69086 CLIMATE CHANGE RISKS AND FOOD SECURITY IN BANGLADESH CLIMATE CHANGE RISKS AND FOOD SECURITY IN BANGLADESH Winston H. Yu, Mozaharul Alam, Ahmadul Hassan, Abu Saleh Khan, Alex C. Ruane, Cynthia Rosenzweig, David C. Major and James Thurlow publishing for a sustainable future London • Washington, DC First published in 2010 by Earthscan © World Bank, 2010 All rights reserved. No part of this publication may be reproduced, stored in a retrieval system, or transmitted, in any form or by any means, electronic, mechanical, photocopying, recording or otherwise, except as expressly permitted by law, without the prior, written permission of the publisher. Earthscan Ltd, Dunstan House, 14a St Cross Street, London EC1N 8XA, UK Earthscan LLC,1616 P Street, NW, Washington, DC 20036, USA Earthscan publishes in association with the International Institute for Environment and Development For more information on Earthscan publications, see www.earthscan.co.uk or write to earthinfo@earthscan.co.uk ISBN: 978-1-84971-130-2 hardback Typeset by JS Typesetting Ltd, Porthcawl, Mid Glamorgan Cover design by Susanne Harris A catalogue record for this book is available from the British Library Library of Congress Cataloging-in-Publication Data Climate change risks and food security in Bangladesh / Winston H.Yu … [et al]. p. cm. Includes bibliographical references and index. ISBN 978-1-84971-130-2 (hbk.) 1. Crops and climate–Bangladesh. 2. Climatic change–Bangladesh. 3. Food security–Environmental aspects– Bangladesh. 4. Agricultural productivity–Environmental aspects–Bangladesh. 5. Agriculture–Economic aspects– Bangladesh. I.Yu, Winston H. S600.64.B3C65 2010 363.8’2095492–dc22 2009053662 At Earthscan we strive to minimize our environmental impacts and carbon footprint through reducing waste, recycling and offsetting our CO2 emissions, including those created through publication of this book. For more details of our environmental policy, see www.earthscan.co.uk. Printed and bound in the UK by the Cromwell Press Group. The paper used is FSC certiï¬?ed. Contents List of Figures and Tables vii 5.3 Incorporating Coastal Inundation Acknowledgements xi Effects 48 Foreword by Isabel M. Guerrero xiii 5.4 Projections of Future Potential Executive Summary xv Unflooded Production (Climate Only) 49 Glossary of Terms xxi 5.5 Projections of Future Projected Flood Acronyms xxiii Damages 52 5.6 Projections of Potential Coastal 1 INTRODUCTION 1 Inundation Damages 53 1.1 Objectives of Study 2 5.7 Projections of Integrated Damages 53 1.2 Literature Review 2 5.8 Using the Crop Model to Simulate 1.3 Integrated Modelling Methodology 3 Adaptation Options 56 1.4 Organization of Study 4 6 ECONOMY-WIDE IMPACTS OF CLIMATE 2 VULNERABILITY TO CLIMATE RISKS 5 RISKS 60 2.1 The Success of Agriculture 6 6.1 Integrating Climate Effects in an 2.2 Living with Annual Floods 10 Economy-wide Model 61 6.2 Economic Impacts of Existing 2.3 Lean Season Water Availability 15 Climate Variability 64 2.4 Sea level Rise in Coastal Areas 17 6.3 Additional Economic Impacts of 2.5 Regional Hydrology Issues 19 Climate Change 72 3 FUTURE CLIMATE SCENARIOS 21 7 ADAPTATION OPTIONS IN THE 3.1 Future Estimated Precipitation and AGRICULTURE SECTOR 82 Temperature 22 7.1 Identifying and Evaluating 3.2 Future Sea level Rise 26 Adaptation Options 83 4 FUTURE FLOOD HYDROLOGY 28 8 THE WAY FORWARD – TURNING IDEAS 4.1 GBM Basin Model Development 28 TO ACTION 105 4.2 National Hydrologic Super Model 30 8.1 A Framework for Assessing the 4.3 Approach to Modelling Future Flood Economics of Climate Change 107 Changes 30 4.4 Future Changes over the Ganges- ANNEX 1 – Using DSSAT to Model Brahmaputra-Meghna Basin 31 Adaptation Impacts 108 4.5 Future Flood Characteristics and ANNEX 2 – Description of the CGE Model 113 Analysis 33 ANNEX 3 – Constructing the Social Accounting Matrix for Bangladesh 119 5 FUTURE CROP PERFORMANCE 41 5.1 Development of the Baseline Period 42 References 133 5.2 Developing Flood Damage Functions 46 Index 139 List of Figures and Tables 4.3 Percentage change in discharges in Figures (a) the 2030s and (b) the 2050s for A2 1.1 Multi-stage integrated framework scenario in August 33 methodology 3 4.4 Percentage change in discharges in 2.1 Agricultural and total GDP growth (a) the 2030s and (b) the 2050s for A2 trends, 1975–2008 7 scenario in May 34 2.2 Historical trends in rice production 4.5 Total change in national flooded area quantities in Bangladesh, 1972–2006 8 for (a) 2030s A2, (b) 2030s B1, 2.3 Historical trends in land area under rice (c) 2050s A2, (d) 2050s B1 36 cultivation in Bangladesh, 1972–2006 8 4.6 Yearly peak levels at Jamuna station 2.4 Decomposition of historical Aman rice for the baseline and model production trends into land area and experiments (2030s) 37 yield contributions, 1972–2006 9 4.7 Average hydrographs (baseline, 2030s, 2.5 Observed yields for major staples 2050s) for MIROC GCM and A2 (kg/ha) 9 scenario on Teesta River 38 2.6 Time-series of flood-affected areas 4.8 Average hydrographs (baseline, 2030s, (km2) in Bangladesh (1954–2004) 11 2050s) for GFDL GCM and A2 scenario 2.7 Annual and seasonal precipitation on Meghna River 38 time-series (mm) averaged across 4.9 Average hydrographs on the Gorai Bangladesh Meteorological River (baseline, 2030s, and 2050s – Department stations 12 for CCSM A2 scenario) and plus/ 2.8 Average discharges in 1998 and 2002 minus one standard deviation bounds 39 for (a) Brahmaputra, (b) Ganges and 5.1 Baseline sub-regional yields with flood (c) Meghna rivers 13 damages applied (as a percentage of 2.9 Cropping calendar corresponding to undamaged yields) 48 flood land type 14 5.2 Percentage change (versus the 2.10 Aman crop production loss curve as a baseline undamaged simulation) in function of combined discharge 16 national potential production of a) aus, 2.11 Locations of coastal water level b) aman, c) boro and d) wheat 50 stations 17 5.3 Percentage change (versus the 2.12 Ganges-Brahmaputra-Meghna river baseline flood-only simulation) in basin 20 national potential production affected 3.1 Monthly, annual and seasonal by basin floods of a) aus and b) aman temperature changes 24 (boro and wheat are assumed to be 3.2 Monthly, annual and seasonal flood-free) 53 precipitation changes 25 5.4 Percentage of production lost to 4.1 Validated discharges from 1998–2007 coastal inundation associated with at (a) Bahadurabad (b) Hardinge Bridge 29 sea level rise in each coastal region 4.2 Temperature changes for A2 scenario sub-region (9–16) for three future over GBM basin (the 2050s) 32 scenarios, as compared to the baseline period (for A2 SRES) 54 viii Climate Change Risks and Food Security in Bangladesh 5.5 Percentage change (versus the 2.4 Peak discharge and timing during baseline flood-affected simulation) in extreme flood years 12 national potential production with the 2.5 Typical crop calendar for four different combined effects of CO2, temperature rice varieties 14 and precipitation, and basin flooding of 2.6 Hydrological regions and flood land a) aus, b) aman, c) boro and d) wheat 55 types 15 5.6 Regional production changes from 2.7 Summary of drought severity areas in baseline (per cent) for 2050s (a) aman, Bangladesh by crop season (in Mha) 16 (b) aus, (c) boro and (d) wheat 57 2.8 Estimated trends in water level of 6.1 Losses in total national rice production different stations along the coastline 18 due to existing climate variability, 2.9 Area affected by low, moderate and 2005–50 66 high salinity level (in 2005) 19 6.2 Losses in national rice production by 3.1 IPCC AR4 global circulation models 22 crop due to existing climate variability, 3.2 Summary precipitation statistics 2005–50, (a) aus, (b) aman, (c) boro 67 averaged across Bangladesh 6.3 Decomposing regional rice production (1960–2001) 26 losses due to existing climate 3.3 Sea level rise impacts on flood land variability, 2005–50 69 types 26 6.4 Losses in national agricultural GDP 4.1 The sub-regions with hydrological due to existing climate variability, region, agro-ecological zone and 2005–50 69 districts 32 6.5 Losses in national total GDP due to 4.2 Estimated average change (per cent) existing climate variability, 2005–50 71 in discharge across all model 6.6 Losses in total national rice production experiments 33 due to climate change, 2005–50 73 4.3 Modelled baseline season flood land 6.7 Losses in national rice production by type distribution for each month (ha) 35 crop due to climate change, 2005–50 75 4.4 Protected areas flood control and 6.8 Deviation in average ï¬?nal year rice drainage infrastructure (FCDI) 35 production from the Existing Variability 4.5 Number of model experiments Scenario under the Average Climate exceeding one standard deviation Change Scenario, 2050 76 bounds on baseline (2030s/2050s) and 6.9 Deviation in average ï¬?nal year rice 2050s average estimated change in production from the Existing Variability area flooded 36 Scenario under different emissions 4.6 Peak water level summary for the 2050s 37 scenarios, 2050 76 5.1 Sub-regional agricultural information 43 6.10 Losses in national agricultural GDP 5.2 Climate information for each sub- due to climate change, 2005–50 77 region: the representative BMD station, 6.11 Losses in national total GDP due to its code and annual mean climate climate change, 2005–50 79 statistics during the 1970–99 baseline 6.12 Cumulative discounted losses due to period 44 climate change as a share of total 5.3 Soil proï¬?le information for each GDP, 2005–50 80 sub-region 44 7.1 Layout of modiï¬?ed sorjan system 88 5.4 Agriculture management options for simulations of the three main rice varieties 45 Tables 5.5 Agriculture management options for 2.1 Production of different crop varieties wheat simulations 46 (metric tons) 6 5.6 Flood damages (percentage yield) 2.2 Flood classiï¬?cations 10 according to submergence depth, 2.3 Comparison of losses resulting from duration and phenological stage 47 recent large floods 11 5.7 Representative hydrographs 48 List of Figures and Tables ix 5.8 Carbon dioxide concentrations (ppm) A1.1 Cultivars available in the DSSAT for baseline period and future climate v4.5 CERES-Rice model 108 scenarios 51 A1.2 Genetic coefï¬?cients in the DSSAT 5.9 Median integrated production change v4.5 CERES-Rice model 109 (per cent) for the 2030s and 2050s 54 A1.3 Cultivars available in the DSSAT 5.10 Sub-regional average production v4.5 CERES-Wheat model 109 changes (per cent) disaggregated by A1.4 Genetic coefï¬?cients in the DSSAT crop (aman, aus, boro, wheat) and v4.5 CERES-Wheat model 109 climate risk for 2050s – all scenarios 58 A1.5 Planting method options in the 6.1 Summary of the Optimal Climate DSSAT v4.5 models 110 Scenario 65 A1.6 Tillage implements available in the 6.2 National rice production losses due DSSAT v4.5 models 110 to existing climate variability, 2005–50 66 A1.7 Irrigation options in the DSSAT v4.5 6.3 Regional rice production losses due models 111 to existing climate variability, 2005–50 68 A1.8 Fertilizer types in the DSSAT v4.5 6.4 Losses in GDP due to existing climate models 111 variability, 2005–50 70 A1.9 Fertilizer and organic amendment 6.5 Losses in national households’ application options in the DSSAT consumption spending due to existing v4.5 models 111 climate variability, 2005–50 72 A1.10 Organic amendments available in 6.6 National rice production losses due to the DSSAT v4.5 models 111 climate change, 2005–50 74 A2.1 Simple CGE model equations 114 6.7 Average GDP losses due to climate A2.2 Simple CGE model variables and change, 2005–50 78 parameters 115 6.8 GDP losses under different climate A2.3 Summary of climate impact change scenarios, 2005–50 78 channels in economy-wide model 6.9 Losses in national households’ simulations 116 consumption spending due to climate A3.1 Basic structure of a SAM 120 change, 2005–50 80 A3.2 Sectors in the 2005 Bangladesh SAM 122 6.10 Losses in regional farm households’ A3.3 Average cultivated crop land consumption spending due to climate allocation across divisions and scale change, 2005–50 81 of production 123 7.1 Sample of past and present A3.4 National and divisional per cent of programmes on adaptation in the gross domestic product (GDP) 124 agriculture sector 84 A3.5 Household factor income shares 7.2 Sample adaptation options in the from the 2005 Bangladesh SAM 125 agriculture sector 84 A3.6 Household factor income shares from 7.3 Estimated costs and beneï¬?ts of the 2005 Household Income and selected adaptation options 85 Expenditure Survey 125 7.4 Common vegetable cultivation A3.7 Household consumption 126 patterns 101 A3.8 2005 macro SAM for Bangladesh 7.5 Common vegetable cropping (millions of Taka) 127 patterns for sorjan system 104 A3.9 Cross-entropy SAM estimation equations 131 Acknowledgements This report was prepared by a team led by (from the Department of Agriculture Exten- Winston H. Yu (Task Team Leader, World Bank). sion), M. Akkas Ali (from the Bangladesh Agri- Speciï¬?c team contributions included: Mozaharul cultural Research Institute), Jiban Krishna Biswas Alam, Rabi Uzzaman, Aminur Rahman (Bangla- (from the Bangladesh Rice Research Institute), desh Center for Advanced Studies) and Sk. Mahmuder Anwar, Zahangir Alam, and Razaul Ghulam Hussain (Bangladesh Agricultural Research Hoque (from the Agricultural Information Serv- Council), who identiï¬?ed and evaluated the adapta- ice), and Ad Spijkers, Dr C.S. Karim, Tommaso tion options present in this study and pro- Alacevich, Ciro Fiorillo, and Z Karim (from the vided agricultural information for crop models; Food and Agriculture Organization). Brooke Ahmadul Hassan, Bhuiya Md. Tamim Al Hossain, Yamakoshi, Michael Westphal, and Siobhan Mohammad Ragib Ahsan, Ehsan Haï¬?z Chowd- Murray (World Bank) also provided valuable hury and Giasuddin Ahmed Choudhury (Center technical assistance during this study. for Environmental and Geographic Information Finally, the authors are grateful for the sup- Systems), who provided an analysis of the historical port of the World Bank management team and hydrology of the country; Abu Saleh Khan, Sardar several of the Bangladesh country ofï¬?ce staff M Shah-Newaz, Sohel Masud and Emaduddin including: Zhu Xian (former Country Director), Ahmad (Institute of Water Modelling), who devel- Robert Floyd, Adolfo Brizzi, Gajan Pathmanathan, oped the models used to project changes in future John Henry Stein, Simeon Ehui, Karin Kemper, flooding; Alex C. Ruane, Cynthia Rosenzweig, Masood Ahmad, Nihal Fernando, S.A.M. Raï¬?quz- David C. Major, Radley Horton and Richard zaman, Shakil Ferdausi and Khawaja Minnatul- Goldberg (Columbia University), with the advice lah. Ryma Aguw, Tarak Chandra Sarker, Venkat of Md Sohel Pervez, who provided the future cli- Ramachandran and Talat Fayziev helped in the mate scenarios and developed the models used administration of this study. The photographs of to project future crop production; and James various adaptation options were provided by the Thurlow and Paul Dorosh (International Food Department of Agricultural Extension, Ministry of Policy Research Institute), who developed the Agriculture, Bangladesh and the Livelihood Adap- computable general equilibrium model. tation to Climate Change (LACC-II) Project. The authors beneï¬?ted enormously from Generous support for this study was pro- the many technical discussions with colleagues vided by the World Bank, the Global Facility for and Government of Bangladesh ofï¬?cials. Spe- Disaster Reduction and Recovery (GFDRR), ciï¬?c reviewers included: Richard Damania, the Trust Fund for Environmentally and Socially Julia Bucknall, Abel Lufafa, Ian Noble, Nagaraja Sustainable Development (TFESSD), the Eco- Harshadeep, and Anna Bucher (from the World nomics of Adaptation to Climate Change Bank), Abu Wali Raghib Hassan, Md Mahsin, (EACC) study team, and the Bank-Netherlands Sanjib Saha, Md Abul Hossain, Mazharul Aziz, Water Partnership Program (BNWPP). We also Mohammad Ataur Rahman, and Bilkish Begum acknowledge the Program for Climate Model xii Climate Change Risks and Food Security in Bangladesh Diagnosis and Inter-comparison (PCMDI) and This study is a product of the staff and con- the WCRP Working Group on Coupled Model- sultants of the World Bank.The ï¬?ndings, interpre- ling (WGCM) for their roles in making available tations, and conclusions expressed in this paper the WCRP CMIP3 multi-model dataset. Sup- do not necessarily reflect the views of the Execu- port of this dataset is provided by the Ofï¬?ce of tive Directors of The World Bank or the govern- Science, US Department of Energy. ments they represent. Foreword This report is an important ï¬?rst step in better wisely to protect its citizenry to ensure growth understanding how climate risks (both cur- and a prosperous nation. This includes invest- rent and future) can undermine food security ments in infrastructure, including embankments in Bangladesh. It identiï¬?es key areas that require and cyclone shelters which have saved count- concerted effort by the government and its many less numbers of lives, in early warning systems to development partners. help the country prepare for imminent disasters, The year 2007 was indicative of the develop- and polders to protect vital agricultural areas to ment challenges that Bangladesh faces. Severe maintain production to feed its population. The flooding from July to September 2007 along the gains from these investments continue to support Ganges and Brahmaputra rivers affected over 13 a growing nation. million people in 46 districts and caused extens- Climate change, however, threatens to offset ive damage to agricultural production and physi- to some degree these important advances. The cal assets. With hardly any time to recover, on 15 prospect of changing temperatures and precipita- November 2007 the deadly Cyclone Sidr, a cate- tion patterns, the uncertainty of the timing and gory IV storm, made landfall across the southern magnitude of extreme events, and rising sea levels coast of the country, causing over 3000 deaths. will have important impacts on the agriculture The economic damages amounted to over US$1 sector. Action is needed today because Bangla- billion, with over a million tons of rice destroyed. desh will continue to depend on the agriculture Then, the increase in international prices of oil sector for growth and poverty reduction. Invest- and food, which Bangladesh imports, put further ments from the public and private sectors will strains on both government budgets and house- have to increase if Bangladesh is to ensure food hold livelihoods. security for its current and future populations. The long-term economic consequences of The challenges that the agriculture sector will these three simultaneous shocks remain to be seen, face as it adapts to climate change coincide well but they have shown the inherent vulnerability with the needs required to address the climate of Bangladesh to climate risks and the degree to variability risks of today. Thus, the adaptation which food security remains a major challenge options identiï¬?ed are no-regret approaches and for the country. With too much water during only a small example of what is possible. I hope the heavy monsoon months and too little water that this report can serve as a useful and mean- during the spring and early summer months, ingful guide for Bangladesh (and other countries) communities have needed to adapt to changing in addressing a future uncertain world. conditions. They have done so by adopting new varieties of crops and new farming practices and by starting small businesses and trades to diver- sify incomes. Furthermore, over the last several decades the government has invested heavily and Executive Summary Background Climate is only one input factor in a sector that is already under pressure Bangladesh is one of the countries most vulnerable to The achievement of food self-sufï¬?ciency remains climate risks a key development agenda for the country. Sig- From annual flooding to a lack of water during niï¬?cant progress has been made in the sector the dry season, from frequent coastal cyclones and since the 1970s, in large part due to the rapid storm surges to changing groundwater aquifer expansion of surface and groundwater irriga- conditions, the importance of adapting to climate tion and the introduction of new high-yielding risks to maintain economic growth and reduce crop varieties. The production of rice and wheat poverty is clear. Households have for a long time increased from about 10 million tonnes/metric needed to adapt to these dynamic conditions to tons (10Mt) in the early 1970s to almost 30Mt maintain their livelihoods. Moreover, substantial by 2001. The challenge now for Bangladesh is to public investment in protective infrastructure enhance productivity, especially as demands for (e.g. cyclone shelters, embankments) and early food increase with the growing population (1.3 warning and preparedness systems has played and per cent growth rate) and improved incomes. will continue to play a critical role in minimizing Moreover, overuse, degradation and changes in these impacts. In the long list of potential impacts resource quality (e.g. salinity) will place addi- from climate change, the risks to the agriculture tional pressures on already constrained available sector stand out as among the most important. land and water resources. Agriculture is a key economic sector in Bangladesh, Climate change is recognized as a key sustainable accounting for nearly 20 per cent of the GDP (gross development issue for Bangladesh domestic product) and 65 per cent of the labour force Future climate change risks will be additional The performance of the sector has considerable to the challenges the country and sector already influence on overall growth, the trade balance, face. Long-term changes in temperatures and the budgetary position of the government, and precipitation have direct implications on evapora- the level and structure of poverty and malnutri- tive demands and consequently on agriculture tion in the country. Moreover, much of the rural yields. Moreover, water-related disasters may population, especially the poor, is reliant on the increase in magnitude and frequency. Finally, sea agriculture sector as a critical source of livelihood level rise may have important implications for the and employment. Many also depend on the agri- sediment balance and may alter the proï¬?le of the culture sector indirectly through employment in area inundated and salinity in the coastal areas. small-scale rural enterprises that provide goods and services to farms and agro-based industries and trades. xvi Climate Change Risks and Food Security in Bangladesh The objective of this study is to examine the implications the primary drivers of declining overall produc- of climate change on food security in Bangladesh and to tion during major flood events (driven mainly by identify adaptation measures in the agriculture sector area changes); these losses, however, are increas- This objective is achieved in the following ways. ingly being compensated for by ‘boro’ (dry sea- First, the most recent science available is used to son rice). As a result, compared to the pre-1990s, characterize current climate and hydrology and agricultural GDP is becoming less sensitive to its potential changes. Second, country-speciï¬?c this climate variability. Finally, droughts and survey and biophysical data is used to derive coastal inundation from sea level rise can have more realistic and accurate agricultural impact consequences for agriculture production as large functions and simulations. A range of climate as those from floods. risks (i.e. warmer temperatures, higher carbon dioxide concentrations, changing characteristics Future Climate (Chapter 3) of floods, droughts and potential sea level rise) Using global climate models (GCMs), a trend toward a is considered to gain a more complete picture warmer and wetter future climate is projected to impact of potential agriculture impacts. Third, while the agriculture sector, particularly if the climate state estimating changes in production is impor- goes beyond the variations found in the historical record tant, economic responses may to some degree buffer against the physical losses predicted, and Median warming of 1.1°C, 1.6°C and 2.6°C by an assessment is made of these. Food security is the 2030s, 2050s and 2080s respectively is pro- dependent not only on production stocks, but jected from a range of plausible scenarios. Median also future food requirements, income levels and annual precipitation increases of 1 per cent, 4 per commodity prices. Fourth, adaptation possibili- cent and 7.4 per cent by the 2030s, 2050s and ties are identiï¬?ed for the sector. The framework 2080s respectively is projected with greater con- established here can be used effectively to test trasts between the wet and dry seasons. Greater such adaptation strategies. Multiple models are model uncertainty (in terms of magnitude and used in this integrated study, and as with all mod- direction) exists with future precipitation than els, parameters may not be known with precision future temperature. Simulated future tempera- and functional forms may not be fully accurate; ture changes signiï¬?cantly separate from the thus, careful sensitivity analysis and a full under- background temperature variations. Precipitation standing of limitations (identiï¬?ed throughout the is subject to large existing inter-annual and intra- study) are required. annual variations. Projections of precipitation changes vary widely amongst models, with small Vulnerability to Climate Risks (Chapter 2) median changes compared to historic variability. Using three scenarios of future sea level rise (15 The performance of the agriculture sector is heavily cm, 27 cm, and 62 cm) the total area that peren- dependent on the characteristics of the annual flood nially floods is projected to increase by 6%, 10%, Regular flooding of various types (e.g. flash, river- and 20% respectively. ine) has traditionally been beneï¬?cial. However, low frequency but high magnitude floods can Future Floods (Chapter 4) have adverse impacts on rural livelihoods and production (e.g. the 1998 flood resulted in a loss Primarily driven by increased monsoon precipitation in of over 2Mt of production). The timing of the the Ganges-Brahmaputra-Meghna (GBM) basin, models peaks of the three major river systems (Ganges, on average demonstrate increased future flows in the Brahmaputra and Meghna) is an important deter- three major rivers into Bangladesh (as much as 20 per cent) minant of the overall magnitude of flooding. The Larger changes are anticipated by the 2050s com- economy-wide impact of these extreme events pared to the 2030s. Larger changes are observed can be substantial. Impacts on the ‘aman’ (mon- on average for the Ganges. The exact magnitude soon season rice) and ‘aus’ (inter-season rice) are is dependent on the month. Given that most Executive Summary xvii GCMs project both an increasing trend of mon- and labour reallocation, price effects). These eco- soon rainfall and greater inflows into Bangla- nomic effects will to some degree buffer against desh, it follows that the flooding intensity would the physical losses predicted. worsen. On average, models simulate increases in flooded area in the future (over 10 per cent by Economy-wide Impacts of Climate Risks 2050).This is primarily located in the central part (Chapter 6) of the country at the confluence of the Ganges and Brahmaputra rivers and in the south. Existing climate variability can have a pronounced Moreover, increases in yearly peak water lev- detrimental economy-wide impact els are projected for the northern sub-regions This is explored using a dynamic computable and decreases are projected for the southern sub- general equilibrium (CGE) model. Compared to regions. Not all estimated changes are statisti- an ‘optimal’ climate simulation in which highest cally signiï¬?cant. Model experiments demonstrate simulated yields are used and sector productiv- more changes that are signiï¬?cant by the 2050s. ity and factor supplies increase smoothly at aver- Changes are in general less than 0.5m from the age long-term growth rates with no inter-annual baseline. Furthermore, across the sub-regions, variations, climate variability is estimated to most GCMs show earlier onset of the monsoon reduce long-term rice production by an average and a delay in the recession of flood waters. 