. levels and brings much of the year’s water, which facilitates agriculture. These water resources are close to fully used in many parts Highly Diverse Climate of South Asia, resulting in strong agricultural The geography of South Asia is extremely productivity (though not at its full potential), varied. This, combined with regional circula- but with high vulnerability to changes in tion patterns, leads to a highly diverse climate water supply or demand. across the region. The glaciated northern Too much water delivered too suddenly can parts—punctuated by the Himalayas, cause significant damage. For example, swaths Karakoram, and Hindu Kush mountains— of South Asia have experienced several cata- have annual average temperatures at or below strophic tropical storms and flooding. freezing (map 2.1). These areas, which contain Bangladesh is highly susceptible to flooding small villages, are much less densely populated because much of the country is close to sea because of the harsh conditions. In contrast, level and because the Ganges River and the much of the Indian subcontinent has average Brahmaputra River—two of South Asia’s three temperatures of 25°C to 30°C, resulting in a great rivers—drain through the country. 19 20   SOUTH ASIA’S HOTSPOTS MAP 2.1  Temperatures Vary Significantly across MAP 2.3  Temperatures Have Been Increasing in South Asia Most of South Asia Source: Harris and others 2014 (Climate Research Unit TS 2.24). Note: Linear trend in average annual temperature from 1951 through 2010. Areas showing 0°C change include locations where trends are not Source: Harris and others 2014 (Climate Research Unit TS 2.24). statistically significant. Note: Annual average for 1981 through 2010. MAP 2.2  Average Monsoon Precipitation in significantly, but unevenly, in recent decades. South Asia Generally Increases from West to East Western Afghanistan and southwestern Pakistan have experienced the largest increases, with annual average temperatures rising by 1.0°C to 3.0°C from 1950 to 2010 (map 2.3). Southeastern India, western Sri Lanka, northern Pakistan, and eastern Nepal have all also experienced increases of 1.0°C to 1.5°C over the same time frame. Although the precise magnitude of these esti- mated historic temperature changes varies depending on the time frame and the observa- tional data set, the fact that temperature changes have been occurring is unambiguous (figure 2.1, panels a through g). Changes in average precipitation are much harder to detect because of large year-to-year and interdecadal variability. From 1950 through 2010, statistically significant trends of increasing monsoon precipitation are found Source: Harris and others 2014 (Climate Research Unit TS 2.24). for parts of eastern Afghanistan and central Note: Average monsoon precipitation for 1981 through 2010. Pakistan, and decreasing monsoon precipita- tion for Uttaranchal and Uttar Pradesh in India, but no statistically significant trends for Unambiguous Historic Temperature other regions (map 2.4). Consequently, there Increases are contradictory scientific findings regarding Average annual temperatures throughout if and how precipitation is changing based on many parts of South Asia have increased analysis of station records. For example, I n c r e a s i n g l y H o t    21 FIGURE 2.1  Unambiguous Temperature Trends in South Asia a. Afghanistan b. Bangladesh 15.0 Historic time series 27.0 Historic time series Historic trend (0.27˚C/decade) Historic trend (0.09˚C/decade) 14.5 26.5 14.0 Temperature (°C) Temperature (°C) 13.5 26.0 13.0 25.5 12.5 25.0 12.0 11.5 24.5 1960 1970 1980 1990 2000 2010 1960 1970 1980 1990 2000 2010 c. Bhutan d. India Historic time series Historic time series 13.0 Historic trend (0.15˚C/decade) Historic trend (0.11˚C/decade) 25.5 12.5 Temperature (°C) Temperature (°C) 25.0 12.0 24.5 11.5 24.0 11.0 23.5 1960 1970 1980 1990 2000 2010 1960 1970 1980 1990 2000 2010 e. Nepal f. Pakistan Historic time series Historic time series 22.0 Historic trend (0.14˚C/decade) Historic trend (0.17˚C/decade) 14.0 21.5 Temperature (°C) Temperature (°C) 13.5 21.0 13.0 20.5 12.5 20.0 19.5 12.0 1960 1970 1980 1990 2000 2010 1960 1970 1980 1990 2000 2010 (continues next page) 22   SOUTH ASIA’S HOTSPOTS FIGURE 2.1  Unambiguous Temperature Trends in South Asia (continued) g. Sri Lanka Historic time series Historic trend (0.17˚C/decade) 28.0 Temperature (°C) 27.5 27.0 26.5 26.0 1960 1970 1980 1990 2000 2010 Source: Calculations based on CRU TS 3.22 (Harris and others 2014). MAP 2.4  No Overall Monsoon Precipitation from 1965 through 1980. Other studies that Trends for Most of South Asia are based on station records have found increasing monsoon precipitation trends in some parts of India and decreasing trends in others (Kumar and others 1992; Pal and Al-Tabbaa 2009; Roy and Balling 2004). The primary reasons for differences between these aforementioned results and those presented in this book are the data sets, time frames, and statistical tests used in the analyses. There is more robust evidence for changes in monsoon wet and dry spells and overall weakening caused by human activities. For most regions in South Asia, the monsoon patterns have remained constant (similar to the findings in map 2.4). However, the core region (western and central India) has experi- enced increases in both intensity of extreme wet periods and the frequency of dry periods (Singh and others 2014). The precise set of reasons for these changes is not known. It has Source: Harris and others 2014 (Climate Research Unit TS 2.24). Note: The change in monsoon precipitation is based on a linear trend been shown, though, that one contributing in the average monsoon precipitation from 1951 through 2010. Areas factor is an increase in the concentration of showing zero percent change include locations where trends are not statistically significant. human-produced aerosols in the region. These aerosols have caused an overall drying (or weakening) of the monsoon in recent on the basis of daily station records, Abbas decades (Bollasina, Ming, and Ramaswamy and others (2014) do not identify any statisti- 2011; Singh 2016). cally significant precipitation time trends in Pakistan, whereas Basistha, Arya, and Goel (2009) find an increasing trend in monsoon Projecting Future Climate precipitation for the Indian Himalayas from Climate change refers to long-term deviations 1902 through 1964 and a decreasing trend in the strength or frequency of weather events I n c r e a s i n g l y H o t    23 relative to a historic baseline. There are many physics (for example, conservation of energy, important aspects of the climate that may mass, and momentum) and interactions change, including for example: among the atmosphere, land, cryosphere, and oceans. The goal is to capture all the relevant •  30-year average annual or seasonal tem- processes governing the Earth’s climate. In peratures at a given location practice, these models have certain deficien- •  Number or average strength of tropical cies, such as not correctly accounting for storms cloud processes (Rosenfeld and others 2014). •  Timing of the onset of the monsoon rains The relative strengths and weaknesses of •  Frequency of droughts climate models mean that they can more Each of the aforementioned types of climatic accurately project change in long-term tem- changes has been observed in recent decades. perature, but they are much less able to proj- The primary driver of current climate ect extreme events and long-term changes in changes is GHG emissions. The addition of precipitation. These relative degrees of uncer- human-emitted GHGs into the atmosphere tainty are borne out in the analysis through- has several effects, including raising tempera- out this book. ture through the greenhouse effect. The green- Of the several existing RCPs, this book house effect is well understood and agreed on uses climate projections corresponding to by the scientific community. The basic princi- RCPs 4.5 and 8.5. Conceptually, a higher ple of the greenhouse effect is that more number associated with an RCP corresponds GHGs in the atmosphere lead to more heat to a scenario with greater overall emissions being trapped by the climate system. This and potential for more severe climate change. trapped heat transfers to many parts of the The landmark global Paris Agreement, in Earth system, resulting in warmer air temper- which countries came together to agree to atures, warmer oceans, and melting glaciers limit GHG emissions, set a target of holding and ice sheets. This means that the response temperature increases at 2°C relative to prein- of long-term average air temperatures to dustrial conditions. The climate change sce- release of GHGs is also well understood and nario corresponding to the Paris Agreement agreed on. Although there are many sources would be even less than RCP 4.5, which of GHGs, it is also well documented and ­ represents a future in which some collective agreed on that emissions from human sources action is taken to limit GHG emissions and are the primary contributor driving recent global annual average temperatures increase observed changes in climate (IPCC 2013). by 2.4°C (range of 1.7°C to 3.2°C) by 2100 Projecting future climatic changes requires relative to preindustrial levels. RCP 8.5 is creating a scenario that represents future con- closer to a scenario in which no actions are ditions, including the amount, timing, and taken to reduce emissions and global annual type of emissions by human activities. average temperatures increase by 4.3°C Multiple scenarios are developed to account (range of 3.2°C to 5.4°C) by 2100 relative to for uncertainty about the path that the world preindustrial levels. Because RCP 4.5 depends will take. The socioeconomic dimensions of on taking collective action, this book refers to each scenario are referred to as shared socio- it as the climate-sensitive scenario; because economic pathways (SSPs) (O’Neill and oth- RCP 8.5 corresponds to emitting significant ers 2014), and the scenarios representing carbon dioxide (and other GHGs), this book atmospheric emissions of greenhouse gases refers to it as the carbon-intensive scenario. (GHGs) are referred to as representative con- The range of projected global temperature centration pathways (RCPs) (Taylor, Stouffer, increases and precipitation changes cited in and Meehl 2012). the previous paragraph stems from differ- Climate models are the primary tools for ences between the approximately 45 climate projecting how a given RCP scenario will models used in the IPCC’s Fifth Assessment affect the Earth’s climate. Climate models are Report (IPCC 2013). In this book, climate designed to approximate fundamental laws of projections for South Asia are based on the 24  SOUTH AS I A ’ S HOTS P OTS 11 climate models which have publicly avail- data sets were considered as possible repre- able data and perform best for South Asia, as sentation of the “true” historical climate. The described in the following section. principal data sets were Aphrodite v1101, a daily gridded data set for monsoon Asia, and CHIRPS v2.0, a daily gridded data set avail- Selecting Appropriate Climate able globally. These two data sets were com- Models pared with the daily gridded data set of the It is well accepted that projecting future Indian Meteorological Department (IMD), changes to climate should be conducted using whose data set is considered to be the most multiple climate models. The reason is that spatially and temporally consistent, but is although all models have imperfections, they available only over India. Therefore, the IMD do not always have the same imperfec- data were used to determine which regionally tions, and random errors will tend to cancel. available data set performs best for South Although a multimodel approach is preferred, Asia, on the basis of agreement over India. adding poorly performing models degrades Overall, the Aphrodite data set best matches the quality of the information. Therefore, the spatial pattern and local magnitudes of model selection is important to developing the IMD data, particularly with respect to climate projection scenarios and assessing variability, trends, and precipitation extremes. uncertainty. Aphrodite is therefore used to assess the Of the approximately 45 climate mod- climate model performance. els participating in the Climate Model The performance of each of the 18 climate Intercomparison Program (CMIP5), 18 were models is displayed in figure 2.2, panels a evaluated for this book because they had pub- and b. In general, the models better reproduce licly available monthly output for the required the spatial patterns of long-term average cli- historic and projection simulations (see mate and year-to-year standard deviation com- the list of models assessed and selected in pared with the spatial pattern of multiyear table B.1). The climate models were examined trends; however, the climate models better for their ability to replicate the observed char- reproduce regionally aggregated trends than acteristics of the regional climate (see the long-term average climate or year-to year vari- more detailed explanation of model assess- ability. In general, though, the climate models ment in appendix D). The two metrics used to better reproduce the regionally aggregated cli- assess model performance are spatial pattern mate than the spatial pattern of climate within correlation and regionally aggregated root South Asia. The method of model selection was mean squared error (RMSE). to eliminate those with the worst performance. A high spatial pattern correlation is Four ­ m odels—CSIRO Mk3.6.0, GFDL desirable because it suggests that models ESM2G, HadGEM2 ES, and MPI ESM-LR— capture the right climate processes respon- were eliminated because of particularly poor sible for that pattern (that is, spatial pattern spatial correlation performance. Three models— of mean climate, variability, or trend). GISS E2R, INM CM4, and MIROC5—were A low RMSE is desirable because it suggests eliminated because of particularly poor region- that model response to the relevant climate ally aggregated performance (see appendix D processes is more accurate. The aforemen- for more details). Therefore, of the 18 climate tioned metrics were calculated for each models assessed, 11 were selected for projecting model’s ability to reproduce the long-term conditions in South Asia (figure 2.2). mean, standard deviation, and trend in tem- These 11 selected climate models are used perature and precipitation relative to grid- to form ensemble climate projections for the ded observational data. climate-sensitive and carbon-intensive Not all observational data sets are of equal scenarios. The three metrics calculated for quality, and errors and uncertainties in obser- each ensemble are the multimodel mean vations are inescapable. Several observational (MMM), low, and high. The MMM is the I n c r e a s i n g l y H o t    25 average of all 11 model values at each loca- estimated for each district for all countries tion and time frame. The ensemble low and in South Asia. Climate projections are calcu- high are the values from the climate model lated for annual (January through that projects the minimum or maximum December), premonsoon (March through future climate aggregated over the country for May), monsoon (June through September), each climate parameter and season (or on an and postmonsoon (October through annual basis). A different climate model can February) seasons. Each projection is a be selected as the low or high ensemble mem- 30-year average centered on the target year. ber for each country and each season. The hotspot analysis is conducted for the Ensemble climate model projections of two climate projection time frames shown the MMM, low, and high values are in table 2.1. FIGURE 2.2  An Illustration of Model Selection Criteria a. Spatial pattern reproduction BCC CSM1.1 CanESM2 CCSM4 0.8 CNRM CM5 CSIRO Mk3.6.0 Correlation (larger values are better) GFDL ESM2G 0.6 GFDL ESM2M GISS E2R 0.4 HadGEM2 CC GCM HadGEM2 ES INM CM4 0.2 IPSL CM5A-LR MIROC ESM MIROC ESM-CHEM 0 MIROC5 MPI ESM-LR –0.2 MPI ESM-MR NorESM1-M SD in monsoon precipitation SD in postmonsoon precipitation Mean premonsoon temperature Mean monsoon temperature Mean postmonsoon temperature SD in monsoon temperature SD in postmonsoon temperature Trend in premonsoon temperature Trend in monsoon temperature Trend in postmonsoon temperature Trend in premonsoon precipitation Trend in monsoon precipitation Trend in postmonsoon precipitation SD in premonsoon precipitation SD in premonsoon temperature Mean premonsoon precipitation Mean monsoon precipitation Mean postmonsoon precipitation Seasonal climate statistics (continues next page) 26  SOUTH AS I A ’ S HOTS P OTS FIGURE 2.2  An Illustration of Model Selection Criteria (continued) b. Regionally averaged error BCC CSM1.1 2.0 CanESM2 CCSM4 1.8 CNRM CM5 CSIRO Mk3.6.0 1.6 RMSE (lower values are better) GFDL ESM2G GFDL ESM2M GISS E2R 1.4 HadGEM2 CC GCM HadGEM2 ES 1.2 INM CM4 IPSL CM5A-LR MIROC ESM 1.0 MIROC ESM-CHEM MIROC5 0.8 MPI ESM-LR MPI ESM-MR NorESM1-M 0.6 SD in monsoon precipitation SD in postmonsoon precipitation Trend in postmonsoon precipitation Mean premonsoon temperature Mean monsoon temperature Mean postmonsoon temperature SD in monsoon temperature SD in postmonsoon temperature Trend in premonsoon temperature Trend in monsoon temperature Trend in postmonsoon temperature Trend in premonsoon precipitation Trend in monsoon precipitation SD in premonsoon precipitation SD in premonsoon temperature Mean premonsoon precipitation Mean monsoon precipitation Mean postmonsoon precipitation Seasonal climate statistics Note: Green rectangles indicate the four climate models that were eliminated because of low spatial pattern correlations and the three climate models that were eliminated because of large RMSE values (see appendix D for more details on model selection). RMSE = root mean squared error; SD = standard deviation. TABLE 2.1  Results from Climate Model Projections of 1.0°C to 2.3°C) under the climate-sensitive for Two Future Time Frames scenario and 2.2°C (range is 1.5°C to 3.1°C) Target year Definition under the carbon-intensive scenario by 2050, relative to 1981–2010 (figure 2.3). Although 2030 Midpoint of projected climate values for 2016 through 2045 the uncertainty range is notable, the magni- 2050 Midpoint of projected climate values for tude of the projected temperature increases is 2036 through 2065 larger than the uncertainty. This indicates high confidence that temperatures in South Asia will continue increasing under both the climate-sensitive and carbon-­ i ntensive South Asia Continues to Get Hotter scenarios. Annual average temperatures in South Asia Unlike temperature, projected changes in are projected to increase 1.6°C (with a range precipitation are very uncertain (figure 2.4). I n c r e a s i n g l y H o t    27 One reason for the high degree of uncertainty FIGURE 2.3  Historic Trends in Annual Temperature Increases is that precipitation patterns highly depend on Are Projected to Increase cloud microphysics, which is difficult to repre- 25.5 sent in current climate models. The projected Historic time series 25.0 Historic average (1981–2010) MMM change in average monsoon precipita- Climate-sensitive (RCP 4.5) tion is ±3.9 percent under the climate-sensitive 24.5 Carbon-intensive (RCP 8.5) Temperature (°C) scenario and ±6.4 percent under the carbon- 24.0 Historic trend (0.14˚C/decade) intensive scenario (the range of projections is 23.5 negative 6.9 percent to positive 25.2 percent, 23.0 depending on the climate model and scenario). 22.5 This large range represents a risk for South 22.0 Asia. If average precipitation increases, some 21.5 areas that have historically experienced low precipitation could benefit. At the same time, it 21.0 1960 1970 1980 1990 2000 2010 2020 2030 2040 2050 is also likely that extreme precipitation events would become more common under a scenario Year of increasing precipitation patterns, especially Sources: Harris and others 2014 (Climate Research Unit TS 2.24); 11 climate models cited because of the large temperature increases. in box 2.1. Note: The black line indicates yearly annual temperature, the gray line indicates average annual Extreme precipitation events would cause an temperature from 1981 through 2010, the dashed purple line indicates multimodel mean under increase in damage. Decreasing precipitation the carbon-intensive scenario, the dashed green line represents multimodel mean under the climate-sensitive scenario, and the shaded areas indicate range of results based on 11 climate would result in less overall water availability models for each scenario. in South Asia, which would also cause prob- lems for people and agricultural yields. FIGURE 2.4  Monsoon Precipitation Varies Considerably Year to The patterns of temperature change are Year, and Projections Are Highly Uncertain not evenly distributed throughout South Asia 210 Historic time series (box 2.1 and map 2.5, panels a and b). Under Historic average (1981–2010) the climate-sensitive scenario, temperatures are 200 Climate-sensitive (RCP 4.5) Precipitation (mm/month) projected to increase the most for the Hindu 190 Carbon-intensive (RCP 8.5) Kush and Karakoram mountains. Under the 180 carbon-intensive scenario, the MMM climate 170 model projection is for annual average temper- atures to increase 2.5°C to 3.0°C for 160 Afghanistan, the portion of Pakistan neighbor- 150 ing Afghanistan, the Karakoram mountains, 140 and the Himalayas, relative to 1981–2010 val- 130 ues. Part of the reason for this spatial pattern of large temperature increases is that these regions 1960 1970 1980 1990 2000 2010 2020 2030 2040 2050 will lose substantial snow and ice cover under Year these climate scenarios. For example, Mosier Sources: Harris and others 2014 (Climate Research Unit TS 2.24); 11 climate models cited (2015) finds that snowfall will decrease more in in box 2.1. Note: The black line indicates yearly monsoon precipitation, the gray line indicates average the Hindu Kush mountains than the monsoon precipitation from 1981 through 2010, the dashed purple line indicates multimodel Karakoram or Himalaya mountains. Snow and mean under the carbon-intensive scenario, the dashed green line represents multimodel mean under the climate-sensitive scenario, and the shaded areas indicate range of results based on ice help to regulate air temperatures because 11 climate models for each scenario. they reflect solar radiation and regulate air tem- peratures through the melting process. Snow In South Asia, temperatures are projected and ice also store water, which gets released to increase the least along the coastal areas of during the hottest portions of the year. India, Bangladesh, and Sri Lanka because the Therefore, losing these important natural water oceans help to moderate the temperature. reservoirs results in feedback that enhances Temperature increases in these areas are still climate change and affects water availability. 1.0°C to 1.5°C under the climate-sensitive 28  SOUTH AS I A ’ S HOTS P OTS BOX 2.1  Understanding Historic and Projected Temperatures for Each Country Temperature patterns vary considerably through- spatially analyzing climate change. At the national out South Asia. This means that in a given year, level, historic temperature trends are lowest for temperatures may be hotter than average in one Bangladesh (0.09°C per decade), India (0.11°C part of the region but cooler than average in a dif- per decade), Nepal (0.14°C per decade), and ferent part. It also means that the trends are non- Bhutan (0.15°C per decade). Trends are the high- uniform. These attributes do not decrease the est for Afghanistan (0.27°C per decade), Pakistan ability to make projections about future climate, (0.17°C per decade), and Sri Lanka (0.17°C per but they do demonstrate the importance of decade) (figure B2.1.1, panels a through g). FIGURE B2.1.1  Annual Temperatures Are Increasing for All Countries, but the Rate of Change Varies a. Afghanistan b. Bangladesh 30 Historic time series Historic time series 17 Historic average (1981–2010) Historic average (1981–2010) Climate-sensitive (RCP 4.5) 29 Climate-sensitive (RCP 4.5) 16 Carbon-intensive (RCP 8.5) Carbon-intensive (RCP 8.5) Historic trend (0.27˚C/decade) Temperature (°C) 28 Historic trend (0.09 °C/decade) Temperature (°C) 15 27 14 26 13 25 12 24 11 1960 1970 1980 1990 2000 2010 2020 2030 2040 2050 1960 1970 1980 1990 2000 2010 2020 2030 2040 2050 c. Bhutan d. India 14.5 Historic time series 29 Historic time series Historic average (1981–2010) Historic average (1981–2010) 14 Climate-sensitive (RCP 4.5) Climate-sensitive (RCP 4.5) Carbon-intensive (RCP 8.5) 28 Carbon-intensive (RCP 8.5) 13.5 Historic trend (0.15°C/decade) Historic trend (0.11°C/decade) Temperature (°C) Temperature (°C) 13 27 12.5 26 14 25 11.5 24 11 10.5 23 1960 1970 1980 1990 2000 2010 2020 2030 2040 2050 1960 1970 1980 1990 2000 2010 2020 2030 2040 2050 (continues next page) I n c r e a s i n g l y H o t    29 BOX 2.1  Understanding Historic and Projected Temperatures for Each Country (continued) FIGURE B2.1.1  Annual Temperatures Are Increasing for All Countries, but the Rate of Change Varies (continued) e. Nepal f. Pakistan Historic time series 26 19 Historic average (1981–2010) Historic time series Climate-sensitive (RCP 4.5) Historic average (1981–2010) 25 Climate-sensitive (RCP 4.5) 18 Carbon-intensive (RCP 8.5) Historic trend (0.11°C/decade) Carbon-intensive (RCP 8.5) 24 Historic trend (0.17 ˚C/decade) Temperature (°C) 17 Temperature (°C) 23 16 22 15 14 21 13 20 12 19 1960 1970 1980 1990 2000 2010 2020 2030 2040 2050 1960 1970 1980 1990 2000 2010 2020 2030 2040 2050 g. Sri Lanka 31 Historic time series Historic average (1981–2010) 30 Climate-sensitive (RCP 4.5) Carbon-intensive (RCP 8.5) Temperature (°C) 29 Historic trend (0.17 ˚C/decade) 28 27 26 25 1960 1970 1980 1990 2000 2010 2020 2030 2040 2050 Sources: Harris and others 2014 (Climate Research Unit TS 2.24); 11 climate models cited in box 2.1. Note: The black line line indicates yearly annual temperature, the gray line indicates average annual temperature from 1981 through 2010, the dashed purple line indicates multimodel mean under the carbon-intensive scenario, the dashed green line represents multimodel mean under the climate-sensitive scenario, and the shaded areas indicate 100 percent confidence interval based on 11 climate models for each scenario. 30   SOUTH ASIA’S HOTSPOTS MAP 2.5  Annual Average Temperature Is Projected to Continue Increasing Dramatically under the Climate-Sensitive and Carbon-Intensive Scenarios Sources: Harris and others 2014 (Climate Research Unit TS 2.24); 11 climate models cited in box 2.1. Note: Changes are by 2050 (average for 2036 through 2065) relative to 1981 through 2010 averages (figure 2.1). scenario and 1.5 to 2.0°C under the carbon- Kumar, K. R., G. B. Pant, B. Parthasarathy, and intensive scenario by 2050, relative to 1981 N. A. Sontakke. 1992. “Spatial and Subseasonal through 2010. Patterns of the Long-Term Trends of Indian Summer Monsoon Rainfall.” International Journal of Climatology 12 (3): 257–68. Mosier, T. M. 2015. “Characterizing Linkages References Between the Climate, Cryosphere, and Impacts Abbas, F., A. Ahmad, M. Safeeq, S. Ali, F. Saleem, on Run-of-River Hydropower in Data-Sparse H. M. Hammad, and W. Farhad. 