7.4 per cent each year over the 2005–50 simu- lation period. This primarily lowers the produc- Future Crop Performance (Chapter 5) tion of the aman and aus crop. Average annual rice production growth is lowered in all sub- The median of all rice crop projections shows declining regions. This simulated variability is projected to national production, with boro showing the largest cost the agriculture sector (in discounted terms) median losses US$26 billion in lost agricultural GDP during Potential future crop production is projected the 2005–50 period. This climate variability has using well-developed crop models considering economy-wide implications beyond simply the multiple climate impacts (temperature and pre- size-effect of the lost agricultural GDP. Existing cipitation changes, CO2 fertilization, flood climate variability is estimated to cost Bangladesh changes, sea level rise). For aus (-1.5 per cent) US$121 billion in lost national GDP during this and aman (-0.6 per cent) the range of model period (US$3 billion per year). This is 5 per cent experiments for the 2050s covers both potential below what could be achieved if the climate were gains and losses and does not statistically sepa- ‘optimal’. rate from zero. However, most GCM projections estimate a potential decline in boro production Climate change exacerbates the negative impacts with a median loss of 3 per cent by the 2030s of existing climate variability by further reducing rice and 5 per cent by the 2050s. Wheat production production by a projected cumulative total of 80Mt over is projected to increase out to the 2050s (+3 per 2005–50 (about 3.9 per cent each year), driven primarily cent). Boro and wheat changes are conservative by reduced boro crop production as it is assumed that farmers have unconstrained This is equivalent to almost 2 years worth of rice access to irrigation. In each sub-region, produc- production lost over the next 45 years as a result tion losses are estimated for at least one crop. of climate change. Uncertainty about future cli- The production in the southern sub-regions is mate change means that annual rice production most vulnerable to climate change. For instance, losses range between 3.6 per cent and 4.3 per average losses in the Khulna region are -10 per cent. Climate change has particularly adverse cent for aus, aman and wheat, and -18 per cent implications for boro rice production and will for boro by the 2050s due in large part to ris- limit its ability to compensate for lost aus and ing sea levels. These production impacts ignore aman rice production during extreme climate economic responses to these shocks (e.g. land events. This will further jeopardize food security xviii Climate Change Risks and Food Security in Bangladesh in Bangladesh, necessitating greater reliance on in rice production due to climate change. This other crops and imported food grains. Rice pro- is for three reasons. First, these regions already duction in the southern regions of Patuakhali experience signiï¬?cant declines in aus and aman and Khulna is particularly vulnerable. rice production due to climate variability, which is expected to worsen under climate change. Sec- Overall, agricultural GDP is projected to be 3.1 per cent ond, boro yields are severely affected by changes lower each year as a result of climate change (US$7.7 in mean rainfall, temperature and mean shifts in billion in lost value-added) the flood hydrographs. Thus, reductions in boro Climate change also has broader economy-wide production limit the ability for these regions to implications. This is estimated to cost Bangladesh compensate for lost aus and aman rice production US$26 billion in total GDP over the 45-year during extreme events. The south is also affected period 2005–50, equivalent to US$570 million the most by rising sea levels, which permanently overall lost each year due to climate change, or reduce cultivable land. The largest percentage alternatively an average annual 1.15 per cent declines in per capita consumption are projected reduction in total GDP. Average loss in agri- in these regions. Finally, the northwest is also vul- cultural GDP due to climate change is projected nerable as the lost consumption is a large fraction to be a third of the agricultural GDP losses asso- of the existing household consumption. Adapta- ciated with existing climate variability. Uncer- tion measures should focus on these areas. tainty surrounding GCMs and emission scenarios means that costs may be as high as US$1 billion Adaptation Options in the Agriculture per year in 2005–50 under less optimistic sce- Sector (Chapter 7) narios. Moreover, these economic losses are pro- Adaptation options can address several different climate jected to rise in later years, thus underlining the risks need to address climate change related losses in Bangladesh will continue to depend on the agri- the near-term. culture sector for economic growth. Rural house- holds will continue to depend on the agriculture These climate risks will also have severe implications for sector for income and livelihoods. Though the household welfare government has made substantial investments to For both the climate variability and climate change increase the resilience of the poor (e.g. new high- simulations, around 80 per cent of total losses fall yielding crop varieties, protective infrastructure, directly on household consumption (cumulative disaster management), existing constraints in the total consumption losses of US$441.7 billion and sector may be exacerbated by long-term effects US$104.7 billion for climate variability and cli- of climate change. The scale of current efforts mate change simulations respectively). Also, about remains limited and is not commensurate with 80 per cent of the economic losses occur outside the probable impacts. A no-regrets strategy is to of agriculture, particularly in the upstream and promote activities and policies that help house- downstream agriculture value-added processing holds build resilience to existing climate risks sectors. This means that both rural and urban today. households are adversely affected. Per capita con- sumption is projected to fall for both farm and Both processes of adapting to climate change and non-farm households. stimulating the agriculture sector to achieve rural growth and support livelihoods align well. The southern and northwest regions are the most This requires, among other things, efforts to: vulnerable diversify household income sources; improve The south sits at the confluence of multiple cli- crop productivity; support greater agricultural mate risks, as shown throughout this study. These research and development; promote education areas are expected to experience the largest decline and skills development; increase access to ï¬?nan- Executive Summary xix cial services; enhance irrigation efï¬?ciency and livelihoods and develop sustainably. As popula- overall water and land productivity; strengthen tions grow, the ability for many countries to climate risk management; and develop protec- meet basic food requirements and effectively tive infrastructure. Moreover, the current large manage future disasters will be critical for sus- gap between actual and potential yields suggests taining long-term economic growth. These are substantial on-farm opportunities for growth and challenges above and beyond those that many poverty reduction. Expanded availability of mod- countries are already currently facing. ern rice varieties, irrigation facilities, fertilizer use The integrated framework used in this analy- and labour could increase average yields at rates sis provides a broad and unique approach to esti- that could more than offset the climate change mating the hydrologic and biophysical impacts of impacts. Signiï¬?cant additional planning and climate change, the macro-economic and house- investments in promoting these types of adapta- hold-level impacts and an effective method for tions are still needed. assessing a variety of adaptation practices and policies. The framework presented here can serve The Way Forward (Chapter 8) as a useful guide to other countries and regions faced with similar development challenges and The precise impact of climate change on coun- objectives of achieving food security. Continued tries in the developing world remains to be seen. reï¬?nements to the assessment approach devel- This much is known, however: climate change oped in this volume will further help to sharpen poses additional risks to many developing coun- critical policies and interventions by the Bangla- tries in their efforts to reduce poverty, promote desh government. Glossary of Terms B. aman: broadcast aman; a rice crop usually Kharif: the wet season (typically March to Octo- planted in March/April under dry land condi- ber) characterized by monsoon rain and high tions, but in areas liable to deep flooding. Also temperatures. known as deep water rice. This crop is harvested from October to December. All varieties are Kharif 1: the ï¬?rst part of the kharif season highly sensitive to day length. (March to June). Rainfall is variable and temper- atures are high. The main crops grown are Aus, T. aman: transplanted aman; a rice crop usually summer vegetables and pulses. Broadcast aman planted in July/August, during the monsoon, in and jute are planted. areas liable to a maximum flood depth of about 0.5m. This crop is harvested from November/ Kharif 2: the second part of the kharif season December. Local varieties are sensitive to day (July to October) characterized by heavy rain and length whereas modern varieties are insensitive floods. T. aman is the major crop grown during or only slightly sensitive. the season. Harvesting of jute takes place. Fruits and summer vegetables may be grown on high B. aus: broadcast aus; a rice crop planted in land. March/April under dry land conditions. Matures on pre-monsoon showers, harvested in June/July, Rabi: The dry season (typically November to and is insensitive to day length. February) with low or minimal rainfall, high evapo-transpiration rates, low temperatures and T. aus: transplanted aus; a rice crop, transplanted clear skies with bright sunshine. Crops grown are in March/April, usually under irrigated condi- boro, wheat, potato, pulses and oilseeds. tions, and harvested June/July. The distinction between late planted boro and early transplanted High yielding variety: introduced varieties aus is academic since the same varieties may be developed through formal breeding programmes, used. Varieties are insensitive to day length. they have a higher yield potential than local varie- ties but require correspondingly high inputs of Boro: a rice crop planted under irrigation dur- fertilizer and irrigation water to reach full yield ing the dry season from December to March and potential. harvested in April to June. Local boro varieties are more tolerant of cool temperatures and are Local varieties developed and used by farm- usually planted early in areas which are subject to ers: Sometimes referred to as inbred varieties or early flooding due to rise in river levels. Improved local improved varieties (LIVs). varieties, less tolerant of cool conditions, are usu- ally transplanted from February onwards. All Net cultivable area: total area which is under- varieties are insensitive to day length. taken for cultivation. Acronyms AIS Agricultural Information Service GBM Ganges-Brahmaputra-Meghna AR4 Fourth Assessment Report GCM global climate model BARC Bangladesh Agricultural Research GDP gross domestic product Council GOB Government of Bangladesh BARI Bangladesh Agricultural Research GTOPO Global Topography Institute HIES Household Income and Expenditure BBS Bangladesh Bureau of Statistics Survey BCAS Bangladesh Centre for Advanced Studies HYV high yielding variety BINA Bangladesh Institute of Nuclear IFPRI International Food Policy Research Agriculture Institute BMD Bangladesh Meteorological Department IMF International Monetary Fund BRRI Bangladesh Rice Research Institute IPCC Intergovernmental Panel on Climate BWDB Bangladesh Water Development Board Change CEGIS Center for Environmental and IWM Institute of Water Modelling Geographic Information Services LACC Livelihood Adaptation to Climate CERES Crop Environment Resource Synthesis Change CGE computable general equilibrium MJO Madden-Julian Oscillation CO2 carbon dioxide Mt million tonnes (million metric tons) DAE Department of Agriculture Extension MPO Master Plan Organization DEM digital elevation model MSL mean sea level DSSAT Decision Support System for NASA National Aeronautics and Space Agency Agrotechnology Transfer NCA net cultivable area ENSO El Niño-Southern Oscillation NGO non-governmental organization FAO Food and Agriculture Organization PCMDI Program for Climate Model Diagnosis FCDI Flood control and drainage infrastructure and Inter-comparison FFWC Flood Forecast and Warning Center RCM regional climate model xxiv Climate Change Risks and Food Security in Bangladesh SAM social accounting matrix TAR Third Assessment Report SRES Special Report on Emissions Scenario TRMM Tropical Rainfall Measuring Mission SRTM Shuttle Radar Topography Mission USGS United States Geologic Survey 1 Introduction Bangladesh is one of the most vulnerable countries the rural population, especially the poor, is reliant to climate risks, both from existing variability and on the agriculture sector as a critical source of future climate change. From annual flooding of all livelihoods and employment. Many may also do types to a lack of water resources during the dry so indirectly through employment in small-scale season, from frequent coastal cyclones and storm rural enterprises that provide goods and services surges to changing groundwater aquifer condi- to farms and agro-based industries and trades. tions, the importance of adapting to these risks to Climate is only one input factor in an agri- maintain economic growth and reduce poverty culture sector that is already under pressure. The is clear. Households have for a long time needed achievement of food self-sufï¬?ciency remains a to adapt to these dynamic conditions to maintain key development goal for the country. Signiï¬?cant their livelihoods. The nature of these adaptations progress has been made in the sector since the and the determinants of success depend on the 1970s, in large part due to the rapid expansion availability of assets, labour, skills, education, and of surface and groundwater irrigation and the social capital. The relative severity of disasters has introduction of new high-yielding crop varieties. decreased substantially since the 1970s, however, The production of rice and wheat increased from as a result of improved macro-economic manage- about 10 million tonnes/metric tons (10Mt) in ment, increased resilience of the poor and signiï¬?- the early 1970s to almost 30Mt by 2001. The cant progress in disaster management. Substantial challenge now for Bangladesh is to enhance pro- public investment in protective infrastructure ductivity, especially as demands for food increase (e.g. cyclone shelters, embankments) and early with the growing population (1.3 per cent growth warning and preparedness systems have played a rate) and improved incomes. Moreover, overuse, critical role in minimizing these impacts. More degradation and changes in resource quality (e.g. investments are still required. In the long list of salinity) will place additional pressures on already potential impacts from climate change, the risks constrained available land and water resources. to the agriculture sector stand out as among the Future climate change risks will be additional most important. to the challenges the country and sector already Agriculture is a key economic sector in Bang- face. Long-term changes in temperatures and pre- ladesh, accounting for nearly 20 per cent of the cipitation have direct implications on evaporative GDP and 65 per cent of the labour force. The demands and consequently on agriculture yields. performance of the sector, here to include crops Increased carbon dioxide concentrations may also (70 per cent of agricultural GDP), livestock (10 impact the rates of photosynthesis and respiration. per cent) and ï¬?sheries (10 per cent), has con- Moreover, water-related disasters may increase in siderable influence on overall growth, the trade magnitude and frequency. In fact, between 1991 balance, the budgetary position of the govern- and 2000, 93 major disasters were recorded, result- ment, and the level and structure of poverty and ing in billions of US$ in losses, most of which malnutrition in the country. Moreover, much of were in the agriculture sector. Sea level rise may 2 Climate Change Risks and Food Security in Bangladesh have important implications on the sediment bal- term production losses and the impacts on long- ance and may alter the proï¬?le of available land term prospects is dependent on many macro and for production in the coastal areas. It is clear that micro factors. Third, the prospects of sea level climate change is a key sustainable development rise in the coastal areas will change the proï¬?le issue for Bangladesh (World Bank, 2000). of available land for agriculture production and potentially the quality of groundwater used for 1.1 Objective of Study irrigation. This is especially critical in land-con- strained countries such as Bangladesh. Increases The objective of this study is to examine the in carbon dioxide concentrations will also impact implications of climate change on food secu- the rates of photosynthesis and respiration. rity in Bangladesh and to identify adaptation Much of the existing analysis on climate measures in the agriculture sector. This objec- change impacts on the agriculture sector has pri- tive is achieved in the following ways. First, the marily been focused on the ï¬?rst driver: changes most recent science available is used to charac- in temperature and precipitation. Several global terize current climate and its potential changes. studies look at these impacts. For instance, Cline Second, country-speciï¬?c survey and biophysical (2007) demonstrates using a range of method- data is used to derive more realistic and accu- ologies and several global circulation models rate agricultural impact functions and simula- (GCMs) that agriculture production may decline tions. A range of climate risks (i.e. warmer tem- in Bangladesh by as much as between 15 and 25 peratures, higher carbon dioxide concentrations, per cent. This study is dependent on global sta- changing characteristics of floods, droughts and tistical production functions. Fischer et al (2002) potential sea level rise) is considered, to gain a derive similar estimates using an agro-ecological more complete picture of potential agriculture approach and the results from four global circula- impacts. Third, while estimating changes in pro- tion models. duction is important, this is only one dimension Several regional level studies also exist which of food security considered here. Food security is show mixed responses to climate change. Lal et al dependent on several socio-economic variables (1998a,b,c) demonstrate that rice yields in neigh- including estimated future food requirements, boring India could decline by 5 per cent under income levels and commodity prices. Fourth, a 2°C warming and CO2 doubling. Karim et al adaptation possibilities are identiï¬?ed for the sec- (1994) indicated a decrease in potential yields for tor. The framework established here can be used aman and boro rice in Bangladesh when only a effectively to test such adaptation strategies. 2°C or 4°C temperature change is considered, but this decrease was nearly offset when the physio- 1.2 Literature Review logical effect of 555 parts per million (ppm) CO2 fertilization was taken into account. More recent Global changes in climate will have important results (Karim et al, 1998; Faisal and Parveen, implications for the economic productivity of the 2003) show overall enhancement of potential agriculture sector.The sector will be impacted by rice yields but declines in potential wheat yields three primary water-related climate drivers. First, when 4°C temperature changes and 660ppm gradual changes in the distribution of precipi- CO2 fertilization are simulated. The offset poten- tation and temperature will impact agriculture tial by carbon fertilization effects remains an area yield through possible changes in water availabil- of active research (Long et al, 2005;IPCC, 2007b; ity and evaporative demands, tolerance of crops Tubiello et al, 2007a,b; Hatï¬?eld et al, 2008; Ains- and incidence of pest attacks. Second, changes in worth et al, 2008). the frequency and magnitude of extreme events Although it is clear that floods can affect (i.e. above-average floods, prolonged droughts) agriculture production signiï¬?cantly, little is may result in additional shocks to the agriculture known about the incremental future damages sector. The ability to recover from these short- from more frequent extreme events or increased Introduction 3 discharges. Economic damages have been cal- picture of the geographic distribution of climate culated after several recent extraordinary flood change impacts on the agriculture sector. Then, events (e.g. almost US$700 million in agriculture the economic implications of these projected losses were reported after floods in 2004). Hus- crop yield changes are assessed using a dynamic sain (1995) developed a methodology to incor- computable general equilibrium (CGE) model. porate yield losses from annual flooding into a The CGE model estimates their economy-wide crop simulation model. Sea level rise and salinity implications, including changes in production intrusion implications on the agriculture sector and household consumption for different sectors, are even less understood. Habibullah et al (1998) household groups and agro-climatic sub-regions calculated that the loss of food-grain due to soil in the country. Additional impacts from extreme salinity intrusion in the coastal districts is about events are also considered here. 200,000 to 650,000 tons. As noted, multiple models are used in the study. These are among the best mathemati- 1.3 Integrated Modelling cal representations available of the physical and economic responses to a variety of exogenous Methodology changes (here, climate). However, like all model- The methodology employed in this study includes ling approaches, uncertainty exists as parameters several stages. Climate and hydrologic models may not be known with precision and functional are used to produce future scenarios of climate forms may not be fully accurate. Thus, careful and land inundation (from floods and sea level sensitivity analysis and an understanding and rise) for various GCMs and emissions scenarios. appreciation of the limitations of these models Then, these are linked to crop models to produce (identiï¬?ed throughout the study) are required. physical estimates of climate- and flood-affected Further collection and analysis of critical input potential crop yield changes for the three main and output observations (e.g. climate data, farm- rice varieties and wheat.These yield estimates are level practices and irrigation constraints) will based on climate and biophysical data for 16 agro- enhance this integrated framework methodology climatic sub-regions in Bangladesh and provide a and future climate impact assessments. Global climate models with emissions Water basin models Crop models Economy-wide model scenarios Changes in the frequency of extreme events Output: Downscaled Output: Land inundation Output: Potential Output: Changes in future climate due to rising sea levels Yield changes production and predictions and changing floods (by (by crop and region) consumption (by crop, (e.g. change in mean region) sector, region and temp, rainfall and household group) CO2) Adaptation Options in the Agriculture Sector Figure 1.1 Integrated modelling framework 4 Climate Change Risks and Food Security in Bangladesh 1.4 Organization of Study models is also presented as all available climate models could not be used. Chapter 5 describes This study is organized into seven further chap- the dynamic biophysical crop production models ters. Chapter 2 sets the historical context of cli- used. Here, various impacts of different climate mate risks in Bangladesh. Past experience with risks (floods, droughts and sea level rise) on agri- floods, droughts, sea level rise and observed culture yields, focusing on rice and wheat, are trends is reviewed. Broader regional issues are also incorporated. Chapter 6 describes a dynamic briefly discussed. Chapter 3 reviews the predicted computable general equilibrium model used to future changes in precipitation and temperature evaluate the macro-economic and household (both at the country level and at the Ganges- welfare impacts of both climate variability and Brahmaputra-Meghna [GBM] river basin level). change-induced yield losses and gains. Chap- Chapter 4 presents an analysis on modelling the ter 7 presents potential adaptation options for hydrology of future floods. This consists of both the agricultural sector including unit costs that descriptions of a regional and national hydrologic are currently being piloted in the ï¬?eld. Finally, models used and an analysis of the characteristics in Chapter 8, the study concludes with general of the future floods both temporally and spatially. recommendations. Annexes provide additional Among other aspects, the extent of the flood and information about using the crop models to test the changes in the peak floods are analysed. A adaptation options and technical details of the procedure for selecting a sub-set of global climate CGE. 2 Vulnerability to Climate Risks Box 2.1 Key messages • Despite the challenging physiography and extreme climate variability, Bangladesh has made signiï¬?- cant progress towards achieving food security. Investments in surface and groundwater irrigation and the introduction of high yielding crop varieties have played and will continue to play a key role in this. • The performance of the agriculture sector is heavily dependent on the characteristics of the annual flood. Regular flooding of various types has traditionally been beneï¬?cial. However, low frequency but high magnitude floods can have adverse impacts on rural livelihoods and production. • The timing of the peaks on the three major river systems (Ganges, Brahmaputra and Meghna) is an important determinant of the overall magnitude of flooding. • The economic toll of these extreme events can be signiï¬?cant, the order of billions of US dollars. • Aman and aus rice are the primary drivers of declining overall production during major flood events, which is increasingly being compensated for by boro rice. Agriculture share of total GDP is declining and is likely to continue to do so, thus increasingly insulating the country from these shocks. • Lean-season water availability, particularly in the northwest, can have consequences on agriculture production comparable to floods. • In coastal areas, agriculture productivity is affected by the surface and groundwater salinity distribution. • Future regional changes in the Ganges-Brahmaputra-Meghna basin will play an important role in the overall timing and magnitude of water availability in Bangladesh. Bangladesh is indeed a hydraulic civilization situ- climate risks, including severe flooding and peri- ated at the confluence of three great rivers – the odic droughts. Ganges, the Brahmaputra and the Meghna. Over Most of Bangladesh consists of extremely low 90 per cent of the Ganges-Brahmaputra-Meghna land. The capital city of Dhaka (population of (GBM) basin lies outside the boundaries of the over 12 million) is about 225km from the coast country. The extensive floodplains at the conflu- but within 8m above mean sea level (MSL). Land ence are the main physiographic feature of the elevation increases towards the northwest and country. The country is intersected by more than reaches a height of about 90m above MSL (Plate 200 rivers; there are 54 rivers that enter Bangla- 2.1). The highest areas are the hill tracts in the desh from India alone. Moreover, more than 80 eastern and Chittagong regions. The lowest parts per cent of the annual precipitation of the coun- of the country are in the coastal areas.These areas try occurs during the monsoon period between are particularly vulnerable to sea level rise and June and September.These hydro-meteorological tidal storm surges. characteristics of the three river basins are unique Bangladesh has a humid sub-tropical climate. and make the country vulnerable to a range of The year can be divided into four seasons: the 6 Climate Change Risks and Food Security in Bangladesh relatively dry and cool winter from December Table 2.1 Production of different crop varieties (tonnes) to February, the hot and humid summer from Crop Variety 1981 1991 2001 March to May, the southwest summer monsoon Local aus 2,176,670 1,630,006 ,980,650 from June to September and the retreating mon- HYV aus 1,044,810 , 690,590 , 934,950 soon from October to November. The southwest B. aman 1,499,430 1,006,230 , 962,520 summer monsoon is the dominating hydrologic HYV aman 1,083,890 3,596,210 6,938,360 driver in the GBM basin. The Tibetan Plateau, Local t. aman 4,309,705 3,923,520 3,348,050 the Great Indian Desert and adjoining areas of Local boro ,630,290 , 406,670 , 367,380 northern and central India heat up considerably HYV boro 1,756,945 5,816,200 11,573,560 Total 12,501,740 17,069,426 25,105,470 during the summers. This causes a low pressure area over the Indian subcontinent and western China which quickly ï¬?lls with moisture-laden winds from the Indian Ocean. The Himalayas act about 180. Table 2.1 shows the production of the like a wall, forcing moist air masses to rise in order different crop varieties of rice. to pass into the Tibetan Plateau. With the gain Plates 2.2 and 2.3 show the spatial distribu- in altitude of the clouds, the temperature drops tion of the aman (speciï¬?cally transplanted aman, and moisture condenses into heavy precipitation. or t. aman) and boro cropped areas respectively in Some areas of the South Asia subcontinent can the country. The total aman rice area cultivated receive up to 10,000mm of rain. was 5,225,058ha in the year 2002. The aman crop is grown mostly in the northern and south- ern regions. The total cropped area dedicated 2.1 The Success of Agriculture to aman rice is also slowing. The total aus rice Despite the challenging physiography and cropped area has declined signiï¬?cantly over the extreme climate variability, enormous success has years. In 1981, it was 3.11 million hectares (Mha) been achieved in the last several decades, with and only 1.33Mha in 2001. The total boro rice the country largely food self-sufï¬?cient. Agricul- area cultivated in Bangladesh was 3,973,414ha ture is the most important sector in the Bangla- (31 per cent of the total country area) in the year desh economy, contributing 19.6 per cent to the 2002, with production concentrated mostly in national GDP and providing employment for 63 the northern regions. Winter season boro crop- per cent of the population. Rice is the dominant ping is reduced in the southwest due to the pres- crop in Bangladesh. There are three major rice ence of saline water. The cropped area under varieties: aman (flood season rice), boro (dry sea- boro has increased signiï¬?cantly over the years. In son rice) and aus (inter-period rice). The over- 1980, it was 1.15Mha and increased to 3.76Mha all production of rice has increased from about in 2000. This is in large part due to the expan- 12Mt in 1981 to over 25Mt in 2001. Note that sion of groundwater irrigation (Plate 2.4). This the population increased from 90 to 129 million has raised some concerns in terms of overall sus- over this same time period. The rice production tainability as water tables have fallen dramatically growth rate from 1981 to 1991 was about 3 per over the decades. cent per annum and increased to 4 per cent per Besides rice, Bangladesh also produces a annum. The introduction of high yielding varie- number of other crops of which wheat, maize, ties of aman and boro and groundwater irriga- different types of pulses, oil seeds, jute, sugar cane, tion (surface and groundwater) have signiï¬?cantly tea and tobacco are signiï¬?cant. It is found that contributed to these gains. The aus crop has production of wheat has increased from 0.97Mt steadily decreased in response. Moreover, pub- in 1981 to 1.67Mt in 2001 (Plate 2.5). Maize lic investment in flood protection and drainage production has also increased from 1.35 thou- works have contributed to an overall increase sand tonnes in 1981 to 3.04 and 10.46 thousand in cropped area. Cropping intensity is at present tonnes in 1991 and 2001 respectively. Maize Vulnerability to Climate Risks 7 (mainly used for poultry feed) is particularly 1990s, reflecting the steadily falling share of agri- popular in the northwestern part of Bangladesh culture to total GDP as the economy becomes where droughts and high temperatures are com- more diversiï¬?ed. Major floods are indicated by mon. Total production of pulses increased from black dots in Figure 2.1. Until the 1990s, major 1981 to 1991 but declined from 1991 to 2001. floods resulted in sharp declines in agricultural Total production of different pulses was 0.20, GDP growth, with similar effects for total GDP. 0.52 and 0.37Mt for the years 1981, 1991 and However, after 1990 the relative effects of major 2001 respectively. Production of sugar cane is floods have diminished. Growth in fact remained typically between 6.5 and 7.5Mt and has showed positive even during the extraordinary flood of a decline in recent years. Sugar cane is the pri- 1998. mary input material for sugar mills operated by The composition of rice production has the public sector. clearly shifted towards greater reliance on boro rice (Figure 2.2). Major flood years are character- Historical climate variability and ized by sharp declines in aman and aus produc- tion. By contrast, boro production is increasingly agricultural production playing a compensating role, rapidly expanding Figure 2.1 below shows historical agricultural production during major flood years.This is most GDP growth and total GDP growth. Despite evident in the 1998 flood (and to a lesser extent the continued growth that Bangladesh has seen in the 1988 flood). Moreover, in years following over the last three decades (i.e. Bangladesh has a major flood, aman and aus production rebound not seen a single year of negative growth since as boro continues to grow. 1975), agricultural GDP growth remains highly The variability in aman production is even erratic. Moreover, total GDP and agricultural more pronounced when looking at the rice area GDP growth track fairly closely until the early under cultivation (Figure 2.3). Aman land area 15 100 Agricultural GDP growth 90 Total GDP growth 10 80 Agricultural GDP share Annual GDP growth rate (%) 70 Share of total GDP (%) 5 60 50 0 40 30 -5 20 10 -10 0 1975 1980 1990 2000 76 77 78 79 81 82 83 84 85 86 87 88 89 91 92 93 94 95 96 97 98 99 01 02 03 04 05 06 07 08 Figure 2.1 Agricultural and total GDP growth trends, 1975–2008 Note: Black dots represent years where the historical climate data indicate major flood occurrences; these are calendar years and represent the second part of a typical crop season (e.g. 1975 calendar year is the crop season 1974–5). Source: Bangladesh Bureau of Statistics, 2009; World Bank, 2009. 8 Climate Change Risks and Food Security in Bangladesh 45 Contribu�on to annual rice area growth Boro contribu�on 35 Aman contribu�on Aus contribu�on (percentage point) 25 Total growth rate 15 5 -5 -15 -25 1972/73 2000/01 73/74 74/75 75/76 76/77 77/78 78/79 79/80 80/81 81/82 82/83 83/84 84/85 85/86 86/87 87/88 88/89 89/90 90/91 91/92 92/93 93/94 94/95 95/96 96/97 97/98 98/99 99/00 01/02 02/03 03/04 04/05 05/06 Figure 2.2 Historical trends in rice production quantities in Bangladesh, 1972–2006 Source: Bangladesh Bureau of Statistics, 2008c. 8 Contribu�on to annual rice area growth 6 4 (percentage point) 2 0 -2 -4 Boro contribu�on -6 Aman contribu�on -8 Aus contribu�on Total growth rate -10 1972/73 2000/01 73/74 74/75 75/76 76/77 77/78 78/79 79/80 80/81 81/82 82/83 83/84 84/85 85/86 86/87 87/88 88/89 89/90 90/91 91/92 92/93 93/94 94/95 95/96 96/97 97/98 98/99 99/00 01/02 02/03 03/04 04/05 05/06 Figure 2.3 Historical trends in land area under rice cultivation in Bangladesh, 1972–2006 Source: Bangladesh Bureau of Statistics, 2008c. drops dramatically during major flood years, declines dominate the yield changes. In contrast, driving almost the entire decline in overall pro- yield improvements dominate the recovery years duction. after floods. Observed yields for rice and wheat Decomposing the historical aman rice pro- in Bangladesh from 1985 to 2000 have improved duction into land area and yield contributions marginally over time (Figure 2.5). shows that both contribute to the decline of Actual yields are much lower than the poten- aman production during major flood years (Fig- tial yields (5–10kg/ha) observed at research plots ure 2.4). However, in relative terms, the land area under controlled ï¬?eld conditions (Sattar, 2000). Vulnerability to Climate Risks 9 Contribu�on to annual Aman rice produc�on growth 35 30 Produc�on 25 Yield contribu�on 20 Area contribu�on (percentage point) 15 Total growth rate 10 aman 5 0 -5 -10 -15 -20 1972/73 2000/01 73/74 74/75 75/76 76/77 77/78 78/79 79/80 80/81 81/82 82/83 83/84 84/85 85/86 86/87 87/88 88/89 89/90 90/91 91/92 92/93 93/94 94/95 95/96 96/97 97/98 98/99 99/00 01/02 02/03 03/04 04/05 05/06 Figure 2.4 Decomposition of historical aman rice production trends into land area and yield contributions, 1972–2006 Source: Bangladesh Bureau of Statistics, 2008c. 3500 Total aus Total aman Total boro Wheat 3000 2500 Observed yields (kg/ha) 2000 1500 1000 500 0 1985/86 1987/88 1989/90 1991/92 1993/94 1995/96 1997/98 1999/00 Figure 2.5 Observed yields for major staples (kg/ha) Source: Bangladesh Bureau of Statistics, 2008c. A major factor for this can be attributed to the management play an important role including ability of existing varieties of rice to withstand sub-optimal time of planting, use of poor qual- the annual variations in climate conditions (unfa- ity seed, unbalanced use of fertilizers and other vourable temperatures, floods and droughts) as inputs, and failure to control weeds. In addition, well as pests and disease pressures which vary many farmers have not yet adopted modern rice from season to season. In addition, low levels of varieties. Soil-related factors include reduced 10 Climate Change Risks and Food Security in Bangladesh organic matter content and the widespread occur- Table 2.2 shows the classiï¬?cation of floods from rence of sulphur and zinc deï¬?ciencies. Mahmood Mirza (2002). About 26 per cent of the country et al (2003) noted a large yield gap in Bangla- is subject to annual flooding and an additional 42 desh, with actual average yields approximately per cent is at risk of floods with varied intensity one-sixth of the potential yields produced under (Ahmed and Mirza, 2000). high-input conditions that were protected from Historical records describe that ï¬?ve major floods. Closing this yield gap would lead to floods occurred in the 19th century (1842, higher average production and enhanced climate 1858, 1871, 1885 and 1892) and 16 such floods resilience. occurred in the 20th century (1900, 1902, 1907, 1918, 1922, 1954, 1955, 1956, 1962, 1968, 1970, 2.2 Living with Annual Floods1 1974, 1984, 1987, 1988, 1998) (Rashid and Paul, 1987; Khalil, 1990; Haque, 1997; Chowdhury, Bangladesh is one of the most flood-prone coun- 2000). Many of these serious floods can affect tries in the world. The literature on floods in the 35–75 per cent of the land area. The catastrophic country is extensive. Due to its location in the flood of 1998 was the worst on record and lasted low-lying deltaic floodplains at the convergence from the ï¬?rst week of July to the third week of of the Himalayan rivers, heavy monsoon rainfall September and was the most severe both in terms concomitant with poor drainage often results in of depth and duration. It inundated more than annual flooding. Exposure to storm surges in the 70 per cent of the total lands and caused severe coastal areas also exacerbates the severity of the damages to lives and properties. This flood alone floods.These river systems drain a catchment area caused 1100 deaths, flooded nearly 100,000km2, of about 1.7 million km2. The intensity of the affected 30 million people and impacted the floods is dependent on the magnitude and pat- property of about 1 million households. It also tern of precipitation in the three river sub-basins. damaged 16,000km and 6000km of roads and Among the peak discharge of the three rivers, the embankments, respectively, and affected 6000km2 Brahmaputra contributes the greatest volume, 58 of standing crop lands. A time-series of total area per cent, while the Ganges and Meghna contrib- affected by floods is shown in Figure 2.6. ute about 32 per cent and 10 per cent respectively. The relative severity of these disasters in These floodplains are home to a large population Bangladesh has decreased substantially since the (most of which is rural and poor) whose life is 1970s as a result of improved macro-economic intricately linked to the flooding regime. Annual management, increased resilience of the poor and regular flooding has traditionally been beneï¬?cial, progress in disaster management and flood pro- providing nutrient-laden sediments and recharg- tection infrastructure. Despite several major dis- ing groundwater aquifers; while low frequency asters, Bangladesh remains among the few coun- but high magnitude floods can have adverse tries that have avoided a single year of negative impacts on rural livelihoods and production. growth since the 1990s. Agricultural damage due to flooding has decreased with changes in crop- ping patterns, particularly the shift from deep- Table 2.2 Flood classiï¬?cations water aman rice (highly susceptible to floods) to Types of Flood Range of flooded Range of Probability boro rice, which is harvested before the monsoon area (km2) percent season starts. Table 2.3 summarizes statistics from inundation some recent large floods in the country. Normal 31,000 21 0.50 Moreover, adequate reserves of food grains Moderate 31,000–38,000 21–26 0.30 and increases in rice imports by both the public Severe 38,000–50,000 26–34 0.10 and private sectors have played a major role in Catastrophic 50,000–57,000 34–38.5 0.05 managing any potential food insecurity following Exceptional >57,000 >38.5 0.05 a flood event. This was evidenced following the Source: Mirza, 2002. 2004 and 2007 flood events which did not impact Vulnerability to Climate Risks 11 12000 10000 8000 Area (km2) 6000 4000 2000 0 1954 1956 1958 1960 1962 1964 1966 1968 1970 1972 1974 1976 1978 1980 1982 1984 1986 1988 1990 1992 1994 1996 1998 2000 2002 2004 Year Figure 2.6 Time-series of flood-affected areas (km2) in Bangladesh (1954–2004) Table 2.3 Comparison of losses resulting from recent large floods major problems. The south-central (SC), south- Item 1988 1998 2004 2007 east (SE), and river and estuarine (RE) regions in the coastal areas are mainly vulnerable to tidal Inundated area of Bangladesh (%) 60 68 38 42 flooding and salinity intrusion. The northwest People affected (million) 45 31 36 14 Total deaths (people) 2,300 1,100 750 1110 (NW) is impacted most from lean-season water Livestock killed (nos) 172,000 26,564 8,318 40,700 availability. Most regions are impacted by riverine Crops fully/partly damaged 2.12 1.7 1.3 2.1 flooding occurring during the monsoon period (million ha) (May–September). Rice production losses (million 1.65 2.06 1.00 1.2 tons) Roads damaged (km) 13,000 15,927 27,970 31,533 Observed historical trends with Number of homes fully/partly 7.2 0.98 4.00 1.1 precipitation damaged (million) Total losses: Using data from 32 rainfall stations (both Bang- Tk (billion) 83 118 134 78 ladesh Meteorological Department [BMD] and US$ (billion) 1.4 2.0 2.3 1.1 Bangladesh Water Development Board [BWDB] Source: World Bank (2007) stations) from 1960 to 2001, the national mean annual rainfall is 2447mm, with a maximum of 4050mm (in Sylhet, northeastern Bangladesh) and minimum of 1450 mm (in Rajshahi, north- overall rice availability despite flood losses of over western Bangladesh). The maximum rainfall a million tons of rice. Some segments of the pop- occurs during the June, July and August mon- ulation (e.g. rural landless and small and marginal soon months (JJA). Neither the annual nor sea- farmers), however, were adversely affected by sonal precipitation time-series show any statisti- changes in production and retail prices. Adequate cally signiï¬?cant changes over this time period access for these households depends on the level (Figure 2.7). of income, purchasing power and available social safety nets. Observed historical trends with discharge A summary of the extreme flood events on record Types of floods and the observed peaks and corresponding dates Bangladesh can be divided into eight primary are given in Table 2.4. In some cases, the dis- hydrological regions (see Plate 2.6).The northeast charges are almost twice the average, highlight- (NE) region is at the foot of the hill catchments ing the extreme inter-annual variability charac- in India. In this region flash floods are one of the terizing these river systems. The 1987 flood was 12 Climate Change Risks and Food Security in Bangladesh 3500 DJF MAM JJA SON ANN 3000 2500 2000 (mm) 1500 1000 500 0 1960 1965 1970 1975 1980 1985 1990 1995 2000 Figure 2.7 Annual and seasonal precipitation time-series (mm) averaged across Bangladesh Meteorological Department stations Note: DJF = December, January, February; MAM = March, April, May; JJA = June, July, August; SON = September, October, November. Table 2.4 Peak discharge and timing during extreme flood years Extreme Brahmaputra Ganges Meghna Return period Return period Years Date m /s 3 Date m /s 3 Date m /s 3 (area) (vol) 1974 7 Aug 91,100 3 Sep 50,700 – 21,100 7.04 6.61 1980 20 Aug 61,200 22 Aug 57,800 7 Aug 12,400 2.31 2.12 1984 20 Sep 76,800 17 Sep 56,500 17 Sep 15,400 1.85 4.20 1987 16 Aug 73,000 20 Sep 75,800 4 Aug 15,600 9.44 9.77 1988 31 Aug 98,300 4 Sep 71,800 18 Sep 21,000 79.34 33.54 1998 9 Sep 103,100 11 Sep 74,280 – 18,600 100.34 51.60 2004 12 Jul 83,900 19 Jul 77,430 – 16,300 9.86 20.14 Average 67,490 51,130 13,370 Min 40,900 31,500 7,940 Max 103,130 77,440 21,070 Source: BWDB. primarily from the Ganges. In 1988, all three riv- be estimated.The 1998 event is the 1 in 100-year ers had peaks within one week of each other.The event from the total area impacted perspective 1998 flood discharge in the Ganges and Brah- and the 1 in 50-year event from the discharge maputra rivers was even higher. This particularly perspective. devastating flood was a result of a simultaneous Hydrographs for a normal year (2002) and peak in both the Brahmaputra and the Ganges an extreme year (1998) are also plotted for these rivers (Mirza, 2003). In 2004, the Ganges and locations in Figure 2.8. The historical water level Brahmaputra peaked early. Moreover, assuming a data shows that the timing of the peak discharges Gumbel Type I distribution, return periods (both on the Ganges, Brahmaputra and Meghna riv- in terms of total area affected and total volume ers on average do not coincide.The Brahmaputra discharge of the Ganges and Brahmaputra) can starts rising in March due to snow melt in the Vulnerability to Climate Risks 13 21 1998 20 2002 19 Water level (mm) 18 17 16 15 11 13 12 Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec 15 1998 2002 13 Water level (mm) 11 9 7 5 Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec 8 1998 7 2002 6 Water level (m) 5 4 3 2 1 Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec Figure 2.8 Average discharges in 1998 and 2002 for (a) Brahmaputra, (b) Ganges and (c) Meghna rivers Source: BWDB. Himalayas while the Ganges starts rising in early Using available long-term records,2 trends June with the onset of the monsoon. Monsoon in peak discharges were statistically analysed. rainfall occurs in the Brahmaputra and Meghna Though records show small increasing trends in basins earlier than the Ganges basin due to the peak discharges, these are not statistically signiï¬?- pattern of progression of the monsoon air mass. cant except for the Ganges. Similarly, shifts in the The flood peaks of the Brahmaputra occur in timing of the peak are not statistically signiï¬?cant July and August, while peak flows occur in the except for the Meghna (over the time period of Ganges in August and September. record, the peak has shifted later by almost two months). 14 Climate Change Risks and Food Security in Bangladesh Flood determinants of agricultural Table 2.5 Typical crop calendar for four different rice varieties performance Crop Seedling Sowing/ Harvesting transplanting date date The performance of the agriculture sector is Start End Start End Start End heavily dependent on the annual floods. If floods unexpectedly arrive early this will affect the har- Aus 20 Mar 20 Apr 20 Apr 20 May 20 Jul 20 Aug vesting of the boro crop while a late recession T. aman 1 Mar 1 Apr 1 Apr 30 Apr 1 Jul 31 Jul 1 Jun 15 Jul 1 Jul 31 Aug 1 Nov 15 Dec delays the transplanting of the aman crop. An B. aman 20 May 30 Jun 15Jul 15 Aug 15 Nov 31 Dec indicator-based classiï¬?cation system for floods – – 15 Mar 15 Apr 1 Nov 15 Dec is used to characterize the primary flood deter- Boro – – 15 Mar 15 Apr 1 Nov 15 Dec minants for agriculture performance (Hassan et 20 Nov 20 Dec 1 Jan 31 Jan 1 May 31 May al, 2007). These include onset and recession of 20 Nov 20 Dec 1 Jan 31 Jan 1 May 31 May flood waters, the observed peak discharge and the – – 1 Jan 31 Jan 1 May 31 May duration above a deï¬?ned danger level. Table 2.5 – – 1 Jan 31 Jan 1 May 31 May represents a typical crop calendar for the major – – 15 Dec 15 Jan 15 Apr 15 May rice crops in Bangladesh.These planting practices are given graphically in Figure 2.9 for various tics of the land. Flood land types were categorized flood land types. more speciï¬?cally by the Master Plan Organization The rice variety grown by farmers in large (MPO, 1987) and are based on a three-day maxi- part depends on the normal flooding characteris- mum flood depth with a return probability of Figure 2.9 Cropping calendar corresponding to flood land type Vulnerability to Climate Risks 15 Table 2.6 Hydrological regions and flood land types Hydrological Region Percentage (%) area Highland Medium Highland Medium Lowland Lowland Very Lowland F0 F1 F2 F3 F4 (0–30cm) (30–90cm) (90–180cm) (180–300cm) (over 300cm) Eastern Hill 85.98 14.98 1.98 0.10 0.10 North Central 24.98 59.98 13.98 4.10 0.10 Northeast 23.98 19.98 10.98 43.10 5.10 Northwest 33.98 57.98 6.98 4.10 0.10 River and Estuary 11.98 69.98 14.98 7.10 0.10 South Central 0.98 73.98 27.98 0.10 0.10 Southeast 18.98 54.98 17.98 8.10 2.10 Southwest 35.98 47.98 18.98 0.10 0.10 Bangladesh Total 29.98 48.39 12.65 8.10 0.88 one in two years. The ï¬?ve main flood land types water availability. Early flooding and flash floods include: F0 (0–30cm), F1 (30–90cm), F2 (90– areas may disrupt the harvesting of the boro. 180cm), F3 (180–300cm) and F4 (over 300cm). F0 is typically classiï¬?ed as flood free. These flood land types represent the average expected depth 2.3 Lean Season Water Availability of inundation during a normal flood season.Table Bangladesh has a distinct dry season which 2.6 describes the percentage distribution of areas occurs from November to May. This is typi- for the eight hydrologic regions described earlier. cally most severe in the northwest portion of On F0 lands the main crop is t. aman during the the country. Agricultural droughts are associated monsoon season and wheat and HYV boro in with the late arrival or the early recession of the the rabi, or dry, season. Many of the same crops monsoon rains and with intermittent dry spells are also grown on F1 lands with the addition of coinciding with critical stages of the t. aman rice some local varieties of aus. On F2 lands the main season. Droughts in May and June also impact crop is b. aman during the monsoon season and broadcast aman and aus. Similarly, boro, wheat similarly, wheat and HYV boro in the rabi. Many and other crops grown during the dry season are of the same crops are grown on F3 lands with the also directly affected by the lack of water avail- exception of wheat. ability (both surface and groundwater). The pro- The aman crop is the main rice crop grown gressive development of groundwater for both during the monsoon season. A major factor rural water supply and agriculture during the last affecting the total production of aman during the several decades has meant that dry season water kharif, or rainy, is the overall magnitude of the availability is not the major threat that it used to floods. Figure 2.10 shows the aman production be. Indeed, dry season agriculture has been the losses (reported by the Bangladesh Bureau of Sta- main source of increased food production over tistics [BBS]) as a function of the combined dis- the past 20 years (Bangladesh Bureau of Statistics, charge in the Ganges and Brahmaputra. A statisti- 1998). However, declining groundwater tables in cally signiï¬?cant positive relationship is observed some places have begun to constrain production. whereby an increase in flood discharges correlates Moreover, if in the future less water is available with an exponential increase in production losses. in the river systems, groundwater tables may The severe floods on record are also shown. The decline even further, increasing the costs of pro- performance of the boro crop is more dependent duction and limiting overall performance in the on the availability of irrigation and lean-season sector. 16 Climate Change Risks and Food Security in Bangladesh 2,500,000 1988 2,000,000 Production loss (metric tons) 1,500,000 1987 1,000,000 1998 500,000 0 70,000 90,000 110,000 130,000 150,000 170,000 190,000 Maximum combined discharge on Ganges and Brahmaputra Figure 2.10 Aman crop production loss curve as a function of combined discharge Bangladesh experienced droughts in 1973, 1978, lion ha are vulnerable to annual drought; there is 1979, 1981, 1982, 1989, 1994 and 1995. The about a 10 per cent probability that 41–50 per droughts in 1973 were in part responsible for the cent of the country experiences drought in a famine in northwest Bangladesh in 1974. The given year. Areas of Bangladesh that are affected 1978–9 drought was one of the most severe, result- by drought during the different crop seasons are ing in widespread damage to crops (rice produc- given in Table 2.7. About 18 per cent of the rabi tion was reduced by about 2Mt), and it directly crops and 9 per cent of the kharif crops are highly affected about 42 per cent of the cultivated land. vulnerable to annual drought conditions. Rice production losses due to drought in 1982 were about 50 per cent more than losses due to Table 2.7 Summary of drought severity areas in Bangladesh by floods that same year. Losses in 1997 were about crop season (in Mha) 1Mt and valued at around US$500 million (Sel- varaju et al, 2006). Drought Class Rabi Pre-Kharif Kharif The Bangladesh Agricultural Research Very Severe 0.446 0.403 0.344 Council (BARC) has identiï¬?ed and mapped the Severe 1.716 1.156 0.746 drought-prone areas of Bangladesh for the main Moderate 2.956 4.766 3.176 cropping seasons in the country (based on esti- Slight 4.216 4.096 2.906 No Drought 3.176 2.096 0.686 mated yield impacts) (Plate 2.7). About 2.7 mil- Vulnerability to Climate Risks 17 2.4 Sea level Rise in Coastal Areas Estimating the changes in area that will be inundated due to sea level rise is complicated by Rising sea levels are one of the most criti- the active river morphology. With over a billion cal climate change issues for coastal areas. The tons of sediment being deposited in the alluvial Intergovernmental Panel on Climate Change fan of Bangladesh (Goodbred and Kuehl, 2000), (IPCC, 2007a) projected that an average rise of a combination of accretion and erosion processes 9 to 88cm could be expected by the end of the will work to both increase and decrease the land century. Recent projections suggest even more area available in the coastal areas. For instance, substantial rises (Copenhagen Diagnosis, 2009). satellite images from the coastal zone reveal that Increasing temperatures result in sea level rise by some land areas have gained while others have the thermal expansion of water and through the eroded over the last several decades (Plate 2.8). In addition of water to the oceans from the melting the Meghna estuary speciï¬?cally, about 86,000ha of continental ice sheets. A 1m sea level rise is of land were lost between 1973 and 2000 (CEGIS, estimated to impact 13 million people in Bangla- 2009). The relative contribution of these com- desh, with 6 per cent of national rice production peting processes is largely unknown and an area lost (Nicholls and Leatherman, 1995). Sea level for future research. rise may also influence the extent of the tides (currently the lower third of the country expe- Observed sea level rise trends riences tidal effects) and alter the salinity qual- ity of both surface and groundwater. Currently, Time-series data of daily mean water levels from because of the low topography in these coastal 13 stations in the coastal zone were statistically areas, about 50 per cent typically becomes inun- examined (the locations are shown in Figure dated during the annual monsoons. 2.11). Between 12 to 42 years of data are available 88°30' 89°45' 91° 92°15' 26°30' 26°30' 25°15' 25°15' 24° 24° Chadpur Nilkamal 22°45' 22°45' Companyganj Daulatkhan Daulatkhan 21°30' 21°30' WL Station Sundarban Chittagong Main river Dasmunia Galachipa Dohazari 88°30' 89°45' 91° 92°15' Khepupara Lemsikhali Hiron Point Cox’s Bazar Figure 2.11 Locations of coastal water level stations 18 Climate Change Risks and Food Security in Bangladesh Table 2.8 Estimated trends in water level of different stations along the coastline Station Location Duration No. of Trend Name of Station years (mm/yr) Hiron Point Passur 1977–2002 26 5.6* Khepupara Nilakhi 1959–86 22 2.9* Galachipa Lohalia 1968–88 21 3.3* Dasmunia Tentulia 1968–86 19 1.3* Kyoyaghat Tentulia 1990–2002 12 3.6* Daulatkhan Lower Meghna 1959–2003 31 4.3* Nilkamal Lower Meghna 1968–2003 33 2.3* Chadpur Lower Meghna 1947–2002 50 0.0* Companyganj Little Feni Dakatia 1968–2002 32 3.9* Chittagong Karnafuli 1968–88 16 3.1* Dohazari Sangu 1969–2003 32 2.0* Lemsikhali Kutubdia Channel 1969–2003 27 2.1* Cox’s Bazar Bogkhali 1968–91 22 1.4* *Statistically signiï¬?cant to p<0.05. Source: BWDB, CEGIS (2006) for these stations. The observed trends for these ges channels pushes the 5ppt saline front towards stations are reported in Table 2.8. These esti- the estuary mouth. mates range from a high of 5.6mm/yr at Hiron In contrast, during the dry season (Decem- Point station to no change at the Chadpur sta- ber to March) saltwater intrusion occurs through tion on the Meghna River. At the southeast cor- various inlets in the western part of the coastal ner of Bangladesh (Cox’s Bazar station) sea level zone and through the Meghna estuary. The max- increased at a rate of 1.4mm/yr. In the middle of imum salinity variation during the dry season is the south coastal zone (Companyganj station) sea shown in Plate 2.10. The 5ppt isohaline intrudes level increased at a rate of 3.9mm/year. Though more than 90km landward at the western part all of the linear trends are positive, only the trend of the coastal area in the Sundarbans. Moreover, at Hiron Point is statistically signiï¬?cant with 95 with decreases in freshwater flow in the Lower per cent conï¬?dence. Meghna the saline front can move by as much as 30–40km from the coast.Table 2.9 shows the total area affected by low, moderate and high salinity Observed salinity changes level for a base condition in 2005 during both the Another important factor affecting agricultural monsoon and dry seasons. During the monsoon, productivity is the surface and groundwater about 12 per cent of the total area is under high salinity distribution. In particular, saline water salinity levels which increases to 29 per cent dur- intrusion along inland rivers is highly seasonal. ing the dry season. With increased sea level rise, Using a coastal model (IWM and CEGIS, 2007), drainage gradients may reduce, thereby decreas- it was determined that during the monsoon ing the flow to the Bay of Bengal and allowing period (June to September) the Meghna estuary riverine salinity to move further inland. is hardly saline. The maximum salinity variation Finally, high salinity groundwater is known during the monsoon season in the coastal zone to threaten drinking water wells in the coastal is presented in Plate 2.9. The 5 parts per thou- zone, particularly at shallow depths, and limit the sand (ppt) isohaline (line of equal salinity level) possibility for groundwater irrigation for crop intrudes more than 70km landward in the west- production. However, recent trends in the pro- ern part of Sundarbans, whereas comparatively motion of aquaculture (shrimp production, for higher freshwater flow through the primary Gan- example) have been one local adaptation meas- Vulnerability to Climate Risks 19 Table 2.9 Area affected by low, moderate and high salinity level (in 2005) Season Total area (km2) Area affected (km2) Percentage of area affected (%) Dry Season 0–1 ppt Low 25,625 54 (Dec–Mar) 1–5 ppt Moderate 7808 17 >5 ppt High 13,712 29 Monsoon Season (Jun–Sep) 0–1 ppt Low 37,455 79 1–5 ppt Moderate 4063 9 >5 ppt High 5707 12 ure for coping with saline surface and groundwa- municipal return flows) may place further pres- ter. Deeper in the aquifer, at depths greater than sures on the gap between supply and demand for 150m, groundwater is typically fresh, thus much Bangladesh. of the groundwater used for drinking water sup- Balancing these future demands against ply is drawn from these depths. future supplies may increasingly become difï¬?cult. Of particular importance to the region are the 2.5 Regional Hydrology Issues greater Himalayas where ten of the largest rivers in Asia begin (Amu Darya, Brahmaputra, Gan- Due to its location at the confluence of the Gan- ges, Indus, Irawaddy, Mekong, Salween, Tarim, ges, Brahmaputra and Meghna rivers, a discussion Yangtze and Yellow rivers). These river basins are of water resources in Bangladesh is not complete inhabited by over 1.3 billion people. Thus, this without consideration of the broader basin region water ‘tower’ is critical to the overall economy of of which it is a part (Figure 2.12). This invariably the region.With rising temperatures, it is reported requires an examination of its relationship with that a rapid reduction in glaciers is being observed its neighbours: India, Nepal, China, Bhutan and (Dyurgerov and Meier, 2005). With this glacier Myanmar. The hydro-climatic, demographic and retreat, the rivers that originate at the glacier socio-economic features that characterize the termini and derive signiï¬?cant volume of annual patterns of water utilization in these shared river flow from glacier melt may experience profound basins in the region have important implica- downstream impacts on water resources. How- tions for the overall quantity and quality of water ever, much more research is needed to determine resources available as well as for the relationships the exact balance among glacier melt, snow and among and within riparian states. precipitation contributions to the available dis- The annual rate of population growth in charge in each of these rivers. Current estimates each of these riparian nations is similar (about of glacier melt contribution to runoff are around 1 per cent). Population increases throughout the 9 and 12 per cent for the Ganges and Brahmapu- basins, coupled with increased demand for agri- tra respectively (Jianchu et al, 2007). cultural production, municipal and industrial A major limitation currently is that these requirements, environmental flows and energy mountain systems are poorly understood. The production, will contribute to increasing water Himalayas are characterized by a complex three- demands. Upstream changes in water demand in dimensional mosaic of meteorological and hydro- the Ganges alone, which currently has a popula- logical environments, ranging from tropical rain- tion of 500 million people and contains 82 large forests to alpine deserts, covering an altitudinal cities with populations of 100,000 people or range of more than 8000m. Essential climate and more, will play a signiï¬?cant role in the provision hydrologic data is not readily available. The lack and timing of water downstream in Bangladesh. of a basic understanding of runoff sources and Moreover, changes in the quality of the water timing in these rivers makes it difï¬?cult to resolve that is returned to the system (e.g. irrigation and questions relating to the overall water budget. 