2014. Mountain Environments.” PhD dissertation, “Changes in Precipitation Extremes over Arid Oregon State University, Corvallis. to Semiarid and Subhumid Punjab, Pakistan.” O’Neill, B. C., E. Kriegler, K. Riahi, K. L. Ebi, Theoretical and Applied Climatology 116 S. Hallegatte, T. R. Carter, R. Mathur, and (3–4): 671–80. D. P. van Vuuren. 2014. “A New Scenario Basistha, A., D. S. Arya, and N. K. Goel. 2009. Framework for Climate Change Research: The “Analysis of Historical Changes in Rainfall in Concept of Shared Socioeconomic Pathways.” the Indian Himalayas.” International Journal Climatic Change 122 (3): 387–400. of Climatology 29 (4): 555–72. Pal, I., and A. Al-Tabbaa. 2009. “Trends in Bollasina, M. A., Y. Ming, and V. Ramaswamy. Seasonal Precipitation Extremes—An Indicator 2011. “Anthropogenic Aerosols and the of ‘Climate Change’ in Kerala, India.” Journal Weakening of the South Asian Summer of Hydrology 367 (1): 62–69. Monsoon.” Science 334 (6055): 502–5. Rosenfeld, D., S. Sherwood, R. Wood, and Harris, I. P. D. J., P. D. Jones, T. J. Osborn, and L. Donner. 2014. “Climate Effects of Aerosol- D. H. Lister. 2014. “Updated High-Resolution Cloud Interactions.” Science 343 (6169): Grids of Monthly Climatic Observations—The 379–80. CRU TS3.10 Dataset.” International Journal Roy, S., and R. C. Balling. 2004. “Trends in of Climatology 34 (3): 623–42. Extreme Daily Precipitation Indices in India.” I n c r e a s i n g l y H o t    31 International Journal of Climatology 24 (4): Extreme Wet and Dry Spells during the South 457–66. Asian Summer Monsoon Season.” Nature Singh, D. 2016. “South Asian Monsoon: Tug of Climate Change 4 (6): 456. War on Rainfall Changes.” Nature Climate Taylor, K. E., R. J. Stouffer, and G. A. Meehl. Change 6 (1): 20. 2012. “An Overview of CMIP5 and the Singh, D., M. Tsiang, B. Rajaratnam, and N. S. Experiment Design.” Bulletin of the American Diffenbaugh. 2014. “Observed Changes in Meteorological Society 93 (4): 485–98. Climate and Living Standards 3 T he climate is changing and will con- expenditures from household-level surveys.1 tinue to do so over the coming There has been a long-standing debate in the decades under a range of emissions literature on whether income or consumption scenarios, but what effect will this have on is a better measure of standards of living. living standards? Addressing this question Especially for low-income countries, a strong requires understanding the relationship case has been made for preferring consump- between today’s weather and living stan- tion expenditures, based on both conceptual dards and then extrapolating this relation- and practical considerations (Deaton and ship to look at future climatic conditions. Grosh 2000). Expenditures are supposed to “Living standards” encapsulate a broad set better reflect “long-term” or “permanent” of conditions that can be expressed both income and are from this point of view con- monetarily and nonmonetarily. For example, sidered to be a better measure of economic a family’s type of housing is easy to express well-being (Atkinson 1987). monetarily, although the sense of well-being In the poverty literature, the standard way received from that housing cannot be easily to determine whether a household is poor expressed in monetary terms. This book uses is to compare its daily expenditures per capita consumption expenditures as a metric for to a minimum consumption threshold, or overall living standards because it encapsu- poverty line (Chatterjee and others 2016; lates the monetary dimensions of living stan- Haughton and Khandker 2009). In his semi- dards and is objectively quantifiable. There nal work on poverty measurements, Deaton are strong precedents for using consumption (2005) fervently argues that although con- expenditures as a proxy for living standards, sumption measures are limited in their although it should be acknowledged that it is scope, they are nevertheless a central compo- an imperfect approximation. nent of any assessment of living standards. The most direct (and popular) measures of Consumption expenditures per capita are living standards in the literature are income used extensively at the World Bank to produce and consumption. In general terms, income poverty diagnostics for countries, including refers to the earnings from productive activi- mapping poverty (Li and Rama 2016). These ties and current transfers, whereas consump- poverty maps succinctly represent average tion refers to resources used. Consumption is household expenditures per capita in real usually measured by looking at household terms across space, at a disaggregated level. 33 34  SOUTH AS I A ’ S HOTS P OTS Building on theories of consumption, they use outputs, which in this case are climate and household surveys—whose samples are small consumption expenditures, respectively. but rich in information—to estimate the rela- Structural models can capture the basic rela- tionship between household expenditures per tionships that govern an interaction, but they capita and household characteristics. The set tend to also be highly uncertain in real-world of characteristics considered are those that situations because of the complex nature of can also be found in population censuses. The the causal relationships. As an example, a estimated relationship is then used to predict structural model may contain an equation to household per capita expenditures at varying estimate agricultural yield as a function of spatial levels, based on local household char- weather, but it likely would not capture the acteristics as reported by population censuses psychological effects of multiple poor yield (Demombynes and others 2002; Elbers, years on a farmer. Lanjouw, and Lanjouw 2003). This book builds and extends this knowl- Using this understanding of current living edge base by delving into the relationship standards, the next step is to estimate the link between changes in average weather—long- between climate change and consumption term seasonal average temperature and expenditures (see box 3.1). The two main cat- precipitation—and consumption expenditures. egories of model formulations for estimating Like many previous studies, the analysis living standards are structural and reduced- here uses a reduced-form model to estimate form. Structural models seek to represent the the relationship between weather and con- causal relationships between inputs and sumption expenditures. BOX 3.1  How Climate Change Affects Consumption Expenditures A changing climate can affect consumption Increasing temperatures and wet conditions can expenditures in many diverse ways (see increase the propagation of vector-borne and figure B3.1.1 for a conceptual representation of ­ other infectious diseases, resulting in lost produc- some of the pathways). For example, increasing tivity and income. Similarly, extreme heat days are temperatures combined with shifting precipitation generally correlated with declining productivity of patterns can dampen agricultural productivity, workers, especially in areas that are already leading to a decline in consumption expenditures warm. A changing climate can force people out of for households dependent on agriculture. their traditional professions, resulting in individu- als taking up occupations not suitable for their FIGURE B3.1.1  Climate Change and Living Standards skills and earning less income. Are Linked through a Diverse Set of Pathways Extreme events also cause major disruptions to consumption. For example, individuals may Health consume less, either because they have no sav- ings and their work is disrupted, or because they Agriculture Living know they need to use their resources carefully Climate standards for recovery efforts. In many cases these indi- change (consumption Migration expenditures) vidual disaster responses are of relatively short duration and consumption rebounds after relief and rehabilitation efforts. In contrast, the effects Productivity of changing average weather will be slow-­ moving and persistent. C l i m a t e a n d Li v i n g S t a n d a r d s    35 Reduced-form models do not make 2014; OECD 2015). There is also evidence of assumptions about the way a system works, a negative effect of climate change, especially but instead seek to capture the aggregate rela- extreme events, on GDP growth (Brown and tionship between the inputs and outputs. others 2013; Burke, Hsiang, and Miguel Therefore, reduced-form models tend to have 2015; Dell, Jones, and Olken 2012; Hsiang greater predictive capability than structural and Jina 2014; Moore and Diaz 2015). Other models under a wider set of conditions. studies have identified negative effects of cli- However, because reduced-form models do mate change on health, agriculture, and labor not represent the underlying causal relation- productivity (Auffhammer and Schlenker ships, they do not capture fundamental 2014; Deschenes 2014; Mani and Wang changes to a given system. This is particularly 2014; Heal and Park 2015; WHO 2015). problematic when a relationship includes a At the local level, studies have investigated threshold response. For example, over the effects of climate change on household living baseline period of 1981 through 2010, cli- standards. Verner (2010) examines the rela- mate change may not have caused very sig- tionship between climate change and income nificant migrations, but it is likely that climate in Latin America using municipality-level change induced migrations will occur in the data, and finds a mixed set of relationships future if conditions become too adverse. In between temperature and income. For Bolivia, addition, it is likely that a community will Brazil, and Peru, the relationship is clearly an have greater resources to recover from a sin- inverted U (that is, higher temperatures are gle shock, compared with a series of indepen- good to a point, and then cause harm); in dent shocks. Such thresholds and cumulative Chile, the relationship is more or less inverse; effects cannot be adequately captured using a and in Mexico, the relationship is not statisti- reduced-form model, such as that used in this cally significant. Skoufias, Katayama, and book. Despite these limitations, reduced-form Essama-Nssah (2012) find that climate models are the best means for assessing the change will lower agricultural productivity linkages between living standards and climate and increase food prices, but expect these because these relationships are not adequately changes to be offset by reductions in poverty represented in current structural models and and economic growth rates. In contrast, reduced-form models explain the maximum Hallegatte and others (2016), in Shock Waves, amount of variation in the data. find that economic growth can play a major role in determining future poverty levels, but that an additional 100 million people could Accumulated Knowledge end up in poverty by 2030 because of climate A growing set of research explores the rela- change without such growth, including tionship between weather or climate and 42 million in India. human activities. Many of these studies use Existing studies have looked at an array of reduced-form models. Studies differ primarily economic impacts. Hallegatte and others in the effect that they seek to quantify and (2017), in Unbreakable, focus on the current level of aggregation. Aggregation refers to effects of natural disasters across all income whether a variable is reflective of conditions groups (they make no projections) and find at a single point (such as a single person or that such events account for varying losses in family) or is representative of a larger group consumption across the world. In South Asia, (such as a province or country). losses are estimated at 0.3 percent in Many studies have estimated the relation- Sri Lanka, 0.4 percent in India, 0.9 percent in ship between weather or climate and societies Pakistan, 1.6 percent in Nepal, and 3.5 percent at the national level. Several identify a nega- in Bangladesh. Jacoby, Rabassa, and Skouas tive relationship between increasing tempera- (2011) investigate the effects of rural con- tures and gross domestic product (GDP) sumption levels in India and estimate that (DARA 2012; Ahmed and Suphachalasai because of climate change, rural households 36  SOUTH AS I A ’ S HOTS P OTS will face a loss of between 6 percent and climate factors (Burke, Hsiang, and Miguel 11 percent by 2040. 2015; Mendelsohn, Nordhaus, and Shaw This book adds to this accumulated knowl- 1994; Schlenker and Roberts 2009). The fun- edge through a combination of granularity damental reason is that linear terms capture and region-specific climate change analyses. only a single trend estimate, which does not The research here uses the household as the account for the impact of spatial differences fundamental unit of analysis. The household in the current climate. Also, there are often results are then aggregated to the district or optimal climates, which can be captured in a province level to appropriately represent the quadratic model, but not a linear one. distribution of households in the given politi- Intuitively this is clear: too little or too much cal administrative unit. The book focuses on precipitation causes problems, and exces- effects of changes in average precipitation and sively cold and excessively hot temperatures temperature because changes in the average affect many important activities (box 3.2). can be projected with greater confidence than The analysis here includes seasonal changes in extreme events. weather in the model because many human activities in South Asia are seasonal (Massetti, Mendelsohn, and Chonabayashi 2016). The Analytical Framework model also includes a set of household, dis- Much of the accumulated knowledge uses trict, and geospatial characteristics, which are reduced-form models to estimate the relation- chosen because they are only weakly corre- ship between weather or climate and a given lated with climate and they potentially activity. Many of these reduced-form models help explain variations in consumption include both linear and quadratic weather or expenditures. BOX 3.2  The Quadratic Relationship between Climate and Economy Burke, Hsiang, and Miguel (2015) investigate to unique adaptive capacities) for individual the relationship between climate and the econ- countries. Whereas significant global economic omy by looking at the effect of annual average production is clustered near the estimated tem- temperature on productivity. They find that perature optimum, individual communities and country-level economic production is smooth, countries exhibit similar—but unique—nonlinear nonlinear, and concave with respect to tempera- responses to temperature. In their model, low- ture, with an optimal temperature of 13°C income tropical countries exhibit larger (figure B3.2.1). Productivity in countries with responses mainly because they are hotter on average temperatures lower than 13°C is esti- average, not because they are poorer. Although mated to benefit until the optimal value is there is suggestive evidence that upper-income reached. This model is produced using data from countries might be somewhat less affected by many countries and is therefore driven by aver- temperature, their response is statistically indis- age relationships. Their model does not allow tinguishable from low-income countries at all different optimal temperatures (corresponding temperatures. (continues next page) C l i m a t e a n d Li v i n g S t a n d a r d s    37 BOX 3.2  The Quadratic Relationship between Climate and Economy (continued) FIGURE B3.2.1  Impacts of Temperature on Productivity Are Well Explained Using a Quadratic Model a. Germany US Brazil Indonesia UK Japan India France China Nigeria Parenthetical information change in ln(GDP per capita) 0 –0.1 –0.2 Global distribution of temperature observations Global distribution of population Global distribution of GDP 0 5 10 15 20 25 30 Annual average temperature (°C) Source: Burke, Hsiang, and Miguel 2015. Note: Blue-shaded region indicates 90 percent confidence interval. After extensive investigations, the reduced- year, Y is the log of average real annual house- form model chosen for the analysis is: hold consumption expenditures, tempit is mean seasonal temperature for the survey year t in ∑ 2 j j Yhit = α + (β1 j tempit + β2 j tempit district i, rainit is mean seasonal precipitation j ∈( s, m , w) (Eq 3.1) for the survey year t in district i, Xhit is a vec- 2 j + β3 j rain it j + β 4 j rain it ) tor of control variables, Wi is a vector of dis- trict characteristics, and t is a vector of + β5 Xhit + β6Wi + τ t + uhit dummy variables representing each survey where h refers to the household surveyed, year; j takes the values s, m, and w represent- i refers to the district, t refers to the survey ing premonsoon (March through May), 38  SOUTH AS I A ’ S HOTS P OTS monsoon (June through September), and demographic characteristics will change over postmonsoon (October through February) the coming decades. Shared socioeconomic seasons, respectively. The selection of control pathway (SSP) scenarios can be used to esti- variables is discussed in the “Control Variable mate changes at the national level, but Selection” section later in this chapter. equation (3.1) is implemented at the house- ­ The reduced-form relationship is esti- hold level, and there is very high uncertainty mated using equation (3.1) separately for translating national scenarios to individual Afghanistan, Bangladesh, India, Nepal, households. It is certain that, on average, edu- Pakistan, and Sri Lanka, using all years of cation, electricity, road, and market access available survey data for each country (see will improve in the future. The net extent of the survey information in table 3.1). The rela- these opposing effects (that is, development tionship is estimated using seasonal weather versus negative climate change impacts) contemporaneous to the household survey. would depend on the growth and develop- Impacts of changes in average weather are ment policies of respective countries. The cor- then estimated by applying the empirical responding qualification is that the results relationship to 30-year average seasonal cli- represent the effects of projected climate mate projections corresponding to the ensem- change if it were to happen today. ble multimodel mean (MMM) under the climate-sensitive and carbon-intensive sce- narios for 2030 and 2050; 2030 and 2050 Data Sources refer to average climate from 2016–2045 and Annual household consumption expenditures 2036–2065, respectively (see chapter 2). The (the proxy for living standards) and several historic baseline is established through apply- household characteristics used as control vari- ing the estimated relationship to the most ables are obtained from country-specific recent available year of household surveys household surveys (table 3.1). These house- for each country and the corresponding hold surveys are designed to represent condi- 30-year seasonal climate (that is, if the most tions at different levels of administrative recent survey year is 2011, then the baseline aggregation, which vary between countries. climate is 1981 through 2010). For example, survey data for Pakistan are These predicted consumption changes designed to represent provincial conditions, assume that only the average weather changes whereas survey data for India are designed to between the baseline and 2030 or 2050. That represent district conditions. Several control is, all household, district, and socioeconomic variables are paired with the household sur- characteristics are held constant. Although vey data and tested for suitability (table 3.2). this may appear to be an unreasonable Historic weather and climate are derived assumption, it is impossible to know how from the Climate Research Unit Time-Series household size, dependency ratio, and other (CRU TS) 3.22 monthly precipitation and TABLE 3.1  Household Surveys in South Asia Afghanistan Bangladesh India Nepal Pakistan Sri Lanka Name ALCS HIES NSS NLSS PSLM HIES Years 2008, 2012 2000, 2005, 2010 2004, 2009, 2011 2003, 2010 2001, 2004, 2005, 2006, 2009, 2012 2007, 2010, 2011 Representative District Division District Region Province Province administrative unit Note: Survey names are ALCS (Afghanistan Living Conditions Survey), HIES (Household Income Expenditure Survey), NSS (National Sample Survey), PSLM (Pakistan Social and Living Standard Measurement), and NLSS (Nepal Living Standard Survey). The survey results for Nepal are representative at the provincial level, but five regions are used to represent results in this book. C l i m a t e a n d Li v i n g S t a n d a r d s    39 TABLE 3.2  Variables Considered for Household, District, and Geospatial Differences Household District Geospatial Variables Access to electricity; age of head of Population density; road density; Distance to coast; elevation; household; agriculture household; seasonality of water availability; latitude; longitude dependency ratio; education of head of travel time to market; water household; female-headed household; availability; water stress household size; rural household Sources Survey data Spatial Database for South Asia; WorldClim Digital Elevation Aqueduct data set by the World Model (Hijmans and Resources Institute (Gassert and others 2005) others 2013); GoN (2013); Uchida and Nelson (2008) temperature data (Harris and others 2014). or climate. For example, a household’s deci- These data are available from 1901 through sion on where to live may be influenced by 2013 at a spatial resolution of 0.5 degrees climatic conditions. Similarly, a farmer’s deci- (each grid cell is square in geographic coordi- sion on when to plant a crop will depend nates, with a side length of approximately greatly on the weather patterns observed in 55 kilometers at the equator). The CRU TS the preceding years. The extent to which data set is produced by statistically interpolat- changes in average weather are appropriately ing available station records to a uniform captured in a model requires carefully consid- grid. Therefore, in regions where station ering the set of modeling choices. observations are sparse or climate variability is large, CRU tends to contain greater uncer- tainty. However, in this book, as in many Control Variable Selection other studies, CRU is used as a reference data This analysis uses a variant of the time-series set (that is, the uncertainty is ignored). 2 research design that relies on different cohort Climate projections are produced from the samples over time in the same region (similar ensemble of climate model simulations to the design used by Maccini and Yang described in chapter 2. District-level climate is 2008). In this context, it might be tempting calculated by area-weighting the gridded to control for all observable and poten- CRU TS and climate model ensemble projec- tially confounding factors. Although well tions. The premonsoon season is defined as intentioned, such an approach can introduce ­ March through May, the monsoon season is bias into the coefficients describing the effect defined as June through September, and the of climate on living standards, because these postmonsoon season is defined as October controls may themselves affect the climate. through February. These season definitions For example, elevation may cause households are consistent with agricultural seasons within to make certain decisions, but elevation also the region. impacts surface air temperature. Therefore, if elevation were included in equation (3.1), it would not be possible to determine the por- Two Methodological Challenges tion of the modeled impact due to household Two challenges to the empirical estimation of choices caused by elevation versus tempera- the net relationship between climate change ture. Such an effect is termed a bad control and consumption are selecting appropriate (Angrist and Pischke 2008) and is undesir- control variables and determining the most able in this setting because climatic variables appropriate model structure. The challenges may affect many of the socioeconomic stem from the fact that many household-level factors commonly included as control variables tend to be correlated with weather variables. 