20 Climate Change Risks and Food Security in Bangladesh Figure 2.12 Ganges-Brahmaputra-Meghna river basin Source: World Bank A ï¬?nal key factor influencing water availability servation of water will also play an increasingly and utilization patterns in the basin are chang- major role in changing practices in these basins, ing water management practices, including the particularly with the adoption of micro-irrigation further development of new sources of supply techniques, promotion of artiï¬?cial groundwater (whether surface or ground) and infrastructure. recharge programmes and the planned reuse of Demands from the urban, agriculture, industrial water. All of these changing practices may impact and energy sectors will drive the development of the quantity and quality of water available to new diversions, inter-basin transfers, storage facil- Bangladesh. ities and other infrastructure, as well as further development of groundwater resources where available. This infrastructure may also be devel- Notes oped to strengthen the ability of individual ripar- 1 For an excellent historical discussion of floods ian countries to manage floods and droughts, and in Bangladesh please refer to Hofer and Mes- to better allocate water to higher value uses (e.g. serli (2007). the transfer of water from agriculture to urban 2 Bahadurabad, 1956–2004; Hardinge Bridge, areas). In addition to supply management, con- 1934–2004; Bhairab Bazar, 1979–93. 3 Future Climate Scenarios Box 3.1 Key messages • Projected temperature changes follow a positive trend for all months and seasons from the 2030s onwards. Median warming of 1.1°C, 1.6°C, and 2.6°C by the 2030s, 2050s and 2080s respectively is simulated from a range of plausible scenarios. • Few months or seasons display clear drying or wetting trends in simulations of the 2030s. By the 2050s, annual and wet season precipitation is projected to trend towards increased precipitation. Only simulations for the dry season do not suggest an increase in precipitation. Across the model experiments, precipitation variations are large. Median annual precipitation increases of 1 per cent, 4 per cent and 7.4 per cent by the 2030s, 2050s and 2080s respectively are projected. • Greater uncertainty (in terms of magnitude and direction) exists with future precipitation than future temperature. • A trend toward a warmer and wetter future climate is projected to impact the agriculture sector, par- ticularly if the climate state goes beyond the variations found in the historical record. Projected future temperature changes signiï¬?cantly separate from the background temperature variations. Precipita- tion is subject to large existing inter-annual and intra-annual variations. Future precipitation projec- tions vary widely amongst models, with small median changes compared to historic variability. • Using three scenarios of sea level rise (15cm, 27cm and 62cm), total flooded area in the coastal areas is projected to increase 6 per cent, 10 per cent and 20 per cent respectively. Several climate change scenarios for Bangladesh Tanner et al (2007) followed a similar approach, have been published in recent years. Agrawala et identifying 10 out of 18 models from the IPCC al (2003) examined the performance of 17 glo- Third Assessment Report (TAR) (IPCC, 2001) bal climate models (GCMs)1 over Bangladesh to project into the future using the IPCC A2 during the 20th century, then used several top- and B1 scenarios. Assessments were made for the performing models to simulate future climate 2020s and 2050s for Bangladesh and cover the using the Intergovernmental Panel on Climate entire GBM basin. Temperatures in 2050 were Change (IPCC) B2 emissions scenario (SRES, projected to increase by an average of 1.6ºC and 2000).2 The resulting mean annual temperature 2.0ºC in the B1 and A2 scenarios respectively, changes were 1.4ºC by 2050 and 2.4ºC by 2100, with corresponding increases in rainfall of 4 per with higher increases in the wintertime. Annual cent and 2 per cent. An additional analysis was precipitation rose by 5.6 per cent by 2050 and conducted to generate probability distribution 9.7 per cent by 2100, with indications of win- functions of climate change in Bangladesh using ter drying and summer rainfall increases. These the climate sensitivities of 23 IPCC TAR models, projected seasonal changes are consistently found assuming that each model was an equally likely across many studies of the South Asian monsoon representation of each future scenario. Again, pro- region. jected warming was more prevalent in winter 22 Climate Change Risks and Food Security in Bangladesh (December–February) than in summer (July– the models used, their known climate sensitivities August) and the seasonal monsoon was intensi- and the resolution of their atmospheric output. ï¬?ed.Work has also been conducted using regional The model output consists of monthly averages climate models (RCMs) to more accurately rep- of simulated precipitation and temperature, with resent local processes and regional variations in this study using a maximum total of 64 scenario Bangladesh. RCM experiments are driven by experiments (16 models x 3 emissions scenarios GCMs under particular emissions scenarios and + 16 baseline states). Resolution varies among therefore reflect, at least in part, GCM biases in these models with about ï¬?ve grid boxes typically simulating current and future climates. Islam et covering the country. A weighted average was al (2005) used the PRECIS regional model to used to calculate national values for the country, compare regional anomalies of temperature and with weights determined by the percentage that precipitation. The use of regional models in cli- each grid box overlaps with the country. Varia- mate projections for Bangladesh remains a key tions in space tend to be small relative to varia- area for future research. tions across models and time. The following time slices are deï¬?ned as follows and referenced by 3.1 Future Estimated Precipitation their central decade: and Temperature • GCM Baseline: 1970 to 1999, Sixteen GCMs were analyzed from the Program • GCM 2030s scenario: 2020 to 2049, for Climate Model Diagnosis and Inter-compar- • GCM 2050s scenario: 2040 to 2069, ison (PCMDI, www-pcmdi.llnl.gov) Coupled • GCM 2080s scenario: 2070 to 2099. Model Intercomparison Project phase 3 (CMIP3) multi-model dataset (Meehl et al, 2007), each run Future changes in temperature and precipitation for three emissions scenarios (A1B,A2 and B1) and for the decades of the 2030s, the 2050s and the a 20th-Century Experiment. Table 3.1 describes 2080s are calculated for each model relative to Table 3.1 IPCC AR4 global circulation models GCM Name Institution Atmospheric Climate Resolution Sensitivity (lat, lon, °) (°C)* bccr_bcm2.0 Bjerknes Centre for Climate Research, Norway 2.8 x 2.8 – cccma_cgcm3.1(T63) Canadian Centre for Climate Modelling and Analysis, Canada 3.75 x 3.75 3.4 cnrm_cm3 CERFACS, Centre National Weather Research, METEO-FRANCE, France 2.8 x 2.8 – csiro_mk3.0 CSIRO Atmospheric Research, Australia 1.88 x 1.88 3.1 gfdl_cm2.0 Geophysical Fluid Dynamics Laboratory, USA 2 x 2.5 2.9 gfdl_cm2.1 Geophysical Fluid Dynamics Laboratory, USA 2 x 2.5 3.4 giss_model_er NASA Goddard Institute for Space Studies, USA 4x5 2.7 inmcm3.0 Institute for Numerical Mathematics, Russia 4x5 2.1 ipsl_cm4 Institut Pierre Simon Laplace, France 2.5 x 3.75 4.4 miroc3.2 (medres) Center for Climate System Research; National Institute for Environmental 2.8 x 2.8 4.0 Studies; Frontier Research Center for Global Change, Japan miub_echo_g Meteorological Institute of the University of Bonn, Germany 3.75 x 3.75 3.2 mri_cgcm2.3.2a Meteorological Research Institute, Japan 2.8 x 2.8 3.2 mpi_echam5 Max Planck Institute for Meteorology, Germany 1.878 x 1.88 3.4 ncar_pcm1 National Center for Atmospheric Research, USA 2.8 x 2.8 2.1 ncar_ccsm3.0 University Corporation for Atmospheric Research, USA 1.4 x 1.4 2.7 ukmo_hadcm3 Hadley Centre for Climate Prediction, Met Ofï¬?ce, UK 2.5 x 3.75 3.3 * Climate sensitivity parameter deï¬?ned as temperature increase for a doubling of atmospheric carbon dioxide. Future Climate Scenarios 23 the same model’s 1970–99 baseline period.These cipitation enhancements of 1 per cent, 4 per cent 30-year periods are used to reduce the effects and 7.4 per cent by the 2030s, 2050s and 2080s of large year-to-year variation that can obscure respectively fall into the range of previously pub- mean climate changes. lished literature for Bangladesh. Figure 3.1 shows the projected monthly, annual and seasonal temperature changes. Tem- Comparison to historical variability perature changes are clearly following a positive trend for all months and seasons from as early as A trend towards a warmer and wetter future the 2030s, but do not show any obvious seasonal climate will impact the agriculture sector in structure. Enhanced warming during the dry Bangladesh, particularly if the climate state goes winter months is evident by the 2050s, although beyond the precedent variations found in the the model simulations are distributed more historical record. Warming is projected to gen- widely in these months. Temperature changes are erally accelerate over the 21st century, although positive for every model experiment and every the model-based probability distribution widens. month by the 2080s, with a clear seasonal vari- By the 2030s, the median temperatures in July, ation in magnitude. Median warming of 1.1ºC, August and September of the future model dis- 1.6ºC and 2.6ºC by the 2030s, 2050s and 2080s tribution surpasses the 90th percentile of the his- respectively fall into the range of previously pub- torical temperature variability. Moreover, looking lished literature for Bangladesh. at the monsoon and dry seasons, by the 2080s The monthly, annual and seasonal precipita- the 10th percentile of the future model distribu- tion change projections for the 2030s, 2050s and tion surpasses the 90th percentile of the historical 2080s compared to 1970–99 is shown in Figure variability. That is, the estimated future tempera- 3.2. Deviations for each time period are displayed ture signiï¬?cantly separates from the background as the percentage change from their baseline aver- variations. age. Despite only small changes in actual mag- The future changes in precipitation are com- nitude, rainfall deviations in the dry months of pared to the inter-annual and inter-seasonal the year appear as very large percentage changes variability determined for the 40-year period due to the low baseline average. These dry season discussed earlier. Precipitation is subject to large totals would not have any noticeable impact on existing variation in the historical record (Table the annual rainfall totals, but could still have sig- 3.2). Differences in the monsoon structure and the niï¬?cant ramiï¬?cations for the severity of droughts. influences of large-scale circulation patterns like Conversely, simulated rainfall deviations in the the Madden-Julian Oscillation (MJO) and the El wet season have to be very large to produce high Niño-Southern Oscillation (ENSO) contribute percentage changes. to this background variability. Despite the con- Few months or seasons display clear drying sistently noted enhancement of the monsoonal or wetting trends in simulations of the 2030s. By circulation pattern that leads to a drying trend the 2050s, however, annual and wet season pre- during the winter months and increased rainfall cipitation trend towards increased precipitation during the monsoons in the climate scenarios in the set of climate change scenarios (though of the 2030s, 2050s and the 2080s, precipitation some models do continue to show decreases in does not separate itself from the historical varia- precipitation). Only simulations for the post- bility for any month or season.These ï¬?ndings are monsoonal rabi dry season (when boro and wheat consistent with the general ï¬?nding that greater are grown) do not suggest a rise in precipitation. uncertainty exists with the estimated magnitude An enhancement of the monsoonal circulation of precipitation change than temperature change (IPCC, 2007a) widens the discrepancy between and that existing rainfall variability is substantial. wet and dry seasons in the 2080s, and by then the annual and monsoon season precipitation changes are clustered around positive trends. Median pre- 24 Climate Change Risks and Food Security in Bangladesh Figure 3.1 Monthly, annual and seasonal temperature changes Note: The box and whiskers diagram consists of a line representing the median value, a box enclosing the inter-quartile range, dashed whiskers extending to the furthest model that lies within 1.5 times the inter-quartile range from the edges of the box, and plus symbols for additional models that are perceived as outliers. Future Climate Scenarios 25 Figure 3.2 Monthly, annual and seasonal precipitation changes Note: The box and whiskers diagram consists of a line representing the median value, a box enclosing the inter-quartile range, dashed whiskers extending the to the furthest model that lies within 1.5 times the inter-quartile range from the edges of the box, and plus symbols for additional models that are perceived as outliers. 26 Climate Change Risks and Food Security in Bangladesh Table 3.2 Summary precipitation statistics averaged across Bangladesh (1960–2001) DJF MAM JJA SON Annual Average (mm) 53 460 1406 530 2447 Standard deviation (mm) 36 149 216 125 306 Coefï¬?cient of variation 68% 32% 15% 24% 12% 90% percentile 118 672 1684 669 2910 75% percentile 72 572 1584 608 2593 10% percentile 21 312 1149 361 2199 Source: Bangladesh Meteorological Department. Table 3.3 Sea level rise impacts on flood land types F0 F1 F2 F3 F4 Flooded Area % of total (0–30cm) (30–60cm) (60–90cm) (90–180cm) (+180cm) (F1+F2+F3+F4) Base 15,920 4753 4517 5899 1759 16,928 52 15cm 14,841 4522 4705 6765 2015 18,007 55 27cm 14,189 4345 4488 7456 2370 18,659 57 62cm 12,492 3967 3818 8977 3594 20,356 62 Note: Note that the flood land type classes used in IWM and CEGIS (2007) are slightly different than the MPO deï¬?nitions. 3.2 Future Sea level Rise cale AR4 GCM simulations to the Bangladesh region using the IPCC four-factor method, This section draws upon a previous coastal which includes: 1) sea level rise components for zone modelling effort (using the MIKE21 two- global thermal expansion; 2) local land processes dimensional estuary model, IWM and CEGIS, including accretion, erosion and subsidence; 3) 2007) for 15cm, 27cm and 62cm sea level rise melt-water from glaciers, ice caps and ice sheets; scenarios. Using these scenarios, the changes in and 4) coastal circulation patterns as affected by total flood land type area is given in Table 3.3. currents, tides and weather. This approach has Of a total 33,000km2 in these coastal areas, over been used by, for example, the University of half is annually flooded. With an extreme rise of Washington Climate Impacts Group (2008) and 62cm, an increase in 10 per cent of flooded area is the New York City Panel on Climate Change anticipated. The geographic distribution of this is (2009). Other approaches include the empirical shown in Plate 3.1. Under the 62cm rise scenario, Rahmstorf (2007) method as applied to AR4 cli- districts where it is projected that the flooded area mate models (Horton et al, 2008) and the rapid will increase by more than 10 per cent include: ice-melt scenarios (New York City Panel on Cli- Bagerhat (22 per cent), Barisal (23 per cent), Bhola mate Change, 2009). These three approaches are (14 per cent), Cox’s Bazar (10 per cent), Khagrach- compared in New York City Panel on Climate hari (13 per cent) and Noakhali (12 per cent). In Change (2009). general, with sea level rise, the total flooded area increases by 6, 10 and 20 per cent for each of the scenarios respectively, with the largest increases Notes observed in the southernmost regions. The largest 1 Global climate models feature interactions percentage increases in area are observed for the F4 between the atmosphere and oceans and (+180cm) flood land class. account for forcings from the sun, natural as While these scenarios could be appropriate, well as anthropogenic sources of greenhouse more detailed local-scale sea level rise estimates gas and aerosols emissions, and internal vari- can be developed. One approach is to downs- ability of the climate system. Future Climate Scenarios 27 2 The A1B, A2 and B1 emissions scenarios used by a gradual decrease in emissions after 2050. in this study each project different develop- A2 – relatively rapid population growth and mental paths for global society by forcing the limited technological change combine to GCMs with greenhouse gas emissions deter- produce the highest greenhouse gas levels by mined by particular developmental storylines. the end of the 21st century, with emissions Each represents a unique blend of demo- growing throughout the entire century. B1 graphic, social, economic, technological and – this scenario features what is considered a environmental assumptions. The three sce- low population projection, combining low narios are briefly described as follows: A1B fertility and mortality. Under this scenario, – rapid economic growth is partially offset by global population peaks at 8.6 billion mid- rapid introduction of new and efï¬?cient tech- century and then declines to 7.1 billion by nologies and decreases in global population 2100. When combined with societal changes after 2050. This trajectory is associated with tending to reduce greenhouse gas emissions, relatively rapid increases in greenhouse gas the net result is relatively low greenhouse gas emissions, and the highest overall CO2 levels concentrations with emissions beginning to for the ï¬?rst half of the 21st century, followed decrease by 2040. 4 Future Flood Hydrology Box 4.1 Key messages • Primarily driven by increased monsoon precipitation in the GBM basin, models on average demon- strate increased flows in the three major rivers into Bangladesh (by as much as 20 per cent). Larger changes are anticipated by the 2050s compared to the 2030s. Larger changes are observed on average for the Ganges. The exact magnitude is dependent on the month. • On average, models demonstrate that the flooded area increases in the future (over 10 per cent by the 2050s). This is primarily in the central part of the country at the confluence of the Ganges and Brahmaputra rivers and in the south. Flood area estimates separate from the background variations primarily in August and September at the height of the monsoon. • Increases in yearly peak water levels are estimated for the northern sub-regions and decreases are estimated for the southern sub-regions. Not all estimated changes are statistically signiï¬?cant. More model experiments demonstrate changes that are signiï¬?cant by the 2050s than by the 2030s. Changes in the peak are in general less than 0.5m from the baseline. • Across the sub-regions, most GCMs show earlier onset of the monsoon and a delay in the recession of flood waters. Given the importance of flooding to overall agri- and Nepal. For the topography of the basin, culture production, special effort was made to data from the Shuttle Radar Topography Mis- model the future flood hydrology under various sion (SRTM) of the National Aeronautics and climate change scenarios for the flood monsoon Space Administration (NASA) was used (Farr et months (i.e. May–September). To do this, three al, 2007). GTOPO30, a global digital elevation sequential steps are taken: (1) a sub-set of cli- model (DEM) from the United States Geological mate change scenarios are selected; (2) flows into Survey (USGS), was used to ï¬?ll in the gaps. The Bangladesh are generated using a Ganges-Brah- horizontal grid spacing for GTOPO30 is 30-arc maputra-Meghna (GBM) river basin model; and seconds (approximately 1km) and for SRTM it is (3) hydrologic changes within the country are approximately 90m. This data was used to delin- generated using a national river network model. eate sub-catchments. In total, the GBM basin model comprises 95 sub-catchments: 33 in the Brahmaputra basin, 55 in the Ganges basin and 7 4.1 GBM Basin Model Development in the Meghna basin. The MIKE BASIN1 model was used for the Rainfall and evaporation data serve as bound- GBM basin. Primary input data include topog- ary conditions for rain-fed sub-catchments. Irri- raphy, meteorology and hydrology information. gation withdrawals and river management con- River alignments for the GBM basin were deter- trols in upstream catchments were not considered mined using available physical maps for India in calibrating the model. Monthly temperature Future Flood Hydrology 29 data was available at several stations within the tion purposes (Hardinge Bridge on the Ganges, basin. Temperature data was also incorporated as Bahadurabad on the Brahmaputra, and Amalshid an additional boundary condition for snow-fed on the Meghna). catchments. Actual daily rainfall data within the GBM basin was limited to a few stations. How- Calibration and validation ever, this information was supplemented with satellite rainfall data (0.25° x 0.25° horizontal Only the calibration results for the Ganges and resolution) measured by the Tropical Rainfall Brahmaputra rivers are shown here as these Measurement Mission (TRMM) and expanded are the primary drivers of flow in the country. to a 30-year record covering 1978–2008 using Simulated discharges at Hardinge Bridge on the a bootstrapping weather generator. Measured Ganges and Bahadurabad on the Brahmaputra evaporation data in the basin was also limited to were calibrated against observed runoff for the a few stations. Plate 4.1 shows the map of the time period 2004–2007. Since these models are modelled GBM basin including the locations of used primarily for analysis during the monsoon meteorological data inputs. season, the model is calibrated to minimize the Average monthly or yearly discharge data is sum of the square errors during this period only. available at several stations along the Ganges and There are 11 primary parameters used in calibra- Brahmaputra rivers. Historical data at three loca- tion (built into the rainfall-runoff model) which tions near the border between India and Bang- are sensitive to the different sub-catchment char- ladesh in the GBM basin are used for calibra- acteristics and storage zones.2 The GBM basin Root mean square error = 5287m3/s Root mean square error = 4262m3/s Figure 4.1 Validated discharges from 1998–2007 at (a) Bahadurabad (b) Hardinge Bridge 30 Climate Change Risks and Food Security in Bangladesh model was then validated against TRMM data output is required while the GCM output is typ- from 1998–2007 (shown in Figure 4.1). The ically monthly; (3) the GCM spatial resolution model reasonably replicates the observed peaks. reduces extreme events and misses sub-grid scale geographic variability; and (4) year-to-year vari- ability at a particular location in the GCM out- 4.2 National Hydrologic Super put tends to be underestimated due to simpliï¬?ed Model greenhouse forcing scenarios and coarse spatial resolution. Moreover, GCM contributions to the Inflows at the validated locations from the GBM IPCC are designed to capture climatic changes basin model are used as boundary inflows into the averaged over a long period of time; not to simu- national hydrologic model. This detailed model late particular events in the future. Therefore, the is the primary tool used by the government of relevant information that may be taken from a Bangladesh to make annual flood forecasts and GCM to generate climate scenarios is actually issue warnings. See Hopson and Webster (2007) drawn from a comparison between a future pro- for an application of this model using satellite jection and a baseline period. imagery to improve the forecast from 3 days to A monthly rainfall and temperature series, 10 days. This model is combined with gridded averaged over each 30-year time period, is deter- precipitation and temperature data and predicts mined for the baseline and 2030s and 2050s sce- water levels and discharges throughout the coun- narios for a particular GCM. By comparing the try. This model, which uses the MIKE 11 plat- baseline and future monthly averages, a ‘delta’ form,3 predicts daily water levels and discharges value for both rainfall and temperature can be throughout the country covering most of the calculated (i.e. percentage change is used for pre- major river networks (except for some parts in cipitation and absolute change in degrees Celsius the coastal areas and eastern hills – see Plate 4.2a). is used for temperature). Because the historical A separate coastal model was used for the south- data captures differences between sub-regions eastern part of the country. Plate 4.2a also shows and day-to-day and year-to-year variability that the network of water levels and discharge points is not presented accurately in the GCM histori- where time-series estimates are produced. With cal output, this monthly change is then applied these daily water levels (at 3800 points), the tem- directly to the actual historical observed 30- poral characteristics of the floods can be analysed. year precipitation and temperature data.4 This Moreover, monthly flood maps can be prepared approach removes much of the bias associated using a three-dimensional Geographic Informa- with each of the climate models, assuming that tion Systems (GIS) tool to interpolate the flood the bias is common to the historical and future surface while taking into account the presence of periods. The sign and magnitude of remaining flood protection works (e.g. roads, embankments, biases are unknown. polders). The area under different flood land type classes (as described in section 2.2) can then be calculated and compared to the baseline. Selecting a sub-set of global climate models 4.3 Approach to Modelling Future Not all models could be tested in the flood sim- ulations because of resource limitations. Thus, Flood Changes a sub-set of GCMs was selected for the flood Rather than use the GCM future scenarios hydrology modelling. In using only a sub-set of directly in these models, a ‘delta’ approach was models, care must be taken in interpreting the taken to overcome four signiï¬?cant obstacles to results. The reduction in the number of climate climate scenarios generation: (1) GCM outputs models results in a loss of information about may contain signiï¬?cant biases both in the base- the characteristics of future climate conditions line period and in the future period; (2) daily and the statistical signiï¬?cance of those ï¬?ndings. Future Flood Hydrology 31 A major goal, therefore, is to select the climate Delineating agro-climatic sub-regions models that best represent the larger distribution of results and model characteristics in Bangladesh. To capture regional variations in both flooding The use of a statistical representation of additional characteristics and overall agriculture perform- models was not used, as it would not necessarily ance, Bangladesh is divided into 16 sub-regions represent a physically consistent realization of cli- (see Plate 4.2b). These sub-regions will be used mate or its potential change. Additionally, com- throughout the study.The criteria used to deline- binations of climate models with extreme values ate these boundaries include: flooding character- in different parameters are especially unlikely to istics (e.g. riverine, tidal, flash), watershed catch- provide sound realizations of future climate, and ments (e.g. Ganges dependent), planning units in fact may lead to the creation of even more (e.g. administrative, crop, flood land type), agro- extreme climate scenarios than their individual climatic (e.g. floodplain, drought areas, coastal components. To achieve a robust climate signal zone, hilly region) and the presence of climate it was critical that multiple years and multiple station data. The sub-regions are deï¬?ned in Table emissions scenarios be simulated. 