40  SOUTH AS I A ’ S HOTS P OTS TABLE 3.3  Control Variables for Each Country Country Controls Afghanistan Rural household, household size, dependency ratio, age of head, female-headed household, access to electricity, years of education, agriculture household, baseline water stress, latitude, blue water availability, seasonal variability of water availability, inverse square of distance to coast, primary road density, access to market, population density Bangladesh Rural household, agricultural household, household size, dependency ratio, age of head, female-headed household, access to electricity, years of education, baseline water stress, elevation, blue water availability, primary road density, population density India Rural household, agricultural household, household size, dependency ratio, age of head, female-headed household, access to electricity, years of education, baseline water stress, blue water availability, seasonal variability of water availability, inverse square of distance to coast, primary road density, population density Nepal Rural household, agricultural household, household size, dependency ratio, age of head, female-headed household, access to electricity, years of education, baseline water stress, blue water availability, seasonal variability of water availability, population density Pakistan Rural household, agricultural household, household size, dependency ratio, age of head, female-headed household, access to electricity, years of education, baseline water stress, blue water availability, seasonal variability of water availability, inverse square of distance to coast, population density Sri Lanka Rural household, agricultural household, household size, dependency ratio, age of head, female-headed household, access to electricity, years of education, baseline water stress, access to market Note: Selected control variables have a correlation of less than 0.5 with all seasonal climate values for the given country. Unfortunately, removing a bad control can between a specific control variable and cli- introduce bias into the results. Taking the mate indicator. Thus, there are some differ- example of elevation, eliminating this param- ences in selected control variables between eter from equation (3.1) causes any effects of the six countries (see table 3.3). As subse- elevation to be manifest in the temperature quently discussed, the analysis also tests cor- parameter, because temperature and elevation relation thresholds of 0.3 and 0.7 to assess are highly correlated. Therefore, there may be the robustness of the findings. biases from including controls and biases from excluding controls. Since the household survey data used in this analysis are not a true Absorbed Climate Effects panel data set, but rather include samples There are many district, provincial, and within a given region in different years, one national characteristics that are not available can argue for the use of some controls to in curated data sets. For example, there is little reduce the chance that the results will be quantifiable data on differences in governmen- biased because of the exclusion of potentially tal policies or implementation of policies important characteristics. across districts or provinces. The omission of To minimize the problem of bad controls, these unobserved or unobservable characteris- the analysis here uses controls that are weakly tics can result in omitted variable bias.3 An correlated with the climate indicators. empirical method for eliminating omitted vari- Specifically, each control variable selected able bias is inclusion of fixed effects in the must have a correlation coefficient of less reduced-form model. Models with fixed effects than 0.5 with all climate variables for the address this challenge by empirically account- country being modeled (Booth, Niccolucci, ing for spatial differences in predictions. and Schuster 1994; Dormann and others A challenge to using a model with 2013; Elith and others 2006; Suzuki, Olson, fixed effects is that the climate data are aggre- and Reilly 2008; Tabachnick and Fidell gated to the district level. This can cause 1996). Country characteristics and geo- unexpected interactions between any included graphic setting influence the correlation fixed effects and the model’s sensitivity C l i m a t e a n d Li v i n g S t a n d a r d s    41 to climate. Thus, fixed effects used to correct variable; and t is a vector of dummy variables for omitted variable bias could result in bias- representing each year of survey. ing climate coefficient estimations as they The reason for simplifying the relation is to absorb the effects of the district-level climate. ascertain how reasonable the overall modeled An alternative formulation could be to use relationship appears and identify possible a model with provincial fixed effects. While temperature thresholds (“inflection points”). this method is better suited to the current anal- As clearly demonstrated by Burke, Hsiang, ysis than district fixed effects, it can also be and Miguel (2015) (box 3.1), temperature problematic because climates of neighboring and productivity have an inverted U–shaped districts within a given province are typically relationship. That is, temperatures that are highly correlated. Thus, models using district- both too cold and too hot result in lower pro- or province-level fixed effects can both lead to ductivity. Burke, Hsiang, and Miguel (2015) biased estimates. use national-level average data for countries Another potential issue with models includ- around the world. It is important to confirm ing fixed effects is that the unobserved vari- if a similar relationship between temperature ables accounted for by the formulation may and consumption expenditures holds at the also be highly correlated with climate. subnational level for individual countries in Therefore, introducing fixed effects may lead South Asia. To do this analysis, equation (3.2) to the same control variable correlation prob- is implemented using annual contemporane- lems discussed in the preceding section. Despite ous temperatures to estimate household con- these concerns, the present analysis includes a sumption expenditures. The results confirm variant of the reduced-form model using pro- that each country in South Asia has the vincial fixed effects to test the robustness of the expected inverted U–shaped relationship to findings (appendix B, table B.3). temperature (figure 3.1, panels a through f). Although the overall relationship is similar across countries, the precise temperature Temperature Inflection Points inflection point—the point at which a mar- The first step in the analysis is to understand ginal change in temperature results in no the conceptual relationship between tempera- change to consumption—differs by country. ture and consumption expenditures. For this For Bangladesh, India, Sri Lanka, and Nepal, exercise, equation (3.1) is modified to include the inflection points are in the expected range only annual temperature (that is, seasonality of 24°C to 27°C; in Pakistan, the model finds and precipitation are removed): the inflection point to be 14°C. These temper- 2 ature inflection points matter in the context of Yhit = α + β1tempit + β2tempit + β3rainit the current climate conditions. 2 + β 4 j rain it ) + θ st + θ s2t 2 + µs + τ t + uhit Nationally, Bangladesh, India, Pakistan, and Sri Lanka are already past their tempera-  (Eq 3.2) ture inflection points. This intuitively makes sense given that these four countries have a where h refers to the household surveyed; warm climate. This means that at the national i refers to the district; t refers to the survey level, any further increase in average tempera- year; Y is the log of average real annual ture will have a net negative effect on con- household consumption expenditures; tempit is sumption expenditures. Temperatures in mean annual temperature for the survey year Afghanistan and Nepal are still less than the t in district i; rainit is mean annual precipita- inflection point, meaning that increases in tion for the survey year t in district i; qst + qs2t2 temperatures are predicted to have a net posi- is country-specific time trend, which accounts tive effect on consumption. This can be also for slowly changing factors within a district expected because Nepal overall has a cold cli- or province such as demographic shifts and mate and could potentially benefit from a institutional capacity; m is a province dummy warmer temperature. 42  SOUTH AS I A ’ S HOTS P OTS FIGURE 3.1  Temperature and Consumption Have an Inverted U–Shaped Relationship for Countries in South Asia a. Afghanistan b. Bangladesh 0 0 –1 –5 Consumption change (%) Consumption change (%) –2 –10 –3 Average temperature Average temperature –15 –4 –20 –5 –25 0 5 10 15 20 25 22 23 24 25 26 27 28 Annual average temperature (°C) Annual average temperature (°C) c. India d. Nepal 0 0 –10 Consumption change (%) Consumption change (%) –5 –20 Average temperature Average temperature –30 –10 –40 –50 –15 0 5 10 15 20 25 30 0 5 10 15 20 25 Annual average temperature (°C) Annual average temperature (°C) e. Pakistan f. Sri Lanka 0 0 –5 Consumption change (%) Consumption change (%) –10 –5 –15 Average temperature Average temperature –20 –10 –25 –30 –15 –35 5 10 15 20 25 30 24 25 26 27 28 29 Annual average temperature (°C) Annual average temperature (°C) Source: World Bank calculations. Note: Blue-shaded region indicates 90 percent confidence interval. C l i m a t e a n d Li v i n g S t a n d a r d s    43 It should be noted, however, that the effects weather on living standards. The framework of increases in temperature will be heteroge- is implemented at the household level and neous throughout each country, with some then aggregated into national predictions areas benefiting slightly from small increases through weighting the household-level results. in temperature and other areas being severely Using this methodology, one can estimate harmed. Temperatures can increase by several aggregate effects of changes in average degrees Celsius in much of Nepal and the weather by 2030 and 2050 under the climate- mountainous areas of India and Pakistan sensitive and carbon-intensive climate change before these areas pass their temperature scenarios. These estimates come with some inflection points. For Bangladesh, Sri Lanka, caveats, as explained in the “Analytical the southern and central portions of India, Framework” and “Two Methodological and Pakistan—where temperatures are already Challeges” sections earlier in this chapter. To relatively hot—the levels of climate change summarize a few of the limitations, the results projected under both the climate-­sensitive and are based on an empirical model that does not carbon-intensive scenarios will have negative directly account for any adaptation to climate effects on consumption. change and assumes that multiple independent shocks do not have cascading negative effects, resulting in larger overall negative effects. National-Level Empirical Findings The analysis finds that changes in average Since the results obtained based on equation weather will have a negative effect on living (3.2) (representing effects of annual tempera- standards in Bangladesh, India, Pakistan, and ture) are similar in form to those obtained in Sri Lanka, but a positive effect on living stan- other studies (for example, Burke, Hsiang, dards in Afghanistan and Nepal (table 3.4). and Miguel 2015), one can conclude that the Although the magnitude of the results varies household and climate data for South Asia on the time frame and climate scenario, the are expressing the previously documented qualitative findings are consistent. As seen in relationship between climate and living stan- the analysis of temperature changes in chapter dards. Therefore, equation (3.1) (representing 2, the results under the climate-­ sensitive and effects of seasonal precipitation and tempera- carbon-intensive scenarios are very similar for ture) is implemented to provide a better under- 2030, but they diverge substantially by 2050, standing of the nuanced relationship between with the carbon-intensive scenario being more climate and living standards in South Asia; extreme. The Maldives and Bhutan are not regression coefficients based on equation (3.1) considered in the economic analysis within are provided in appendix B, table B.2). The this book because the required household sur- analytical framework is then used to estimate vey and climate data are not available or the effect of projected changes in average adequate for these countries. TABLE 3.4  Changes in Average Weather Predicted to Have Mostly Negative Effects under Both Scenarios  Time frame 2030 2050 Scenario Climate-sensitive Carbon-intensive Climate-sensitive Carbon-intensive Afghanistan 5.1 5.8 8.3 11.9 Bangladesh –1.3 –2.3 –2.9 –6.7 India –1.3 –1.5 –2.0 –2.8 Nepal 2.1 2.3 3.2 4.1 Pakistan –1.3 –1.5 –2.0 –2.9 Sri Lanka –3.2 –3.7 –4.9 –7.0 Source: World Bank calculations. Note: The model described by equation (3.1) is implemented for two time frames (2030 and 2050) and two projection scenarios (climate-sensitive and carbon- intensive). The national-level results are aggregated from the household predictions. Percentage change is calculated relative to the historic baseline. 44  SOUTH AS I A ’ S HOTS P OTS The results do not include effects of highlights the benefits of taking actions extreme event shocks, natural disasters, or to reduce greenhouse gas (GHG) emissions. changes in water resources (for example, Actions to limit GHG emissions are com- because of overwithdrawal of groundwater, monly referred to as mitigation, or those glacier melt, or changes in snowpack). 4 that reduce the net carbon footprint. The To highlight this caveat, it is well understood predominant international accord for that coastal areas—for example, Bangladesh—​ achieving emissions reductions is the 2015 will experience strong negative effects from Paris Climate Agreement. The results sea-level rise and a probable increase in the provide a further line of economic reasoning severity of storms, neither of which is cap- for continuing to work toward the emis- tured in the results. Similarly, mountain sions targets established under the Paris areas are known to be highly vulnerable to Agreement. increases in natural disasters, which are not The positive modeled effects of changes in considered. Consequently, the results should average weather on living standards in be interpreted as complementary to existing Afghanistan and Nepal may be explained by studies that capture the effect of extreme the countries’ historic climates: Afghanistan is events on household living standards (for a very water-limited country and the mean example, Hallegatte 2017). The findings climate change projection is for an increase are also more or less consistent with the in precipitation. In addition, much of findings of other recent studies for the region Afghanistan and all of Nepal are historically (table 3.5). very cold, which means that temperature The finding that effects of changes in increases may have some positive benefits. average weather increase with time and are These results are consistent with global find- stronger under the carbon-intensive scenario ings, showing that climate change may cause TABLE 3.5  Comparison between This Book’s Results and Those of Other Studies in the Same Time Frame Climate events   studied Bangladesh India Nepal Pakistan Sri Lanka Hallegatte (2017) Disasters –3.5 –0.4 –1.6 –0.9 –0.3 ∆ consumptiona Hallegatte (2016) Temp. and prec. –4.2 –4.5 –3.0 –4.0 –2.7 total changes, disasters ∆ income Disasters –1.2 –0.6 –0.3 –0.5 –0.5 Other –3.0 –3.9 –2.7 –3.6 –2.2 Ahmed and Temp. and prec. –2.0 –1.8 –2.2 n.a. –1.2 Suphachalasai changes, sea-level rise (2014) (ADB) ∆ GDP DARA (2012) Temp. and prec. –6.4 –4.3 –2.7 –4.2 –7.2 ∆ GDP changes, disasters, sea-level rise Jacoby, Rabassa, and Temp. and prec. n.a. –5.0 n.a. n.a. n.a. Skouas (2011) changes ∆ rural consumption Source: World Bank calculations. Note: GDP = gross domestic product; n.a. = not applicable. a. Focuses only on extreme events. C l i m a t e a n d Li v i n g S t a n d a r d s    45 BOX 3.3  Why the Positive Results for Nepal Are Not an Anomaly The findings presented in table 3.4 suggest a net observed climate variables on yields of major positive effect of changes in average weather on food crops in Nepal—rice, wheat, maize, millet, Nepal, whereas the rest of South Asia (except barley, and potatoes—based on a regression Afghanistan) is projected to be adversely affected. model for historical (1978–2008) climate and Although this may be surprising, the differenti- food crop yield data. Although the temperature ated effect of climate across regions and countries increased by 0.7°C during the period, there are has been well documented in the literature. no significant trends in precipitation patterns. Burke, Hsiang, and Miguel (2015) find that The study finds that during this period, (a) the productivity in cold countries increases as annual growth in the yield of most food crops is posi- temperature increases. This result continues tive; (b) increases in summer rain and maximum until the temperature increases pass an optimum temperature contribute positively to rice yields; inflection point, after which further temperature (c) increases in summer rain and minimum increases result in negative marginal effects on temperature have positive effects on potato productivity. Although Burke, Hsiang, and yields; (d) increases in summer rain and maxi- Miguel (2015) find that much of the global eco- mum temperature adversely affect the yield of nomic production is clustered near the estimated maize and millet; and (e) increases in winter rain temperature optimum, both upper-income and and temperature have a positive effect on wheat low-income countries exhibit similar nonlinear and barley yields. responses to temperature. The results of Burke, Acharya and Bhatta (2013) model climate Hsiang, and Miguel (2015) are based on pooling change and its effect on the agricultural value all countries together, and therefore they esti- addition, taking into consideration annual agri- mate a single temperature inflection point. This cultural gross domestic product (AGDP), rainfall, book finds unique inflection points for each temperature, seeds, and fertilizer distribution country, indicating that adaptation to climate data for the 36 years from 1975 to 2010. Although change is possible. annual average temperatures show an increasing Similarly, Mendelsohn and Reinsborough trend during this period, precipitation trends are (2007) find that the climate responses of Canada mostly mixed. The findings are that (a) rainfall and the United States are similar but statistically has a significant positive effect on AGDP; (b) one different, even though the two countries are unit of rainfall causes agricultural output to neighbors. Comparing the marginal effects of increase by 9.6 percent; (c) since the AGDP con- climate change, they find that Canadian agricul- tribution to the GDP is high in Nepal, the authors ture is unaffected by warmer temperatures, and infer that more rainfall will result in a higher GDP would benefit from more precipitation. U.S. growth rate; and (d) although increases in tem- farms, on the other hand, are much more sensi- perature may also affect AGDP, the authors find tive to higher temperatures and benefit relatively the relationship to be statistically insignificant. less from increased precipitation. The authors Poudel and Shaw (2016) explore the effects of conclude that these marginal results are antici- climate change on major crop yields in the moun- pated given that Canadian farms are generally tainous parts of Nepal using a regression model cooler and drier than American farms. between 30 years of historical climate data and A growing literature describes a positive rela- yield records for food crops. Their climate analy- tionship between climate and the economy in sis shows an increase in temperature of approxi- Nepal, consistent with the findings of this book. mately 0.02°C to 0.07°C per year (varying Joshi, Lall, and Luni (2011) assess the effect of by season) and a mixed trend in precipitation. (continues next page) 46  SOUTH AS I A ’ S HOTS P OTS BOX 3.3  Why the Positive Results for Nepal Are Not an Anomaly (continued) During this period rice yields increase by surveys and information on crop production 4.7 kilograms per hectare per year, maize yields adaptation practices collected through focus increase by 16.0 kilograms per hectare per year, group discussions, interviews, and direct and wheat yields increase by 26.8 kilograms per observations. Over their 30-year study period, hectare per year. Over the same period millet annual average rainfall in the area decreases yields increase steadily, but barley yields decrease. 10.21 millimeters per year and annual mean While these results indicate correlation rather temperature increases 0.02°C per year. During than causation, they are consistent with the this period, yields of rice, maize, wheat, sugar- results found in this book. cane, potatoes, and pulses all have an increas- Dhakal, Sedhain, and Dhakal (2016) study ing trends. The surveys and focus group the effect of climate change and adaptation discussions suggest that farmers achieved these practices on agriculture in the Rautahat increases using climate change adaptation District of central Nepal by analyzing temper- measures. The measures included using high- ature, rainfall, soil moisture, and agriculture yielding varieties of crops, enhanced irrigation surveys. Their study uses primary data on systems, switching to hybrid seeds, and increas- crop production collected through household ing pesticide use. productivity increases in colder parts of the temperature cooling trends evident in chapter world (Burke, Hsiang, and Miguel 2015; 2 (figure 2.3). For these and potentially other Mendelsohn and Reinsborough 2007). These reasons, the reduced-form model relating sea- results for Nepal are consistent with some of sonal weather to consumption expenditures in the literature (box 3.2). The positive results Afghanistan has a much lower predictive abil- for Afghanistan and Nepal do not provide a ity than the models for other countries exam- complete picture of climate change effects in ined (appendix B, table B.2). these locations, though. For example, other studies that project negative climate change effects for Nepal include a greater emphasis Dealing with Uncertainty on extremes, including water scarcity The three main sources of uncertainty in the and escalating electricity prices (Ahmed and findings produced using equation (3.1) are the Suphachalasai 2014), which are not captured precision of the empirical model, differences in this book. in climate projections between climate models, The results for Afghanistan should be and unknown future socioeconomic condi- viewed with some skepticism because of the tions. Empirical modeling errors are repre- low density of meteorological stations for sented by the standard error, which is a widely much of the country’s history and the associ- used metric for representing modeling confi- ated unknown quality of the spatially aggre- dence and is standard in the econometrics lit- gated climate data used in this book. erature. Another category of model error is In Afghanistan, there are only two stations epistemic, which relates to deficiencies in the that report climate observations to the inter- model’s ability to capture processes and national community. Further, these two sta- responses. These errors relate to the caveats tions have significant gaps in their observations discussed in the “Analytical Framework” and and the data quality is not known. These fac- “Two Methodological Challenges” sections tors have the potential to create spurious in this chapter, but are not possible to repre- statistical results, such as the historical sent quantitatively. Considering only the C l i m a t e a n d Li v i n g S t a n d a r d s    47 MMM climate scenario, the prediction uncer- Uncertainty regarding which emissions tra- tainty stemming from uncertainty in the jectory will manifest is represented through empirical model is statistically different from providing results for both the climate-­ sensitive zero for all countries and all future climate and carbon-intensive scenarios. Highlighting projections except Nepal (figure 3.1, panels a this cause of uncertainty provides perspective through f). This finding is consistent with the on the ability of the global community to results shown in table 3.4. impact future well-being. The robustness of the model predictions is The climate model and econometric esti- checked by estimating similar empirical mation uncertainties add nuance to the main models, using different control variables and empirical findings (figure 3.2). For example, including provincial fixed effects (see discus- under the climate-sensitive scenario, the mod- sion in the “Temperature Inflection Points” eling uncertainties include the possibility that section of this chapter). Four alternative effects in 2030 will be positive for all six control variable specifications are imple- countries (figure 3.1, panels a through f). The mented: (a) with “all” control variables; (b) a magnitude of these positive effects decreases correlation threshold of 0.7; (c) a correlation for all countries except Nepal from 2030 to threshold of 0.3; and (d) with no control 2050 and from the climate-­ s ensitive to variables. These alternate specifications pro- carbon-­ intensive scenarios. Most of the posi- duce results that are qualitatively similar but tive living standards responses in figure 3.2 with slightly reduced estimates of the effects correspond to model predictions using the of weather on consumption (see appendix B, high precipitation ensemble member. Given table B.3). Estimations using provincial fixed that climate model projections of precipita- effects tend to reduce the sensitivity of living tion are highly uncertain (figure 2.3), this standards to changes in average weather. For highlights a challenge in using climate models example, estimates for Pakistan under the to make predictions about relationships carbon-intensive scenario by 2050 change involving precipitation, and an area in which from –2.9 percent to –0.9 percent when more research is needed to fundamentally provincial fixed effects are included. improve climate model simulations. There is substantial uncertainty in project- One could argue that the main findings are ing how the climate, especially precipita- not valid because of the possibility that a tion, will respond to atmospheric GHG climate projection estimate exists for which the concentrations (as shown in chapter 2). The effects would be positive. This view would be main findings in this book are based on the shortsighted because it ignores the equally MMM for an ensemble of 11 climate models likely possibility that effects will be very strong selected for use in this study. Uncertainty in and negative (the “high” case in figure 3.2, modeling how the climate will respond to a panels a through d). For example, under the specific emissions scenario is evaluated using high case, which generally corresponds to four combinations of climate models: model predictions using the high temperature and low precipitation ensemble members, •  Low temperature and low precipitation average weather impacts on consumption •  Low temperature and high precipitation expenditures would be approximately •  High temperature and low precipitation –11 percent for Bangladesh, –13 percent for •  High temperature and high precipitation Nepal, and –12 percent for Sri Lanka. These From these four climate uncertainty esti- results are much worse than the MMM results. mates, the two that represent the highest and Furthermore, it is well accepted that MMM lowest effects are chosen to bracket the uncer- scenarios are more likely than any of the cli- tainty due to climate models. This is an overrep- mate model end member projections (that is, resentation of the climate modeling uncertainty high or low in figure 3.2). Therefore, the living because it is based on 100 percent confidence standards effects based on the MMM climate intervals, which could include outliers. projections are also the most likely. 48  SOUTH AS I A ’ S HOTS P OTS FIGURE 3.2  Uncertainties of the Predicted Consumption Changes Arise from Differences between Climate Models and Economic Modeling a. Climate-sensitive, 2030 b. Carbon-intensive, 2030 20 20 15 15 Consumption change (%) Consumption change (%) 10 10 5 5 0 0 –5 –5 –10 –10 an sh a l n ka an sh a l n ka pa pa di di ta ta de an de an ist ist In In Ne Ne kis kis iL iL an an la la Pa Pa ng ng Sr Sr gh gh Ba Ba Af Af c. Climate-sensitive, 2050 d. Carbon-intensive, 2050 20 20 15 15 Consumption change (%) Consumption change (%) 10 10 5 5 0 0 –5 –5 –10 –10 an sh a l n ka an sh a l n ka pa pa di di ta ta de an de an ist ist In In Ne Ne kis kis iL iL an an la la Pa Pa ng ng Sr Sr gh gh Ba Ba Af Af Low Medium High 90% con dence interval Source: World Bank calculations. Note: The “medium” bar refers to results based on the MMM climate model projection. “High” and “low” depict uncertainty based on choice of climate model. Confidence intervals indicate econometric model uncertainty estimated using a robust standard error formulation. Notes 2. For example, CRU TS data are used to assess the historic performance of climate 1. There is an argument that consumption models in many studies referenced in the expenditures exclude consumption that is Intergovernmental Panel on Climate Change’s not based on market transactions. But given Fifth Assessment Report (IPCC 2013). the difficulties associated with collecting 3. Omitted variable bias is the effect on model information on these nonmarket values, con- prediction of not including a characteristic sumption expenditures are often used as a that explains significant differentiation best proxy for household living standards. between samples in a data set. C l i m a t e a n d Li v i n g S t a n d a r d s    49 4. Some effects of high temperatures or levels of Poor World).” Review of Economics and precipitation may have been captured if the Statistics 87 (1): 1–19. years of survey have such temperature or pre- Deaton, A., and M. Grosh. 