4.1. Several criteria were used to narrow the sub- set of GCMs. First, the selected GCM must per- 4.4 Future Changes over the form well in the GBM region and adequately capture the dynamics of the monsoon. Second, Ganges-Brahmaputra-Meghna the sub-set of models must capture the range Basin of climate sensitivity found in the IPCC mod- Across the models and scenarios, a clear consen- els. Third, the resolution of the GCM must be sus on a trend towards warming (between 1 to adequate for this hydrologic application. Fourth, 3°C) in the larger GBM basin is observed (Figure the subset of models should capture the range 4.2).This is consistent with the estimated national of IPCC changes. Fifth, the selected GCMs must changes. Greater warming during the dry winter have a substantial basis in the literature. Based on months is estimated. Moreover, the incremental this, it was decided that two future time periods increase in temperatures between the 2030s and (each for a 30-year record), two emission sce- the 2050s is less than 1°C. Temperature increases narios (A2 and B1 – A1B are omitted because are greater for the A2 than B1 scenarios across for the time period of analysis, there is not much all models. difference between A1B and A2), and the follow- For precipitation in the GBM basin, esti- ing ï¬?ve GCMs would be most suitable for this mated changes differ widely across models. Plates analysis (i.e. ten model experiments for each time 4.3 and 4.4 show the estimated monthly per- period): centage change in precipitation over the Ganges and Brahmaputra sub-basins respectively for the • University Corporation for Atmospheric 2050s for all ten model experiments (5 GCMs x Research – CCSM 2 IPCC Special Report on Emissions Scenarios • Max Planck Institute for Meteorology – [SRES]). Note that the large percentage changes ECHAM5 estimated during the non-monsoon months pri- • Hadley Centre for Climate Prediction – marily reflect little baseline rainfall during this UKMO period. Most GCMs estimate increases in rainfall • Center for Climate System Research – during the monsoon season (both in the 2030s MIROC and the 2050s) – up to 20 per cent more from • Geophysical Fluid Dynamics Laboratory July to September. Large changes at the onset of – GFDL the monsoon (during May and June) particularly in the Ganges may reflect an earlier arrival of the monsoon season. During the dry season, some 32 Climate Change Risks and Food Security in Bangladesh Table 4.1 The sub-regions with hydrological region, agro-ecological zone and districts Sub-region Area (km2) Hydrologic region Agro-ecological zone* Districts SR-01 13,157 NW 10, 11a, 11b, 11c, 1a, 1b, 1c, 25a, 26, 27a, Dinajpur, Joypurhat, Naogaon, Natore, 27b, 3a, 3b, 3e, 3f, 3g, 5, 6, River Nawabganj, Panchagarh, Rajshahi, Thakurgaon SR-02 17,301 NW 11a, 11b, 12b, 2, 25a, 25b, 27a, 27b, 27c, Bogra, Dinajpur, Gaibandha, Joypurhat, 3a, 3b, 3c, 3d, 3e, 3f, 3g, 4a, 4b, 4c, 5, 7, Kurigram, Lalmonirhat, Naogaon, Natore, 8a, River Nilphamari, Rangpur, Sirajganj SR-03 3336 NW 10, 11a, 12a, 12b, 4a, 4b, 4c, 5, 7 Natore, Pabna, Sirajganj SR-04 7794 NC 10, 12b, 15, 19f, 28a–e, 4b, 7, 8a, 8c, 8d, 9a, Dhaka, Jamalpur, Manikganj, Munshiganj, 9b, 9c, 9d, 9e, River Sirajganj, Tangail SR-05 2302 NC 19f, 19g, 28a–e, 28f, 8d, 9b, 9e, River Dhaka, Gazipur, Narayanganj SR-06 12,424 NC, NE 16, 19f, 19g, 21a, 22a, 22b, 22d, 28a-e, 29b, Gazipur, Jamalpur, Kishoreganj, Kurigram, 29c, 7, 8a, 8b, 8d, 9a, 9b, 9d, 9e, River Munshiganj, Mymensingh, Narayanganj, Narsingdi, Netrakona, Sherpur SR-07 12,158 NE 16, 19b, 19c, 19h, 19i, 20, 21a, 21b, 21c, Brahamanbaria, Habiganj, Kishoreganj, 22a, 22b, 22c, 22d, 29a, 29b, 29c, 8b, 9b, Maulvibazar, Netrakona, Sunamganj, Sylhet 9d, 9e, River SR-08 2559 NE 20, 21b, 22b, 22c, 29a, 29b, 29c Habiganj, Maulvibazar SR-09 8756 SE 10, 16, 17a, 17b, 17c, 17d, 18f, 19a, 19b, Brahamanbaria, Chandpur, Comilla, Feni, 19c, 19d, 19e, 19i, 21b, 22b, 22c, 22d, 23a, Habiganj, Lakshmipur, Noakhali 23b, 29a, 29b, 29c, 30, 7, 8d, River SR-10 4574 EH 18e, 18f, 23a, 23b, 23c, 23d, 29a, 29b, 29c Chittagong, Cox’s bazaar SR-11 15,217 EH 23a, 29a, 29b, 29c, KP-LK, River Bandarban, Chittagong, Khagrachhari, Rangamati SR-12 2521 SE 17d, 18f, 19a, 19e, 23b Feni, Lakshmipur, Noakhali SR-13 8443 SW 10, 11a, 12a, 12b, 14a, 14b Chuadunga, Jessore, Jhenaida Kushtia, Magura, Meherpur, Satkhira SR-14 10,580 SC, SW 10, 11a, 12a, 12b, 13a, 13b, 13c, 13d, 13e, Bagerhat, Barisal, Faridpur, Gopalganj, Jessore, 14a, 14b, 19f, 19j, 7 Khulna, Kushtia, Madaripur, Magura, Narail, Pirojpur, Rajbari, Shariatpur SR-15 9346 SC 10, 12b, 13a, 13b, 13d, 14a, 18a, 18b, 18c, Barguna, Barisal, Bhola, Jhalokathi, Patuakhali, 18d, 18e, 18f, 19f Pirojpur, Shariatpur SR-16 9346 SW 11a, 13a, 13c, 13d, 13e, 13f, 13g, 14a, River Bagerhat, Khulna, Pirojpur, Satkhira *Agro-ecological zones as used by the BBS. These are deï¬?ned on the basis of physiography, soils, land levels in relation to flooding and agro-climatology. There are 30 agro-ecological zones in the country. Figure 4.2 Temperature changes for A2 scenario over GBM basin (the 2050s) Future Flood Hydrology 33 models show increased precipitation while others show decreased precipitation. Moreover, there is not even necessarily agreement on the direction of rainfall change between emissions scenarios for individual models (e.g. in January, ECHAM A2 estimates decreases, ECHAM B1 estimates increases). 4.5 Future Flood Characteristics and Analysis Future estimated discharges The future transboundary inflows of the three major rivers (Ganges, Brahmaputra and Meghna) during the monsoon period are simulated. For all three rivers, across the different global circulation models, inflows into Bangladesh are on average projected to increase over the monsoon period (driven primarily from increased basin precipita- tion). Not much difference is observed between the A2 and B1 scenarios. Larger changes are anticipated by the 2050s compared to the 2030s. Larger changes are observed on average for the Ganges.The magnitude of change from the base- line is dependent on the month (Table 4.2). For the Ganges and Brahmaputra, the average dis- charges increase for all months. Figure 4.3 Percentage change in discharges in (a) the 2030s and Not all model experiments predict increases (b) the 2050s for A2 scenario in August in discharge. For example (Figure 4.3), for the month of August the largest increases are esti- mated on average for the Ganges River (9 per per cent by the 2050s. In contrast, for the month cent to 13 per cent for the 2030s and the 2050s of May (Figure 4.4) larger average increases are respectively). The GFDL model, though, esti- observed for the Meghna and Brahmaputra flows mates a reduction in Ganges flow of almost 13 by the 2050s (17 per cent and 20 per cent respec- Table 4.2 Estimated average change (per cent) in discharge across all model experiments* 2030s 2050s Brahmaputra Ganges Meghna Brahmaputra Ganges Meghna May 7.5 9.3 –0.0 17.4 11.8 12.3 June 5.4 11.9 –3.1 10.9 16.7 7.7 July 3.4 13.5 –0.0 6.9 15.0 3.6 August 5.5 8.8 –3.7 9.5 12.0 7.8 September 3.7 7.3 –2.0 9.7 12.5 5.9 * 5 GCM x 2 SRES = 10 model experiments 34 Climate Change Risks and Food Security in Bangladesh For instance, all the model experiments indi- cate that the monsoon flow of the Teesta River will increase by the 2050s (Plate 4.5). For some river locations, direction of change varies among experiments (e.g. the Kushiyara River – Plate 4.6). Changes in spatial extent of land flooding Given that most model experiments indicate an increasing trend of monsoon rainfall and greater inflows into Bangladesh, if all else is equal the extent of flooding is likely to increase. Among the 16 sub-regions described earlier, only 11 were covered by the national hydrologic model (SR-01 to 08, SR-13 to 15). Using the gener- ated water level time-series at each grid point in these regions, average monthly water levels are calculated for the baseline and two future time periods.With these, the distribution of flood land types (F0, F1, F2, F3 and F4 described in section 2.2) can be determined. The locations of flood protection infrastructure (e.g. roads, embank- ments, polders) are incorporated. The national baseline flood maps are shown in Plate 4.7 and summary statistics given in Table 4.3. These maps and statistics are produced for every month and for every sub-region. It is important to note that this baseline dis- Figure 4.4 Percentage change in discharges in (a) the 2030s and (b) the 2050s for A2 scenario in May tribution of flood land types is different from that reported in the MPO (1987). This is in part due to the fact that since the early MPO analysis, the tively).The model range is largest for the Meghna government has invested substantially in polders (55 per cent increase to 32 per cent decrease) in and flood protective works (Table 4.4). this scenario, however, reflecting less baseline dis- Of a total modelled area of about 9940km2, charge at the onset of the monsoon. the total area that is flooded ranges from 6.9 per Changes in discharge could also be estimated cent (in May) to 36.7 per cent (in August). In at each of the water level points in the hydrologic general, the total flooded area peaks in August, model. Four geographically diverse locations coinciding with the peaks of the major rivers.This were selected to illustrate the ï¬?ndings. These are: pattern of flooding varies across sub-regions. the Old Brahmaputra River at Mymensingh, the Comparison of total change in flooded area Kushiyara River at Sherpur, the Teesta River at (sum of F1, F2, F3 and F4) is presented in Figure Kaunia and the Gorai River at Gorai Railway 4.5. Under climate change, across most models Bridge. Similar to the major inflows in Bangla- the flooded area is estimated to increase for most desh, most of the models and scenarios predict of the flood season.An average increase of flooded average increases in discharges ranging from 2 area of 3 per cent in 2030s and 13 per cent in to 50 per cent by the 2050s. Larger changes are 2050s for A2 scenario is projected (with the larg- anticipated by the 2050s compared to the 2030s. est changes simulated during the months of May Future Flood Hydrology 35 Table 4.3 Modelled baseline season flood land type distribution for each month (ha) Flood land type May Jun Jul Aug Sep F0 (flood free) 9,300,316 8,665,915 7,491,199 7,270,645 7,391,137 F1 , 292,437 , 490,887 ,835,497 ,835,533 ,828,990 F2 ,207,657 ,407,457 ,825,183 ,923,472 ,899,145 F3 ,118,944 ,315,819 ,645,453 ,762,957 ,710,262 F4 , 20,628 ,59,904 ,142,650 ,147,375 ,110,448 Total flooded ,639,666 1,274,067 2,448,783 2,669,337 2,548,845 (F1+F2+F3+F4) % Flooded 6.9% 14.7% 32.7% 36.7% 34.5% Table 4.4 Protected areas flood control and drainage infrastruc- the country at the confluence of the Ganges and ture (FCDI) Brahmaputra rivers and in the south. Moreover, Sub-region FCDI area to determine whether or not these estimated changes are large in comparison to the year-to- km2 % area year variations a one-standard deviation flood sur- SR-01 5860 45 face was also calculated. Table 4.5 shows for each SR-02 8314 48 sub-region the number of model experiments SR-03 2541 76 (total of 10) for the 2030s (numerator) and the SR-04 1179 15 SR-05 214 9 2050s (denominator) that exceed these bounds in SR-06 2167 17 each month. This is an indication of signiï¬?cance. SR-07 3337 27 In many sub-regions (1, 2, 6, 7 and 8), though SR-08 570 22 increases in flooded areas are estimated, these fall SR-09 3900 45 mostly within one-standard deviation bounds. SR-10 1407 31 By the 2050s, more model experiments estimate SR-11 319 2 changes that exceed these bounds. Most signiï¬?- SR-12 1461 58 SR-13 3498 41 cant changes occur later in the flood season, pri- SR-14 4064 38 marily in August and September at the height of SR-15 4467 48 the monsoon, and in the south and central parts SR-16 3246 35 of the country. Changes in temporal flood characteristics and June). Some models indicate a decrease in flooding during the month of May, June, and To compare the characteristics of the future and July. The GFDL model, for instance, shows the baseline hydrographs, 36 locations were selected largest decrease of 17 per cent in flooded area (at least 3 locations in each of the 11 sub-regions) in the month of May for the A2 scenario in the for temporal analysis (locations shown in Plate 2050s.The maximum observed increase in flood- 4.8). Using the approach outlined in Hassan et ing occurs during May (about 50 per cent) for al (2007), the time-series of the annual peak val- the MIROC model under the A2 scenario in the ues, onset date of the flood (with respect to May 2050s. The B1 scenarios show smaller changes 15) and recession date of the flood (with respect than the A2 scenarios, though in the same direc- to September 15) were analysed. Here, only tion. mean characteristics will be compared. Note Increases in flooded area vary by sub-region that incorporating future variability changes is (Table 4.5). Most models demonstrate agree- addressed through a Monte Carlo simulation in ment in changes in sub-regions 3, 4, 5, 13, 14 the economic modelling section of this study and and 15. These are primarily in the central part of is exogenous to the hydrologic modelling. 36 Climate Change Risks and Food Security in Bangladesh (a) 2030s – Scenario A2 (b) 2030s – Scenario B1 (a) (b) (c) 2050s – Scenario A2 (d) 2050s – Scenario B1 (c) (d) Figure 4.5 Total change in national flooded area for (a) 2030s A2, (b) 2030s B1, (c) 2050s A2, (d) 2050s B1 Table 4.5 Number of model experiments exceeding one-standard deviation bounds on baseline (2030s/2050s) and 2050s average esti- mated change in area flooded Sub-region 2030s/2050s no. of model experiments 2050s average change (%) exceeding one-standard deviation* in flooded area May June July Aug Sept May June July Aug Sept SR-01 2/4 0/1 0/0 0/1 0/0 – 23 25 14 19 19 SR-02 0/2 0/0 0/0 0/2 0/2 –23 33 18 15 8 SR-03 5/9 0/5 5/5 4/5 1/6 –460 77 15 15 56 SR-04 2/6 2/5 2/6 3/7 0/2 –76 49 17 14 16 SR-05 1/3 0/5 6/8 5/8 0/5 –44 28 10 7 9 SR-06 0/2 0/1 1/4 3/5 0/3 –20 19 10 12 14 SR-07 0/1 0/0 1/3 0/2 0/0 –11 5 1 2 3 SR-08 0/0 0/0 0/0 0/0 1/0 4 2 1 2 2 SR-13 8/6 3/2 4/6 7/7 2/7 –10 13 44 65 63 SR-14 3/1 0/5 3/3 3/4 0/2 –6 15 11 9 10 SR-15 0/9 7/9 8/9 9/9 9/9 – 54 69 38 31 24 * numerator 2030s/denominator 2050s. Future Flood Hydrology 37 Annual peak flows Table 4.6 Peak water level summary for the 2050s To estimate whether or not the peak flows are Sub-region No. model experiments* Average change (m) statistically changing under the climate change SR-01 2 –0.47 scenarios, the peak values at each of the 36 loca- SR-02 10 –0.33 tions are recorded for each year for the baseline SR-03 17 –0.27 and future time periods. Summary statistics (max, SR-04 6 –0.31 mean, min, standard deviation) are calculated. SR-05 11 –0.28 SR-06 8 –0.30 Hypothesis testing is performed to determine SR-07 10 –0.43 whether or not observed differences are statis- SR-08 6 –0.87 tically signiï¬?cant (null hypothesis that average SR-09 2 –0.27 peaks in the baseline and future time periods are SR-13 10 –0.41 not different). This is necessary because peak lev- SR-14 8 –0.19 els vary naturally from year to year. For instance, SR-15 22 –0.28 Figure 4.6 shows the observed peak water levels SR-16 8 –0.28 during the hydrologic baseline period (1978– *Number of model experiments where the estimated change is statistically different from the historical average peak level. Total number of model 2008) at one location on the Jamuna in addition experiments is 30 (5 GCMs x 2 SRES x 3 stations per sub-region). to the estimated peak water levels for the different model experiments for the 2030s. The baseline inter-annual variation is itself almost 0.5m (aver- age value of 11.6m). Thus, any observed change cally signiï¬?cant and the average change antici- in average yearly peak levels must be considered pated for the 2050s. The greater the number of in relation to this background variability. model experiments, the greater the agreement. Not all estimated changes in peak water lev- Changes are in general less than 0.5m from the els are statistically signiï¬?cant. More model exper- baseline. Estimated changes in sub-regions 3 and iments demonstrate changes that are statistically 15 are statistically the most robust. Estimated signiï¬?cant (at the p < 0.01 level) by the 2050s. changes in sub-regions 1 and 9 are statistically By the 2050s, many of the northern sub-regions the least robust. (2–9) show statistically signiï¬?cant increases in the annual peak while many in the southwest Onset and recession times (13–16) show decreases. Table 4.6 summarizes Average hydrographs were generated for each the observed changes in yearly peak levels, the of the 36 locations for the baseline period and number of model experiments that are statisti- two future climate change futures to compare the Figure 4.6 Yearly peak levels at Jamuna station for the baseline and model experiments (2030s) Note: Solid line is historical baseline and dashed lines are all future model experiments. 38 Climate Change Risks and Food Security in Bangladesh timing of the onset and recession of the yearly by ï¬?ve days (2030s) and ten days (2050s). Simi- monsoon. Using 15 May and 15 September as larly, the baseline water level on 15 September the baseline onset and recession dates respectively, is 50.97mm. This same water level occurs later the future dates when the baseline water level is under the two future scenarios – around 19 Sep- reached can be determined. Figure 4.7 shows an tember in the 2030s and 21 September in the example for a location on the Teesta River in SR- 2050s. That is, the recession of the flood waters 02 using the MIROC GCM for the A2 scenario. is delayed by about ï¬?ve to six days. For this loca- The baseline water level on 15 May is 49.55m. tion, this delay is consistent across the range of This same water level occurs earlier under the scenarios. two future scenarios – around 6 May in the 2030s In some locations the onset of the flood and 28 April in the 2050s.That is, the onset of the is delayed. For instance, for a location on the flood is early by almost two weeks by the 2050s. Meghna River in SR-15 the 15 May water level Averaged over all of the model experiments for is 1.79m (Figure 4.8). This same water level for this location, the estimated date of onset is earlier the GFDL GCM A2 scenario does not occur till Figure 4.7 Average hydrographs (baseline, 2030s, 2050s) for MIROC GCM and A2 scenario on Teesta River Figure 4.8 Average hydrographs (baseline, 2030s, 2050s) for GFDL GCM and A2 scenario on Meghna River Future Flood Hydrology 39 Figure 4.9 Average hydrographs on the Gorai River (baseline, 2030s, and 2050s – for CCSM A2 scenario) and plus/minus one-standard deviation bounds around 4 June in both the 2030s and 2050s.When and environmental modelling package from averaged over all the model experiments, a delay DHI Water and Environment. MIKE BASIN of approximately ten days is estimated.The reces- represents all elements of water resource sion of the flood waters at this location shows no modelling: users, reservoirs, hydropower, sur- signiï¬?cant change from the baseline. face water, groundwater, rainfall-runoff and Across the sub-regions, most model experi- water quality. ments estimate an earlier onset (as compared to 2 Umax = Maximum water content in surface the baseline) of the monsoon and a delay in the storage; Lmax= Maximum water content in recession.This is more apparent by the 2050s and root zone storage; CQ = Overland flow driven in large part due to the increased flows and runoff coefï¬?cient; CKIF = Time constant flooding under these climate change scenarios. for routing interflow; CKOF = Time constant Some caution must be exercised, however, when for routing overland flow; TOF = Root zone interpreting these results as the range of dates threshold value for overland flow;TIF = Root across the model experiments can be as much as zone threshold value for interflow; TG = 1–2 weeks. Moreover, in many cases the year-to- Root zone threshold value for groundwater year variation in the annual hydrograph is larger recharge; CKBF = Time constant for routing than the predicted changes. Figure 4.9 shows, for base flow; Csnow = Constant degree day coef- a location on the Gorai River, the baseline and ï¬?cient; T0 = Base temperature (snow/rain). two future estimated average hydrographs for 3 MIKE 11 is a system for the one-dimen- the CCSM A2 scenario. Here, the two estimated sional, dynamic modelling of rivers, channels future time-series fall within the one-standard and irrigation systems, including rainfall- deviation bounds, thus not separating from the runoff, advection-dispersion, morphological historical variability. and water quality. The complete St Venant equations can be solved, so the model can be applied to any flow regime where the flow Notes can be assumed one-dimensional. Diffusive 1 MIKE BASIN is a versatile Geographic wave, kinematic wave and quasi-steady state Information Systems-based water resource options are also available. Flow over weirs, 40 Climate Change Risks and Food Security in Bangladesh through culverts and user-deï¬?ned structures, 4 The baseline climate period (1979–99) differs and over the flood plain can be simulated. slightly from the baseline hydrologic period Output from the hydrodynamic module can (1978–2008) which introduces slightly higher be routed to additional modules that simulate baseline conditions for the ‘delta’ method. the transport of cohesive and non-cohesive However, changes between 1999 and 2008 sediment, dissolved oxygen, nutrients, heavy are the smallest of the 20th century. metals and eutrophication. 5 Future Crop Performance Box 5.1 Key messages • Elevated CO2 concentrations are projected to have a substantial positive effect on crop yield for all crops and locations. • Considering only temperature, precipitation and CO2 changes, aus and aman median production is projected to increase by 2 per cent and 4 per cent by the 2030s and the 2050s respectively. Wheat also increases, reaching a maximum of 4 per cent by 2050 before following a downward trend. These distributions range approximately +/– 2 per cent. • Boro production is projected to decline under climate change scenarios, around 8 per cent by the 2080s. Changes for boro and wheat are conservative as it is assumed that farmers have uncon- strained access to irrigation. • Shifts in the average floods are projected to reduce production of aus and aman by between 1 per cent and 4 per cent. The narrow model distribution of flood impacts projected by different GCMs suggest a robust change, although changes are small in comparison to year-to-year variability. • The area lost to production due to sea level rise can be substantial. Maximum crop losses of nearly 40 per cent are projected by the 2080s for the south. • Considering all climate impacts for the 2050s, the median of all rice crop projections show declining national production, with boro showing the largest median losses. However, for aus (–1.5 per cent) and aman (–0.6 per cent), the range of model experiments covers both gains and losses and does not statistically separate from zero. Most GCM projections estimate decline of boro production with a median loss of 3 per cent by the 2030s and 5 per cent by the 2050s. Wheat production increases up to the 2050s (3 per cent). • In each sub-region, production losses are estimated for at least one crop. The production in the southern sub-regions is most vulnerable to climate change. For instance, average losses in SR-16 (containing Khulna) are –10 per cent for aus, aman and wheat, and –18 per cent for boro by the 2050s. • The current large gap between actual and potential yields suggests substantial on-farm opportunities for growth and poverty reduction. Expanded availability of modern rice varieties, irrigation facilities, fertilizer use and labour could increase average yields at rates that could potentially more than offset the climate change impacts. This section describes the use of a Decision Sup- The DSSAT model covers 16 sentinel locations port System for Agrotechnology Transfer model in Bangladesh (to correspond to each sub-region (DSSAT v4.5, Hoogenboom et al, 2003; Jones et described in the previous section) and focuses on al, 2003) to simulate agricultural yields under a yields of rice (primarily three seasons) and wheat. range of climate change scenarios in Bangladesh. Simulations of changes in crop yield include 42 Climate Change Risks and Food Security in Bangladesh impacts from climate only (CO2, temperature ration rates. A soil proï¬?le provides information and precipitation), floods and coastal inundation, about available nutrients and root-zone moisture both separately and in combination. For the cli- processes. Cultivar genetics determine the type mate only simulations, probabilistic distributions of crop that is grown, including biophysical char- of yield changes are generated across the set of acteristics determining development and vulner- global climate models, several emissions sce- ability to environmental stresses. Management narios (A1B, A2, and B) and future time slices practices dictate the date, method and geometry (2030s, 2050s and 2080s).Together, these 16 loca- of planting, as well as any applications of irriga- tions, 3 growing seasons, 3 emissions scenarios, tion, fertilizer or chemicals. 16 GCMs, 3 time-slices and 30-year periods As discussed in section 2.1, actual yields are for each simulation required the simulation of much lower than the potential yields observed more than 200,000 individual DSSAT cropping at experimental stations under controlled condi- seasons. The sensitivity of various individual cli- tions.The model outputs for this study are poten- mate parameters was also extensively tested. The tial yields as simulated by the CERES models number of individual runs for the flood-damaged under recommended agricultural practices, and yields is signiï¬?cantly reduced as only a sub-set so validation of the potential yields is complex. of model experiments are used (as described in Insects and rodents, which may severely damage section 4.3). The results of these simulations are infested areas, were not modeled.Thus, crops were reported here. assumed to be disease- and weed-free. Histori- cally, shortages or prohibitive costs of irrigation, fertilizer and labour reduce yields, while variation 5.1 Development of the Baseline in management practices exists across the country. Period In addition, many farmers have not yet adopted modern rice varieties. For this study, as the main concern is to estimate the changes from the base- CERES crop models line, replicating the actual observed yields is of Crop simulations for this project utilized the secondary importance. This, of course, assumes Crop Environment Resource Synthesis (CERES) that crop response functions will be similar for Rice and Wheat models, which are components high-input and low-input cropping systems. of the DSSAT cropping systems model (Hoog- The ability of the CERES models to accu- enboom et al, 2003). These dynamic biophysical rately represent the agricultural impacts of anthro- crop models simulate plant growth on a per hec- pogenic climate change is hindered by consider- tare basis, maintaining balances for water, carbon able uncertainty in the magnitude of CO2 effects and nitrogen. CERES models have been applied (Easterling et al, 2007; Long et al, 2006; Tubiello previously in Bangladesh to model rice (Hussain, et al, 2007a, b; Ainsworth et al, 2008; Hatï¬?eld et 1995; Mahmood et al, 2003) and rice-wheat sys- al 2008) and the location of temperature thresh- tems (Timsina and Humphreys, 2006a,b; Timsina olds for crop damage. CO2 is a primary element et al, 1998). Studies examining climate change of photosynthesis, and plants respond to elevated impacts on agricultural production in Bangladesh levels by increasing the rate of primary produc- have also employed the CERES models (Karim tion. High CO2 concentrations also increase root et al, 1994; Hussain, 2006). densities and allow a plant to make more efï¬?cient CERES models require information about gaseous transfers with its environment, collect- the plant environment (weather and soils), cultivar ing sufï¬?cient CO2 in shorter periods of open leaf genetics and agricultural management practices. stomata. This has the added effect of increasing Daily maximum and minimum temperatures, water use efï¬?ciency in the plant as the duration precipitation, carbon dioxide concentrations and that stomata are open is lessened and stomatal solar radiation determine respiration and photo- resistance is increased, reducing the loss of mois- synthesis rates, available water and evapotranspi- ture to transpiration. Biophysical crop models, Future Crop Performance 43 controlled chamber experiments, and Free Air Climate data used CO2 Enrichment (Hendry and Kimball, 1994) experiments have demonstrated these effects, but Each sub-region was required to have a Bang- the extent to which large-scale ï¬?eld crops will ladesh Meteorological Department (BMD) respond to CO2 is uncertain. The CERES-Rice observation station covering a 1970–99 baseline and CERES-Wheat models use a simple look- period.Table 5.2 shows the BMD station selected up table to relate growth coefï¬?cients with CO2 for each sub-region, as well as the mean climate levels, and produce responses that are relatively conditions for variables required by the CERES optimistic. In addition, high temperatures can models. The weather generator component of damage crop development if sharp events occur DSSAT was used to ï¬?ll in gaps in the observa- during key phenological stages (particularly the tional climate record and to convert the BMD grain setting period), but these stresses are not sunshine hours measurements to solar radiation modelled in the CERES simulations. data needed by the CERES models Sub-region production totals Soil data The CERES models simulate crop develop- Soil proï¬?le data were not available at all of the ment and yield on a single hectare. Production sentinel BMD locations, but suitable matches totals are determined by multiplying the yield were located for every sub-region (see Table 5.3). by cropped area in a particular sub-region. The Soil proï¬?les for some regions are available in agricultural areas in each sub-region producing DSSAT-compatible formats from Hussain (1995) aus, aman, boro and wheat are presented in Table and others were generated drawing from Bram- 5.1 below. mer (1996). If multiple proï¬?les existed, the pro- ï¬?le closest to each sentinel site was selected, pro- vided it had soil information down to at least 1m depth. When no proï¬?les existed within a region, proï¬?les from neighbouring sub-regions or sub- regions with similar surface soil conditions were Table 5.1 Sub-regional agricultural information used. Following Mahmood et al (2003), paddy Sub- Aus area Aman area Boro area Wheat area rice percolation rates for all sub-regions were set region (ha) (ha) (ha) (ha) at 4mm/day. 1 65,401 613,853 543,438 174,969 2 42,491 815,334 976,956 149,176 Cultivar information 3 21,165 116,985 86,369 53,228 4 46,826 239,834 261,113 48,131 Genetic information for the CERES models 5 13,458 51,661 32,107 8225 was drawn from existing and estimated coefï¬?