2000. Designing cipitation. But the surveys are usually selected Household Survey Questionnaires for to take place outside such periods, so the Developing Countries: Lessons from Ten extent to which they are covered is limited. Years of LSMS Experience. Washington, DC: World Bank. Dell, M., B. F. Jones, and B. A. Olken. 2012. “Temperature Shocks and Economic Growth: References Evidence from the Last Half Century.” Acharya, S. P., and G. R. Bhatta. 2013. “Impact of American Economic Journal: Macroeconomics Climate Change on Agricultural Growth in 4 (3): 66–95. Nepal.” NRB Working Paper 15, Research Demombynes, G., C. Elbers, J. Lanjouw, Department, Nepal Rastra Bank, Kathmandu. P. Lanjouw, J. Mistiaen, and B. Özler. 2002. Ahmed, M., and S. Suphachalasai. 2014. “Producing an Improved Geographic Profile of “Assessing the Costs of Climate Change and Poverty: Methodology and Evidence from Adaptation in South Asia.” Asian Development Three Developing Countries.” WIDER Bank, Mandaluyong City, Philippines. Discussion Paper 2002/39, UNU-WIDER Atkinson, A. 1987. “On the Measurement of Helsinki. Poverty.” Econometrica 55: 749–64. Deschenes, Olivier. 2014. “Temperature, Human Angrist, J. D., and J. S. Pischke. 2008. Mostly Health, and Adaptation: A Review of the Harmless Econometrics: An Empiricist’s Empirical Literature.” Energy Economics 46: Companion . Princeton, NJ: Princeton 606–19. University Press. Dhakal, S., G. K. Sedhain, and S. C. Dhakal. 2016. Auffhammer, M., and W. Schlenker. 2014. “Climate Change Impact and Adaptation “Empirical Studies on Agricultural Impacts Practices in Agriculture: A Case Study of and Adaptation.” Energy Economics 46: Rautahat District, Nepal.” Climate 4 (4): 63. 555–61. Dormann, C. F., J. Elith, S. Bacher, C. Buchmann, Booth, G. D., M. J. Niccolucci, and E. G. Schuster. G. Carl, G. Carré, J. R. G. Márquez, B. Gruber, 1994. “Identifying Proxy Sets in Multiple B. Lafourcade, P. J. Leitão, T. Münkemüller, Linear Regression: An Aid to Better Coefficient C. McClean, P. E. Osborne, B. Reineking, Interpretation.” US Department of Agriculture, B. Schröder, A. K. Skidmore, D. Zurell, and Forest Service, Intermountain Research Station, S. Lautenbach. 2013. “Collinearity: A Review Ogden, UT. of Methods to Deal with It and a Simulation Brown, C., R. Meeks, Y. Ghile, and K. Hunu. Study Evaluating Their Performance.” 2013. “Is Water Security Necessary? An Ecography 36 (1): 27–46. Empirical Analysis of the Effects of Climate Elbers, C., J. O. Lanjouw, and P. Lanjouw. 2003. Hazards on National-Level Economic “Micro-Level Estimation of Poverty and Growth.” Philosophical Transactions of the Inequality.” Econometrica 71 (1): 355–64. Royal Society 371: 1–19. Elith, J., C. H. Graham, R. P. Anderson, M. Dudík, Burke, M., S. M. Hsiang, and E. Miguel. 2015. S. Ferrier, A. Guisan, R. J. Hijmans, “Global Non-Linear Effect of Temperature on F. Huettmann, J. R. Leathwick, A. Lehmann, Economic Production.” Nature 527 (7577): J. Li, L. G. Lohmann, B. A. Loiselle, G. Manion, 235–39. C. Moritz, M. Nakamura, Y. Nakazawa, Chatterjee, U., R. Murgai, A. Narayan, and J. M. Overton, A. T. Peterson, S. J. Phillips, M. Rama. 2016. Pathways to Reducing Poverty K. Richardson, R. Scachetti-Pereira, R. E. and Sharing Prosperity in India. Washington, Schapire, J. Soberón, S. Williams, M. S. Wisz, DC: World Bank. and N. Zimmermann. 2006. “Novel Methods DARA. 2012. “Climate Vulnerability Monitor Improve Prediction of Species’ Distributions (2nd edition): A Guide to the Cold Calculus of from Occurrence Data.” Ecography 29: a Hot Planet.” DARA. http://daraint.org​ 129–51. /­climate-vulnerability-monitor/climate-vulnera​ Hallegatte, S., M. Fay, M. Bangalore, T. Kane, and bility-monitor-2012/data. L. Bonzanigo. 2016. Shock Waves: Managing Deaton, A. 2005. “Measuring Poverty in a the Impacts of Climate Change on Poverty . Growing World (or Measuring Growth in a Washington, DC: World Bank. 50  SOUTH AS I A ’ S HOTS P OTS Hallegatte, S., C. Green, R. J. Nicholls, and Mani, M., and L. Wang. 2014. “Climate Change J. Corfee-Morlot. 2013. “Future Flood Losses and Health Impacts: How Vulnerable Is in Major Coastal Cities.” Nature Climate Bangladesh and What Needs to Be Done.” Change 3 (9): 802–6. World Bank, Washington, DC. Hallegatte, S., A. Vogt-Schilb, M. Bangalore, and Massetti, E., R. Mendelsohn, and S. Chonabayashi. J. Rozenberg. 2017. Unbreakable: Building 2016. “How Well Do Degree Days over the the Resilience of the Poor in the Face of Growing Season Capture the Effect of Climate N a t u r a l D i s a s t e r s . Wa s h i n g t o n , D C : on Farmland Values?” Energy Economics 60: World Bank. 144–50. Harris, I. P. D. J., P. D. Jones, T. J. Osborn, and Mendelsohn, R., W. D. Nordhaus, and D. Shaw. D .   H . L i s t e r. 2 0 1 4 . “ U p d a t e d H i g h - 1994. “The Impact of Global Warming on Resolution Grids of Monthly Climatic Agriculture: A Ricardian Analysis.” American Observations—The CRU TS3.10 Dataset.” Economic Review 84 (4): 753–71. International Journal of Climatology 34 (3): Mendelsohn, R., and M. Reinsborough. 2007. 623–42. “A Ricardian Analysis of US and Canadian Haughton. J., and S. Khandker. 2009. Handbook Farmland.” Climate Change 81: 9–17. on Poverty and Inequality. Washington, DC: Moore, F. C., and D. B. Diaz. 2015. “Temperature World Bank. Impacts on Economic Growth Warrant Heal, G., and J. Park. 2015. “Goldilocks Stringent Mitigation Policy.” Nature Climate Economies? Temperature Stress and the Direct Change 5: 127–31. Impacts of Climate Change.” NBER Working OECD (Organisation for Economic Co-operation Paper 21119, National Bureau of Economic and Development). 2015. The Economic Research, Cambridge, MA. Consequences of Climate Change. Paris: OECD. Hsiang, S. M., and A. S. Jina. 2014. “The Causal Poudel, S., and R. Shaw. 2016. “The Relationships Effect of Environmental Catastrophe on Long- between Climate Variability and Crop Yield in Run Economic Growth: Evidence from 6,700 a Mountainous Environment: A Case Study in Cyclones.” NBER Working Paper 20352, Lamjung District, Nepal.” Climate 4 (13). National Bureau of Economic Research, Schlenker, W., and M. J. Roberts. 2009. “Nonlinear Cambridge, MA. Temperature Effects Indicate Severe Damages IPCC (Intergovernmental Panel on Climate to US Crop Yields under Climate Change.” Change). 2013. Climate Change 2013: The Proceedings of the National Academy of Physical Science Basis—Contribution of Sciences 106 (37): 15594–8. Working Group I to the Fifth Assessment Skoufias, E., R. S. Katayama, and B. Essama- Report of the Intergovernmental Panel on Nssah. 2012. “Too Little Too Late: Welfare Climate Change . Cambridge: Cambridge Impacts of Rainfall Shocks in Rural Indonesia.” University Press. Bulletin of Indonesian Economic Studies 48 (3): Jacoby, H. G., M. Rabassa, and E. Skoufias. 2011. 351–68. “Distributional Implications of Climate Change Suzuki, N., D. H. Olson, and E. C. Reilly. 2008. in India.” Policy Research Working Paper 5623, “Developing Landscape Habitat Models for World Bank, Washington, DC. Rare Amphibians with Small Geographic Joshi, N. P., M. K. Lall, and P. Luni. 2011. “Effect Ranges: A Case Study of Siskiyou Mountains of Climate Variables on Yield of Major Food- Salamanders in the Western USA.” Biodiversity Crops in Nepal—A Time-Series Analysis.” and Conservation 17 (9): 2197–218. Journal of Contemporary India Studies: Space Tabachnick, B. G., and L. S. Fidell. 1996. Using and Society 1: 19–26. Multivariate Statistics. New York: HarperCollins. Li, Y., and M. Rama. 2016. “Households or WHO (World Health Organization). 2015. Locations? Cities, Catchment Areas and Quantitative Risk Assessment of the Potential Prosperity in India.” Policy Research Working Effects of Climate Change on Health. Geneva: Paper 7473, World Bank, Washington, DC. WHO. Mapping Hotspots 4 A lthough climate change is occurring standards is implemented at the household throughout South Asia (chapter 2), level. These results are then aggregated at the the effects of these changes are district, province, or national levels to deter- expected to be heterogeneous. For example, mine aggregate effects. The use of the term long-term temperature increases are predicted hotspot predominantly refers to the impact of to have a positive effect on household living changes in average weather aggregated to the standards in Nepal, even though the aggre- district level in each of the six countries for gate effects will be negative for Bangladesh, which survey data are available. To qualify the India, Pakistan, and Sri Lanka (chapter 3). magnitude of a hotspot, the predicted changes The effects of changes in average weather– in living standards are calculated in per capita long-term average seasonal temperature and percentage terms and defined in three hotspot precipitation–on household living standards levels: mild, moderate, and severe (see defini- also vary by household and location within tions in table 4.1). countries, resulting in some hotspots—that is, South Asian megacities such as Chennai, locations predicted to be negatively affected Dhaka, Karachi, Kolkata, and Mumbai are by changes in average weather. As a result, often identified as being climate hotspots understanding the granularity—the spatial or vulnerable to extreme events and sea- locations and d ­ istributions—of climate level rise, including coastal flooding and impacts on living standards is necessary storm surges. The hotspots defined in this to understand the challenges faced in book only account for changes in average these locations and to design appropriate weather and therefore are complementary interventions. to locations categorized as hotspots as a result of the effects of extreme events and sea-level rise. What Is a Hotspot? Hotspots are the result of two interrelated In this book, a hotspot is defined as a location factors: (a) the magnitude and seasonality of where changes in average weather will have a climate change; and (b) the relationship negative effect on living standards (the model between climate and living standards at a is described in chapter 3). The model linking given location. Since the model used to pre- changes in average weather and living dict hotspots is unique for each country in 51 52  SOUTH AS I A ’ S HOTS P OTS TABLE 4.1  Hotspot Labels and Definitions number and greater the severity of hotspots (maps 4.1, panels a and b, and 4.2). This indi- Hotspot label Definition cates that efforts to decrease the development Severe Living standards decline of more than of hotspots—for example, through global 8 percent mitigation of climate change or location-­ Moderate Living standards decline of 4 percent to 8 percent specific hotspot interventions—can posi- Mild Living standards decline of 0 percent to tively affect living standards throughout 4 percent the region. Note: Hotspots are locations where living standards are negatively Even by 2030 most locations in South affected by changes in average weather. Asia are predicted to become mild or moder- ate hotspots (map 4.1). The overall picture of the region is that of concentric rings, with South Asia, the relationship between changes the outer ring (the coastal areas of India and in average weather and living standards varies districts in the mountains along the northern between countries. This means that the model border of South Asia) not emerging as implicitly captures the effects of differences in hotspots, whereas the areas closer to the cen- policies or regulations between countries as ter of India are more affected. In a broad well as some degree of inherent adaptive sense, therefore, low-lying inland areas capability by households and communities to appear to be more fragile to changes in aver- changes in average weather. age weather than regions along the coast or Identifying hotspots is not as simple as in the mountains, where the historic climate identifying the regions where changes in aver- is relatively cold. However, coastal areas are age weather are projected to be the largest. susceptible to rising oceans, stronger storms, Even if climate were to change by similar and other climate-related effects (see box 4.1 magnitudes in two locations, the response for a further exploration of these issues). (including whether the locations become The pattern of hotspots does not appear to hotspots) depends on the historic relationship correspond significantly to major river between climate and living standards at the basins in a manner not explained by differ- two locations. For example, if two countries ences between countries (appendix C, are identical except that one relies more heav- maps C.7 and C.8). ily on agriculture for household income, the The predictions for 2030 are very simi- magnitude of the coefficients for climate vari- lar between the climate-sensitive and car- ables during the growing season(s) are bon-intensive scenarios (map 4.1, panels a expected to be larger in the agriculture-heavy and b). The reason for the similarity is that country, all other factors held equal. the cumulative emission differences of the Therefore, the emergence of hotspots should climate-­ sensitive and carbon-intensive sce- vary relatively smoothly within a given coun- narios take time to accumulate and result try, but may include discontinuities across in different magnitudes of climate change. borders of different countries. The climate change scenarios diverge around 2050, with effects beginning to level off under the climate-sensitive The Carbon-Intensive Scenario scenario (IPCC 2013). Under the carbon- Leads to More Severe Hotspots intensive scenario, climate changes con- The precise effects of changes in average tinue accruing through the end of the weather on living standards vary depending century. on the climate change scenario, model, and By 2050, many severe hotspots emerge time frame. However, a general theme is that under the carbon-intensive scenario, while the the larger the magnitude of climate changes— climate-sensitive scenario primarily contains either from looking further into the future or moderate hotspots (map 4.2, panels a and b). a more extreme scenario—the higher the Under the carbon-intensive scenario, M a p p i n g H o t s p o t s    53 MAP 4.1  Mild and Moderate Hotspots Are Prevalent Throughout South Asia by 2030 Source: World Bank calculations. Note: These results are based on the mean of the 11 climate model ensemble used in this book. MAP 4.2  Moderate and Severe Hotspots Cover a Significant Portion of South Asia by 2050 Source: World Bank calculations. Note: Achieving the climate-sensitive scenario (panel a) would mostly prevent the emergence of severe hotspots through 2050, compared to the carbon-intensive scenario (panel b). These results are based on the mean of the 11 climate models used in this book. Grey areas are those where insufficient data are available. 54  SOUTH AS I A ’ S HOTS P OTS BOX 4.1  Will Mountain and Coastal Areas Benefit from Climate Change? The hotspots analyses conducted in this book pre- headwaters. The effects, however, will be stron- dict that many mountains and coastal areas will gest in and around the mountains. In mountain not be negatively affected by changes in average communities where all water resources come weather. This is not the same as finding that these from melting snow and glaciers, the effects areas will benefit overall from climate change. For could be devastating to personal well-being and example, the predictions do not account for spe- agriculture. cific aspects of climate change that will have nega- tive effects. The following is a breakdown of Coastal Areas general negative dimensions of climate change that are not accounted for in the analysis. Climate change is leading to rising sea levels (Asuncion and Lee 2017; Hallegatte and others 2013; Nicholls and others 2007). This is an Mountain Areas existential threat to several coastal areas in Climate change will likely affect the frequency of South Asia, including all of the Maldives, sig- natural disasters in mountain areas (Chen and nificant portions of Bangladesh, and selected others 2010; Keiler, Knight, and Harrison 2010; regions of coastal India. Rising sea levels can Stoffel and Huggel 2012). This includes increas- submerge certain areas and worsen storm ing the likelihood of events such as landslides and surges, leading to more flooding during extreme glacier lake outburst floods. Such events can have weather events. devastating effects, such as—in extreme Climate change will likely increase the sever- instances—destroying entire villages. ity of tropical storms, which will increase dam- Climate change will have devastating effects ages in affected coastal areas (Mendelsohn and on biodiversity in mountain areas (Bellard and others 2012). Because of the large natural vari- others 2012; Chakraborty and Newton 2011; ability of extreme storm events, no conclusive Siraj and others 2014; Xu and others 2009). This statistical evidence exists indicating that they means that the natural ecosystems that people in will become more frequent in the future. mountains are accustomed to will potentially However, there is a good physical basis for pre- change more rapidly than communities and life- dicting that storm events will at least become styles are able to adapt. For example, certain stronger, if not more frequent. The main reason pests and disease vector organisms will flourish, is that sea surface temperatures are warming. causing a negative effect on agriculture. Warmer oceans mean there is more energy avail- Climate change will affect the timing and able to fuel storms once they materialize. Cities stability of snow and glacier melt (Immerzeel, in coastal areas are also rapidly growing around Van Beek, and Bierkens 2010; Miller, Immerzeel, the world. This increases potential economic and Rees 2012; Xu and others 2009). These damage from a storm event of a given magni- changes will affect all communities that rely on tude, since there is simply more property in the freshwater resources originating in mountain storm’s path. moderate and severe hotspots are predomi- with inland areas predicted to be more nantly in central India, northern Sri Lanka, affected than coastal or mountain areas. The and southeastern Bangladesh. The overall spatial pattern of these hotspots predictions is pattern of hotspots predicted for 2050 under very similar to the recent estimate of heat vul- both scenarios is similar to that for 2030, nerability in India (box 4.2). M a p p i n g H o t s p o t s    55 BOX 4.2  Heat Vulnerability Index for India High heat vulnerability index districts are MAP B4.2.1  Central India Is the Most those for which a larger portion of individuals Vulnerable to Heat experience heat-related medical incidents. Interestingly, districts with high heat vulnera- bilities are not necessarily those with the high- est temperatures (map B4.2.1 compared to map 2.1). One reason for this is that people have the capacity to adapt to changes in their environment, and this capacity is a by-product of their environment. Therefore, the most vul- nerable are those with low adaptive capacities and sufficient temperatures to trigger a health problem. Azhar and others (2017) also find that districts with high heat vulnerability index values are less urbanized, have lower literacy rates, have less access to water and sanitation, and have fewer household amenities. Source: Azhar and others 2017. Note: Larger values indicate high vulnerability. Mountainous regions do not emerge as 45 percent of the region’s population. hotspots under this analysis because these Under the climate-sensitive scenario, the regions are the coldest; therefore, some degree number of people affected is 375 million— of warming may be beneficial for them. That or 21 percent of the population. These does not mean that climate change will numbers do not account for Bhutan or the always have positive effects in these zones Maldives, which will also be affected by cli- (see box 4.1). For example, the economies of mate change. The solutions will vary based these mountain regions rely extensively on on the location and country contexts, but streamflow from snow and glaciers. Warming clearly this is a challenge that must be will affect the timing and availability of these addressed. water resources, which could have profound effects (Bolch and others 2012; Immerzeel, Van Beek, and Bierkens 2010). In addition, Hotspots Tend to Have Less mountain regions are often highly vulnerable Infrastructure and Services to natural disasters. Hotspots tend to have less infrastructure Although the conditions leading to and worse integration with the broader soci- hotspots vary within countries and across ety (table 4.3, appendix C, maps C.1 the region, the estimated effects are unam- through C.6). For example, the average biguous: approximately 800 million people household residing in a severe hotspot by in South Asia today live in locations that 2050 under the carbon-intensive scenario has could become moderate or severe hotspots an average road density of 1.5 kilometers (per by 2050 under the carbon-intensive sce- 10 square kilometers of area) compared with nario (table 4.2). This is equivalent to the overall density of 2.1 kilometers of road. 56   SOUTH ASIA’S HOTSPOTS TABLE 4.2  Millions of South Asians Are Living in Areas Projected to Become Hotspots Hotspot category Afghanistan Bangladesh India Nepal Pakistan Sri Lanka South Asia Severe 0.0 26.4 148.3 0.0 0.0 3.6 178.4 Moderate 0.0 107.9 440.9 0.0 48.7 14.9 612.4 Mild 0.0 20.4 399.9 0.0 144.5 2.6 567.4 Total population 34.7 163.0 1,324.2 29.0 193.2 21.2 1,765.2 Sources: World Bank calculations based on WDI (population data); World Bank 2016. Note: Estimates are based upon the carbon-intensive scenario by 2050. Data show that around 800 million people live in moderate or severe hotspots. TABLE 4.3  Locational Characteristics, by Hotspot Category Hotspot Living Average Average category / standards road density population Travel time to Country overall Households (%) change (%) (km/10 km2) (per km2) market (hours) Afghanistan Severe 0.1 –10.4 0.0 1.0 36.3 Afghanistan Moderate n.a. n.a. n.a. n.a. n.a. Afghanistan Mild n.a. n.a. n.a. n.a. n.a. Afghanistan Overall 100.0 11.9 2.9 951.6 5.2 Bangladesh Severe 16.2 –14.4 6.1 1,043.1 2.2 Bangladesh Moderate 66.2 –6.4 6.1 1,539.1 1.9 Bangladesh Mild 12.5 –1.5 4.4 795.7 1.7 Bangladesh Overall 100.0 –6.7 5.8 1,320.6 2.0 India Severe 11.2 –9.8 0.8 231.3 2.7 India Moderate 33.3 –5.6 1.7 1,005.8 2.1 India Mild 30.2 –2.3 1.6 1,119.1 2.7 India Overall 100.0 –2.8 1.6 840.7 2.7 Pakistan Moderate 25.2 –4.6 0.7 205.1 3.1 Pakistan Mild 74.8 –2.4 1.6 448.4 3.7 Pakistan Overall 100.0 –2.9 1.4 387.0 3.6 Sri Lanka Severe 17.2 –10.5 10.4 254.6 2.6 Sri Lanka Moderate 70.3 –7.1 13.2 865.8 2.5 Sri Lanka Mild 12.4 –4.0 19.5 451.7 2.6 Sri Lanka Overall 100.0 –7.0 13.5 708.9 2.6 South Asia Severe 11.6 –10.2 1.6 335.6 3.3 South Asia Moderate 32.6 –5.6 2.1 987.1 2.1 South Asia Mild 32.9 –2.3 2.2 990.2 2.9 South Asia Overall 100.0 –3.2 2.1 831.3 2.8 Source: World Bank calculations. Note: Under the carbon-intensive scenario in 2050. n.a. = not applicable. Severe hotspots are also less densely popu- individually. In India and Pakistan, water- lated than average (319 people per square stressed areas will also be more adversely kilometer for hotspots, compared with 829 affected compared with the national average. people per square kilometer overall). Travel time to market is not related in any Moreover, these population, infrastructure, clear way to the severity of hotspots. Although and integration differences hold for this is based on observing the correlation and Bangladesh, India, Pakistan, and Sri Lanka not attributing any causality, one could M a p p i n g H o t s p o t s    57 potentially look more deeply at other social compared with 10.2 percent overall). and economic characteristics of these hotspots In Bangladesh, the households that will be the to see whether better services and infrastruc- most affected are more likely to be headed by tures would reduce the effect of climate a woman (9.9 percent, compared with change on living standards. 7.6 percent), whereas the likelihood is approx- imately equal in Nepal and Sri Lanka. This needs further analysis, but could occur The Most Vulnerable Households because female-headed households find it Changes in average weather will affect house- harder to survive and move out of severely holds to different degrees. Although the affected areas. hotspot analysis investigates the overall rela- In all countries except Bangladesh, the tionship between household living standards households most affected by changes in aver- and changes in average weather throughout age weather are more likely to be engaged in the region, understanding the characteristics agriculture as their main livelihood (table 4.4). of vulnerable households can be informative This is particularly important to note because for developing targeted policies. The results agriculture will be affected by climate change discussed in this section are for 2050 hotspots in multiple ways not accounted for by the predictions under the carbon-intensive analysis. For example, higher temperatures scenario. increase the likelihood of droughts, which can Taking all households in South Asia have devastating effects on crops, yet these together, those that are most affected by types of events are not fully captured in the changes in average weather are less likely to methodology. be headed by a woman (table 4.4). This result The relationship between hotspots and agri- appears to be driven mainly by India culture varies between countries. For South (7.4 percent headed by a woman in severe Asia as a whole, the heads of 46.9 percent of hotspots, compared with 10.8 percent overall) the households in severe hotspots derive their and Pakistan (2.1 percent in severe hotspots, primary income from agriculture, compared TABLE 4.4  Characteristics of the Most Affected Households Compared with National Averages Living Education of Severe / standards Female-headed Agriculture head of Electricity Country overall change (%) household (%) head (%) household (years) (%) Afghanistan Severe –10.4 0.0 70.0 2.7 0.0 Afghanistan Overall 11.9 0.7 31.2 3.2 27.0 Bangladesh Severe –12.9 8.2 27.4 4.2 60.3 Bangladesh Overall –6.7 7.6 39.1 3.9 54.9 India Severe –9.8 7.4 51.0 5.7 91.3 India Overall –2.8 10.8 39.8 5.7 79.8 Nepal Severe n.a. n.a. n.a. n.a. n.a. Nepal Overall 4.1 26.7 52.6 3.8 70.0 Pakistan Severe n.a. n.a. n.a. n.a. n.a. Pakistan Overall –2.9 10.2 24.0 5.3 13.6 Sri Lanka Severe –9.7 22.9 26.1 8.5 88.2 Sri Lanka Overall –7.0 22.5 28.6 8.3 90.6 South Asia Severe –10.2 7.6 48.6 5.5 86.1 South Asia Overall –3.2 10.7 38.0 5.4 69.6 Source: World Bank calculations. Note: Under the carbon-intensive scenario in 2050. n.a. = not applicable. 