- 6 100,018 610,184 561,919 39,339 cients for cultivars used in Bangladesh (BRRI, 7 124,288 350,173 589,900 10,034 2007). For the aus season, the Bangladesh Rice 8 24,304 64,558 19,294 541 Research Institute (BRRI) BR3 cultivar was 9 129,498 330,897 426,250 33,260 selected. Known locally as biplab, BR3 is a com- 10 38,209 23,194 74,850 39 bination of a foreign rice, the International Rice 11 30,320 14,628 71,881 17 Research Institute (IRRI) IR-506-1-133 and a 12 45,945 121,579 26,894 103 local rice variety. BR3 grows quickly and may 13 98,834 359,053 333,375 67,317 14 167,592 351,018 286,413 49,575 be planted late, leading to productive aus seasons. 15 240,201 198,407 162,463 6399 Cultivar information was not available for the 16 14,809 215,747 78,225 1794 more popular BR24, 26 and 27 varieties. BR11, Note: Aus and wheat areas come from the 2003–2004 season, based on known locally as mukta, was used for the aman Bangladesh Bureau of Statistics (2005), while 2003 aman and 2008 boro areas season. Cultivar information was not available come from CEGIS. for the more popular BR31 variety. CERES 44 Climate Change Risks and Food Security in Bangladesh Table 5.2 Climate information for each sub-region: the representative BMD station, its code and annual mean climate statistics during the 1970–99 baseline period Sub- BMD Station Location BMD Station Mean Tmax Mean Tmin Mean Rainfall Mean Sunshine region Code (°C) (°C) (mm) (MJ/m2/day) 1 Dinajpur 10,120 30.1 19.7 2003 16.9 2 Rangpur 10,208 29.7 19.9 2239 17.2 3 Ishwardi 10,910 31.7 20.3 1652 17.3 4 Tangail 41,909 30.3 20.8 1902 16.7 5 Dhaka 11,111 30.6 21.6 2148 17.6 6 Mymensingh 10,609 30.7 20.5 2255 16.4 7 Sylhet 10,705 29.6 20.2 4150 16.7 8 Srimangal 10,724 30.4 19.4 2421 17.7 9 Comilla 11,313 30.1 20.9 2054 17.2 10 Chittagong 11,921 30.2 21.6 2931 18.7 11 Rangamati 12,007 30.2 21.4 2532 17.7 12 Maijdee Court 11,809 29.8 21.6 3103 16.9 13 Jessore 11,407 31.4 20.9 1600 17.2 14 Faridpur 11,505 30.4 21.1 1967 17.2 15 Patuakhali 12,103 30.3 21.9 2704 15.6 16 Khulna 11,604 31.1 21.6 1812 17.4 Table 5.3 Soil proï¬?le information for each sub-region Sub- Soil Location Soil Description Soil Type Soil Depth Percolation Initial Initial NO3 Initial NH4 region (cm) Rate Moisture (kg N/ha) (kg N/ha) (mm/day) (mm) 1 Dinajpur Aeric Endoaquepts 130 4 243 44.7 4.11 2 Rangpur Aeric Endoaquepts 84 4 142 47.2 2.16 3 Jessore Aeric Endoaquepts 137 4 332 44.5 4.78 4 Karatia Silty Loam Aeric Endoaquepts 107 4 290 46.6 2.86 5 Ghatail Typic Dystrudepts 122 4 375 42.6 7.53 6 Phulpur Loam Aeric Endoaquepts 116 4 187 44.1 5.86 7 Biani Bazar Typic Dystrudepts 107 4 162 42.9 7.04 8 Srimangal Very ï¬?ne sandy loam Udic Ustochrept 185 4 397 41.1 8.49 9 Shalpur Hyperthermic Typic 160 4 353 43.1 5.99 Endoaquept 10 Chittagong Aeric Endoaquepts 216 4 378 39.6 9.78 11 Srimangal Very ï¬?ne sandy loam Udic Ustochrept 185 4 397 41.1 8.49 12 Hatiya Silty Aeric Fluvaquent 165 4 543 41.6 8.01 13 Jessore Aeric Endoaquepts 137 4 332 44.5 4.78 14 Jessore Aeric Endoaquepts 137 4 332 44.5 4.78 15 Satkhira Clay Loam Typic/Aeric Hapluquept 142 4 603 36.8 13.1 16 Satkhira Clay Loam Typic/Aeric Hapluquept 142 4 603 36.8 13.1 coefï¬?cients for BR11 transplanted aman and the most common variety of wheat grown in BR3 are distributed with DSSAT v4.5 (Hoog- Bangladesh, is not packaged with DSSAT v4.5. enboom et al, 2003). BR29, introduced in 1994, However, kanchan genetic coefï¬?cients exist for was selected as a more current variety for the the CERES-Wheat model in DSSAT v3.0 and boro season, with genetic information provided for the CropGro Wheat model in DSSAT v4.5. by Dr Sk. Ghulam Hussain (at BARC). Kanchan, Using the coefï¬?cients from these model versions, Future Crop Performance 45 genetic coefï¬?cients were estimated for CERES- Characteristics of farm-level management Wheat in DSSAT v4.5 format, with modiï¬?ca- practices for the cultivation of rice and wheat tions necessary to capture appropriate season were selected for the CERES models according length.These cultivars have been commonly used to the recommendations of the Bangladesh Rice in crop modelling studies of Bangladesh. Research Institute (BRRI, 2007), annual reports from the Bangladesh Agricultural Research Insti- Agricultural management practices tute (BARI) and published studies examining rice and wheat systems in Bangladesh. Rice seedlings assumed are commonly raised in a seedbed before they are CERES-Rice and CERES-Wheat require infor- transplanted to the wider ï¬?eld, with aus plants mation about the management practices govern- spending 25 days in a seedbed, aman 30 days and ing crop cultivation during the growing season. boro 35 days (BRRI, 2007). The transplant envi- Planting dates, planting geometry and planting ronment temperature was set as the mean tempera- environment are necessary, as well as any fertilizer ture for Bangladesh for the 30 days prior to trans- or irrigation applications. These simulations were planting, approximating the seedbed temperature initiated with 75 per cent initial moisture for that determines initial development. Aus and aman two weeks of fallow period before the planting/ are dependent on local rainfall, but boro and wheat transplanting date, with no crop residue left on are aided in the simulations by an automatic irriga- the ï¬?eld. Nitrogen and water cycle processes and tion routine that applies irrigation whenever the limitations were included in these experiments. soil moisture in the top 30cm of soil falls below 50 Tables 5.4 and 5.5 show the management char- per cent of saturation. Water availability for irriga- acteristics selected for the rice and wheat simu- tion was assumed to be limitless. 120kg/ha of urea lations respectively. Management practices may fertilizers were added in the simulations to all rice vary considerably across any given sub-region crops, and 100kg/ha were added to wheat. These and between sub-regions across Bangladesh. totals far exceed the current average application, For the purposes of this study, management prac- but fertilizer use is expected to expand with devel- tices for a given crop were assumed to be the opment in future periods and thus optimistic sce- same across all locations to allow sub-regional narios are used. Finally, these management options comparisons. were maintained for future scenarios, allowing a direct comparison of the climate impacts on yield. Table 5.4 Agriculture management options for simulations of the three main rice varieties Aus Aman Boro Source Cultivar BR3 BR11 BR29 Hussain, 1995 Local name Biplab Mukta BRRI, 2007; Hussain, personal communication, 2008 Simulation date 1 Apr 1 Jul 1 Dec Two weeks before planting Planting date 15 Apr 15 Jul 15 Dec Hussain, 1995; Mahmood, 1997; Gomosta et al, 2001; Hussain, personal communication, 2008 Plant population (plants/m2) 50 50 50 Latif et al, 2005; BRRI, 2007 Row spacing (cm) 20 20 20 Hussain, 1995 Sowing depth (cm) 6 6 6 Mahmood et al, 2003; Hussain, 1995 Transplant age (days) 25 30 35 BRRI, 2007; Mahmood et al, 2003 Transplant temperature (ºC) 27 28.5 21.9 BMD observations Bund height (mm) 100 100 n/a Hussain, 1995 Irrigation Rainfed Rainfed Automatic BBS, 2005b Fertilizer type Urea Urea Urea Hussain, 1995 Fertilizer amount (kg N/ha) 40, 40, 40 40, 40, 40 40, 40, 40 Latif et al, 2005; BRRI, 2007 Applications (days after transplanting) 1, 30, 60 1, 30, 60 1, 30, 60 46 Climate Change Risks and Food Security in Bangladesh Table 5.5 Agriculture management options for wheat simulations than water with a high silt content, and do not Wheat Source account for damages from flood water currents (Hussain, 1995). Similar approaches to estimating Cultivar Kanchan Timsina et al, 1998 flood damages based on comparisons between Simulation date 1 Dec water level and rice plant height are reported in Planting date 15 Dec Two weeks Yoshida (1981) and Kotera and Nawata (2007). before planting To characterize the phenological stage and Plant population (plants/m2) 220 Timsina et al, height of rice crops during the flooding season, 1998 daily CERES-Rice output was analysed for the Row spacing (cm) 20 baseline period. Although dates of some pheno- Sowing depth (cm) 3.5 logical transitions are available, plant height is not Irrigation Automatic Fertilizer type Urea recorded as a variable in CERES-Rice. Dates for Fertilizer amount (kg N/ha) 66, 34 key developmental milestones (transplanting, end Applications (days after transplanting) 1, 21 of juvenile development, panicle initiation, head- ing, beginning of grain ï¬?lling and maturity) were calculated across all crops, with results showing only minor changes between years and sub- 5.2 Developing Flood Damage regions. The date of maximum tillering, which is required by the Hussain method, was not avail- Functions able and therefore was estimated. Plant height The CERES-Rice model simulates water stress at each developmental milestone was estimated for both photosynthesis and growth during according to published reports and the height at water shortages, but assumes that excess rainfall maturity for BR3 and BR11 cultivars, with daily (that cannot be absorbed by the soil or puddle height in any given stage interpolated from its in a bund) is lost as runoff without any damages endpoints (Yoshida, 1981; Hussain, 1995; Chen et inflicted. As simulations occur on a single hec- al, 2007; Kotera and Nawata, 2007). tare, water that flows on to agricultural areas from For each of the 11 sub-regions where future flash floods or rising rivers cannot be seen by floods were modelled, flood damages were deter- the model. Thus, the CERES-Rice model does mined separately for crops grown on each flood not model the flood damages that affect Bang- land type (see Table 2.6). To determine flood ladesh. Instead, in this work flood damages were depths, the mean baseline hydrograph of a rep- separately assessed and applied to CERES-Rice resentative location (typically the upstream-most output using information gleaned from potential point; see Table 5.7) was used as a reference to rice yield simulations and the hydrologic models determine the base of the rice plants (to ensure, described in Chapter 4. for example, that F1 rice plants really flooded The assessment of flood damages for aus and between 30 and 90cm on average). Each daily aman rice was based on the methodology devel- hydrograph was then compared to plant heights oped for Bangladesh by Hussain (1995; herein estimated according to phenological stages from referred to as the Hussain method). Boro rice and the CERES model output and known harvest wheat, which are grown during the dry season, heights for the BR3 and BR11 cultivars (Yosh- are assumed to be free from flood impacts. Flood ida, 1981; Hussain, 1995; Chen et al, 2007; Kotera damages are assessed according to the depth of and Nawata, 2007). The Hussain (Hussain, 1995) a flood (as a percentage of the plant height), the damages were then assessed according to flood duration of the flood and the developmental stage depths as a percentage of plant height, flood dura- of the rice plant when the flood occurs.Table 5.6 tions and phenological stage (Table 5.6). The rise presents the crop damage inflicted under vari- and fall of flood waters to different percentages of ous flood scenarios. These damages assume clear the plant heights often led to overlapping dam- flood waters, which are slightly less damaging aging events from long-duration low floods and Future Crop Performance 47 Table 5.6 Flood damages (percentage yield) according to submergence depth, duration and phenological stage Submergence of 25–50% of plant height Days of Submergence Phenological Stage 3 to 6 7 to 9 10 to 14 15 or greater 10 days after transplanting 10 15 20 30 Maximum tillering 10 15 25 40 Panicle initiation 0 0 30 40 Heading 0 0 30 40 Early grain ï¬?lling 0 0 30 40 Maturity 0 25 40 40 Submergence of 50–75% of plant height Days of Submergence Phenological Stage 3 to 6 7 to 9 10 to 14 15 or greater 10 days after transplanting 10 40 50 60 Maximum tillering 5 50 60 70 Panicle initiation 15 40 50 60 Heading 15 40 50 60 Early grain ï¬?lling 20 40 60 70 Maturity 20 40 60 70 Submergence of 75% or more of plant height Days of Submergence Phenological Stage 3 to 6 7 to 9 10 to 14 15 or greater 10 days after transplanting 40 80 100 100 Maximum tillering 40 60 80 100 Panicle initiation 50 70 100 100 Heading 40 80 100 100 Early grain ï¬?lling 30 60 80 100 Maturity 30 60 80 100 Source: from Hussain, 1995. brief high waters. In such cases only the more maximum hydrograph reading for each season. damaging element was recorded. If flood waters Therefore, years with higher than average annual receded entirely before a later flood occurred, maximum floods would have more land classiï¬?ed that event could inflict further damages. as flooded (and deeply flooded). This approach Sub-regional damages for a particular year produced slightly larger flood damages than were then aggregated according to the area of those estimated by simply examining the damage each flood land type in that sub-region, which caused by the average annual hydrograph over a could change from season to season as the flooded given scenario. The baseline flood damage fac- areas are larger during high flood seasons. Inter- tors (between 0 and 100 per cent) for each sub- annual variations in flooded areas were deter- region is given in Figure 5.1. mined from monthly maps of the mean flooded area (e.g. Plate 4.7 for the baseline period) and a Uncertainties with flood damages +/- one standard deviation flood map. For each sub-region, a second-order polynomial was used There are several uncertainties that lead to likely to describe the land area covered by each flood biases in the flood damage assessment approach land type according to the anomaly of the annual conducted for this study. These biases are offset 48 Climate Change Risks and Food Security in Bangladesh Table 5.7 Representative hydrographs situated without regard to flood land type leads Sub-region River Name to a likely overestimate of flood damages. Sec- ond, taking only the maximum damage during 1 Upper Karatoya 2 Teesta an event where waters rise underestimates flood 3 Atrai damages that would accrue from lower water 4 O Brahmaputra levels and inundation periods. Finally, deï¬?ning 5 Turag water levels in each flood classiï¬?cation according 6 O Brahmaputra to the maximum annual flood extent underes- 7 Surma timates the damages caused by early-season and 8 Juri late-season floods that rise to a deeper flood clas- 13 Ganges siï¬?cation during peak flow. 14 Gorai 15 Southern Meghna 5.3 Incorporating Coastal to an extent, but there are areas where additional Inundation Effects information could improve projections. First, the As discussed in section 3.3, coastal zone mod- location of agricultural areas in each sub-region els were used to simulate baseline conditions, as is only measured on a thana (precinct) administra- well as 15cm, 27cm and 62cm of mean sea level tive level, which is not ï¬?ne enough to determine rise. These future simulations were designed to the actual amount of agricultural area that falls represent the B1 2080s, A2 2050s and A2 2080s under each flood classiï¬?cation. Farms are located respectively. In this study we also attributed the disproportionately in regions that are slightly ele- 15cm level to the A2 2030s. B1 and A1B sea lev- vated to avoid flood damages and low-lying areas els would be lower. For the B1 scenario, which may grow taller rice varieties in anticipation of was given a 15cm rise for the 2080s, we estimated floods, so the assumption that agricultural land is a 5cm and 8cm rise for the 2030s and 2050s Figure 5.1 Baseline sub-regional yields with flood damages applied (as a percentage of undamaged yields) Light grey bars represent sub-regions that were not simulated by the hydrologic model. Flood damages for the boro and wheat seasons were not modelled. Future Crop Performance 49 respectively. Since these lower values were not observational data. This adjusted observational explicitly modelled, we estimated that their inun- climate data is used directly in DSSAT. Thus, dations were approximately linear proportions of these scenarios are plausible climate conditions the 15cm inundation maps. Since the publication that retain the day-to-day and year-to-year char- of the IPCC Fourth Assessment Report (AR4), acteristics of the 1970–99 baseline period, with new research into ice dynamics in a warming monthly mean temperatures and precipitation climate suggest that sea level rise may occur that reflect the mean climate changes simulated more rapidly than previously thought (Alley et by the GCM/emissions scenarios for the 2030s, al, 2005; Copenhagen Diagnosis, 2009). As noted 2050s and 2080s. By adjusting according to in Chapter 3, improved methods are available for monthly means, changes in seasonality simulated re-estimating sea level rise. by a given model experiment are also captured. Even with current sea levels, many coastal The entire set of GCMs (Table 3.1), all three areas are periodically flooded with seawater due emissions scenarios and all future time slices are to tidal oscillations and river floods (see Plate 3.1). utilized. Areas that are classiï¬?ed as experiencing coastal In the initial simulations of future changes floods of 30cm or more during the baseline in potential unflooded yield (before the applica- period are assumed to have already abandoned tion of flood and sea level damages), crops are grain production. Thus, impacts of sea level rise affected by CO2 concentration, temperature and on agriculture can be assessed by removing sub- precipitation changes. These factors interact in a regional production according to the proportion non-linear manner, but an estimate of the relative of additional land lost under the future flooding contribution of the CO2 concentration com- scenarios. pared to the effects of changing temperature and Future agricultural impacts of sea level rise precipitation may be attained through test simu- are likely biased by somewhat offsetting factors lations. Both rice and wheat are C3 crops, and that were not modelled. Only mean sea level thus both react strongly to changes in tempera- rise and tidal fluctuations were considered for ture and CO2 concentrations (Kimball and Ber- this study, representing an optimistic scenario nacchi, 2006; Hatï¬?eld et al, 2008). The isolated of coastal inundation. Additional land lost to impacts of changing CO2 concentrations, as well extreme tidal and storm surges that can penetrate as of changing precipitation and temperature, are further up the distributaries along the Bang- described in this section, along with their com- ladesh coast were not considered, nor were the bined direct effect of changing climate in each effects of salinity increases on soil, ground and sub-region. Figure 5.2 summarizes the results as irrigation water. However, agricultural land in changes in national potential production due to these areas was assumed to be distributed without isolated and combined direct climate effects. regard for potential coastal inundation, leading to a likely overestimate of sea level rise damage. Also CO2 impact experiments not considered was the salinity impact of farm- ers who convert their ï¬?elds to aquaculture (e.g. To isolate the agricultural effects of increased shrimp) by inundating their land with saltwater. CO2 concentration predicted by climate change, future climate scenarios were generated that used the baseline temperature and precipitation data 5.4 Projections of Future Potential but used CO2 levels to match the future scenarios (see Table 5.8).Thus, the only difference between Unflooded Production (Climate Only) the future experiments and the baseline condi- As discussed in section 4.3, a ‘delta’ approach tions was an elevated CO2 concentration. Because is used for generating future climate scenarios the same CO2 levels were used for all GCMs in whereby climate changes from a particular model a particular emissions scenario, a single 30-year experiment are applied directly to the BMD simulation for each crop in each sub-region 50 Climate Change Risks and Food Security in Bangladesh National Aus Production Change 20 10 0 % –10 –20 2030s 2050s 2080s 2030s 2050s 2080s 2030s 2050s 2080s 2030s 2050s 2080s National Aman Production Change 20 10 0 % –10 –20 2030s 2050s 2080s 2030s 2050s 2080s 2030s 2050s 2080s 2030s 2050s 2080s National Boro Production Change 20 0 % –20 2030s 2050s 2080s 2030s 2050s 2080s 2030s 2050s 2080s 2030s 2050s 2080s National Wheat Production Change 20 0 % –20 2030s 2050s 2080s 2030s 2050s 2080s 2030s 2050s 2080s 2030s 2050s 2080s Figure 5.2 Percentage change (versus the baseline undamaged simulation) in national potential production of a) aus, b) aman, c) boro and d) wheat Note: Each panel has four sections, each containing the three future time periods and presenting (from left to right) the combination of all emissions scenarios, the A1B, the A2 and the B1 scenario. The results of the CO2 impact experiments are displayed as upside-down triangles, while the median of the temperature and precipitation impact experiments is shown as an upside triangle. The distribution of undamaged potential yields projected by GCMs are presented as a box and whiskers diagram, consisting of a line representing the median value, a box enclosing the inter-quartile range (the middle 50 per cent of models), dashed whiskers extending to the furthest model that lies within 1.5 times the inter-quartile range from the edges of the box, and plus symbols for additional models that are perceived as outliers. Future Crop Performance 51 Table 5.8 Carbon dioxide concentrations (ppm) for baseline is driven by temperature increases. Although the period and future climate scenarios medians are all negative, some GCMs produce B1 A1B A2 slightly positive changes even when the CO2 is 1980s 345 345 345 ï¬?xed to baseline levels, as temperature or pre- 2030s 450 472 470 cipitation changes may be favourable for a par- 2050s 498 552 556 ticular season in a given GCM. Regardless of the 2080s 541 667 734 uncertain (although positive) magnitude of CO2 effects, including future CO2 concentrations increases potential production from the baseline allowed the contribution of CO2 enhancement levels seen in these experiments. to be explored. Elevated CO2 concentrations have a positive Unflooded (climate only) potential effect on crop yield for all crops and locations in production projections the simulations. Presented as upside-down trian- Simulations of crop production with changes gles in Figure 5.2, the CO2 impact experiments in temperature, precipitation and CO2 concen- simulate an effect that can raise potential produc- trations reveal large differences between crops, tion by up to 20 per cent for 2080s A2 aus and 12 GCMs and emissions scenarios. Results are pre- per cent in the 2080s B1 aus. CO2 effects gener- sented as box-and-whisker diagrams of national ally follow an upward trajectory in the future for production changes in Figure 5.2, capturing the each scenario. Potential production increases are range of changes produced from the full set of least for aman, not quite reaching 10 per cent for IPCC GCM outputs relevant to Bangladesh. the A2 2080s. These simulations may be optimis- Changes largely depend on the interplay between tic about the beneï¬?cial effects of CO2 on rice CO2 enhancement and the detrimental effects of and wheat production, as research continues into temperature and precipitation. Their combined the impact of CO2 on open ï¬?eld crops, particu- effect is clearly complex. larly in tropical areas (Hatï¬?eld et al, 2008). Aus production Temperature and precipitation impact Aus production increases under climate change experiments scenarios, although the range of GCMs indicates To examine the effects of future temperature some uncertainty in these projections. There and precipitation on potential yield without is also a clear shift in trend in the latter half of the influence of elevated CO2, future scenarios the 21st century that suggests that the effects of for each GCM were created using the ‘delta’ temperature and precipitation stresses are more approach but with CO2 levels ï¬?xed at their base- pronounced. For all scenarios and time periods line level (345ppm). The results are summarized at least one GCM produces a decline in produc- as the upside triangles in Figure 5.2, representing tion, but median production changes rise from the median change in national potential produc- 1.1 per cent in the 2030s to 2.2 per cent in the tion (across all GCMs) due only to changes in 2050s and 2.4 per cent in the 2080s. The central temperature and precipitation. half of the GCM distribution suggests a range Without the beneï¬?cial effect of CO2, future of about +/- 2 per cent in the 2030s and 2050s, climate changes reduce crop production. The with the centre of the distribution expanding most strongly affected crop is wheat, with poten- to +/- 3 per cent in the 2080s. The 2030s A1B tial production declining almost 30 per cent in results are approximately distributed around zero the A2 2080s and 15 per cent in the B1 2080s. change, but the 2080s A1B are the most positive Large potential production losses are also seen of any scenario (4 per cent). Both the A2 and B1 in the boro and aus crops. Because they are irri- scenarios begin with a larger change. Only the gated, the decline in wheat and boro production 2050s period has the entire central range of each 52 Climate Change Risks and Food Security in Bangladesh scenario simulation higher than zero change. Aus in the 2080s A1B.The central portion of the boro production changes are fairly evenly distributed distributions ranges around +/- 2 per cent, and between the CO2 impact and the temperature the distributions grow tighter at the end of the and precipitation impact experiment results. 21st century. Boro production changes are more responsive to the temperature changes than the Aman production CO2 enhancement, with irrigation offsetting any precipitation change. Note that these estimates Aman production rises under all future time are conservative given that groundwater irriga- periods, emissions scenarios and GCM simu- tion is assumed to be unconstrained. Currently, lations. Following rising CO2 concentrations, with evidence of declining groundwater tables in median production changes increase with time many parts of the northwest, future water avail- to 2 per cent in the 2030s to 4 per cent in the ability may pose a serious additional constraint 2050s and to 5 per cent in the 2080s. Aman rice on boro production. has the most robust changes of any crop, show- ing remarkably low variability between different scenarios. Overall changes are relatively small, Wheat production however, with only a few GCM scenarios caus- Wheat production increases under most climate ing rises of 10 per cent or more. The centre half change scenarios, although there is a clear shift of the simulation distributions range around +/- in trend at the end of the 21st century towards 0.5 per cent in the 2030s, rising to +/- 1 per lower yields. Median increases across all scenar- cent by the 2080s when emissions scenarios are ios are 2 per cent, 4 per cent and 3 per cent for combined. This increase in range is due mostly 2030s 2050s, and 2080s respectively. The largest to the diverging emissions scenarios, which drive median increases in yield for all emission scenar- signiï¬?cantly different CO2 enhancement impacts. ios occur in the 2050s, with A2 production rising Aman rice is grown at the height of the mon- 4 per cent, B1 production rising 3 per cent and soon season, which provides plenty of rain even the centre of the distribution ranging +/- 2 per in simulations where mean rainfall decreases cent. Potential production tracks more closely slightly. Monsoon clouds and high humidity also with the CO2 enhancement than with the detri- keep temperatures lower than in the pre-mon- mental temperature increases (irrigation negates soon period. These factors reduce the tempera- precipitation changes), but toward the end of the ture and precipitation impacts of climate change, century production begins to move away from and cause the production to be influenced most the CO2 influence toward the much lower tem- strongly by the beneï¬?cial CO2 enhancement. perature impacts. Some scenarios even produce better production gains than the CO2 enrichment alone, suggesting that the temperature and precipitation character- 5.5 Projections of Future Projected istics are more favourable for aman production. Flood Damages These results are in the absence of flood impacts Production changes due to flood damages are discussed in subsequent sections. presented in Figure 5.3 for aus and aman rice (boro and wheat are assumed to be un-impacted Boro production by floods). Flood damages are projected to Boro production declines under climate change increase in most scenarios, particularly for the scenarios, with median national production losses 2050s time period and for the aman crop grown across all emissions scenarios reaching 3 per cent at the height of the monsoon. Median additional in the 2030s, 5 per cent in the 2050s and 8 per losses across all scenarios are 1 per cent and 2 per cent in the 2080s. Several simulations project cent in the 2030s and 2050s respectively. Maxi- decreases in production of more than 10 per cent, mum median flood losses occur in the 2050s A2, with one GCM projecting a 15 per cent decline with national aman production falling 4 per cent Future Crop Performance 53 National Aus Production Changes – Food Damages Only 10 5 0 % –5 –10 ALL 2030s ALL 2050s A2 2030s A2 2050s B1 2030s B1 2050s National Aman Production Changes – Food Damages Only 10 5 % 0 –5 –10 ALL 2030s ALL 2050s A2 2030s A2 2050s B1 2030s B1 2050s Figure 5.3 Percentage change (versus the baseline flood-only simulation) in national potential production affected by basin floods of a) aus and b) aman (boro and wheat are assumed to be flood-free) Note: Each panel has three sections, each containing the two future time periods and presenting (from left to right) the combination of all emissions scenarios, the A2 scenario and the B1 scenario. The distribution of flood-damaged potential yields projected by GCMs is presented as a box and whiskers diagram, consisting of a line representing the median value, a box enclosing the inter-quartile range (the middle 50 per cent of models), dashed whiskers extending to the furthest model that lies within 1.5 times the inter-quartile range from the edges of the box, and plus symbols for additional models that are perceived as outliers. and national aus production dropping 2.4 per sub-regions experience impacts that increase cent. More modest crop losses are projected for with time as sea levels rise, with maximum crop the B1 scenario, reaching only 1 per cent of aus losses of nearly 40 per cent projected for the production and 3 per cent of aman production 2080s in sub-region 16 in the southwest. Sub- in the 2050s. The narrow distribution of flood region 15 also loses a substantial portion of agri- damages projected by different GCMs suggests a cultural production, while more modest losses robust change, although changes are small com- are simulated in sub-regions 10 and 12. Even pared to the year-to-year variability in each time without considering the effects of increased fre- period. These results are likely to be optimistic, quency of cyclones and increasing groundwater as changes in inter-annual variability between salinity, these simulations still project consider- the baseline and future time periods are likely to able production losses. produce larger flood damages and several sub- regions were not modelled. 5.7 Projections of Integrated Damages 5.6 Projections of Potential Coastal Inundation Damages National production changes The percentage of production lost to coastal The combined impacts of climate change on inundation associated with sea level rise in each potential national cereal production are pre- sub-region is presented in Figure 5.4. Coastal sented in Figure 5.5, including the effects of CO2 54 Climate Change Risks and Food Security in Bangladesh SR-09 SR-10 SR-11 SR-12 SR-13 SR-14 SR-15 SR-16 nario where CO2 levels are higher. The added 0 effects of basin and coastal flooding result in –5 production losses in national aus. This is despite the CO2 enhancement slightly exceeding the Percentage production change –10 –15 damage caused by temperature and precipita- tion changes (see Figure 5.2). Losses reach 3.5 –20 per cent in the 2050s A2, although the range in –25 GCM projections covers both positive gains and –30 2030s (15 cm) losses of nearly 10 per cent. In all, the aus produc- –35 2050s (27 cm) tion changes do not separate themselves convinc- 2080s (62 cm) ingly from zero, suggesting that the aus crop, on –40 balance, will not be strongly affected by climate Figure 5.4 Percentage of production lost to coastal inundation change up to the 2050s. The role of aus in total associated with sea level rise in each coastal region sub-region production is expected to decline over time. (9–16) for three future scenarios, as compared to the baseline period (for A2 SRES) Aman production Note: Depths in the legend refer to the mean sea level rise associated with Aman production is substantially impacted by each future scenario. basin and coastal flood effects where projected yield changes are overwhelmingly negative by the 2050s. The tight distribution between GCM enhancement, temperature and precipitation projections allows a more deï¬?nitive assessment changes, river basin flooding and coastal flood- that losses are expected, but the magnitude of ing. The median of all rice crop projections show the median change is only losses of 0.4 per cent declining national potential production in future in the 2030s and 0.6 per cent in the 2050s. The decades, with boro production showing the larg- largest median decrease is -1.5 per cent projected est median losses.Wheat production increases out for the 2050s A2 scenario. Thus, aman produc- to the 2050s period. Median value production tion will also not experience strong effects due to losses are given in Table 5.9. climate change. Note that this only reflects mean changes in flood impacts and not changes in the Aus production future frequency of extreme events and inter- Changes in aus production are mostly negative, annual variability. but median losses are only 1.5 per cent of the baseline potential production by the 2050s for Boro production all scenarios. Signiï¬?cant projections, however, Projections for boro production are entirely neg- indicate anywhere from 9 per cent losses to 3.5 ative by the 2050s, with substantial losses likely. per cent gains. A substantial number of GCMs Boro is not affected by river floods, and most project slight increases, especially in the A2 sce- production occurs away from the coastal sub- regions; thus, integrated damages are similar to Table 5.9 Median integrated production change (per cent) for the climate-only production estimates (Figure the 2030s and 2050s 5.2). Boro production declines with time, with 2030s 2050s 2030s 2050s 2030s 2050s median losses across all emissions scenarios reach- All SRES All SRES A2 A2 B1 B1 ing 3 per cent by the 2030s and 5 per cent by Aus –0.27 –1.52 –1.11 –3.51 –0.14 –0.01 the 2050s, although some GCMs project losses Aman –0.37 –0.62 –0.42 –1.49 –0.37 –0.40 of greater than 10 per cent. The robust loss pro- Boro –3.06 –4.74 –1.68 –5.54 –3.76 –3.54 jections suggest that boro is the major Bangla- Wheat –2.05 –3.44 –2.23 –3.74 –1.33 –3.03 deshi crop that is most at-risk to climate change Note: All SRES refers to both the A2 and B1 emissions scenarios impacts. Future Crop Performance 55 National Aus Production Changes – All Impacts 10 0 % –10 ALL 2030s ALL 2050s A2 2030s A2 2050s B1 2030s B1 2050s National Aman Production Changes – All Impacts 10 0 % –10 ALL 2030s ALL 2050s A2 2030s A2 2050s B1 2030s B1 2050s National Boro Production Changes – All Impacts 10 0 % –10 ALL 2030s ALL 2050s A2 2030s A2 2050s B1 2030s B1 2050s National Wheat Production Changes – All Impacts 10 0 % –10 ALL 2030s ALL 2050s A2 2030s A2 2050s B1 2030s B1 2050s Figure 5.5 Percentage change (versus the baseline flood-affected simulation) in national potential production with the combined effects of CO2, temperature and precipitation changes, basin flooding and coastal flooding Note: Each panel has three sections, each containing two future time periods and presenting (from left to right) the combination of all emissions scenarios, the A2 scenario and the B1 scenario. The distribution of potential yields projected by GCMs are presented as a box and whiskers diagram, consisting of a line representing the median value, a box enclosing the inter-quartile range (the middle 50 per cent of models), dashed whiskers extending to the furthest model that lies within 1.5 times the inter- quartile range from the edges of the box, and plus symbols for additional models that are perceived as outliers. 56 Climate Change Risks and Food Security in Bangladesh Wheat production production is lower. For the boro crop, all but Wheat production increases under climate one sub-region, Sylhet (SR-7), show decreases in change, reaching nearly 4 per cent in the 2050s yields (primarily due to CO2 fertilization effects). A2. Some GCMs project gains as high as 7 The largest decrease in yield from basin flood- per cent in that period, but some B1 scenarios ing is observed in sub-region 7 (approximately project production losses when CO2 levels are 9 per cent for both the aman and aus crop). For not as high. Wheat does not experience river coastal flooding, the largest decreases in yield flooding and most production occurs away from are observed in SR-12, SR-15 and SR-16 (7 coastal regions affected by sea level rise, so inte- per cent, 9 per cent and 13 per cent reductions grated estimates are very similar to the undam- respectively). The yield impacts of temperature aged production estimates above (Figure 5.2). In and precipitation changes vary by sub-region all, wheat production is projected to be positively depending on the crop. The largest declines in affected by climate change out to the 2050s, but aman, aus, boro and wheat are in SR-3 (4 per strong temperature effects (represented as upside cent), SR-4 (13 per cent), SR-8 (16 per cent) and triangles in Figure 5.2d) and uncertain beneï¬?ts of SR-1 (19 per cent) respectively. enhanced CO2 concentrations suggest that these wheat gains may be overly optimistic. These sim- ulations also indicate that wheat production may 5.8 Using the Crop Model to decline rapidly as temperature changes pass key Simulate Adaptation Options thresholds. Vulnerabilities to and potential adaptation strat- egies for climate change have been identiï¬?ed Sub-regional production changes in Bangladesh for agriculture and many other Figure 5.6 and Table 5.10 show the production sectors (Karim et al, 1994; Huq et al, 1999; Ali, changes by crop, by climate impact factor and 1999; NAPA, 2005; Ahmed, 2006; Tanner et al, by sub-region for the 2050s for a combined A2 2007). The IPCC also devoted chapters to glo- and B1 scenarios. Production losses compared to bal climate change impacts on the agricultural the baseline are estimated for at least one crop sector (Easterling et al, 2007) and coastal regions in each sub-region. For sub-regions including (Nicholls et al, 2007) as well as adaptation options Maijdee Court (SR-12), Patuakhali (SR-15) and (Adger et al, 2007). Thomalla et al (2005) detail Khulna (SR-16), the yields for all four crops are some of the historical efforts to implement wide- reduced. These southern areas are the most vul- spread adaptation strategies in Bangladesh utiliz- nerable to climate change due primarily to both ing local, regional, governmental and non-gov- sea level rise and riverine flooding. The larg- ernmental entities. Much of the international est observed decreases in yields are in Khulna effort for adaptation has focused on reducing the (SR-16) (approximately 10 per cent reduction threats of floods and tropical cyclones follow- in aman, aus and wheat yields, and 18 per cent ing devastating events in the 1980s and 1990s. reduction in boro yields). For ï¬?ve sub-regions Some agricultural adaptation is already being – Tangail (SR-4), Sylhet (SR-7), Maijdee Court tested in the ï¬?eld, e.g. dried hyacinth used as a (SR-12), Patuakhali (SR-15) and Khulna (SR- basis for floating agriculture during flooding in 16) – both the aman and aus crops demonstrate Barisal and salinity-tolerant cultivars introduced negative changes in yields. That is, any poten- to adapt to salinity intrusion (Sarwar, 2005). But tial gains from CO2 effects on the aus and aman widespread efforts in the agricultural sector are will be more than offset by the negative impacts not yet prominent. Moreover, the current large of temperature and precipitation changes, and gap between actual and potential yields suggests inland and coastal flooding. For the wheat crop, substantial on-farm opportunities for growth and only ï¬?ve sub-regions show decreases in yields. poverty reduction. Expanded availability of mod- These are concentrated in the south where wheat ern rice varieties, irrigation facilities, fertilizer use Future Crop Performance 57 Figure 5.6 Regional median potential production changes from baseline (per cent) for 2050s (a) aman, (b) aus, (c) boro and (d) wheat Note: Numbers refer to sub-region. Table 5.10 Sub-regional average production changes (per cent) disaggregated by crop (aman, aus, boro, wheat) and climate risk for 2050s – all scenarios Aman Aus Boro1 Wheat1 SR CO2 T&P FLD SLR2 COM CO2 T&P FLD SLR COM CO2 T&P FLD SLR COM CO2 T&P FLD SLR COM 1 2.72 –1.46 –1.57 0 3.5 8.37 –8.89 –1.12 0 1.47 4.66 –12.43 0 0 –5.43 14.04 –19.26 0 0 3.96 2 4.11 –0.64 –2.69 0 2.71 8.31 –7.4 –1.95 0 –0.75 5.16 –12.57 0 0 –6.81 9.69 –13.24 0 0 3.88 3 5.83 –4.43 –2.8 0 –0.78 11.06 –8.23 –1.91 0 2.93 5.91 –9.4 0 0 –2.78 3.25 0.15 0 0 4.76 4 5.2 –2.21 –8.24 0 –4.57 9.54 –11.74 –2.05 0 –0.55 8.22 –8.75 0 0 –1.49 4.48 –6.44 0 0 1.91 5 3.7 –0.16 –4.83 0 0.5 8.37 –7.7 –3.89 0 –0.12 5.73 –8.96 0 0 –4.83 11.78 –12.23 0 0 5.2 6 4.8 –1.42 –5.38 0 –2.37 12.35 –9.55 –2.28 0 0.68 9.18 –10.37 0 0 –4.32 16.59 –16.15 0 0 7.61 7 4.59 –1.22 –9.16 0 –8.31 8.39 –11.23 –8.94 0 –10.97 10.25 –4.88 0 0 4.83 9.98 –10.59 0 0 4.11 8 5.63 –2.51 –0.31 0 2.68 8.56 –8.73 –0.6 0 –1.07 5.88 –15.57 0 0 –9.2 9.87 –9.02 0 0 4.73 9* 4.38 –1.66 0 –0.41 3.39 8.14 –9.96 0 –0.41 –0.3 9.31 –12.74 0 –0.41 –7.08 4.49 –6.49 0 –0.41 –0.94 10* 3.24 –0.57 0 –3.56 1.68 8.32 –7.92 0 –3.56 –1.22 6.97 –10.46 0 –3.56 –7.38 9.13 –8.65 0 –3.56 0.61 11* 5.06 –2.17 0 –0.02 3.49 9.76 –9.5 0 –0.02 1.67 8.23 –11.59 0 –0.02 –5.41 17.28 –17.34 0 –0.02 5.14 12* 3.44 0.78 0 –7.09 –0.66 8.53 –12.48 0 –7.09 –5.09 8 –9.21 0 –7.09 –10.49 6.38 –6.6 0 –7.09 –6.68 13 6.69 –1.82 –1.25 –0.03 2.79 11.99 –10.36 –0.04 –0.03 3.44 6.63 –12.27 0 –0.03 –5.85 4.37 –4.46 0 –0.03 2.19 14 4.56 –2.62 –5.17 –0.12 –1.83 8.43 –6.58 –0.75 –0.12 2.43 7.44 –11.39 0 –0.12 –5.98 3.19 –1.85 0 –0.12 3.64 15 4.19 –2.51 –3.67 –8.73 –9.64 15.58 –7.09 –0.04 –8.73 –3.05 14.37 –13.09 0 –8.73 –12.05 27.97 –17.51 0 –8.73 0.96 16* 4.61 –3.4 0 –13.26 –10.76 14.3 –9.71 0 –13.26 –10.5 9.07 –14.3 0 –13.26 –20.79 14.26 –14.76 0 –13.26 –12.56 Notes: CO2 = Carbon dioxide enhancements are estimated from the CO2 impact experiments and differ from the baseline period only in their atmospheric CO2 concentration; T&P = Temperature and precipitation effects are estimated from the range of scenarios simulated by the temperature and precipitation impact experiments and differ from the baseline period only in their temperature and precipitation series; FLD = Basin flood effects are estimated from changing flood magnitudes and extent without regard for direct climate effects (CO2, temperature, or precipitation); SLR = Sea level rise damages are estimated from the changing coastlines projected as mean sea levels rise of the A2 scenario, with no other factors differing from the baseline period; COM = Combined effects include all climate risks described above. *Sub-regions 9–12 and 16 were not modelled in the future flood analysis. **Sea level rise only impacts sub-regions 9–16. ***Boro and wheat crop is assumed not impacted by the flood. Future Crop Performance 59 and labour could increase average yields at rates tion; adjustments in planting and harvesting dates that dwarf the climate change impacts. and crop sequence; changes in planting systems, An important follow-on to the present study including crop density; and input adjustments is the use of the crop simulation approach devel- including irrigation, fertilizer and farm-level oped here to evaluate adaptive responses them- environmental modiï¬?cations. Studying these selves. There are many elements of cropping elements will enable the development of addi- for which adaptations can be studied; these are tional adaptation options such as those detailed described in detail in Annex 1. The elements in Chapter 7. for potential evaluation include: cultivar selec- 6 Economy-wide Impacts of Climate Risks Box 6.1 Key messages • Economy-wide adjustments will to some extent mitigate the physical losses predicted. • Existing climate variability has a pronounced detrimental economy-wide impact. Compared to an ‘optimal’ climate simulation with the highest simulated yield and no inter-annual variations, climate variability is estimated to reduce long-term rice production by an average 7.4 per cent each year during 2005–50, primarily by lowering production of the aman and aus crops. Average annual rice production growth is lowered in all sub-regions. • Simulated climate variability is projected to cost the agriculture sector (in discounted terms) US$26 billion in agricultural GDP during the 2005–50 simulation period (US$0.57 billion per year in 2005 US$) compared with optimal production and growth. This climate variability has economy-wide impli- cations beyond simply the size-effect of the lost agricultural GDP. Existing climate variability is esti- mated to cost Bangladesh US$121 billion in lost national GDP during the same period (US$3 billion per year). • Climate change exacerbates the negative impacts of existing climate variability for food security by further reducing rice production by a projected cumulative total of 80Mt over the 2005–50 simulation period (3.9 per cent each year), driven primarily by reduced boro crop production. This is equivalent to almost two years’ worth of rice production lost over the next 45 years as a result of climate change. Uncertainty about future climate change means that annual rice production losses range between 3.6 per cent and 4.3 per cent. • Climate change primarily impacts boro rice and thus limits its ability to compensate for lost aus and aman rice production during extreme events. • Agricultural GDP is projected to be 3.1 per cent lower each year as a result of climate change (US$8 billion in lost value-added in 2005 US$). Average loss in agricultural GDP due to climate change is estimated to be a third of the agricultural GDP losses associated with existing climate variability. This is projected to cost Bangladesh US$26 billion in total GDP over the 2005–50 period. This is equiva- lent to US$570 million overall lost each year to climate change, or alternatively an average annual 1.15 per cent reduction in total GDP by 2050. • Uncertainty surrounding GCMs and emission scenarios means that costs may be as high as US$1 billion per year over 2005–50 under less optimistic scenarios. • These climate risks will have severe implications for household welfare. For both the climate variability and climate change simulations, around 80 per cent of these losses fall directly on household con- sumption (cumulative total consumption losses of US$441.7 billion and US$104.7 billion for climate variability and climate change respectively). • About 80 per cent of the projected economic losses from existing climate variability and climate change occur outside of the agriculture sector (from a national accounts perspective), particularly in the upstream and downstream agriculture value-added processing sectors. This means that both rural and urban households may be adversely affected. Economy-wide Impacts of Climate Risks 61 Chapter 5 provides estimates of changes in pro- major floods in Bangladesh will become more duction for four different crops (aus, aman, boro frequent, thus exacerbating economic losses and wheat) due to various climate risks. Though through heightened climate variability. Finally, informative, this information should be supple- though sea level rise impacts on yields were mented by economic responses to these produc- determined earlier (see section 5.6) changes in tion shocks (e.g. land and labour reallocation, cultivable land are explicitly incorporated in the price effects). These economic adjustments will CGE. The CGE analysis, therefore, builds on the to some degree mitigate the physical losses pre- hydro-crop modelling analysis by incorporat- dicted. What is described in this section is the ing the predicted crop yield changes, while also development and use of a dynamic computable extending the analysis by including the impact general equilibrium model to assess the econ- and frequency of extreme climate events. omy-wide impacts from these projected losses. The focus here is on rice production economic Climate simulations impacts only since this dominates agricultural and household food consumption. Three sets of simulations are run using the CGE model. 6.1 Integrating Climate Effects in an Optimal Climate Economy-wide Model The ï¬?rst simulation is the ‘Optimal Climate Simulation’, in which Bangladesh is unaffected Conceptual framework of the by existing climate variability or future climate methodology change. This means that the highest simulated crops yields are used and sector productivity and A dynamic computable general equilibrium factor supplies increase smoothly at long-term (CGE) model was developed to estimate the growth rates with no inter-annual variations. impacts of existing climate variability and future ‘Optimal’ is deï¬?ned as the best simulated crop climate change in Bangladesh on the agriculture yield achieved in each sub-region during the sector.The yield change estimates from the hydro- 30-year baseline period 1970–99. This scenario crop models described earlier are passed down to reflects a hypothetical trajectory for Bangladesh the CGE model to estimate their economy-wide in which there are no yield losses caused by cli- implications, including changes in production mate variability. This simulation provides a hypo- and household consumption for different sectors, thetical baseline scenario against which other household groups and sub-regions in the coun- climate-affected simulations can be compared. try. A detailed description of the model is pro- Since climate conditions are always assumed to vided in Annex 2. be ‘optimal’ (i.e. no crop yield losses or major Three impact channels, apart from crop yield floods), it is not necessary to run multiple base- changes estimated from the hydro-crop models, line simulations to account for climate uncer- are captured in the CGE model. First, the CGE tainty (i.e. there is only one optimal scenario). model includes the additional impacts that occur under extreme climatic events, such as during the major floods of 1988 and 1998. These comprise, Existing Variability for example, major losses of cultivatable land due The second set of simulations is the ‘Existing to floodwater inundation, which occurs over and Variability Simulation’.These simulations include above the average flood yield losses described in the yield losses associated with the historical cli- previous sections. The second additional impact mate data series. The CGE model is run forward channel considered in the CGE analysis is the over 45 years (2005–50) and for each year a change in the frequency of these extreme events random observation is drawn from the histori- caused by climate change. It is expected that cal data series (1970–99). The crop yield changes 62 Climate Change Risks and Food Security in Bangladesh estimated by the hydro-crop models (relative These three sets of simulations can be used to to the potential yields) for that particular base- decompose the impacts of existing climate vari- line historical year are imposed on the CGE ability and future climate change. The difference model, which then estimates the economy-wide between the results from the CGE model for the impacts. Moreover, years in the historical record Existing Variability Simulation and the Optimal during which major floods took place (i.e. 1970, Climate Simulation is the estimated economic 1974, 1984, 1987, 1988 and 1998) are identiï¬?ed. impact of existing climate variability. Similarly, If one of these years is drawn during the random the difference between the results for the Existing selection, then additional impacts are imposed Variability and Climate Change simulations is the on the CGE model (discussed later in this sec- estimated economic impact of climate change. tion). Together these random selections from the historical climate data produce a single 45- Bangladesh CGE model year climate scenario based on existing climate variability and patterns (i.e. without the effects A CGE model is a representation of the struc- of climate change). This Monte Carlo process is ture and workings of an economy. CGE models repeated 50 times in order to produce a series of are often called ‘economy-wide’ models because 45-year climate scenarios.The average economy- they include all sectors and households as well as wide outcomes for these 50 simulations provide a country’s government and its interactions with an estimate of the economic impact of existing the rest of the world (i.e. imports and exports). climate variability since all Monte Carlo runs are They are also called ‘macro-micro’ models because considered equally likely. they estimate how changes in macro-level condi- tions, such as the external shocks caused by cli- mate variation, influence micro-level outcomes, Climate Change including sector production and household The third set of simulations is the ‘Climate incomes and spending.This macro-micro linkage Change Simulation’. As described earlier, the is achieved by simulating the functioning of fac- hydro-crop models estimate crop yield changes tor and commodity markets, and thus captures for future 30-year time slices around the 2030s how changes in economic conditions are medi- and 2050s based on a range of model experi- ated through prices. Economic decision-making ments. As the CGE model is run forward over in CGE models is the outcome of decentralized the 2005–50 period, yield impacts are drawn ini- optimization by producers and consumers within tially from the historical series and then gradually a coherent economy-wide framework. The out- from the future series. For example, there are 30 comes of a CGE model are therefore deter- years separating the 2005 base year of the CGE mined by the structure of the economy and by model and the mid-point of the 2030s time slice. the behavioural assumptions. This section briefly Thus, in 2010, which is year 5, 25/30 of the yield describes the main characteristics of the Bang- impact from the randomly selected year in the ladesh model and the way in which the results historical dataset and 5/30 of the yield impact from earlier sections are incorporated within this from the same year in the 2030s time slice are analytical framework to assess climate variability used1. A similar linear transition from the 2030s and change. to the 2050s time-series is used. As with the In order to capture the heterogeneity of pro- Existing Variability Simulation, this Monte Carlo ducers and households, the Bangladesh CGE process is repeated 50 times and the average is model is based on a highly disaggregated 2005 taken to provide an overall estimate of economic social accounting matrix (SAM).2 The model dis- outcomes under climate change. Finally, this tinguishes between 36 productive activities/com- whole process is repeated for the two emissions modities (17 in agriculture, 14 in industry and 5 scenarios and ï¬?ve GCMs described earlier. in services). Agricultural production in each crop or sub-sector is further disaggregated across the Economy-wide Impacts of Climate Risks 63 16 sub-regions described in earlier sections. Each Private and public savings are pooled and used to of the 36 sectors in each of the 16 sub-regions ï¬?nance investment. is represented by a production function, which The CGE model is run over the simula- combines factor and intermediate inputs to pro- tion period 2005–50. During this time the duce a certain level of output.This output is sup- model’s parameters are updated based on long- plied to either domestic or foreign markets based term demographic trends and rates of techni- on relative prices. cal change. For example, population and labour In Bangladesh the size of agricultural land- supply growth is assumed to diminish over time holdings is an important determinant of the from 2 per cent per year in 2005. Agricultural activities and technologies that are available to land expansion also declines over time. Long-run farmers. Agricultural land in the model is thus growth in total factor productivity (TFP) starts disaggregated into marginal, small, medium and at 2 per cent per year and falls to 0.5 per cent by large-scale holdings based on the 2005 Agricul- 2050. However, land availability and technology tural Census (BBS, 2006a). Similarly, education outcomes vary from year to year depending on is important in determining employment oppor- climate conditions. In the CGE model, climate tunities for workers. The CGE model therefore variability and future change affect the growth separates workers into four education-based cat- and development of Bangladesh through three egories taken from the 2005 Household Income primary mechanisms: and Expenditure Survey (HIES) (BBS, 2006b). Labour in each category is assumed to be fully 1 Crop yield changes. The impact of climate mobile across sectors and regions. A flexible wage variables on agricultural productivity are then adjusts to ensure total labour demand equals obtained from the hydro-crop models, which supply. Agricultural land, by contrast, is region- estimate yield changes for different crops and speciï¬?c, but can be reallocated to different crops sub-regions (relative to a potential yield). and sub-sectors depending on their relative prof- Speciï¬?cally, the CGE model ï¬?rst determines itability. Finally, capital in the model is immobile how much land, labour, capital and interme- and earns region/sector-speciï¬?c returns. The diate inputs are allocated to a crop. This gives model’s detailed treatment of producers and fac- an estimated level of production under the tors allows it to capture Bangladesh’s unique pro- assumption of ‘optimal’ climatic conditions. duction structure and resource constraints, as well The hydro-crop models then determine as some of the ‘autonomous’ adaptation to cli- deviations from this level as a consequence of mate variation that is driven by economic forces realized climate. These short-term deviations (i.e. prices and proï¬?tability). are imposed on the technology parameters of The Bangladesh model separates households the production functions. Together the long- into 52 groups based on the region where they term resource allocations determined by the are situated; whether they are engaged in farming; CGE model and the short-term deviations the size of farmers’ land holdings; and, for non- in crop yields obtained from the hydro-crop farm households, their land-ownership status models determine the level of production in and the educational attainment of the household each sector and region during a particular head. Households in the model earn incomes year. from producers’ use of the factors of production. 