58   SOUTH ASIA’S HOTSPOTS with 38.1 percent overall (table 4.4). This pat- households compared with the national tern exists in India, Nepal, Pakistan, and average. Seven of the 10 top hotspot districts Sri Lanka, whereas the reverse pattern is seen are in Chittagong Division, with Cox’s Bazar in Bangladesh. Because this is based on predicted to experience the largest negative observed correlation, one cannot attribute effects (table 4.6). any causality between the socioeconomic char- Although low-lying coastal areas in acteristics of households and climate vulnera- Chittagong have received a lot of attention in bility. However, one can safely say that climate Bangladesh due to weather events, hill tracts adds another dimension to the existing vulner- in Chittagong also emerge as vulnerable to ability of households living in hotspots. changes in average weather. Over the years, the hill tracts have become hotspots for out- breaks of vector-borne diseases. In addition, Country Hotspots deforestation and hill-cutting have affected A strength of this hotspot analysis is being the hill slopes considerably, resulting recently able to explain the spatial dimensions of in major landslides and destruction of prop- changes in average weather effects on living erty. Cox’s Bazar has gone through a major standards throughout South Asia. This sec- environmental upheaval in recent years and is tion lists the most affected states or provinces now also embroiled in a social crisis due to and districts for Afghanistan, Bangladesh, the influx of Rohingya refugees from neigh- India, Nepal, Pakistan, and Sri Lanka, and boring Myanmar. Chittagong city, which describes some of the relevant country emerges as the third-most-vulnerable city in context. Bangladesh, is also the second-largest city in County context matters in terms of the country. It is the busiest seaport in the the development of hotspots (maps 4.1, panels region and a major economic hub, attracting and b, and 4.2). For example, Ramanathapuram strong inflows of foreign investment to pro- District in India and Jaffna District in Sri Lanka duce apparel, decommission ships, and refine are separated by only about 100 kilometers, oil. Going forward, climate vulnerability will meaning that their climates are relatively simi- have huge economic implications for the lar. Yet Ramanathapuram does not emerge as a growing city. hotspot, whereas Jaffna emerges as a moderate to severe hotspot (depending on the time hori- India zon and climate change scenario). Given that the underlying regression model is empirical, States in the central, northern, and northwest- the modeled relationship implicitly reflects the ern parts of India emerge as the most vulner- aggregate differences in how climate affects able to changes in average weather. household living standards in the two Chhattisgarh and Madhya Pradesh, which are countries. predicted to experience a decline in living standards of more than 9 percent, are the top two hotspot states, followed by Rajasthan, Bangladesh Uttar Pradesh, and Maharashtra (table 4.7). In Bangladesh, Chittagong Division emerges In addition to being poverty hotspots, as the most vulnerable to changes in average Chhattisgarh and Madhya Pradesh are home weather, followed by Barisal and Dhaka to large tribal populations. Coastal areas in divisions (table 4.5). Chittagong is relatively India receive a lot of attention due to extreme more developed in terms of infrastructure storms and flooding. However, here the compared with the national average and is inland areas emerge as hotspots due to characterized by fewer households in which changes in average weather. the head of the household is engaged in agri- Seven out of the top 10 most affected culture. It is relatively densely populated, hotspot districts belong to the Vidarbha with a greater number of female-headed region of Maharashtra State, with TABLE 4.5  Predicted Change in Living Standards and Characteristics of Divisions in Bangladesh Living Average length Average Travel time Female- Share of standards of road in population to market Water headed Agriculture Years of Electricity Division households(%) change (%) km/10 km2 density per km2 (hours) availability household (%) head (%) education (%) Chittagong 16.2 −14.4 6.1 1,043.1 2.2 31.3 10.0 34.4 4.2 58.5 Barisal 6.0 −7.4 8.5 680.6 4.1 34.4 6.4 31.8 4.1 39.4 Dhaka 32.9 −6.9 7.1 2,330.9 1.5 0.4 6.9 29.0 4.3 67.5 Khulna 13.1 −6.7 3.7 686.0 2.3 0.7 6.1 44.7 3.9 54.8 Rajshahi 14.2 −4.6 5.2 851.9 1.6 1.9 7.1 53.0 3.4 52.7 Rangpur 12.5 −1.5 4.4 795.7 1.7 1.0 7.0 54.8 3.7 30.7 Sylhet 5.1 0.8 4.4 645.7 2.5 3.5 12.6 36.1 2.9 44.9 Overall 100.0 −6.7 5.8 1,320.6 2.0 7.9 7.6 39.1 3.9 54.9 Source: World Bank calculations. Note: Under the carbon-intensive scenario in 2050. TABLE 4.6  Predicted Change in Living Standards and Characteristics of the Top 10 District Hotspots in Bangladesh Living Average length Average Travel time Female- Share of standards of road in population to market Water headed Agriculture Years of Electricity District Division households (%) change (%) km/10 km2 density per km2 (hours) availability household (%) head (%) education (%) Cox’s Bazar Chittagong 0.9 −20.2 8.3 812.9 2.7 5.8 7.6 43.5 2.6 25.9 Bandarban Chittagong 1.1 −18.4 1.9 73.6 4.5 3.4 8.5 55.4 2.8 39.0 Chittagoun Chittagong 4.8 −18.1 6.7 1,395.4 1.6 55.5 10.7 14.7 5.9 79.3 Rangamati Chittagong 1.1 −15.8 1.3 91.9 3.6 1.9 10.7 54.8 4.0 29.1 Noakhali Chittagong 1.3 −14.8 4.9 926.4 2.3 62.0 6.9 38.9 3.7 44.9 Feni Chittagong 0.8 −13.5 8.0 1,312.6 1.5 3.1 9.4 33.8 4.8 72.0 Khagrachhari Chittagong 1.0 −12.6 6.0 190.3 3.2 2.1 7.2 42.6 3.4 43.3 Barguna Barisal 0.9 −12.5 4.3 524.3 5.7 0.9 5.8 30.2 4.3 22.6 Bagerhat Khulna 1.1 −12.0 3.2 368.1 3.6 1.2 5.6 35.2 4.5 36.8 M a p p i n g H o t s p o t s    Satkhira Khulna 1.3 −11.5 2.7 490.3 3.1 1.9 5.4 42.9 4.6 41.4 Overall 100.0 −6.7 5.8 1,320.6 2.0 7.9 7.6 39.1 3.9 54.9 Source: World Bank calculations. Note: Under the carbon-intensive scenario in 2050. 59 60  SOUTH AS I A ’ S HOTS P OTS TABLE 4.7  Predicted Change in Living Standards and Characteristics of the 10 Most Affected States in India Living Average length Average Travel time Female- Share of standards of road in population to market Water headed Agriculture Years of Electricity State households change (%) km/10 km2 density per km2 (hours) availability household (%) head (%) education (%) Chhattisgarh 2.1 −9.4 1.0 212.7 2.9 0.3 6.3 60.7 5.5 89.5 Madhya pradesh 5.7 −9.1 1.0 237.0 2.6 0.4 6.0 48.5 5.4 88.4 Rajasthan 5.9 −6.4 0.7 229.4 2.6 0.1 9.4 36.8 4.8 82.7 Uttar Pradesh 15.8 −4.9 1.4 801.3 1.9 0.9 10.5 42.9 5.1 51.7 Maharashtra 10.0 −4.6 1.0 325.6 2.7 0.4 9.4 40.3 7.1 94.2 Jharkhand 2.0 −4.6 1.6 482.4 2.0 3.5 8.2 30.6 5.2 74.3 Haryana 2.1 −4.3 2.3 480.5 2.6 0.2 7.4 36.2 6.6 96.6 Andhra Pradesh 10.2 −3.4 2.1 1,831.3 2.6 2.3 14.3 41.2 5.2 98.2 Punjab 1.8 −3.3 2.1 464.6 3.5 0.2 12.4 23.5 5.7 98.6 Chandigarh 0.1 −3.3 5.1 4,529.6 1.5 0.1 6.2 0.2 8.9 97.9 Overall 100.0 −2.8 1.6 840.7 2.7 2.0 10.8 39.8 5.7 79.8 Source: World Bank calculations. Note: Under the carbon-intensive scenario in 2050. TABLE 4.8  Predicted Change in Living Standards and Characteristics of the Top 10 District Hotspots in India Living Average Average Travel time Female- Share of standards length of road population to market Water headed Agriculture Years of Electricity District State household(%) change (%) in km/10 km2 density per km2 (bours) availability household (%) head (%) education (%) Chandrapur Maharashtra 0.2 −12.4 1.2 161.6 1.7 3.1 8.7 50.6 6.8 84.6 Bhandara Maharashtra 0.1 −11.9 0.8 219.7 2.5 0.3 5.3 51.9 7.2 93.1 Gondiya Maharashtra 0.1 −11.8 0.8 215.9 2.5 0.2 9.5 51.2 7.0 96.6 Wardha Maharashtra 0.1 −11.8 0.5 172.0 2.6 0.1 9.8 53.1 8.3 93.6 Nagpur Maharashtra 0.5 −11.7 0.2 379.9 2.3 0.1 7.7 17.7 8.8 97.2 Raj Nandagaon Chhattisgarh 0.1 −11.4 1.5 153.0 3.8 0.1 1.8 59.2 4.4 98.0 Durg Chhattisgarh 0.3 −11.4 0.5 314.4 2.3 0.2 10.6 43.7 7.1 94.3 Hoshangabad Madhya Pradesh 0.1 −11.3 1.3 144.1 3.6 0.6 0.2 40.0 5.8 91.2 Yavatmal Maharashtra 0.3 −11.1 0.3 169.3 2.3 0.1 4.4 67.7 5.4 83.0 Garhchiroli Maharashtra 0.1 −11.1 0.8 61.8 2.5 7.7 9.1 74.0 5.1 81.1 Overall 100.0 −2.8 1.6 840.7 2.7 2.0 10.8 39.8 5.7 79.8 Source: World Bank calculations. Note: Under the carbon-intensive scenario in 2050. M a p p i n g H o t s p o t s    61 the remaining three districts located in Punjab Province, which is the most Chhattisgarh and Madhya Pradesh densely populated province, is also the (table 4.8). s econd-most vulnerable. Punjab has the ­ largest economy in Pakistan (contributing ­ 53.3 percent of Pakistan’s GDP), and over- Sri Lanka all has the lowest rate of poverty of all the The Northern and North Western provinces provinces. However, the prosperity is of Sri Lanka emerge as the top two hotspots, unevenly distributed throughout the prov- followed by the much less densely populated ince, with the northern portion being rela- North Central Province (table 4.9). Northern tively well off economically and the Province is home to a large number of poor southern portion among the most impover- and displaced people. The effects of climate ished in the country. Long-term climate vul- change will add a challenge to this long-term nerability has implications for both growth recovery. The highly urbanized and densely and poverty reduction for Punjab. populated Western Province, which includes Hyderabad District in Sindh emerges as Colombo, is also predicted to experience a the top hotspot followed by the districts of 7.5 percent decline in living standards by Mirpur Khas and Sukkur (table 4.12). Some 2050. This has huge economic implications of the densely populated cities in Punjab, for the country, especially since the province including Lahore, Multan, and Faisalabad, contributes more than 40 percent of Sri emerge among the top 10 hotspot districts. Lanka’s gross domestic product (GDP). This highlights the importance of addressing Among the districts, Jaffna emerges as the changes in average weather in the economi- top hotspot, followed by Puttalam in North cally important Punjab and Sindh We s t e r n P r o v i n c e a n d M a n n a r a n d provinces. Kilinochchi in Northern Province (table 4.10). Given that 5 of the 10 most vulnerable dis- Nepal tricts of Sri Lanka are in Northern Province, changes in average weather and vulnerability The Mid-Western, Western, and Far-Western must be considered for future planning and development regions of Nepal will benefit development activities there. Gampaha, the most from changes in average weather which is among the 10 most vulnerable dis- (table 4.13). Almost all districts in the Mid- tricts, is also the second-most-populous dis- Western Development, Western Development, trict in the country and was declared one of and Far-Western Development regions are at the worst-affected districts in the recent relatively high altitudes and are part of the droughts. trans-Himalayan corridor. The more densely populated Eastern Development and Central Development regions will benefit less from Pakistan changes in average weather because they are Sindh Province emerges as the most vulnera- at lower altitudes and currently have temper- ble hotspot in Pakistan, followed by Punjab atures closer to the optimum. (table 4.11). Sindh has the second-largest Mugu, Rasuwa, Solukhumbu, and economy, with a per capita GDP of US$1,400, Taplejung districts will be negatively which is 35 percent more than the national affected by changes in average weather average. The province has a highly diversified (table 4.14). The rest of the districts in economy ranging from heavy industry and Nepal are predicted to experience either finance centered in and around Karachi to a neutral or positive effects from warming substantial agricultural base along the Indus and change in long-term precipitation River. Changes in average weather will add p atterns. In contrast, Nepal is considered ­ another dimension to the future growth of extremely fragile to natural disasters and Sindh, given its high vulnerability. extreme climate events. 62   SOUTH ASIA’S HOTSPOTS TABLE 4.9  Predicted Change in Living Standards and Characteristics of Provinces in Sri Lanka under the Carbon-Intensive Scenario in 2050 Average Change Average length population Travel time Female Share of in living of road in density to market Water household Agriculture Years of Electricity Province households (%) standards (%) (km/10 km2) (per km2) (hours) availabilitya head (%) head (%) education (%) Northern 4.9 −11.2 9.3 264.8 2.5 0.1 21.0 34.7 8.0 65.8 North Western 12.3 −10.3 10.9 250.6 2.6 0.5 22.9 30.5 8.2 89.5 North Central 6.5 −8.0 5.2 98.8 2.8 0.3 23.4 50. 7.9 87.4 Western 28.0 −7.5 18.1 1,764.8 1.1 0.5 21.5 8.8 9.6 97.8 Eastern 7.4 −7.2 4.3 124.8 3.3 0.3 23.3 27.9 7.1 81.4 Southern 12.3 −7.1 13.1 446.7 3.2 0.6 23.4 34.7 8.0 94.9 Sabaragamuwa 9.7 −6.8 13.0 341.6 3.9 0.6 20.4 38.6 7.8 89.3 Uva 6.4 −4.6 11.1 176.6 4.5 0.2 21.1 53.4 7.3 83.9 Central 12.4 −4.0 19.5 451.7 2.6 0.1 25.0 32.0 7.9 92.9 Overall 100.0 −7.0 13.5 708.9 2.6 0.4 22.5 28.6 8.3 90.6 a. “Water availability” refers to the ratio of surface water use to groundwater use. A large value is good because it indicates that water use is more likely to be sustainable. TABLE 4.10  Predicted Change in Living Standards and Characteristics of the Top 10 District Hotspots in Sri Lanka under the Carbon-Intensive Scenario in 2050 Average Change Average length population Travel time Female Share of in living of road in density to market Water household Agriculture Years of Electricity District Province households (%) standards (%) (km/10 km2) (per km2) (hours) availabilitya head (%) head (%) education (%) Jaffna Northern 2.7 −11.9 13.4 433.0 1.6 0.1 22.4 30.3 8.2 76.2 Puttalam North Western 4.0 −10.4 8.1 196.0 2.8 0.4 23.9 31.3 7.5 86.5 Manar Northern 0.4 −9.8 2.9 41.1 3.6 0.1 14.0 51.7 7.7 71.0 Kilinochchi Northern 0.5 −9.5 6.9 89.0 3.8 0.1 14.1 38.6 7.7 26.1 Kurunegala North Western 8.3 −9.4 12.2 276.8 2.5 0.5 22.5 30.1 8.6 91.0 Trincomalee Eastern 1.8 −9.3 3.3 107.6 3.0 0.2 23.4 23.8 7.6 78.8 Gampaha Western 11.3 −8.9 28.2 1,401.2 0.8 0.7 23.0 6.9 9.5 97.9 Κegalle Sabaragamuwa 4.2 −8.7 16.9 427.6 3.8 0.4 24.4 26.2 8.4 90.7 Mullaitivu Northern 0.4 −8.3 2.6 41.0 4.5 0.1 19.4 53.3 7.0 35.0 Vavuniya Northern 0.8 −8.3 4.9 76.6 3.0 0.1 25.4 27.2 7.9 71.4 Overall 100.0 −7.0 13.5 708.9 2.6 0.4 22.5 28.6 8.3 90.6 a. “Water availability” refers to the ratio of surface water use to groundwater use. A large value is good because it indicates that water use is more likely to be sustainable. TABLE 4.11  Predicted Change in Living Standards and Characteristics of Provinces in Pakistan under the Carbon-Intensive Scenario in 2050 Average Share of Change Average length population Travel time Female households in living of road in density to market Water household Agriculture Years of Electricity Province (%) standards (%) (km/10 km2) (per km2) (hours) availabilitya head (%) head (%) education (%) Sind 25.2 −4.6 0.7 205.1 3.1 0.9 3.9 19.0 6.6 8.0 Punjab 59.0 −2.6 1.6 464.3 2.4 0.9 11.9 26.6 4.9 17.4 Khyber Pakhtukhwa 12.9 −1.7 1.6 455.6 9.1 0.2 16.5 21.4 4.3 9.2 Baluchistan 2.8 −1.3 0.1 79.5 7.1 0.0 0.7 25.2 4.5 5.6 Overall 100.0 −2.9 1.4 387.0 3.6 0.8 10.2 24.0 5.3 13.6 a. “Water availability” refers to the ratio of surface water use to groundwater use. A large value is good because it indicates that water use is more likely to be sustainable. TABLE 4.12  Predicted Change in Living Standards and Characteristics of the Top 10 District Hotspots in Pakistan under the Carbon-Intensive Scenario in 2050 Average Average Share of Change length of population Travel time Female households in living road in density to market Water household Agriculture Years of Electricity District Province (%) standards (%) (km/10 km2) (per km2) (hours) availabilitya head (%) head (%) education (%) Hyderabad Sindh 4.3 −6.0 0.0 175.5 3.9 0.4 1.3 31.1 4.5 2.8 Mirpur Khas Sindh 2.3 −5.7 0.0 151.2 4.6 0.0 2.2 41.8 3.9 1.8 Sukkur Sindh 6.9 −4.1 0.1 183.0 3.7 0.9 2.7 20.2 6.7 5.8 Larkana Sindh 11.8 −4.0 1.5 239.2 2.2 1.4 6.0 9.5 7.9 12.3 Bahawalpur Punjab 5.4 −3.2 0.1 187.8 4.3 0.6 7.8 49.6 2.6 2.6 Faisalabad Punjab 8.2 −2.8 2.7 581.6 1.6 0.1 11.4 30.4 5.2 7.8 Lahore Punjab 4.3 −2.7 2.5 1,088.2 1.4 0.4 9.0 21.2 4.5 3.1 Multan Punjab 8.1 −2.6 0.9 506.7 1.6 0.0 8.4 39.7 3.7 28.6 Dera Ghazi Khan Punjab 4.9 −2.6 0.5 197.5 3.9 2.3 10.6 35.0 3.2 36.7 M a p p i n g H o t s p o t s    Sargodha Punjab 9.0 −2.5 2.5 232.9 2.4 4.0 10.9 17.8 5.1 15.4 Overall 100.0 −2.9 1.4 387.0 3.6 0.8 10.2 24.0 5.3 13.6 a. “Water availability” refers to the ratio of surface water use to groundwater use. A large value is good because it indicates that water use is more likely to be sustainable. 63 64   SOUTH ASIA’S HOTSPOTS TABLE 4.13  Predicted Change in Living Standards and Characteristics of Regions in Nepal under the Carbon-Intensive Scenario in 2050 Average Share of Change Average length population Travel time Female households in living of road in density to market Water household Agriculture Yeras of Electricity Region (%) standards (%) (km/10 km2) (per km2) (hours) availabilitya head (%) head (%) education (%) Eastern 23.5 3.5 0.9 297.8 10.4 2.6 24.7 58.8 3.7 69.1 Centeral 35.6 3.9 1.8 843.7 7.8 0.9 21.2 44.9 4.1 78.9 Far-Western 8.6 4.4 0.0 124.4 11.6 1.8 34.6 53.1 3.7 51.8 Western 19.8 4.5 1.4 223.2 10.8 0.7 34.3 54.1 4.0 78.8 Mid-Western 12.4 4.9 0.5 126.6 8.5 0.7 28.7 60.2 3.1 45.0 Overall 100.0 4.1 1.2 441.5 9.4 1.3 26.7 52.6 3.8 70.0 a. “Water availability” refers to the ratio of surface water use to groundwater use. A large value is good because it indicates that water use is more likely to be sustainable. TABLE 4.14  Predicted Change in Living Standards and Characteristics of the Top 10 District Hotspots in Nepal under the Carbon-Intensive Scenario in 2050 Average Share of Change Average length population Travel time Female households in living of road in density to market Water household Agriculture Years of Electricity District Region (%) standards (%) (km/10 km2) (per km2) (hours) availabilitya head (%) head (%) education (%) Solukhumbu Eastern 0.5 −1.2 0.8 29.3 11.1 0.2 22.0 52.1 2.3 53.3 Mugu Mid-Western 0.2 −0.8 0.0 13.7 25.5 0.0 0.0 66.7 4.4 0.0 Rasuwa Central 0.2 −0.7 2.0 23.3 10.4 0.1 0.0 72.7 2.3 0.0 Taplejung East 0.2 −0.6 0.5 27.9 12.3 127.1 16.7 66.7 3.6 0.0 Sankhuwasabha East 0.9 0.0 1.2 36.6 8.8 0.2 25.2 61.6 4.0 69.9 Dolakha Central 1.0 0.3 1.2 69.3 8.4 0.1 30.2 70.1 2.6 87.3 Bajhang Far-Western 0.7 0.8 0.0 46.3 12.8 0.0 19.6 64.0 4.2 0.0 Sindhupalchok Central 1.1 1.4 2.0 92.5 10.4 0.2 32.5 59.8 2.0 96.3 Darchula Far-Western 0.5 2.1 0.1 47.2 9.8 10.0 21.7 43.5 4.5 43.5 Gorkha West 1.2 2.2 1.0 64.1 18.3 0.2 43.1 72.2 2.7 66.9 Overall 100.0 4.1 1.2 441.5 9.4 1.3 26.7 52.6 3.8 70.0 a. “Water availability” refers to the ratio of surface water use to groundwater use. A large value is good because it indicates that water use is more likely to be sustainable. M a p p i n g H o t s p o t s    65 Afghanistan Nonmonetary Indicators of Well-Being Only Wakhan District in northeastern Afghanistan is projected to emerge as a hotspot Although the focus of the book has been on by 2050 under either climate scenario living standards—as measured through con- (table 4.15). The districts with the least positive sumption expenditures—there are also non- effects of climate change are spread throughout monetary effects of climate change the country, with many in the central, moun- on well-being (Carleton and Hsiang 2016). tainous portions of the country (for example, in A growing literature links changes in Bamyan, Wardak, and Ghazni provinces). The temperature and precipitation patterns to ­ spatial pattern of climate change effects is simi- increasing crime rates, civil conflict, inter- lar at the provincial level, with many of the group riots, migration, and mortality least positively affected provinces being in the (box 4.3). It is possible that these could Hindu Kush mountains (table 4.16). be triggered by monetary effects such as The lack of infrastructure in the most negative rain shocks that lower income, affected districts and provinces is staggering which, in turn, increase the likelihood of (tables 4.15 and 4.16). For example, whereas violence. Nonmonetary effects could be a 27 percent of people in Afghanistan overall complement to living standards measures have access to electricity, between 0 and that focus on income and expenditures. This 5 percent have access in these areas; the is especially relevant as the climate hotspots extremely low density of paved primary roads identified in the book could also potentially (0 kilometers in many districts) and long aver- become hotspots for crime, violence, and age travel time to markets (more than 36 hours civil conflict. in Wakhan District) is similarly profound. 66   SOUTH ASIA’S HOTSPOTS TABLE 4.15  Predicted Change in Living Standards and Characteristics of the Top 10 Most Affected Districts in Afghanistan under the Carbon-Intensive Scenario in 2050 Average Average Change length of population Travel time Female Household in living road in density to market Water household Agriculture Years of Electricity District Province (%) standards (%) (km/10 km2) (per km2) (hours) availabilitya head (%) head (%) education (%) Wakhan Badakhshan 0.0 −10.4 0.0 1.0 36.3 3.0 0.0 70.0 2.7 0.0 Bamyan City Bamyan 0.3 2.2 0.0 32.0 7.3 0.3 0.6 39.2 2.5 0.0 Nawur Ghazni 0.3 2.8 0.0 16.0 10.2 0.7 0.0 40.0 5.9 0.0 Shighnan Badakhshan 0.0 2.8 0.0 9.0 11.8 2.1 0.0 70.0 7.2 0.0 Yakawlang Bamyan 0.2 2.8 0.0 10.0 13.9 0.2 0.0 47.4 1.9 0.0 Shibar Bamyan 0.1 2.9 0.0 18.0 9.1 0.0 1.5 37.9 1.2 0.0 Hisa-i-Awali-Bihsud Wardak 0.3 3.0 0.0 18.0 6.8 0.2 0.0 81.5 4.0 0.0 Markazi Bihsud Wardak 0.6 3.1 0.0 26.0 6.7 1.1 0.5 54.6 4.4 0.0 Kohistanat Sari Pul 0.2 3.3 0.0 11.0 8.5 0.4 0.0 21.1 1.0 4.1 Ajristan Ghazni 0.2 3.3 0.0 16.0 11.3 0.0 0.0 49.3 1.5 0.0 Overall 100.0 11.9 2.9 951.6 5.2 0.5 0.7 31.2 3.2 27.0 a. “Water availability” refers to the ratio of surface water use to groundwater use. A large value is good because it indicates that water use is more likely to be sustainable. TABLE 4.16  Predicted Change in Living Standards and Characteristics of the Top 10 Province Hotspots in Afghanistan under the Carbon-Intensive Scenario in 2050 Average Change Average length population Travel time Female in living of road in density to market Water household Agriculture Years of Electricity Province Household (%) standards (%) (km/10 km2) (per km2) (hours) availabilitya head (%) head (%) education (%) Bamyan 1.4 3.2 0.0 25.6 9.8 0.3 0.5 46.1 1.7 0.2 Ghor 2.8 4.5 0.0 15.2 6.3 0.1 0.2 72.9 0.9 4.2 Wardak 2.7 4.7 1.1 48.6 6.4 1.0 0.1 61.2 7.0 2.3 Panjshir 0.6 5.2 0.0 39.9 17.3 0.0 0.1 25.2 7.8 1.8 Daykundi 1.7 5.4 0.0 21.8 8.8 0.0 3.1 41.4 1.3 0.3 Ghazni 5.2 5.6 1.5 112.6 6.7 0.1 0.0 31.3 5.0 5.0 Logar 1.7 6.4 1.1 91.2 5.2 0.1 0.5 48.7 6.3 1.2 Paktika 1.4 6.7 0.0 28.9 6.3 0.1 0.4 25.2 4.4 0.8 Kapisa 1.6 7.1 0.0 356.4 5.1 0.0 0.7 16.4 4.8 0.4 Paktya 1.8 7.2 0.7 82.5 4.5 0.7 0.0 29.2 3.0 2.5 Overall 100.0 11.9 2.9 951.6 5.2 0.5 0.7 31.2 3.2 27.0 a. “Water availability” refers to the ratio of surface water use to groundwater use. A large value is good because it indicates that water use is more likely to be sustainable. M a p p i n g H o t s p o t s    67 BOX 4.3  Other Dimensions of Hotspots: Tracking Nonmonetary Effects of Climate Change Although the focus of this book has been on doc- are currently trying to reproduce the results of umenting declining living standards from climate Hsiang, Burke, and Miguel (2013). change, there is now a growing body of research Analysis of crime, agriculture, and weather that is attempting to capture adverse human data from India from 1971 to 2000 shows that dimensions of climate change, including increased drought and heat exert a strong effect on virtu- incidences of suicides, violent crimes, civil con- ally all types of crimes, with the effect on prop- flict, and riots (Carleton 2017; Carleton and erty crimes being greater than for violent crimes Hsiang 2016; and figure B4.3.1, panels a through f). (Blakeslee and Fishman 2017). This relationship Through a meta-analysis of the literature, is relatively stable over three decades of eco- Hsiang, Burke, and Miguel (2013) find strong nomic development. They also find the effects of causal evidence linking climatic events to human income shocks on crime are highly nonsymmet- conflict across a range of spatial and temporal ric: although negative agriculture shocks consis- scales and across all major regions of the world. tently lead to increases in crime, positive They find the magnitude of climate’s influence is agriculture shocks do not result in a decline in substantial: for each 1°C increase in long-term crime. The researchers conclude that despite the average temperatures or one standard deviation effects that accompany economic growth— increase in extreme rainfall, interpersonal vio- higher incomes, greater access to consumption- lence rises by 4 percent and the frequency of smoothing instruments, and reduced susceptibility intergroup conflict rises by 14 percent. Because of agriculture to climatic variability—there is locations throughout the inhabited world are little evidence that crime has become less respon- expected to warm 2°C to 4°C by 2050, the sive to extreme rainfall than it was before the researchers argue that the amplified rates of improvements. This may be taken as evidence human conflict could represent a large and criti- that despite India’s remarkable gains in human cal impact of anthropogenic climate change. The and economic development, the poorest mem- relationship between climate change and conflict bers of society continue to remain highly vulner- is an active area of research, and several groups able to aggregate economic shocks. FIGURE B4.3.1  Climate Has Diverse Monetary and Nonmonetary Effects on Well-Being a. Agricultural income (Brazil) b. Labor supply (U.S.) c. Math test scores (U.S.) Hidalgo et al. (2010) Gra Zivin and Neidell (2014) Gra Zivin et al. (2015) 0.05 40 Change in child’s test score Change in labor (minutes/day) 2 Log agricultural income 1 0 0 (percentiles) 0 –1 –0.05 –2 –3 –40 –4 –0.10 –5 –6 –0.15 –80 –3 –2 –1 0 1 2 3 0 10 20 30 40 15 20 25 30 Rainfall (standard deviations) Daily maximum temperature (˚C) Daily average temperature (˚C) (continues next page) 68   SOUTH ASIA’S HOTSPOTS BOX 4.3  Other Dimensions of Hotspots: Tracking Nonmonetary Effects of Climate Change (continued) FIGURE B4.3.1  Climate Has Diverse Monetary and Nonmonetary Effects on Well-Being (continued) d. Gross domestic product per capita e. Total income capita (U.S.) f. Total factor productivity (China) (global) Burke et al. (2015) Deryugina and Hsiang (2014) Zhang et al. (2013) Log total factor productivity Log annual total income 0.02 0 Annual growth rate 0 0 per capita × 100 –0.02 –0.1 0.05 –0.04 –0.06 –0.2 –0.10 –0.08 0 5 10 15 20 25 30 –10 –5 0 5 10 15 20 25 >30 –12–7 <1–4 10–16 21–27 >32 Annual average temperature (˚C) Daily average temperature (˚C) Daily average temperature (˚C) Change in annual migration probability g. Rape (U.S.) h. Civil con ict risk (global tropics) i. Household migration (Indonesia) Ranson (2014) Hsiang et al. (2011) Bohra-Mishra et al. (2014) Change in crime rate capita (%) 8 0.06 20 Civil con ict risk 0.04 (% prob/yr) 10 6 0.02 0 4 –10 0 –20 2 –0.02 –10 –5 0 5 10 –1 0 1 2 22 24 25 28 Avg. daily maximum temperature anomaly (˚C) Multiyear average temperature (˚C) Source: Carleton and Hsiang 2016. References Blakeslee, D. S., and R. Fishman. 2017. “Weather Shocks, Agriculture, and Crime: Evidence from Asuncion, R. C., and M. Lee. 2017. “Impacts of India.” Journal of Human Resources 52 (3). Sea-Level Rise on Economic Growth in Bolch, T., A. Kulkarni, A. Kääb, C. Huggel, F. Paul, Developing Asia.” ADB Economics Working J. G. Cogley, H. Frey, J. S. Kargel, K. Fujita, Paper Series 507, Asian Development Bank, M. Scheel, S. Bajracharya, and M. Stoffel. Manila. https://www.adb.org/sites/default/files​ 2012. “The State and Fate of Himalayan /­publication/222066/ewp-507.pdf. Glaciers.” Science 336 (6079): 310–14. Azhar, G., S. Saha, P. Ganguly, D. Mavalankar, and Carleton, T. A. 2017. “Crop-Damaging J. Madrigano. 2017. “Heat Wave Vulnerability Temperatures Increase Suicide Rates in India.” Mapping for India.” International Journal PNAS 114 (33): 8746–51. of Environmental Research and Public Health Carleton, T. A., and S. Hsiang. 2016. “Social and 14 (4): 357. Economic Impacts of Climate.” Science 353 Bellard, C., C. Bertelsmeier, P. Leadley, W. Thuiller, (6304). and F. Courchamp. 2012. “Impacts of Climate Chakraborty, S., and A. C. Newton. 2011. “Climate Change on the Future of Biodiversity.” Ecology Change, Plant Diseases and Food Security: An Letters 15 (4): 365–77. Overview.” Plant Pathology 60 (1): 2–14. M a p p i n g H o t s p o t s    69 Chen, Y., C. Xu, Y. Chen, W. Li, and J. Liu. Physical and Engineering Sciences 368 2010. “Response of Glacial-Lake Outburst (1919): 2461–79. Floods to Climate Change in the Yarkant Mendelsohn, R., K. Emanuel, S. Chonabayashi, and River Basin on Northern Slope of Karakoram L. Bakkensen. 2012. “The Impact of Climate Mountains, China.” Quaternary International Change on Global Tropical Cyclone Damage.” 226 (1): 75–81. Nature Climate Change 2 (3): 205–9. Hallegatte, S., C. Green, R. J. Nicholls, and Miller, J. D., W. W. Immerzeel, and G. Rees. 2012. J. Corfee-Morlot. 2013. “Future Flood Losses “Climate Change Impacts on Glacier in Major Coastal Cities.” Nature Climate Hydrology and River Discharge in the Hindu Change 3 (9): 802–6. Kush-Himalayas: A Synthesis of the Scientific Harris, I. P. D. J., P. D. Jones, T. J. Osborn, and Basis.” Mountain Research and Development D. H. Lister. 2014. “Updated High-Resolution 32 (4): 461–7. Grids of Monthly Climatic Observations—The Nicholls, R. J., P. P. Wong, V. R. Burkett, J. O. CRU TS3.10 Dataset.” International Journal of Codignotto, J. E. Hay, R. F. McLean, Climatology 34 (3): 623–42. S. Ragoonaden, and C. D. Woodroffe. 2007. Hsiang, S. M., M. Burke, and E. Miguel. 2013. “Coastal Systems and Low-Lying Areas.” In “Quantifying the Influence of Climate on Climate Change 2007: Impacts, Adaptation Human Conflict.” Science 341. and Vulnerability , edited by M. L. Parry, Immerzeel, W. W., L. P. Van Beek, and M. F. O. F. Canziani, J. P. Palutikof, P. J. van der Bierkens. 2010. “Climate Change Will Affect Linden, and C. E. Hanson, 315–56. Cambridge: the Asian Water Towers.” Science 328 (5984): Cambridge University Press. 1382–5. Siraj, A. S., M. Santos-Vega, M. J. Bouma, IPCC (Intergovernmental Panel on Climate D. Yadeta, D. R. Carrascal, and M. Pascual. Chante). 2013. Climate Change 2013: The 2014. “Altitudinal Changes in Malaria Physical Science Basis: Contribution of Incidence in Highlands of Ethiopia and Working Group I to the Fifth Assessment Colombia.” Science 343 (6175): 1154–58. Report of the Intergovernmental Panel on Stoffel, M., and C. Huggel. 2012. “Effects of Climate Change. edited by T. F. Stocker, D. Qin, Climate Change on Mass Movements in G.-K. Plattner, M. Tignor, S. K. Allen, Mountain Environments.” Progress In Physical J. Boschung, A. Nauels, Y. Xia, V. Bex, and Geography 36 (3): 421–39. P. M. Midgley. Cambridge: Cambridge World Bank. 2016. World Development Indicators University Press. doi:10.1017/CBO​ 9 7811​ Database, World Bank, Washington, DC. 07415324. Xu, J., R. E. Grumbine, A. Shrestha, M. Eriksson, Keiler, M., J. Knight, and S. Harrison. 2010. X. Yang, Y. U. N. Wang, and A. Wilkes. 2009. “Climate Change and Geomorphological “The Melting Himalayas: Cascading Effects of Hazards in the Eastern European Alps.” Climate Change on Water, Biodiversity, and Philosophical Transactions of the Royal Livelihoods.” Conservation Biology 23 (3): Society of London A: Mathematical, 520–30. Toward Greater Resilience 5 C limate change is one of the most sig- less severe if countries implement their national nificant threats facing the world strategies and the Paris Agreement’s goal of today. The adverse impacts of climate limiting average global temperature increases change are affecting all countries, especially to 2°C is achieved. In contrast, hotspots are developing countries, including persistent expected to be more severe if the Paris drought and extreme weather events, rising Agreement fails, leading to a future climate sea levels, and coastal erosion, and further more consistent with the carbon-intensive sce- threatening food security, water, energy and nario. Under the carbon-intensive scenario, the health, and more broadly efforts to eradicate effects to living standards will be widespread poverty and achieve sustainable develop- throughout the region: more than 800 million ment. The global nature of the problem calls people, or 45 percent of the region’s current for the widest possible cooperation by all population, live in locations projected to countries and their participation in an effec- become moderate to severe climate hotspots tive and appropriate international response. because of changes in average weather by 2050 Also, it is critical to continue mobilizing (table 4.1). Furthermore, living standards of financing from a variety of sources, public more than 80 percent of the overall population and private, bilateral and multilateral, includ- could be adversely affected. ing innovative sources of finance. Given the From a public policy perspective, these critical importance of resilience to addressing granular findings point to the importance of climate change impact and risk, what type of the geographical, political, and household strategies and actions can be adopted at all context in developing interventions to assist levels to make sure that resilience is incorpo- people living in climate hotspots. For exam- rated and mainstreamed in international and ple, some inland areas in India emerge as national planning and budgeting processes, severe hotspots, whereas in Sri Lanka, as well as informing investment and develop- the postconflict northern coastal areas are ment cooperation strategies and decisions? most vulnerable. The household characteris- Using extensive climate and household-level tics of these areas also differ from one another, data, this book shows the effects of changes in as do the characteristics of the locations them- average weather—long-term changes in aver- selves, so the interventions must be tailored to age seasonal temperature and precipitation— the specific context. This granular analysis can to be significant, but with substantial variations also inform decision making on the locations across South Asia. Hotspots are expected to be and households most in need of resources. 71 72  SOUTH AS I A ’ S HOTS P OTS TABLE 5.1  Changes in Average Weather Projected under the Carbon-Intensive Scenario Will Disproportionately Impact Severe Hotspots Bangladesh India Sri Lanka Entire country with no climate change (GDP per capita in PPP$) 13,365 21,148 28,632 Entire country under the carbon-intensive climate scenario 12,470 20,555 26,628 (GDP per capita in PPP$) Change for entire country due to carbon-intensive climate scenario (%) –6.7 –2.8 –7.0 Severe hotspots with no climate change (GDP per capita in PPP$) 13,231 21,782 29,491 Severe hotspots under the carbon-intensive climate scenario 11,326 19,647 26,394 (GDP per capita in PPP$) Change for severe hotspots under the carbon-intensive climate scenario (%) –14.4 –9.8 –10.5 Source: Calculations based on World Development Indicators, SSPs, and results in chapters 3 and 4. See explanation of calculations in Appendix E. Notes: Only the countries where severe hotspots are projected to emerge are shown. Severe hotspots correspond to those identified in chapter 4 under the carbon-intensive scenario by 2050. Methodology described in appendix E. GDP = gross domestic product; PPP = purchasing power parity; SSP = shared socioeconomic pathway. Dollars are US dollars. Money Worth Spending TABLE 5.2  Changes in Average Weather Projected under the Carbon-Intensive Scenario The findings of this book suggest that good Will Reduce Total GDP development outcomes are the best adapta- Loss of GDP by 2050 (US$, billions) tion: investing in skills, health, knowledge, better infrastructure, and a more diversified Country Severe hotspots Entire country economy should reduce climate hotspots at Bangladesh 58.7 171.1 the household, district, and country levels. India 403.9 1,177.8 This supports earlier conclusions on the Sri Lanka 12.2 49.9 importance of growth in reducing the poten- Source: Calculations based on World Development Indicators, SSPs, and tial negative effects of climate change results in chapters 3 and 4. Notes: Only countries where severe hotspots are projected to emerge (Hallegatte and others 2016; Skoufias, are shown. Severe hotspots correspond to those identified in chapter 4 Katayama, and Essama-Nssah 2012). under the carbon-intensive scenario by 2050. Methodology described in appendix E. Includes population changes corresponding to SSPs The climate hotspots presented in this (see description in appendix E) and per capita calculations in table 5.1. book will impact future gross domestic GDP = gross domestic product; SSP = shared socioeconomic pathway. product (GDP). Underlying each of the climate scenarios is a shared socioeconomic pathway Thus, the most affected hotspot regions in (SSP) scenario, which includes projections of Bangladesh, India, and Sri Lanka would dis- national-level GDP and many other macro- proportionately suffer, unless bolstered with economic variables (O’Neill and others 2014; additional growth. descriptions provided in appendix E). Cast in The national decreases in living standards as terms of GDP projections, the hotspots pre- a result of changes in average weather are sub- dicted under the carbon-intensive climate sce- stantial and tend to be concentrated in the nario will reduce projected per capita GDP severely affected regions (table 5.2). The costs 6.7 percent in Bangladesh, 2.8 percent in of inaction expressed as amount of total GDP India, and 7.0 percent in Sri Lanka by 2050 losses in severe hotspots are significant— compared to a scenario in which further US$59 billion in Bangladesh, US$404 billion in climate change does not occur (table 5.1). India, and US$12 billion in Sri Lanka by 2050 The estimated GDP losses are even greater under the carbon-intensive scenario. The total for the regions identified in this book as costs for the entire countries are even larger— severe hotspots (table 5.1). By 2050 in these US$171 billion in Bangladesh, US$1,178 billion areas, per capita GDP is predicted to be in India, and US$50 billion in Sri Lanka by 14.4 percent lower in Bangladesh than with- 2050 under the carbon-intensive scenario. out further climate change, 9.8 percent lower As discussed in the following section, this in India, and 10.5 percent lower in Sri Lanka. potential damage can be reduced through good T o w a r d G r e a t e r R e s i l i e n c e    73 development policies, even if the carbon-­ also increase the likelihood that a household intensive climate change scenario manifests. or community will experience negative out- comes. The benefits from continued invest- ments in basic infrastructure—such as Reducing Hotspots in Vulnerable improving access to electricity or density of Communities and Vulnerable primary roads (as identified in the literature)—​ Households could outweigh the climate-related loss in liv- Reducing hotspots could involve a portfolio ing standards for households that lack access of actions aimed at making affected places to these infrastructure services. Similarly, and the households located in them more technological advances, coupled with resilient. Potential actions include improving expanded irrigation systems, work to make infrastructure, introducing market reforms, agriculture less sensitive to climate change in and building individual and institutional the long-term (Taheripour and others 2016). capacity. In this context, some of the difficult The analysis suggests that the risks associ- questions that governments often grapple ated with changes in average weather with are: Which interventions are most war- can increase over time when combined ranted? Where? And when? with poverty, lack of education, and poorly The hotspot analysis provides some inter- maintained infrastructure. Table 5.3 shows esting insights on locations that are particu- the profiles of the most resilient households larly vulnerable to changes in average relative to the overall country profiles. 1 weather. In the case of India, the top hotspots Resilient households are those that face are not often talked about as being particu- smaller reductions in living standards from larly vulnerable to climate change, but are changes in average weather. In India, for frequently identified as critical from a devel- ­ example, such households have higher levels opment perspective. For example, central of education and enjoy higher rates of electri- India—including states such as Madhya fication. The impact of changes in average Pradesh, Chhattisgarh, Rajasthan, and Uttar weather could presumably be attenuated if the Pradesh—emerges as being highly vulnerable less resilient households acquired the charac- to changes in average weather. These states teristics of their more resilient counterparts. are also home to many poor and tribal peo- ple. In contrast, coastal areas in India—often identified as being most vulnerable to extreme Policy Agenda events and sea-level rise—are found to be rel- Although increasing temperatures and chang- atively more resilient to changes in average ing precipitation patterns present unique and weather compared with nearby areas in Sri sometimes hard-to-predict challenges, house- Lanka. holds, communities, and governments can This book underscores how the vulnerabil- take actions to improve resilience. Decisions ity of communities and households to climate about investment in adaptation strategies, risks depends on local, social, and economic development of human skills, and engage- factors. To explain the implications of ment options with communities will signifi- changes in average weather, the book helps cantly affect this generation and the next identify the factors that increase both vulner- generation’s quality of life. With more ability and resilience to changes in average knowledge about how these changes will weather. Although not necessarily causal, affect communities and households, espe- these elements can be important indicators, cially poor and vulnerable populations, gov- and understanding them can help shape poli- ernments will be better able to design policies cies and programs to strengthen communities and interventions that best serve specific seg- and households, and their capacities to adapt. ments of society. Targeting resources effi- Vulnerability factors (for example, elevation, ciently to the most vulnerable communities education, electricity access, and water stress) and groups should be a priority. The 74  SOUTH AS I A ’ S HOTS P OTS TABLE 5.3  Profile of the Top 10 Percent Resilient Households Living Average length Average Travel time Female- Top 10 percent / standards of road in km / population to market Water headed Agriculture Years of Electrification Country overall change (%) 10 km2 density per km2 (hours) availability household (%) head (%) education (%) Afghanistan Top 10% 17.1 3.8 713.3 3.6 0.4 1.5 30.8 2.6 37.5 Afghanistan Overall 11.9 2.9 951.6 5.2 0.5 0.7 31.2 3.2 27.0 Bangladesh Top 10% 0.9 5.7 680.3 2.9 3.0 7.6 44.6 3.5 39.2 Bangladesh Overall −6.7 5.8 1320.7 2.0 7.9 7.6 39.1 3.9 54.9 India Top 10% 4.4 2.2 596.9 3.1 0.8 17.9 28.0 6.5 91.8 India Overall −2.8 1.6 840.7 2.7 2.0 10.8 39.8 5.7 79.8 Nepal Top 10% 6.5 0.5 125.0 9.5 0.2 34.0 61.2 3.2 36.2 Nepal Overall 4.1 1.2 441.5 9.4 1.3 26.7 52.6 3.8 70.0 Pakistan Top 10% −1.3 0.6 181.9 11.8 0.1 13.8 26.8 4.4 7.9 Pakistan Overall −2.9 1.4 387.0 3.6 0.8 10.2 24.0 5.3 13.6 Sri Lanka Top 10% −2.9 22.2 437.8 3.4 0.2 22.7 38.8 7.5 91.4 Sri Lanka Overall −7.0 13.5 708.9 2.6 0.4 22.5 28.6 8.3 90.6 South Asia Top 10% 3.6 2.6 553.2 4.1 0.9 16.6 30.1 5.9 76.2 South Asia Overall −3.2 2.1 831.3 2.8 2.3 10.7 38.0 5.4 69.6 Source: World Bank calculations. T o w a r d G r e a t e r R e s i l i e n c e    75 measures that reduce climate hotspots also FIGURE 5.1  Good Development Outcomes Reduce Hotspots have strong overall development benefits. (Effects of Various Interventions on Living Standards in South Asia under the Therefore, policy makers should think of Carbon-Intensive Scenario, by 2050) these investments as win-win decisions that a. Afghanistan can sustainably break the downward spiral 12.9 12.3 12.3 of poverty and inequality, at the same time 11.9 12.0 12.0 12.0 12 driving growth and sustainable Change in living standards (%) development. The book identifies and highlights climate 8 hotspots where communities and households are likely to be particularly vulnerable to changes in average weather. The positive 4 effects of reducing hotspots can be amplified through efforts focused on the most vulnera- ble locations and population groups. The 0 hotspot analysis undertaken here can better o ity en l n s ss bs m a es qu io in on re l jo ns t cc at inform policy through refining our under- st ta ti us de fic ta ra at uca er at ltu tri ke ad at St standing of the underlying reasons that peo- Ed ec icu ar w Ro El M g gr cin na ple in specific hotspot areas are particularly du No Re vulnerable. Of the six countries investigated, living b. Bangladesh standards are predicted to be adversely Status quo Nonagricultural jobs affected by changes in average weather in 6.0 four: Bangladesh, India, Pakistan, and 3.9 Change in living standards (%) Sri Lanka. Afghanistan and Nepal are esti- 4.0 mated to benefit from these changes in 2.0 average weather. The broader growth-and- 0 development agenda includes investing in human capital (such as through education) –2.0 and infrastructure (such as through electrifi- –4.0 cation and construction of roads). The ques- –6.0 tions then arise: What are the co-benefits of –6.7 these strategies for climate resilience? What –8.0 effect would a given policy have in a specific setting? Using the analysis from chapters 3 c. India and 4, this book investigates investment and Nonagricultural Reducing Educational policy options that countries could consider 0 Status quo jobs water stress attainment Change in living standards (%) to reduce the negative consequences of changes in average weather under the carbon- –0.5 intensive climate scenario (figure 5.1, panels a –1.0 through f; see appendix A for a description of –1.5 the methodology used). Although all policies may not work for all the countries, the analy- –2.0 sis here illustrates promising avenues that –2.5 –2.5 –2.4 could be explored at the national and subna- –2.6 –3.0 –2.8 tional levels. Several development interventions could assist Afghanistan in leveraging projected (continues next page) increases in temperature and changes in 76  SOUTH AS I A ’ S HOTS P OTS FIGURE 5.1  Good Development Outcomes Reduce Hotspots by 30 percent could reduce negative impacts (continued) on living standards by roughly 1 percent, (Effects of Various Interventions on Living Standards in South Asia under the whereas the other two aforementioned devel- Carbon-Intensive Scenario, by 2050) opment strategies would each provide a d. Nepal 0.4 percent benefit. Improving education and 8 primary road density are also projected to 7.1 weakly—but positively—increase net improve- Change in living standards (%) 7 ments from changes in average weather. 6 For Bangladesh, the analysis suggests that 5.0 5 4.1 enhancing nonagricultural employment 4 opportunities could potentially reduce the liv- 3 ing standards burden of changes in average weather (figure 5.1b). A 15 percent increase 2 in nonagricultural employment opportunities 1 would lead to a reduction in the impact of 0 average weather changes on living standards Status quo Nonagricultural jobs Electri cation from –6.7 percent to –1.4 percent. Similarly, a 30 percent increase in nonagricultural e. Pakistan employment would not only negate all the Status quo Electricity negative effects of changes in average weather Change in living standards (%) 0 but also result in a 3.9 percent increase in –0.5 living standards. In India, the analysis identifies three pos- –1.0 sible avenues to offset the effects of changes –1.5 in average weather, including improving edu- –2.0 cational attainment, reducing water stress, –2.5 and improving nonagricultural employment –2.5 opportunities (figure 5.1c). The analysis pre- –3.0 –2.9 dicts that increasing the average educational attainment by 30 percent (or 1.5 additional f. Sri Lanka years of schooling) would reduce the impact Educational Market Nonagricultural of changes in average weather on living stan- Status quo attainment access jobs dards from –2.8 percent to –2.4 percent. Change in living standards (%) 1.0 0.1 Reducing water stress and enhancing nonag- 0 ricultural employment by 30 percent could –1.0 –2.0 yield similar benefits. Therefore, multiple –3.0 –2.6 –2.1 actions could be taken simultaneously to –4.0 maximally reduce hotspots. Conversely, –5.0 these results also indicate that the wrong –6.0 policy actions or worsening water stress –7.0 –8.0 –7.0 could exacerbate the effects of changes in average weather on living standards. Source: World Bank calculations. Although the analysis indicates that Nepal Note: Impacts of interventions are estimated using the method described in appendix A. 15% and 30% increases in market access are defined as 1.5% and 3% decreases in travel time to major will on average benefit from changes in aver- cities, respectively. age weather, the country can further leverage climatic changes (figure 5.1d). The analysis precipitation (figure 5.1a): (a) increasing shows that living standards increase when access to electricity; (b) reducing water stress; there is access to electricity and nonagricul- and (c) providing nonagricultural employ- tural employment opportunities. Based on ment opportunities. Based on the correlations the findings, increasing access to electricity observed today, increasing access to electricity by 30 percent could improve living standards T o w a r d G r e a t e r R e s i l i e n c e    77 by approximately 3 percent. Although drought-resistant crops, or providing weather warming temperatures may open up more forecasts and climate risk assessments that can areas for agriculture, the analysis highlights leverage adaptive actions. In addition, the gov- that people must have access to nonagricul- ernment can play a key role through establish- tural job opportunities to leverage the effects ing the policy framework for adaptation, of changes in average weather for maximum which sets the incentives for private action. increases in living standards. This could include (a) regulatory and insur- In Pakistan, the analysis reveals that ance instruments that convey the correct expanding electrification by 30 percent could incentives for adaptation; (b) pricing and other reduce the impact of average weather on policies that encourage efficient use of energy, living standards from –2.9 percent to water, agriculture, and other natural resources; –2.5 percent (figure 5.1e). Thus, electrifica- and (c) facilitating market access and provid- tion alone may not completely overcome the ing fiscal incentives for research and develop- negative effects of changes in average weather ment to exploit existing technologies or on living standards. This indicates that addi- develop new ones in the energy, water-supply, tional analysis could be warranted to better agricultural, forestry, and livestock sectors. understand how to prevent the emergence of Hotspots tend to have lower living stan- hotspots within the country. dards compared to the national average. In For Sri Lanka, the policy choices consid- this respect, it seems right to conclude that ered include enhancing education, improving changes in average weather will hurt poor market access, and increasing nonagricultural households disproportionately and therefore employment (figure 5.1f). The analysis sug- increase poverty and inequality. While this is gests that increasing nonagricultural employ- true on average, the granularity of the analy- ment by 30 percent relative to current levels sis in this book provides a more nuanced could entirely eliminate the burden of changes profile of the households that stand to lose in average weather on living standards; the the most. As seen in table 5.3, in Nepal and overall impact would shift from –7 percent to Sri Lanka, the top 10 percent of the resilient 0.1 percent. On the other hand, reducing time households are more rural than the average traveled to the market by 3 percent and household in the country. In a relatively large increasing eductional attainment by 30 percent country like India, poor and rich households would respectively change the impact on living are spread evenly across all climate zones. standards from –7 percent to –2.1 percent and Therefore, poor households living in cooler –2.6 percent, respectively. If these interven- areas may in fact benefit from changing tions were implemented together, living stan- average weather compared with those living dards would most likely increase under the in warmer areas. It should be noted, how- climate change scenario. ever, that this book investigates only changes These national-level policy choices may in average weather, not differences in climate mask some of the subtle regional differences variability or shocks caused by extreme in terms of the benefits. In Pakistan, for events. The reason for the focus on changes example, increasing education may reduce in the averages is that changes in the vari- hotspots in some regions, even though the ability are not as well captured in the current effect is not significant at the national level. generation of climate models. As shown This analysis therefore should be taken as an by Hallegatte and others (2017), natural illustration of an array of various comple- disasters tend to affect poor households mentary policy and investment choices avail- the most. able for decision makers. In the future, economic growth and struc- Resilience in communities and households tural changes will lead people to migrate from can also be built through policies that enable rural areas to cities, leaving behind many of effective private actions on adaptation. their agricultural and other climate-sensitive Examples include boosting research and devel- practices. Although this could potentially opment on new technologies, such as make the migrants more climate-resilient, it 78  SOUTH AS I A ’ S HOTS P OTS may also create new climate risks. For exam- References ple, urban populations will face several health Hallegatte, S., M. Fay, M. Bangalore, T. Kane, and risks exacerbated by climate change, such as L. Bonzanigo. 2016. Shock Waves: Managing heatwaves (enhanced by heat island effects) the Impacts of Climate Change on Poverty . and flood-related challenges. To the extent Washington, DC: World Bank. that economic growth is noninclusive and cer- Hallegatte, S., A. Vogt-Schilb, M. Bangalore, and tain segments of the population are left J. Rozenberg. 2017. Unbreakable: Building the behind, there is always a danger that climate Resilience of the Poor in the Face of Natural change will deepen poverty in some parts of Disasters. Washington, DC: World Bank. the region. These results, along with sug- Harris, I. P. D. J., P. D. Jones, T. J. Osborn, and gested costs of inaction, point in the direction D. H. Lister. 2014. “Updated High-Resolution of resilience policies that are more targeted Grids of Monthly Climatic Observations—The toward poorer populations and areas and CRU TS3.10 Dataset.” International Journal of Climatology 34 (3): 623–42. households that have high vulnerability. O’Neill, B. C., E. Kriegler, K. Riahi, K. L. Ebi, S. Hallegatte, T. R. Carter, R. Mathur, and D. P. van Vuuren. 2014. “A New Scenario Note Framework for Climate Change Research: The 1. Table 5.3 compares the overall impacts of Concept of Shared Socioeconomic Pathways.” climate change with those for the 10 percent Climatic Change 122 (3): 387–400. most resilient households. Table 4.2 does the Skoufias, E., R. S. Katayama, and B. Essama- same for the 10 percent least resilient house- Nssah. 2012. “Too Little Too Late: Welfare holds. Thus, the two tables provide a comple- Impacts of Rainfall Shocks in Rural Indonesia.” mentary picture of the effects at the two ends Bulletin of Indonesian Economic Studies 48 (3): of the scale. 351–68. Methodology for Policy Cobenefits A E xtra resources made available through efficiency and reduce baseline water stress in specific policies of governments and a district may make households in that dis- nongovernmental organizations trict more resilient to changes in average may facilitate the mitigation of the risks of weather. Last, improving primary road den- changes in average weather on living stan- sity and access to markets may make new dards. Six variables that can be influenced by resources available to households, which policy actions are considered: (a) nonagricul- might allow better protection against changes tural households, (b) households’ access to in average weather. electricity, (c) years of education of the head Not all policy-relevant variables could be of the household, (d) baseline water stress, analyzed in the context of all the countries (e) primary road density, and (f) access to since the country-specific models include only market. The impact of these was explored the variables that are weakly correlated with using a variant of the model specification the seasonal climate indicator variables with provided in equation (3.1). a ­c orrelation coefficient of less than 0.5. Agricultural households (in general) and For example, access to market fails the weak rain-fed agriculture (in particular) are most correlation criterion in all the countries except susceptible to changes in average weather. Sri Lanka. Similarly, primary road density is Policies that help households move from agri- used only in Bangladesh and India, because culture to nonagricultural occupations may this policy-relevant variable is highly corre- help mitigate the ill effects of changes in aver- lated with seasonal climate indicators in other age weather on living standards. Similarly, countries. access to electricity may help in coping with To analyze how policy-relevant variables long-term increases in average temperature. If mediate the effects of changes in average household heads have more education, then weather on living standards, the original they may be better equipped to deal with equation (3.1) is rewritten with a few changes changes in average weather. This would be in in notation, leading to equation (A.1). In addition to the direct effects of higher educa- equation (3.1) household variables are tion on living standards, through higher denoted by X; here, they are denoted as H. income. Policies that improve water use Similarly, originally district and locational 79 80  SOUTH AS I A ’ S HOTS P OTS variables in equation (3.1) are W; here, they living standards because of changes in aver- are denoted as L: age weather. Because it is expected that an increase in any of the policy-relevant variables ∑ 2 Yhit = α + j (β1 j tempit j + β2 j tempit will improve the resilience of households and j ∈( s, m , w ) help them better cope with long-term changes j j 2 in average weather, the marginal effects + β3 j rain it + β 4 j rainit ) on consumption, ∆Yi/∆Wi, are expected to be + β5 Hhit + β6Li + τ t + uhit positive. These positive effects are expected to be in addition to the main benefits of improv-  (Eq. A.1) ing any of the policy-relevant variables, such The main equation can be rewritten as as education, for consumption. The main follows: benefits, which can be substantial, are not Yhit = a + aWit + bXhit + tt + uhit (Eq. A.2) included in this analysis. To understand the marginal effects of where: policy relevant variables on changes in con- sumption stemming from changes in average  β1 j tempitj j2  weather, the effects are plotted around the + β2 j tempit αWit = ∑   j  j2  j ∈( s, m , w )  + β3 j rain it + β 4 j rainit  mean predicted changes in consumption and the respective average policy-relevant indica- tor (see figure A.1). and To plot more than one policy variable on the X axis, the indicators are rescaled to be β Xhit = β5 Hhit + β5Li in the range of 0 to 100 (for example,  X POL − Min  To capture the interaction between the policy POL  hit  ). Here, all Rescaled Xhit = actions and effects of changes in average weather POL [ Max − Min ] on living standards, let Xhit be one of these variables (termed “development outcomes”) six policy variables described above (that is, are improved by 30 percent to calculate their POL =1, 2…6). Then up to six interaction mod- impact on living standards in the context of els, one for each POL, are estimated for each average changes in precipitation and temper- country. The interaction model is represented as: ature. The only exception is “market access,” POL which is only increased by 3 percent. Yhit = α + αW + β Xhit + γ Xhit Wit + τ t + uhit  (Eq. A.3) FIGURE A.1  Effects of Development Outcomes on Hotspots in Sri Lanka under the Carbon-Intensive Note that g is a vector of coefficients for Scenario by 2050 interactions with each of the seasonal and 30% improvement quadratic components of W. These interaction models—and associated Status Market Nonagricultural marginal changes in consumption expenditure— quo Education access jobs 0.1 are computed as: 1.0 0 Consumption change (%) ∆Yi –1.0 = γ IXiPOL , ∆Wi –2.0 –3.0 –2.1 –2.6 where I is an identity vector of the dimension –4.0 of g. In other words: –5.0 –6.0 12 ∆Yi ∑γ –7.0 = XiPOL j –8.0 –7.0 ∆Wi j =1 Source: World Bank calculations. Note: Impacts of interventions are estimated using the method described This captures only the additional effects of in this appendix. 15% and 30% increases in market access are defined as the policy-relevant variables on changes in 1.5% and 3% decreases in travel time to major cities, respectively. Supplementary Tables B M ore than 40 climate models have this book is publicly available for 18 (names been developed and used by scien- given in table B.1). These 18 climate models tists around the world in the CMIP5 are assessed as described in appendix D. Eleven climate modeling experiment to help under- of the climate models are selected as best repre- stand the Earth’s climate system (Taylor, senting climate conditions best in South Asia. Stouffer, and Meehl 2012). Of these models, These 11 models are used throughout the the climate output needed for the analysis in report to project future climate conditions. TABLE B.1  18 Climate Models Assessed Climate model Included Reference ACCESS1.0 Yes Bi and others 2013 BCC CSM1.1 Yes Wu and others 2008 CanESM2 Yes Arora and others 2011 CCSM4 Yes Gent and others 2011 CNRM CM5 Yes Voldoire and others 2012 CSIRO Mk3.6.0 No Rotstayn and others 2012 GFDL ESM2G No Freidenreich and others 2004 GFDL ESM2M Yes Freidenreich and others 2004 GISS E2R No Schmidt and others 2014 HadGEM2 CC Yes Collins and others 2011 HadGEM2 ES No Collins and others 2011 INM CM4 No Volodin and others 2010 IPSL CM5A-LR Yes Dufresne and others 2013 MIROC ESM Yes Watanabe and others 2011 MIROC ESM-CHEM Yes Watanabe and others 2011 MIROC5 No Watanabe and others 2010 MPI ESM-LR No Giorgetta and others 2013 MPI ESM-MR Yes Giorgetta and others 2013 NorESM1-M Yes Kirkevåg and others 2013 Note: The models are the subset of those participating in CMIP5 that include publicly available simulations for the historical, RCP 4.5, and RCP 8.5 experiments. 81 82  SOUTH AS I A ’ S HOTS P OTS Table B.2 shows the regression results Table B.3 shows the results of a robustness for the reduced-form model in equation test for changes in consumption expenditures (3.1). The estimated coefficients from this under different model specifications using dif- model are used to predict the changes in ferent sets of control variables and with and consumption expenditures resulting from without provincial fixed effects. The selection the long-term changes in average weather of control variables is based on different in table 3.4. correlation coefficient threshold criteria. TABLE B.2  Regression Results Used for Consumption Predictions Afghanistan Bangladesh India Nepal Pakistan Sri Lanka Summer temperature 0.311*** –0.469*** 0.302*** 0.109*** 0.077*** 0.604 (0.028) (0.1100) (0.0130) (0.0360) (0.0100) (0.4220) Summer temperature squared –0.008*** 0.009*** –0.006*** –0.001* –0.001*** –0.017** (0.001) (0.0020) (0.0002) (0.0010) (0.0002) (0.0070) Summer precipitation 0.001* 0.0004* –0.001*** 0.005*** 0.0001 –0.002*** (0.001) (0.0002) (0.0001) (0.0010) (0.0004) (0.0004) Summer precipitation squared 0.00002*** 0.0000 0.00001*** –0.00003*** 0.00001*** 0.00000*** (0.00000) (0.0000) (0.0000) (0.0000) (0.0000) (0.0000) Monsoon temperature –0.330*** –0.328 –0.124*** –0.023 –0.174*** –0.426** (0.027) (0.2860) (0.0140) (0.0800) (0.0110) (0.1980) Monsoon temperature squared 0.005*** 0.005 0.003*** 0.001 0.002*** 0.010*** (0.001) (0.0050) (0.0002) (0.0020) (0.0002) (0.0030) Monsoon precipitation –0.017*** –0.001*** –0.0002*** 0.002*** –0.001*** 0.001 (0.001) (0.0003) (0.0001) (0.0010) (0.0001) (0.0004) Monsoon precipitation squared 0.0001*** 0.00000*** 0.00000*** –0.00000** 0.00000*** –0.00000** (0.00001) (0.0000) (0.0000) (0.0000) (0.0000) (0.0000) Winter temperature –0.037*** 1.210** –0.193*** –0.119*** 0.040*** 0.064 (0.009) (0.4990) (0.0100) (0.0350) (0.0060) (0.3920) Winter temperature squared 0.005*** –0.028** 0.005*** 0.001* –0.001*** 0.003 (0.001) (0.0110) (0.0002) (0.0010) (0.0002) (0.0070) Winter precipitation 0.003 –0.001 0.0004** 0.031** –0.001** –0.004*** (0.002) (0.0020) (0.0002) (0.0130) (0.0010) (0.0010) Winter precipitation squared –0.00005** 0.00002 –0.00001*** –0.001* 0.0000 0.00000*** (0.00002) (0.0000) (0.0000) (0.0004) (0.0000) (0.0000) Rural household 0.016 0.044*** –0.046*** –0.044*** –0.172*** –0.145*** (0.011) (0.0080) (0.0060) (0.0150) (0.0040) (0.0070) Household size –0.033*** –0.050*** –0.068*** –0.060*** –0.054*** –0.117*** (0.001) (0.0020) (0.0010) (0.0030) (0.0010) (0.0020) Dependency ratio –0.102*** –0.126*** –0.101*** –0.120*** –0.072*** –0.058*** (0.003) (0.0060) (0.0020) (0.0060) (0.0020) (0.0040) Age of household head 0.001*** 0.007*** 0.005*** 0.008*** 0.004*** 0.004*** (0.0002) (0.0003) (0.0002) (0.0004) (0.0001) (0.0002) Female-headed household –0.066*** 0.080*** 0.047*** 0.206*** 0.199*** –0.001 (0.025) (0.0140) (0.0060) (0.0150) (0.0070) (0.0070) Household has electricity 0.110*** 0.227*** 0.189*** 0.293*** 0.171*** 0.258*** (0.012) (0.0080) (0.0050) (0.0130) (0.0050) (0.0070) Years of education of head 0.018*** 0.042*** 0.045*** 0.054*** 0.035*** 0.059*** (0.001) (0.0010) (0.0004) (0.0010) (0.0004) (0.0010) Agricultural household –0.023*** 0.021*** –0.046*** –0.062*** 0.135*** –0.059*** (0.006) (0.0070) (0.0040) (0.0120) (0.0040) (0.0060) Baseline water stress –0.159*** 1.356*** 0.022*** –0.362** –0.001 0.080* (continues next page) S u pp l e m e n t a r y T a b l e s    83 TABLE B.2  Regression Results Used for Consumption Predictions (continued) Afghanistan Bangladesh India Nepal Pakistan Sri Lanka (0.032) (0.1660) (0.0050) (0.1500) (0.0010) (0.0410) Latitude –0.089*** n.a. n.a. n.a. n.a. n.a. (0.007) n.a. n.a. n.a. n.a. n.a. Elevation n.a. 0.0004** n.a. n.a. n.a. n.a. n.a. (0.0002) n.a. n.a. n.a. n.a. Water availability normalized 0.001 0.0004** –0.0001 0.001** –0.005*** n.a. (0.002) (0.0001) (0.0002) (0.0010) (0.0010) n.a. Water availability seasonal –0.103** n.a. 0.023** 0.175*** 0.03 n.a. variability (0.047) n.a. (0.0090) (0.0610) (0.0180) n.a. Coast distance inverse squared –203,172.200*** n.a. 0.00004 n.a. 38.497*** n.a. (26,586.490) n.a. (0.0003) n.a. (13.4370) n.a. Road density: primary 0.010*** –0.001 0.008*** n.a. n.a. n.a. (0.002) (0.0020) (0.0010) n.a. n.a. n.a. Market access –0.00000 n.a. n.a. n.a. n.a. –0.001*** (0.00002) n.a. n.a. n.a. n.a. (0.0001) Population density: 2010 –0.00003*** 0.00004*** 0.00002*** 0.00004*** 0.00004*** n.a. (0.00000) (0.0000) (0.0000) (0.0000) (0.0000) n.a. Constant 13.113*** 4.136 5.926*** 4.593*** 8.577*** 6.614*** (0.292) (4.5510) (0.1340) (0.6530) (0.1070) (1.3400) Observations 38,579 19,508 240,206 9,600 75,635 55,639 R2 0.188 0.38 0.446 0.558 0.43 0.349 Adjusted R2 0.188 0.379 0.446 0.557 0.429 0.349 Residual standard error 5.669 21.776 20.069 15.598 13.970 8.174 (df = 38549) (df = 19480) (df = 240177) (df = 9574) (df = 75604) (df = 55614) F statistic 308.672*** 442.151*** 6,908.214*** 482.813*** 1,897.374*** 1,242.781*** (df = 29; (df = 27; (df = 28; (df = 25; (df = 30; (df = 24; 38549) 19480) 240177) 9574) 75604) 55614) Source: World Bank calculations. Note: Dependent variable: ln consumption. Robust standard errors are in parentheses. Models include survey year dummies. *p < 0.10. **p < 0.05. ***p < 0.01. TABLE B.3  Changes in Consumption from Base Year 2011 to 2030 and 2050 for Climate-Sensitive and Carbon-Intensive Scenarios from Model Specifications Percent Model specification Climate in 2030 Climate in 2050 Provincial Correlation Climate- Carbon- Climate- Carbon- fixed effect threshold Country sensitive intensive sensitive intensive No 0.3 Afghanistan 4.9 5.6 8.0 11.38 No 0.3 Bangladesh –2.3 –3.7 –4.9 –10.4 No 0.3 India –1.3 –1.5 –2.0 –2.9 No 0.3 Nepal 2.0 2.1 3.1 3.8 No 0.3 Pakistan –1.3 –1.5 –2.0 –2.9 No 0.3 Sri Lanka 1.6 1.7 2.8 3.9 No 0.5 Afghanistan 5.1 5.8 8.3 11.9 No 0.5 Bangladesh –1.3 –2.3 –2.9 –6.7 (continues next page) 84  SOUTH AS I A ’ S HOTS P OTS TABLE B.3  Changes in Consumption from Base Year 2011 to 2030 and 2050 for Climate-Sensitive and Carbon-Intensive Scenarios from Model Specifications (continued) Percent Model specification Climate in 2030 Climate in 2050 Provincial Correlation Climate- Carbon- Climate- Carbon- fixed effect threshold Country sensitive intensive sensitive intensive No 0.5 India –1.3 –1.5 –2.0 –2.8 No 0.5 Nepal 2.1 2.3 3.2 4.1 No 0.5 Pakistan –1.3 –1.5 –2.0 –2.9 No 0.5 Sri Lanka –3.2 –3.7 –4.9 –7.0 No 0.7 Afghanistan 5.1 5.8 8.3 11.9 No 0.7 Bangladesh –4.2 –7.3 –10.0 –21.7 No 0.7 India –0.5 –0.6 –0.7 –1.1 No 0.7 Nepal 2.7 3.0 4.1 5.4 No 0.7 Pakistan –1.3 –1.6 –2.0 –3.0 No 0.7 Sri Lanka –0.9 –0.6 –0.8 0.3 No All controls Afghanistan 3.7 4.2 6.1 8.8 No All controls Bangladesh 0.7 –1.2 –0.9 –6.8 No All controls India 1.2 1.4 1.7 2.4 No All controls Nepal 2.2 2.3 3.2 3.9 No All controls Pakistan –2.4 –2.9 –3.7 –5.4 No All controls Sri Lanka –0.5 0.0 0.0 1.5 No No control Afghanistan 7.4 8.3 12.1 17.3 No No control Bangladesh 12.5 15.6 20.7 32.1 No No control India –2.0 –2.2 –3.0 –4.1 No No control Nepal 5.4 5.1 8.2 9.5 No No control Pakistan 1.7 2.2 2.7 4.1 No No control Sri Lanka 5.6 5.7 8.9 11.6 Yes 0.3 Afghanistan 5.6 6.3 9.1 12.9 Yes 0.3 Bangladesh –0.7 –1.5 –2.6 –7.5 Yes 0.3 India –1.5 –1.9 –2.4 –3.5 Yes 0.3 Nepal 4.1 4.9 5.9 8.8 Yes 0.3 Pakistan –0.3 –0.5 –0.5 –0.9 Yes 0.3 Sri Lanka –9.4 –9.4 –11.1 –12.4 Yes 0.5 Afghanistan 5.5 6.2 8.8 12.6 Yes 0.5 Bangladesh 0.0 –0.6 –1.2 –4.9 Yes 0.5 India –1.5 –1.8 –2.3 –3.3 Yes 0.5 Nepal 4.9 5.9 7.1 10.8 Yes 0.5 Pakistan –0.3 –0.5 –0.5 –0.9 Yes 0.5 Sri Lanka –9.0 –9.0 –10.9 –12.3 Yes 0.7 Afghanistan 5.5 6.2 8.8 12.6 Yes 0.7 Bangladesh –2.6 –4.4 –7.7 –17.6 (continues next page) S u pp l e m e n t a r y T a b l e s    85 TABLE B.3  Changes in Consumption from Base Year 2011 to 2030 and 2050 for Climate-Sensitive and Carbon-Intensive Scenarios from Model Specifications (continued) Percent Model specification Climate in 2030 Climate in 2050 Provincial Correlation Climate- Carbon- Climate- Carbon- fixed effect threshold Country sensitive intensive sensitive intensive Yes 0.7 India –0.9 –1.1 –1.5 –2.2 Yes 0.7 Nepal 7.4 8.9 10.5 15.4 Yes 0.7 Pakistan 0.1 0.0 0.1 –0.1 Yes 0.7 Sri Lanka –7.3 –7.2 –8.5 –8.9 Yes All controls Afghanistan 4.2 4.8 6.9 9.9 Yes All controls Bangladesh –1.9 –3.3 –5.3 –12.3 Yes All controls India –1.4 –1.7 –2.2 –3.2 Yes All controls Nepal 3.9 4.0 5.7 6.9 Yes All controls Pakistan –3.9 –4.8 –5.9 –8.4 Yes All controls Sri Lanka –6.1 –6.0 –7.2 –7.6 Yes No control Afghanistan 7.7 8.7 12.7 18.1 Yes No control Bangladesh 15.1 19.2 24.1 37.2 Yes No control India –3.5 –4.0 –5.4 –7.4 Yes No control Nepal 2.9 3.3 5.8 9.1 Yes No control Pakistan –1.0 –1.2 –1.6 –2.3 Yes No control Sri Lanka –12.3 –12.6 –14.2 –16.1 Source: World Bank calculations. Note: “Correlation threshold” is a threshold used for a correlation coefficient of control variables with all climate variables. References Y. Ma, J. Noonan, L. Stevens, P. Uhe, H. Zhu, S. M. Griffies, R. Hill, C. Harris, and K. Puri. Arora, V. K., J. F. Scinocca, G. J. Boer, J. R. 2013. “The ACCESS Coupled Model: Christian, K. L. Denman, G. M. Flato, and Description, Control Climate and Evaluation.” W. J. Merryfield. 2011. “Carbon Emission Australian Meteorological and Oceanographic Limits Required to Satisfy Future Representative Journal 63 (1): 41–64. Concentration Pathways of Greenhouse Gases.” Collins, W. J., N. Bellouin, M. Doutriaux-Boucher, Geophysical Research Letters 38 (5). N. Gedney, P. Halloran, T. Hinton, . . ., and Bellouin, N., J. Rae, A. Jones, C. Johnson, G. Martin. 2011. “Development and Evaluation J. Haywood, and O Boucher. 2011. “Aerosol of an Earth-System Model-HadGEM2.” Forcing in the Climate Model Intercomparison Geoscientific Model Development 4 (4): 1051. Project (CMIP5) Simulations by HadGEM2-ES Dufresne, J. L., M. A. Foujols, S. Denvil, and the Role of Ammonium Nitrate.” A. Caubel, O. Marti, O. Aumont, Y. Balkanski, J. Geophys. Res., 116, D20206, doi:10.1029​ S. Bekki, H. Bellenger, R. Benshila, S. Bony, /2011​JD016074, 2011. L. Bopp, P. Braconnot, P. Brockmann, P. Cadule, Bi, D., M. Dix, S. J. Marsland, S. O’Farrell, H. A. F. Cheruy, F. Codron, A. Cozic, D. Cugnet, Rashid, P. Uotila, A. C. Hirst, E. Kowalczyk, N. de Noblet, J. P. Duvel, C. Ethé, L. Fairhead, M. Golebiewski, A. Sullivan, H. Yan, T. Fichefet, S. Flavoni, P. Friedlingstein, N. Hannah, C. Franklin, Z. Sun, P. Vohralik, J. Y. Grandpeix, L. Guez, E. Guilyardi, I. Watterson, X. Zhou, R. Fiedler, M. Collier, D. Hauglustaine, F. Hourdin, A. Idelkadi, 86  SOUTH AS I A ’ S HOTS P OTS J. Ghattas, S. Joussaume, M. Kageyama, Simulations.” Atmospheric Chemistry and G. Krinner, S. Labetoulle, A. Lahellec, M. P. Physics 12 (14): 6377. Lefebvre, F. Lefevre, C. Levy, Z. X. Li, J. Lloyd, Schmidt, G. A., M. Kelley, L. Nazarenko, F. Lott, G. Madec, M. Mancip, M. Marchand, R. Ruedy, G. L. Russell, I. Aleinov, M. Bauer, S. Masson, Y. Meurdesoif, J. Mignot, I. Musat, S. E. Bauer, M. K. Bhat, R. Bleck, V. Canuto, S. Parouty, J. Polcher, C. Rio, M. Schulz, Y. Chen, Y. Cheng, T. L. Clune, A. Del Genio, D. Swingedouw, S. Szopa, C. Talandier, P. Terray, R. de Fainchtein, G. Faluvegi, J. E. Hansen, N. Viovy, and N. Vuichard. 2013. “Climate R. J. Healy, N. Y. Kiang, D. Koch, A. A. Lacis, Change Projections Using the IPSL-CM5 Earth A. N. LeGrande, J. Lerner, K. K. Lo, E. E. System Model: From CMIP3 to CMIP5.” Matthews, S. Menon, R. L. Miller, V. Oinas, Climate Dynamics 40 (9-10): 2123–65. A. O. Oloso, J. P. Perlwitz, M. J. Puma, Freidenreich, S. M., S. T. Garner, R. G. Gudgel, W. M. Putman, D. Rind, A. Romanou, M. Sato, C. T. Gordon, and I. M. Held. 2004. “The New D. T. Shindell, S. Sun, R. A. Syed, N. Tausnev, GFDL Global Atmosphere and Land Model K. Tsigaridis, N. Unger, A. Voulgarakis, M. Yao, AM2–LM2: Evaluation with Prescribed SST and J. Zhang. 2014. “Configuration and Simulations.” Journal of Climate 17: 4641–73. Assessment of the GISS ModelE2 Contributions Gent, P. R., G. Danabasoglu, L. J. Donner, M. M. to the CMIP5 Archive.” Journal of Advances in Holland, E. C. Hunke, S. R. Jayne, and P. H. Modeling Earth Systems 6 (1): 141–84. Worley. 2011. “The Community Climate Taylor, K. E., R. J. Stouffer, and G. A. Meehl. System Model Version 4.” Journal of Climate 2012. “An Overview of CMIP5 and the 24 (19): 4973–91. Experiment Design.” Bulletin of the American Giorgetta, M. A., J. Jungclaus, C. H. Reick, Meteorological Society 93 (4): 485–98. S. Legutke, J. Bader, M. Böttinger, V. Brovkin, Volodin, E. M., N. A. Dianskii, and A. V. Gusev. T. Crueger, M. Esch, K. Fieg, K. Glushak, 2010. “Simulating Present-Day Climate V. Gayler, H. Haak, H. D. Hollweg, T. Ilyina, with the INMCM4.0 Coupled Model of S. Kinne, L. Kornblueh, D. Matei, T. Mauritsen, the Atmospheric and Oceanic General U. Mikolajewicz, W. Mueller, D. Notz, Circulations.” Izvestiya, Atmospheric and F. Pithan, T. Raddatz, S. Rast, R. Redler, Oceanic Physics 46 (4): 414–31. E. Roeckner, H. Schmidt, R. Schnur, Watanabe, M., T. Suzuki, R. O’ishi, Y. Komuro, J. Segschneider, K. D. Six, M. Stockhause, S. Watanabe, S. Emori, T. Takemura, C. Timmreck, J. Wegner, H. Widmann, K. H. M. Chikira, T. Ogura, M. Sekiguchi, K. Takata, Wieners, M. Claussen, J. Marotzke, and D. Yamazaki, T. Yokohata, T. Nozawa, B. Stevens. 2013. “Climate and Carbon Cycle H. Hasumi, H. Tatebe, and M. Kimoto. 2010. Changes from 1850 to 2100 in MPI-ESM “Improved Climate Simulation by MIROC5: Simulations for the Coupled Model Mean States, Variability, and Climate Intercomparison Project Phase 5.” Journal of Sensitivity.” Journal of Climate 23 (23): Advances in Modeling Earth Systems 5 (3): 6312–35. 572–97. Watanabe, S., T. Hajima, K. Sudo, T. Nagashima, Kirkevåg, A., T. Iversen, Ø. Seland, C. Hoose, T. Takemura, H. Okajima, T. Nozawa, J. E. Kristjánsson, H. Struthers, A. M. L. Ekman, H. Kawase, M. Abe, T. Yokohata, T. Ise, S.  Ghan, J. Griesfeller, E. D. Nilsson, and H. Sato, E. Kato, K. Takata, S. Emori, and M. Schulz. 2013. “Aerosol–Climate Interactions M. Kawamiya. 2011. “MIROC-ESM 2010: in the Norwegian Earth System Model— Model Description and Basic Results of NorESM1-M.” Geoscientific Model CMIP5-20c3m Experiments.” Geoscientific Development 6 (1): 207–44. Model Development 4 (4): 845. Rotstayn, L. D., S. J. Jeffrey, M. A. Collier, Wu, T., R. Yu, F. Zhang, Z. Wang, M. Dong, S. M. Dravitzki, A. C. Hirst, J. I. Syktus, and L. Wang, X. Jin, D. Chen, and L. Li. 2010. K. K. Wong. 2012. “Aerosol-and Greenhouse “The Beijing Climate Center Atmospheric Gas-Induced Changes in Summer Rainfall and General Circulation Model: Description and Its Circulation in the Australasian Region: A Performance for the Present-Day Climate.” Report Using Single-Forcing Climate Climate Dynamics 34 (1): 123–47. Supplementary Maps C MAP C.1  Percentage of People in Each MAP C.2  Average Years of Education of the Head Administrative Unit Who Live in Rural Environments of Household in Each Administrative Unit Source: Based on household data referenced in table 3.2. Source: Based on household data referenced in table 3.2. Note: This classification is based on the most recent year of survey data Note: This classification is based on the most recent year of survey data available, as outlined in table 3.1. available, as outlined in table 3.1. 87 88   SOUTH ASIA’S HOTSPOTS MAP C.3  Percentage of People in Each MAP C.5  Average Density of Roads Administrative Unit Who Have Access to Electricity Source: Based on household data referenced in table 3.2. Source: Based on district data referenced in table 3.2. Note: This classification is based on the most recent year of survey data Note: This classification is based on the most recent year of survey data available, as outlined in table 3.1. available, as outlined in table 3.1. MAP C.4  Average Travel Time to Market in Hours MAP C.6  Average Population Density per Square Kilometer Source: Based on district data referenced in table 3.2. Note: This classification is based on the most recent year of survey data Source: Based on household data referenced in table 3.2. available, as outlined in table 3.1. Note: This classification is based on the most recent year of survey data available, as outlined in table 3.1. S u p p l e m e n t a r y M a p s   89 MAP C.7  Climate-Sensitive Scenario by 2050: MAP C.8  Carbon-Intensive Scenario by 2050: Hotspots Do Not Clearly Overlap with Major Basins Hotspots Do Not Clearly Overlap with Major Basins Source: Based on household data referenced in table 3.2. Source: Based on household data referenced in table 3.2. Note: Same as map 4.2, but with basin boundaries. Note: Same as map 4.2, but with basin boundaries. Climate Model Selection D M ultimodel approaches to estimate The monsoon is the most important future climate are superior com- climatic feature of the region because it regu- pared to approaches using individ- lates the seasonality of temperature and brings ual models. The reason is that multimodel the rain that allows agriculture to thrive in mean is typically more representative than any many locations with the region. The seasonal individual model. Additionally, the multimodel fractions of annual average precipitation in ensemble can be used to estimate uncertainty. each season are presented in map D.1, panels a The conceptual basis for this approach is that through c. This confirms that most precipita- although all models have imperfections, they tion occurs during the monsoon season do not always have the same imperfections. for most of South Asia. However, this is not For multimodel approaches to perform as uniformly the case because Afghanistan, expected, each of the included models must Sri Lanka, Southwestern Pakistan, and perform adequately by itself. Although it is dif- Southeastern India receive significant portions ficult to identify the best-performing model, it of their precipitation during the postmonsoon is often possible to identify the models that do season. not perform as well as the others and to dis- The performance of climate models is card those. This is the approach taken here. assessed using spatial pattern correlation and This book evaluates 18 Coupled Model root mean squared error (RMSE). A high Intercomparison Project Phase 5 (CMIP5) cli- pattern correlation suggests that models ade- mate models (Taylor, Stouffer, and Meehl 2012) quately capture the underlying climate pro- that have publicly available simulation output cesses controlling that pattern. A low RMSE for the historic period, RCP 4.5 and RCP 8.5 suggests that the model represents the correct (appendix B, table B.1). The region used for amplitude of response to the relevant climate evaluation includes Afghanistan, Bhutan, processes. Bangladesh, India, Maldives, Nepal, Pakistan, The baseline climatological period for all and Sri Lanka. The evaluation is conducted for mean and standard deviation calculations is three seasons based on the monsoon: 1981 through 2000. All the trend calcula- tions, however, are performed based on the •  Premonsoon: March through May longest common period between models and •  Monsoon: June through September observations. In the case of precipitation, the •  Postmonsoon: October through February trends are also normalized to a 30-year 91 92   SOUTH ASIA’S HOTSPOTS MAP D.1  Percentage of Annual Precipitation Contained in the Study Seasons Source: Yatagai and others 2012 (Aphrodite v 1101). Note: Based on average conditions during 1981 through 2000. period and compared as a change in percent- For temperature extremes, Aphrodite data age terms. do not contain maximum and minimum temperatures, which are important for the calculation of those extremes. Furthermore, Choice of Observations comparison of representations of extreme The climate models are compared to an events in observational data sets such as IMD observational data set over a common his- (Rajeevan and others 2006), Berkeley Earth toric period. This provides a direct compari- (Rohde and others 2013), and HADEX2 son of how the climate models represent (Donat and others 2013) indicates that there actual climate. Several observational data are large discrepancies between observational sets are considered as possible representation estimates of extremes. Therefore, there is not of the true historical climate. The principal a strong quantitative basis against which to data sets are Aphrodite v1101, a daily grid- compare climate model representations of ded data set for monsoon Asia (Yatagai and extremes. As noted in chapter 2, there is also others 2011), and CHIRPS v2.0, a daily grid- substantial disagreement between climate ded data set available globally (Funk and models in terms of their representation of others 2015). extremes. These are the two primary reasons These two data sets are compared with the that this book focuses on long-term mean Indian Meteorological Department (IMD) climate and does not investigate extreme daily gridded data set (Rajeevan and others events. 2006) to determine which one is most pre- ferred for the South Asian analysis. It is deter- mined that overall, the Aphrodite data better Weather Measurements Are match the spatial pattern and local magni- Uneven across South Asia tudes of the IMD data, particularly with Map D.2 shows the number of weather sta- respect to variability, trends, and precipitation tions contributing data to the Aphrodite data extremes. set. The density of stations contributing to the C l i m a t e M o d e l S e l e c t i o n    93 calculation of precipitation is relatively temperature measurements (maps D.3). The good across the region, particularly over spatial pattern in map D.3 is similar to that in India, Bangladesh, and Nepal. Afghanistan map D.2, with India, Bangladesh, and Nepal effectively has no stations contributing pre- having the most complete records for precipi- cipitation measurements. Temperature mea- tation. For temperature, Nepal appears to surements are sparser for most countries, have the most thorough coverage, though the except Nepal, but this should be sufficient stations reporting for India also appear to be because temperature has a more stable consistent over time. regional structure (map D.2). Afghanistan is again notably data poor. The Maldives does not have station data provided for tempera- Comparison of Climate Models to ture in Aphrodite, and it is challenging to iso- Aphrodite late for the models given that land-sea masks Aphrodite is used as the observational in most models would consider these small data set for evaluating the h ­ istoric islands as ocean grid points. ­ performance of climate models. Prior to com- Precipitation station measurements are parison, Aphrodite is regridded to a 1-degree more temporally complete relative to spatial resolution. Each climate model is used MAP D.2  Spatial Density of Station Measurements’ Contribution to the Aphrodite Data Set, 1981 through 2000 (continues next page) 94   SOUTH ASIA’S HOTSPOTS MAP D.2  Spatial Density of Station Measurements’ Contribution to the Aphrodite Data Set, 1981 through 2000 (continued) Sources: Yatagai and others 2011 (“rstn” variable in Aphrodite v1101 data set, for precipitation); v1204R1 (for temperature). Note: Values are presented as the percentage of 0.05 degrees latitude/longitude grid boxes contained in a 1 degree latitude/longitude grid box. The highest value, 100 percent, would indicate that there is a station in each of the 400 0.05 degree boxes contained in a single 1 degree grid box. Therefore, an average of four stations (or four 0.05 degree boxes) would appear as an average of 1 percent. Standard deviation indicates the temporal variation in availability of station measurements within each grid cell. at its original latitude/longitude resolution. GCMs. Most models simulate colder condi- The quantitative summary of this analysis is tions over Nepal than are observed. Some present in table 2.2.1 models tend to overestimate temperatures in northeast India, Bangladesh, and Bhutan, such as CSIRO Mk3.6.0, and GISS E2R; Temperature others, such as INM CM4, have a cooler bias over most of India. Exploration of the year- March through May (Premonsoon) to-year variability (characterized by standard For the premonsoon season (March through deviation) indicates an even larger disagree- May), all of the climate models (or GCMs) ment between observations and GCMs. capture the main spatial features of the mean The temperature variability tends to be temperature field, including the sharp gradi- overestimated throughout the study region. ent created by the Himalaya mountains and a The observations suggest that most of the relatively cooler west Indian coast because of variability occurs over Pakistan, Afghanistan, the Western Ghats. Other fine-scale details and northwest India, and decreases sharply are not properly represented, mainly because toward the south. However, this region is of the relatively coarse resolution of the not well covered by station observations C l i m a t e M o d e l S e l e c t i o n    95 MAP D.3  Seasonal and Temporal Consistency of Station Measurements’ Contribution to the Aphrodite Data Set, 1979 through 2005 Source: Yatagai and others 2011 (“rstn” variable in Aphrodite v1101 data set). Note: Fraction of daily station data reported for each year during the period 1979 through 2005. Spatial explanation of grid cell configuration is the same as with map D.2. MAP D.4  Average Monsoon Precipitation, 1981 through 2000 Sources: Funk and others 2015 (CHIRPS); Rajeevan and others 2006 (IMD); Yatagai and others 2011 (Aphrodite). 96  SOUTH AS I A ’ S HOTS P OTS (maps D.2 and D.3). A large dispersion is seen India. CanESM2 and the two GFDL models among GCMs, both in the intensity of the are the more extreme cases. variability and the location of its maximums. The observed trend pattern for July With the exception of CSIRO Mk3.6.0, through September is very similar to that of they all overestimate variability in the March through May, suggesting that a weak study region. cooling occurred over 1961–2005. Most The pattern of observed temperature trends models show a slightly larger cooling trend in for March through May shows a lot of small- most of the domain than is observed, with the scale details, probably because of the complex HadGEM2 and MIROC5 models being the topography of the study region. A slight most extreme. warming signal dominates, interrupted by a cooling region along the mountain ranges to October through February the north of the study region. A stronger (Postmonsoon) warming is observed just north of Nepal. The GCMs are unable to reproduce the observed For the postmonsoon season (October trend pattern, though many of them exhibit through February), the models exhibit a range reasonable magnitudes of temperature change of differences in the spatial pattern of mean in the study countries. Most are dominated by temperatures. Most models include a strong a warming through much of the study region, bias of cooler temperatures over the northern and the most extreme cases are the IPSL, portion of the study region, with the CM5A-LR, and MPI ESM-MR models. HaDGEM2 and INM CM4 as the worst per- formers. The GCMs also differ considerably on both the intensity and the patterns of vari- June through September (Monsoon) ability. Most of the models tend to overesti- During the monsoon season (June through mate the magnitude of the variability September), the average temperatures over throughout the study region. the study region yield a more complex pattern The observed trend pattern for October than seen in March through May, with more through February temperatures is very similar regional characteristics such as a cooler west to the one observed for the two previous coast, to the west of the Western Ghats, which seasons. Most of the models, though not all, correlates with the increased precipitation in again tend to overestimate the cooling trends that area. The GCMs tend to have a more slightly. A couple of the models (the worst widespread minimum in the whole southern performers are IPSL and MPI ESM-MR) portion of the peninsula, probably because of show overall warming and fail to represent the resolution of the simulated topographies. the cooling trend associated with the moun- In addition, some models have very strong tain ranges at the north of the study region. warm biases in the northern part of India and extending into Pakistan, such as GISS E2R and CSIRO Mk3.6.0. Precipitation The year-to-year variability for July through March through May (Premonsoon) September is similar in magnitude to that in March through May, but slightly larger for Precipitation is a very challenging variable for central India. The sharp difference in the tem- climate models to capture, and is also affected perature variability between Afghanistan and by the complex topography and circulation of Pakistan might be due to the complex topogra- the region. The premonsoon season (March phy of that region, but also may be due to the through May) is very dry over the study region, lack of station data over Afghanistan. Most with the exception of the southwesternmost GCMs show notable differences with the tip of India and Sri Lanka and the topography observed pattern, with very large overestima- to the north and northeast of the domain. The tions of the temperature variability in north GCMs capture the main structures of the C l i m a t e M o d e l S e l e c t i o n    97 March through May field, but fail to represent coherent patterns of trend—such as GFDL the small-scale details, particularly over south- ESM2G, GISS ER, and HadGEM2—are west India, probably because of their deficient likely the more unreliable. representations of topography. For the year-to-year variability (quantified October through February by the standard deviation), the differences (Postmonsoon) between observations and GCMs are most notable along the Himalayan region and in During the postmonsoon season (October southern India. Many models also show through February), the wettest conditions are weakened variability compared with the found over the southernmost parts of India observations along much of coastal India and and Sri Lanka. Many of the GCMs show a over Bangladesh. distinct dry bias over this seasonally wet area. The trends for March through May show As in the previous seasons, many of the very small values for most of the region. models overestimate the year-to-year variabil- In the southern part of the region, there is a ity, particularly over northern India and Nepal. dipolar structure that includes both a drying Some models—such as CCSM4, GFDL trend in southwest India and wettening trend ESM2M, and INM CM4—overestimate over Sri Lanka. Some other significant wet- ­ precipitation across nearly the entire region. tening trends are present near Bangladesh, The wettening over Sri Lanka is the main Bhutan, and Nepal. No GCM captures these feature of the trend field, with weaker localized details that are seen in the observa- increases seen in central India. The CCSM4 tion data. In terms of percentage change, model is the only one that reproduces this which considers the trends relative to wettening feature over Sri Lanka to a similar the local magnitude of the rainfall, the degree as found in the observations. The BCC GCMs exhibit obvious differences from the model shows the most extreme trends over observed trends. India, which are not consistent with the observations. Cast as a percentage, the model and observed trends show very little agree- June through September (Monsoon) ment in the pattern or magnitude of changes. The agreement between the Aphrodite and GCM simulation mean values is much better for the monsoon season (June through Summary of Climate Model September), despite differences in the small- Selection: Interpretation of scale details due to topography. Some Figure 2.2 models—such as GFDL and MIROC ESM— In general, all the climate models tend to over- overestimate the precipitation over most of estimate the year-to-year variability. They also central, southern, and eastern India. Also, few tend to overestimate the precipitation trends GCMs capture the localized rainfall maxi- and underestimate the temperature trends in mum over the western coast of India. the main monsoon season (June through For the standard deviation, only some September) over most of the study region. models—such as CCSM4, HadGEM, Figure 2.2 summarizes the regionally MIROC5, MPI, and NorESM—simulate integrated measures of pattern correlation the structure of the observed variability, (figure 2.2, panel a) and RMSE (figure 2.2, which shows some resemblance to the mean panel b) as a single number for each model, rainfall pattern. season, and variable. The displayed RMSE The observed monsoon trend is noisy values have been normalized by the average and weak, with localized drying over south- magnitude in each column, in order to plot ernmost India and wettening over Sri Lanka. the values with a more uniform color scale. The GCMs also have weak and noisy patterns It is noted, however, that the magnitude of of trend in this season. Models showing more errors in the trend fields, especially the 98  SOUTH AS I A ’ S HOTS P OTS temperature trends, are typically larger than Note for the mean or standard deviation. 1. The analysis in this section was provided Higher pattern correlations and lower by the International Research Institute for RMSE values indicate better model perfor- Climate and Society at Columbia University. mance. It is generally preferable to have mod- Further details of its analysis are available els that do a fair job overall, rather than an upon request. excellent job in a few selected instances and terrible performance in other instances. Examination of figure 2.2, panel a, con- References firms many of the statements made in the Donat, M. G., L. V. Alexander, H. Yang, preceding discussion. The observed pattern I. Durre, R. Vose, R. J. H. Dunn, and of the mean seasonal climate is generally B. Hewitson. 2013. “Updated Analyses of well represented by the models, even if the Temperature and Precipitation Extreme magnitudes may differ. Although the pat- Indices since the Beginning of the Twentieth tern of standard deviation in precipitation Century: The HadEX2 Dataset.” Journal of may be well captured by the models, several Geophysical Research: Atmospheres 118 (5): models do a poor job. In particular, CSIRO 2098–118. Mk3.6.0, HadGEM2 CC, and MPI Funk, C., P. Peterson, M. Landsfeld, D. Pedreros, J. Verdin, S. Shukla, G. Husak, J. Rowland, ESM-LR all have a near-zero correlation to L. Harrison, A. Hoell, and J. Michaelsen. “The the spatial pattern of the observations. Climate Hazards Infrared Precipitation with These models have a weakly positive to Stations: A New Environmental Record for strongly negative pattern correlation for Monitoring Extremes.” Scientific Data 2: trend in both temperature and precipitation. 150066. GFDL ESM2G does a particularly poor job Harris, I., P. D. Jones, T. J. Osborn, and of reproducing the pattern of temperature D. H. Lister. 2014. “Updated High-Resolution and precipitation trends. Grids of Monthly Climatic Observations—The Figure 2.2 B highlights the relative CRU TS3.10 Dataset.” International Journal of magnitude of regionally aggregated errors, Climatology 34 (3), 623–42. rather than the spatial pattern of errors. Rajeevan, M., J. Bhate, J. D. Kale, and B. Lal. 2006. “High Resolution Daily Gridded Rainfall Particularly notable are the large errors in Data for the Indian Region: Analysis of Break the precipitation fields of GISS E2R and and Active Monsoon Spells.” Current Science, the temperature fields of INM CM4. 296–306. MIROC 5 shows relatively larger errors Rohde, R., R. Muller, R. Jacobsen, S. Perlmutter, overall. These three models do not perform A. Rosenfeld, J. Wurtele, and S. Mosher. 2013. well in their patterns of variability and “Berkeley Earth Temperature Averaging change either. Process.” Geoinformatics & Geostatistics: An Of the 18 CMIP5 models available for Overview 13: 20–100. this study (appendix B, table B.2), seven Taylor, K. E., R. J. Stouffer, and G. A. Meehl. models—CSIRO Mk3.6.0, GFDL ESM2G, 2012. “An Overview of CMIP5 and the GISS-E2R, HadGEM2 ES, INM CM4, Experiment Design.” Bulletin of the American Meteorological Society 93 (4): 485–98. MIROC5, and MPI ESM-LR—were Yatagai, A., K. Kamiguchi, O. Arakawa, excluded from the climate model ensemble. A. Hamada, N. Yasutomi, and A. Kitoh. 2012. The reasons for exclusion are those pro- “APHRODITE: Constructing a Long-Term vided in the preceding paragraphs. The Daily Gridded Precipitation Dataset for Asia remaining 11 models are used to formulate Based on a Dense Network of Rain Gauges.” the MMM, low, and high values used Bulletin of the American Meteorological Society throughout this book. 93 (9): 1401–15. Calculating Gross Domestic Product Based on Shared E Socioeconomic Pathways and Hotspots Results T he calculations in tables 5.1 and 5.2 to national-level impacts of climate change use predicted changes in population on consumption expenditures per capita and gross domestic product (GDP). (table 3.4). This assumes that percentage The predictions used for these properties are changes in consumption expenditures per calculated as the average of values associated capita are equal to percentage changes in with shared socioeconomic pathway 1 GDP per capita. (SSP1), SSP3, and SSP5 in the International •  Row 4—GDP estimates for “severe Institute for Applied Systems Analysis hotspots with no climate change” are (IIASA) database, developed in coordination derived by calculating the portion of esti- with the OECD (see the description of SSPs mated historic baseline consumption in the following section). expenditures in severe hotspots relative to Methods used to calculate values in the entire country and using this propor- table 5.1 are: tion to scale the SSP GDP projections (row 1 of table 5.1). It is assumed that the •  Row 1—GDP estimates for the “entire ratio of per capita GDP in the two areas is country with no climate change” are cal- the same as the ratio of per capita con- culated as the average of GDP projections sumption in the two areas. in SSP1, SSP3, and SSP5. It is assumed •  Row 5—GDP estimates for “severe that percentage changes in GDP per capita hotspots under the carbon-intensive cli- are equal to percentage changes in con- mate scenario” are calculated by assuming sumption expenditures per capita. that reductions in consumption expendi- •  Row 2—GDP estimates for the “entire coun- tures per capita within severe hotspots will try under the carbon-intensive climate sce- be equivalent to reductions in GDP per nario” are calculated such that percentage capita, given the estimated proportion of decreases in GDP are equivalent to corre- historical baseline GDP generated in areas sponding decreases in consumption expendi- projected to become severe hotspots tures per capita for that year (see chapter 3 for (row 4 of table 5.1). consumption expenditures estimate details). •  Row 6—GDP estimates of “change for •  Row 3—Percentage changes in GDP for severe hotspots under the carbon-intensive the “entire country due to the carbon-­ climate scenario” are calculated as the per- intensive climate scenario” are equivalent centage difference between rows 5 and 4. 99 100  SOUTH AS I A ’ S HOTS P OTS TABLE E.1  Population Projections for Countries Three SSPs used in chapter 5 of this with Severe Hotspots book are: Millions Severe hotspots Entire country •  SSP 1, sustainability. This pathway is char- acterized by reduced inequality globally, Country 2016 2050 2016 2050 and within countries, as low-income coun- Bangladesh 26.4 30.8 163.0 190.1 tries develop at a rapid rate and a high level India 148.3 189.2 1,324.2 1,689.3 of education is achieved globally. The low Sri Lanka 3.6 3.9 21.2 23.2 global population growth present in the Source: O’Neill and others 2014. scenario is associated with consumption Note: Severe hotspots correspond to those identified to occur under the carbon-intensive scenario by 2050. Projections of population growth are oriented toward low-energy intensity based on the country-level average of SSP1, SSP3, and SSP5. Population in goods, partly enabled by fast-paced and severe hotspots by 2050 under the carbon-intensive scenario is projected to remain in the same proportion to the total as it is today. SSP = shared environmentally friendly technologi- socioeconomic pathway. cal development. Reduced fossil fuel dependency and rapid clean energy techno- The total GDP losses in table 5.2 are the logical development are concurrent with product of population projection estimates high levels of environmental awareness. and projected changes in per capita GDP Environmental governance is successful at (table 5.1). The total GDP estimates in table achieving globally implemented agree- 5.2 assume that population grows as pre- ments. The Millennium Development Goals dicted by the average of SSP1, SSP3, and SSP5 are achieved within the next decade or two. (table E.1). The SSP scenarios estimate popu- •  SSP 3, fragmentation. This world is frag- lation changes only for the entire country. mented into marginalized and poor Population changes for severe hotspots are regions, countries struggling to maintain calculated by assuming that the national-level their living standards, and pockets of mod- population change projections apply evenly erate wealth. There is little progress toward across the country and using these projections achieving the Millennium Development to scale observed population during the Goals, lowering energy and material inten- historic baseline period. sity consumption, or reducing fossil fuel dependency. Inequalities between countries and populations are increasing. Economic Note on Shared Socioeconomic growth is slowed by low levels of invest- Pathway Scenarios ment in education and clean technologies, A shared socioeconomic pathway (SSP) is along with policies oriented toward secu- essentially a storyline describing a future rity and barriers to trade. Population development scenario (O’Neill and others growth is high and drives up emissions. 2014). There are five SSPs produced and Global governance is weak and interna- agreed on by the international community tional aid is low, leaving some populations and used by the IPCC.1 Each of the SSPs vulnerable to climate change. includes three drivers, projected in five-year •  SSP 5, conventional development. This intervals at the country level: (a) overall pop- pathway illustrates a world where conven- ulation growth, (b) urban population tional development (economic growth and growth, and (c) economic growth. These pursuit of self-interest in a liberalized parameters vary between SSP narratives and world) is perceived as the solution to country. social and economic challenges. As a The SSPs exist in parallel with the RCP result, fossil fuel dependency deepens and emissions scenarios. Multiple SSPs can lead to mitigation challenges are high. The a given RCP since different socioeconomic Millennium Development Goals are changes can result in similar greenhouse gas attained, and robust economic growth, (GHG) concentrations. engineered solutions, and highly managed C a l c u l at i n g G r o s s D o m e s t ic P r o d u c t B a s e d o n S h a r e d S o ci o e c o n o m ic Pat h w ay s a n d H o t s p o t s R e s u lt s    101 ecosystems provide a certain level of Reference adaptive capacity. O’Neill, B. C., E. Kriegler, K. Riahi, K. L. Ebi, S. Hallegatte, T. R. Carter, R. Mathur, and D. P. van Vuuren. 2014. “A New Scenario Note Framework for Climate Change Research: 1. For details, see https://secure.iiasa.ac.at/web​ The Concept of Shared Socioeconomic -apps/ene/SspDb/dsd?Action=htmlpage​ Pathways.” Climatic Change 122 (3): &page=about. 387–400. ECO-AUDIT Environmental Benefits Statement The World Bank Group is committed to reducing its environmental footprint. In support of this commitment, we leverage electronic publishing options and print- on-demand technology, which is located in regional hubs worldwide. Together, these initiatives enable print runs to be lowered and shipping distances decreased, resulting in reduced paper consumption, chemical use, greenhouse gas emissions, and waste. We follow the recommended standards for paper use set by the Green Press Initiative. The majority of our books are printed on Forest Stewardship Council (FSC)–certified paper, with nearly all containing 50–100 percent recycled content. The recycled fiber in our book paper is either unbleached or bleached using totally chlorine-free (TCF), processed chlorine–free (PCF), or enhanced elemental chlorine–free (EECF) processes. More information about the Bank’s environmental philosophy can be found at http://www.worldbank.org/corporateresponsibility. South Asia Development Matters South Asia is highly vulnerable to climate change. Average temperatures have been rising throughout the region, and rainfall has become more erratic. These changes are projected to continue accruing over the coming decades. South Asia’s Hotspots: The Impact of Temperature and Precipitation Changes on Living Standards is the first book of its kind to provide granular spatial analysis of the long-term impacts of changes in average temperature and precipitation on one of the world’s poorest regions. South Asia’s Hotspots finds that higher temperatures and shifting precipitation patterns will reduce living standards in communities across South Asia—locations that the book terms “hotspots.” More than 800 million people in South Asia currently live in communities that are projected to become hotspots under a carbon-intensive climate scenario. Global action to reduce greenhouse gas emissions will reduce the severity of hotspots. Diverse and robust development is the best overall prescription to help people in hotspots. The book also suggests actions tailored to each country in the region—such as increasing employment in nonagricultural sectors, improving educational attainment, and expanding access to electricity— that would offset the declines in living standards associated with hotspots. South Asia’s Hotspots complements previous studies detailing the impacts of sea-level rise and extreme events on the people of South Asia. Together, these bodies of work create a sound analytical basis for investing in targeted policies and actions to build climate resilience throughout the region. www.worldbank.org/SouthAsiaHotspots ISBN 978-1-4648-1155-5 SKU 211155