2 Extreme events. Additional impacts occur These returns are paid to households based on during extreme climate events, such as the their factor endowments, which are drawn from major floods of 1988 and 1998. During major the 2005 household survey. Households use their flood years it is assumed that long-term rates incomes to pay taxes, save, and purchase domes- of land expansion and technology accumula- tic and imported goods in national commodity tion cease and there is a short-term decline in markets. Tax revenues are paid to the govern- available agricultural lands due to particularly ment where they are used for recurrent spending. severe and persistent flooding. These land 64 Climate Change Risks and Food Security in Bangladesh losses are based on historical production data. and hence savings. Reduced savings translate into While crop yields and agricultural lands may lower levels of investment, which in turn lower return to ‘normal’ after an extreme event, the potential future production. Extreme events, such loss in productive assets and forgone techni- as flooding, can destroy assets and infrastructure cal improvements will have lasting implica- in the period in which the event occurs and with tions in the CGE model. Climate change is lasting effects. Generally, even small differences in also predicted to increase the frequency of rates of accumulation can lead to large differences extreme events and this is also captured in the in economic outcomes over long time periods. model. The return periods for the 1988 and The CGE model used in this section is designed 1998 flood years are reduced by one-third. In to capture these accumulation effects. other words, 1988 and 1998 are character- ized as the 1/33 and 1/50 year floods respec- Limitations of the CGE model tively (in relation to water discharges). The frequency of these floods in the sample for As with any economic modelling there is uncer- the random selection of years for the future tainty over the accuracy of underlying data and climate change sequences is increased to the values of behavioural parameters. For exam- 1/25 and 1/33 for the 1988 and 1998 floods ple, the social accounting matrix (SAM) to which respectively. the CGE model is calibrated captures current 3 Sea level rise. Certain parts of Bangladesh production technologies and linkages. While the are particularly vulnerable to rising sea levels, CGE model allows for some endogenous change including crop land salinization from tropical from existing technologies, it cannot predict the cyclones. This is captured in the CGE model emergence of entirely new technologies or eco- by permanently reducing the supply of cul- nomic sectors. Similarly, the model uses estimated tivable land in the affected sub-region. These elasticities for various functions, such as factor land losses are based on the results from the substitution possibilities in the production func- hydrological models described in earlier sec- tion, or the ease at which consumers can shift tions. For all climate change scenarios, the between domestic and foreign goods depending CGE model simulates a gradual 15cm sea on relative prices. Although the CGE model is level rise by the 2030s and a 27cm sea level based on the best available data on Bangladesh’s rise by the 2050s. economic structure and institutional behaviour, both of these characteristics could change sub- Climate change is projected to take place over stantially over the long time periods simulated the course of the next century.The analysis in this in this chapter. Thus, while the analysis in Chap- section only considers the implications of climate ter 6 provides the best estimate based on exist- change to 2050 even though climate change is ing knowledge on Bangladesh’s economy, some expected to be most severe towards the end of caution should be exercised when interpreting the century. Nevertheless the relatively long time- the absolute magnitudes of estimated economic frame considered (45 years into the future) means losses. that dynamic processes are important. Economic development is in many ways about the accumu- lation of factors of production such as physical 6.2 Economic Impacts of Existing capital, human capital and technology. These fac- Climate Variability tors, combined with the necessary institutional frameworks to make them productive, determine An optimal climate scenario without the material wellbeing of a country. The CGE model captures these dynamic processes. To the climate variability extent that climate change reduces agricultural Economic growth in the Optimal Climate Sce- output in a given year, it also reduces income nario (Table 6.1) is driven by assumptions about Economy-wide Impacts of Climate Risks 65 Table 6.1 Summary of the Optimal Climate Scenario Average annual growth rate (%) Share of total GDP (%) 2005–50 2005–25 2025–40 2040–50 2005 2025 2040 2050 Total GDP 4.65 4.47 4.80 4.78 100.00 100.00 100.00 100.00 Agriculture 3.46 3.50 3.47 3.39 20.17 16.71 13.79 12.06 All rice crops 3.03 3.21 2.97 2.75 6.61 5.18 3.98 3.27 Aus variety 3.07 3.24 3.02 2.80 0.38 0.30 0.23 0.19 Aman variety 3.02 3.20 2.97 2.75 2.63 2.06 1.58 1.30 Boro variety 3.03 3.21 2.97 2.74 3.60 2.82 2.17 1.78 Industry 5.15 4.88 5.40 5.29 29.34 31.72 34.57 36.27 Rice processing 2.94 3.17 2.87 2.59 1.99 1.55 1.17 0.95 Services 4.71 4.59 4.81 4.79 50.48 51.57 51.63 51.66 the accumulation of factors of production and from 20.17 per cent in 2005 to 12.06 per cent technical change with no inter-annual variations. by 2050. By contrast, industry’s share rises from A gradually declining rate of population growth 29.34 to 36.27 per cent during the same period. from around 2 per cent per year and a constant This declining role of agriculture has implications dependency ratio, such that the labour force for estimating the economic cost of climate vari- and the population grow at the same rate, are ability and change, since the sector is expected to assumed. However, the supply of higher skilled be the primary impact channel.Thus, any adverse labour grows faster than the supply of illiter- impacts to the agricultural sector will be offset by ate and unskilled workers, reflecting expected the sector’s declining importance in the overall improvements in the Bangladesh educational economy. system and changing labour demands over time. Land expansion rates are initially set at 1 per cent Production losses from existing variability per year, but declines to 0.5 per cent per year by 2050.This is below rural population growth, thus As described in section 6.1, the Existing Variabil- capturing existing and future increases in agri- ity Scenario imposes crop yield losses observed cultural land scarcity. Finally, it is assumed that during the 1970–99 baseline period. These yields total factor productivity growth rates are higher reflect ‘sub-optimal’ climate conditions (i.e. rain- in industry and services than in agriculture, with fall, temperature and flooding).The average of the the former set at 2.5 per cent per year and the lat- Monte Carlo economic outcomes is termed the ter at 2 per cent per year. Together these assump- Variability Scenario. Figure 6.1 shows the esti- tions determine the Optimal Climate Scenario mated losses in national rice production caused and provide a benchmark growth path against by existing climate variability. Under the Opti- which the economic losses from existing climate mal Climate Scenario, national rice production variability can be measured. grows at 3.03 per cent per year during 2005–50. Under the Optimal Climate Scenario, total In physical terms, rice production rises from GDP grows at an average rate of 4.65 per cent per 22.36Mt in 2005 (162kg per capita) to 85.56Mt year during 2005–50, with a slight acceleration by 2050 (255kg per capita). The impact of exist- in the average growth rate from the beginning to ing climate variability is a reduction in rice pro- the end of the period (see Table 6.1). As observed duction, with its average annual growth rate fall- in most countries’ development paths, economic ing from 3.03 to 2.71 per cent per year. Under growth is not evenly balanced across all sectors, the Variability Scenario, rice production rises with a declining contribution of agriculture to to 74.6Mt by 2050 (222kg per capita), which total GDP and a rising contribution from indus- is almost 11Mt (33kg per capita) below what try. Thus, agriculture’s share of total GDP falls would have been achieved without the adverse 66 Climate Change Risks and Food Security in Bangladesh 100,000 Average annual growth rates, 2005–50 Total national rice production quantity (1000mt) 90,000 Optimal scenario: 3.03% Variability scenario: 2.71% 80,000 Worst scenario: 2.67% 70,000 60,000 50,000 40,000 30,000 20,000 Optimal scenario 10,000 Variability scenario 0 2005 10 15 20 25 30 35 40 45 50 Figure 6.1 Losses in total national rice production due to existing climate variability, 2005–50 Table 6.2 National rice production losses due to existing climate variability, 2005–50 Average annual growth rate (%) Rice production quantities (1000 tonnes) 2005–50 2005–25 2025–40 2040–50 2005 2025 2040 2050 All rice crops Optimal Scenario 3.03 3.21 2.97 2.75 22,355 42,038 65,226 85,563 Variability Scenario 2.71 2.84 2.69 2.51 22,355 39,123 58,232 74,596 Worst Case Scenario 2.67 2.43 2.98 2.69 22,355 36,122 56,105 73,186 Aus rice Optimal Scenario 3.07 3.24 3.02 2.80 897 1697 2652 3497 Variability Scenario 2.46 2.49 2.47 2.37 897 1467 2115 2673 Worst Case Scenario 2.34 2.21 1.36 4.12 897 1388 1699 2544 Aman rice Optimal Scenario 3.02 3.20 2.97 2.75 11,687 21,950 34,042 44,668 Variability Scenario 2.41 2.53 2.39 2.22 11,687 19,262 27,443 34,196 Worst Case Scenario 2.27 1.81 2.74 2.51 11,687 16,725 25,079 32,125 Boro rice Optimal Scenario 3.03 3.21 2.97 2.74 9772 18,392 28,532 37,398 Variability Scenario 3.05 3.21 3.00 2.78 9772 18,393 28,674 37,728 Worst Case Scenario 3.09 3.10 3.30 2.76 9772 18,010 29,328 38,517 impacts of climate variability (see Table 6.2). This production averages 2.67 per cent growth per means that Bangladesh will lose on average 7.4 year. By 2050, this implies an additional loss of per cent of its optimal rice production each year 1.41Mt. if the existing climate variability patterns remain Changes in national rice production hide unchanged into the future. Moreover, this share differential impacts for speciï¬?c rice crops (see of lost production increases throughout the Figure 6.2). Under the Variability Scenario, most period, as the effects of climate variability are of the lost rice production is due to reduction compounded. in production for the aus and aman crops. These The Worst Case Scenario is deï¬?ned as the crops are adversely affected by yield declines from randomly drawn climate series resulting in the flooding during the wet season. For example, the largest overall economic losses for the country average annual growth rate for aman rice produc- as a whole.3 Under the Worst Case Scenario, rice tion is 3.02 per cent under the Optimal Climate Economy-wide Impacts of Climate Risks 67 4,000 Aus rice 3,500 Average annual growth rates, 2005–50 National Aus rice production quantity (1000mt) Optimal scenario: 3.07% 3,000 Variability scenario: 2.46% Worst scenario: 2.34% 2,500 2,000 1,500 1,000 Variability scenario 500 Optimal scenario 0 2005 10 15 20 25 30 35 40 45 50 60,000 Aman rice Average annual growth rates, 2005–50 National Aman rice production quantity (1000mt) 50,000 Optimal scenario: 3.02% Variability scenario: 2.41% Worst scenario: 2.27% 40,000 30,000 20,000 10,000 Optimal scenario Variability scenario 0 2005 10 15 20 25 30 35 40 45 50 45,000 Boro rice 40,000 National Boro rice production quantity (1000mt) Average annual growth rates, 2005–50 Optimal scenario: 3.03% 35,000 Variability scenario: 3.05% Worst scenario: 3.09% 30,000 25,000 20,000 15,000 10,000 Optimal scenario 5,000 Variability scenario 0 2005 10 15 20 25 30 35 40 45 50 Figure 6.2 Losses in national rice production by crop due to existing climate variability, 2005–50, (a) aus, (b) aman, (c) boro 68 Climate Change Risks and Food Security in Bangladesh Scenario. This falls to 2.41 per cent under the ity. This role was also empirically observed in the Variability Scenario and to 2.34 per cent under historical production data (see Figure 2.2). the Worst Case Scenario. By 2050, this reduced growth means that aman production is 23 per cent Production losses across sub-regions below the production levels achieved without the The economy-wide model also captures rice pro- effects of climate variability (see Table 6.2). duction losses for the 16 different sub-regions. By contrast, the impact of existing climate var- Two factors determine the overall loss in rice pro- iability on boro rice production is negligible.This duction at the sub-region level. First, some regions is because the boro is an irrigated crop and largely face more severe climate variability causing pro- independent of the annual floods. Moreover, the duction of speciï¬?c crops to decline more than CGE model allows economic forces to shift farm- ers’ incentives away from producing those rice elsewhere in the country. Second, some regions crops that face the largest yields declines. More rely more heavily on crops that are severely speciï¬?cally, existing climate variability greatly affected by climate variability. This can be seen in reduces aus and aman production, which reduces Table 6.3, which shows changes in average annual overall rice supply in the country and causes the rice crop production growth from the Optimal average price of rice to rise. Farmers thus shift Climate Scenario during 2005–50. Figure 6.3 production towards the boro rice to take advan- shows the weighted contribution of each crop to tage of higher rice prices. These economic forces the overall change in regional rice production. coupled with smaller yield impacts encourage Mymensingh (6) and Tangail (4) in the cen- greater boro production. By 2050, boro rice pro- tral region are the worst affected sub-regions duction is 0.881Mt higher under the Variability since they face amongst the largest declines in Scenario than it was under the Optimal Climate aus and aman production due to climate vari- Scenario. These model results highlight the cru- ability effects, while also being regions that are cial compensating role that dry season boro rice most reliant on aus and aman for their overall rice plays in Bangladesh as a result of climate variabil- production. Table 6.3 Regional rice production losses due to existing climate variability, 2005–50 Deviation in average annual production growth Share of rice crop in total regional rice production (%) rate from optimal scenario (%-point) Aus Aman Boro All crops Aus Aman Boro All crops National –0.61 –0.61 –0.02 –0.31 4.01 52.28 43.71 100.00 Dinajpur (SR-1) –0.57 –0.57 –0.04 –0.31 5.21 55.87 38.92 100.00 Rangpur (SR-2) –0.68 –0.52 –0.07 –0.27 5.67 54.04 40.29 100.00 Ishwardi (SR-3) –0.56 –0.50 –0.00 –0.29 5.48 53.84 40.67 100.00 Tangail (SR-4) –0.66 –0.77 –0.01 –0.42 4.57 54.14 41.30 100.00 Dhaka (SR-5) –0.70 –0.63 –0.02 –0.34 2.82 56.18 41.00 100.00 Mymensingh (SR-6) –0.77 –0.76 –0.01 –0.40 2.71 55.51 41.79 100.00 Sylhet (SR-7) –0.52 –0.53 –0.03 –0.28 1.39 56.85 41.76 100.00 Srimangal (SR-8) –0.65 –0.66 –0.05 –0.37 1.92 52.98 45.10 100.00 Comilla (SR-9) –0.62 –0.66 –0.04 –0.32 2.78 52.86 44.36 100.00 Chittagong (SR-10) –0.58 –0.61 –0.01 –0.32 4.06 52.49 43.44 100.00 Rangamati (SR-11) –0.66 –0.64 –0.02 –0.33 4.03 51.58 44.39 100.00 Maijdee Court (SR-12) –0.57 –0.51 –0.05 –0.23 4.19 49.09 46.71 100.00 Jessore (SR-13) –0.60 –0.55 –0.03 –0.26 4.83 46.60 48.58 100.00 Faridpur (SR-14) –0.53 –0.62 –0.01 –0.27 4.71 43.67 51.61 100.00 Patuakhali (SR-15) –0.66 –0.64 –0.07 –0.27 6.18 44.91 48.90 100.00 Khulna (SR-16) –0.54 –0.65 –0.00 –0.30 5.37 44.96 49.66 100.00 Economy-wide Impacts of Climate Risks 69 Majdee Court Mymensingh Chittagong Rangamati Patuakhali Srimangal Rangpur Dinajpur Ishwardi Faridpur National 5 Jessore Comilla Tangail Khulna Dhaka Sylhet Contribution to the total change in rice production 0 (percentage point) –5 –10 –15 Boro contribution Amam contribution Aus contribution –20 Total growth rate Figure 6.3 Decomposing regional rice production losses due to existing climate variability, 2005–50 Note: Total percentage changes in rice production are weighted by each rice crop’s contribution to total regional rice production (see Table 6.3). 60 Agricultural GDP Average annual growth rates, 2005–50 50 Optimal scenario: 3.46% Agricultural GDP (Billions 205 $US) Existing variability scenario: 3.13% Worst scenario: 3.01% 40 Cumulative losses, 2005–50 Existing variability scenario: US$25.78 bil. Worst scenario: US$40.49 bil. 30 20 Optimal scenario 10 Variability scenario 0 2005 10 15 20 25 30 35 40 45 50 Figure 6.4 Losses in national agricultural GDP due to existing climate variability, 2005–50 70 Climate Change Risks and Food Security in Bangladesh Table 6.4 Losses in GDP due to existing climate variability, 2005–50 Agricultural GDP Total GDP 2005–50 2005–25 2025–40 2040–50 2005–50 2005–25 2025–40 2040–50 Average annual growth rate (%) Optimal Scenario 3.46 3.50 3.47 3.39 4.65 4.47 4.80 4.78 Variability Scenario 3.13 3.11 3.13 3.19 4.44 4.20 4.59 4.69 Worst Case Scenario 3.01 2.78 3.12 3.29 4.36 4.01 4.47 4.91 Cumulative economic loss (2005 US$ billion) Variability Scenario 120.96 14.78 45.80 60.38 594.06 61.81 213.64 318.61 Worst Case Scenario 189.24 20.76 81.27 87.20 929.98 77.35 397.88 454.76 Discounted cumulative economic loss (2005 US$ billion) Variability Scenario 25.78 7.17 10.71 7.90 120.66 29.59 49.53 41.54 Worst Case Scenario 40.49 10.09 18.94 11.46 187.74 36.14 91.96 59.64 Average annual discounted economic loss (2005 US$ billion) Variability Scenario 0.57 0.36 0.71 0.79 2.68 1.48 3.30 4.15 Worst Case Scenario 0.90 0.50 1.26 1.15 4.17 1.81 6.13 5.96 Discounted economic loss average share of total optimal GDP (%) Variability Scenario 1.10 0.64 1.43 1.62 5.14 2.66 6.61 8.51 Worst Case Scenario 1.72 0.91 2.53 2.35 7.99 3.25 12.28 12.21 Agricultural GDP impacts from existing 2005–50, or an annual loss of US$0.57 billion. variability This means that 1.10 per cent of agricultural GDP is lost on average each year as a result of Rice production accounted for about one-third existing climate variability. However, this average of total agricultural GDP in Bangladesh in 2005. hides compounding economic losses over time. Reductions in rice production will therefore Economic losses resulting from existing climate have a signiï¬?cant impact on overall value-added variability average 0.64 per cent of agricultural results in the sector. Model results estimate that GDP during 2005–25, rising to 1.62 per cent the agricultural GDP growth rate will decline during 2040–50. from 3.46 per cent per year during 2005–50 The agricultural GDP growth rate falls even under the Optimal Scenario to 3.13 per cent per further under the Worst Case Scenario to 3.01 year under the Variability Scenario (see Figure per cent per year, implying the climate variabil- 6.4). This drop in the growth rate causes sub- ity reduces agricultural GDP by almost 0.5 per stantial economic losses over the 45-year period cent each year during 2005–50. The discounted 2005–50. For example, existing climate vari- cumulative loss in agricultural GDP under this ability results in a loss of US$120.96 billion in scenario reaches US$40.49 billion or an annual agricultural GDP during 2005–50 (measured in average loss of US$0.9 billion (both measured in 2005 prices), which is an average economic loss 2005 prices). This is equivalent to 1.72 per cent of US$2.63 billion per year (see Table 6.4).4 If we of agricultural GDP lost each year. Existing cli- discount future economic losses at 5 per cent per mate variability will therefore have a profoundly year,5 then the total loss in agricultural GDP due negative impact on the future growth of Bangla- to climate variability is US$25.78 billion during desh’s agricultural GDP. Economy-wide Impacts of Climate Risks 71 500 Total GDP 450 Average annual growth rates, 2005–50 Optimal scenario: 4.65% Total GDP (Billions 2005 $US) 400 Variability scenario: 4.44% Worst scenario: 4.36% 350 Discounted cumulative losses, 2005–50 Existing variability scenario: US$120.66 bil. 300 Worst scenario: US$187.74 bil. 250 200 150 Optimal scenario 100 Variability scenario 50 0 2005 10 15 20 25 30 35 40 45 50 Figure 6.5 Losses in national total GDP due to existing climate variability, 2005–50 National GDP impacts from existing Model results estimate that climate variability variability reduces total GDP annual average growth rate by 0.21 percentage points each year during 2005–50 Agriculture is a key sector in Bangladesh, account- (i.e. from 4.65 per cent per year under the Opti- ing for one-ï¬?fth of total GDP in 2005. However, mal Scenario to 4.44 per cent under the Variabil- the impact of climate variability on the agri- ity Scenario). Average GDP growth rates decline culture sector has economy-wide implications further under the Worst Case Scenario. Over the beyond simply the size-effect of the lost agri- 45-year 2005–50 period, climate variability will cultural GDP. For example, declining rice pro- cost Bangladesh US$594.06 billion in lost real duction causes a contraction of upstream rice- GDP at the national level, or an annual average milling industries, which lowers manufacturing decline of US$12.91 billion (both measured in GDP. Since much of the value-addition for rice 2005 prices). Again, if future losses are discounted production occurs during processing, a signiï¬?- at 5 per cent, then climate variability will generate cant share of these impacts occurs in manufac- a real economic loss of US$120.66 billion during turing rather than agriculture. Climate variabil- 2005–50, or an annual loss of US$2.68 billion. ity also has direct impacts on non-agricultural This substantial decline in national income is, on sectors through depreciated capital assets during average, equal to 5.14 per cent of the national major flood years. Falling farm incomes also GDP that could be achieved under optimal cli- reduce households’ demand for non-agricultural mate conditions (see Table 6.4). products and hence production in these sectors. Finally, by reducing overall economic growth, Household consumption impacts from these climate effects lower investment and capi- existing variability tal accumulation, which affects all sectors of the economy, especially those in more capital-inten- The impact of climate variability on household sive non-agriculture sectors. The impact of total per capita consumption is shown in Table 6.5. GDP is therefore expected to be signiï¬?cantly Climate variability reduces private consumption larger than the impact on agriculture alone. Fig- spending by a cumulative US$89.8 billion during ure 6.5 shows that this is true for Bangladesh. 2005–50 discounted at 5 per cent per year.6 This 72 Climate Change Risks and Food Security in Bangladesh Table 6.5 Losses in national households’ consumption spending due to existing climate variability, 2005–50 Average Deviation Discounted Average annual discounted consumption National Per capita annual from cumulative losses as a share of discounted average population consumption growth rate optimal losses, annual total consumption spending (%) share in in 2005 (%) (%-point) 2005–50 2005 (%) (2005 US$) (2005 US$ Optimal Variability 2005–50 2005–25 2025–40 2040–50 billion) scenario scenario All Households 2.37 –0.21 89.8 2.54 1.17 3.20 4.40 100.0 ,365 Agricultural Households 2.38 –0.20 56.1 2.51 1.17 3.16 4.33 72.4 ,320 Farm Households 2.44 –0.21 48.4 2.62 1.23 3.30 4.52 57.3 ,332 Marginal Farms 2.26 –0.17 7.4 2.12 0.98 2.68 3.68 20.3 ,181 Small-scale Farms 2.42 –0.21 25.0 2.63 1.23 3.32 4.55 28.7 ,340 Large-scale Farms 2.59 –0.23 16.1 2.91 1.38 3.66 4.99 8.2 ,675 Landless Workers 2.08 –0.17 7.7 1.98 0.90 2.51 3.46 15.2 ,275 Non-farm Households 2.35 –0.21 33.7 2.59 1.18 3.27 4.50 27.6 ,486 Low Education 2.23 –0.18 11.8 2.21 1.00 2.79 3.85 19.3 ,289 Some Education 2.23 –0.24 12.4 2.94 1.35 3.72 5.09 5.7 ,784 High Education 2.65 –0.22 9.6 2.75 1.26 3.47 4.79 2.7 1,277 Note: Marginal Farms are less than 0.5 acres; Small-scale Farms are between 0.5 and 2.5 acres; and Large-scale Farms are larger than 2.5 acres. Low Education households have household heads that are illiterate or have completed some primary schooling; Some Education households’ heads have completed primary schooling and some secondary schooling; and High Education households’ heads have completed secondary schooling. For more information see Annex 2. Source: Results from the Bangladesh CGE model. is an average annual reduction of US$2 billion amongst non-farm households it is the higher- or 2.54 per cent of consumption spending each educated households that are hurt the most by year. Economic losses compound themselves, climate variability, since these households earn a starting at 1.17 per cent per year in 2005–25 and greater share of the returns to economic growth rising to 4.4 per cent in 2040–50. The welfare and hence suffer more when the size of the econ- losses caused by existing climate variability there- omy contracts. However, it is lower-educated fore become more pronounced over time, thus non-farm households that are likely to be more underlining the importance of addressing climate vulnerable to small income changes than higher- variability in the near-term in order to avoid educated households. long-term welfare losses. Both farm and non-farm households expe- rience declining real per capita consumption 6.3 Additional Economic Impacts of compared to the Optimal Scenario (see column 2 in Table 6.5). Large-scale farm households are Climate Change the worst affected amongst households engaged The additional economic costs of climate change in agricultural production, due in part to their over and above the costs of existing climate vari- greater reliance on the returns from agricul- ability is estimated here. As described in section tural land and capital as sources of incomes. 6.1, the Climate Change Scenario imposes crop In contrast, marginal farmers and landless farm yield losses from the hydro-crop models. The workers rely more heavily on non-farm labour average of the Monte Carlo economic outcomes incomes, and are thus less adversely affected by is termed the Climate Change Scenario. To iso- existing climate variability. Marginal farmers late the economic impact of climate change we and landless farm workers do, however, have the compare the results of the Climate Change Sce- lowest average per capita incomes and are there- nario to the Variability Scenario described in fore likely to be more vulnerable to even small the previous section. To account for climate and changes in per capita consumption. Similarly, model uncertainty, the Monte Carlo process of Economy-wide Impacts of Climate Risks 73 90,000 All rice crops Total national rice production quantity (1000mt) 80,000 Average annual growth rates, 2005–50 Optimal scenario: 3.03% 70,000 Existing variability scenario: 2.71% Average climate change scenario: 2.55% 60,000 Cumulative production loss, 2005–50 Average climate change scenario: 80.4 mil. mt 50,000 40,000 30,000 20,000 Existing variability scenario Average climate change scenario 10,000 Optimal scenario 0 2005 10 15 20 25 30 35 40 45 50 Figure 6.6 Losses in total national rice production due to climate change, 2005–50 randomly constructing future climate patterns tive loss is 3.9 per cent of total rice production is repeated for ï¬?ve GCMs and two emissions each year during 2005–50 (i.e. relative to the rice scenarios. production achieved under the Variability Sce- Figure 6.6 shows total national rice produc- nario). This suggests that climate change will tion for the Variability Scenario and the average exacerbate food availability and security in Bang- outcomes for each of the GCMs and emissions ladesh over the coming decades. scenarios (light curves). In other words, they are Even though national rice production falls the average outcomes for each of these Climate under all GCMs and emissions scenarios there are Change Scenarios after conducting 50 Monte signiï¬?cant differences in outcomes across these Carlo simulations for each scenario. The darker Climate Change Scenarios (see Table 6.6). First, broken curve is then the simple average or arith- as expected, the A2 emissions scenarios lead to metic mean of all the various Climate Change larger rice production losses on average than do Scenarios. The ï¬?gure shows that national rice the B1 scenarios. However, this is not the case for production declines under all of the Climate two of the GCMs: MPI ECHAM5 and UKMO Change Scenarios and that the annual growth HADCM3. Second, some GCMs produce much rate is reduced from 2.71 per cent under the Vari- larger impacts than others. For example, the rice ability Scenario to 2.55 percent under the Aver- production losses from GFDL 2.1 under the less age Climate Change Scenario. This reduction in severe B1 emission scenario (i.e. 6.7Mt) is larger the annual rice production growth rate by 0.17 than the economic losses obtained for most of percentage points causes ï¬?nal year rice produc- the other GCMs even under the more severe A2 tion to be 5.243Mt below what would have been emission scenario. UKMO HADCM3 similarly achieved under existing variability and without produces larger impacts relative to the remaining the additional negative effects of climate change three GCMs considered in this analysis. There- (see Table 6.6). This is equivalent to an average fore, reduced national rice production by an 8kg per capita reduction in rice production (i.e. average 3.9 per cent per year hides considerable 4.9 per cent reduction of current per capita pro- model uncertainty. For example, just considering duction levels). Moreover, the average cumula- emission scenario uncertainty, the average annual 74 Climate Change Risks and Food Security in Bangladesh Table 6.6 National rice production losses due to climate change, 2005–50 Deviation in average annual growth rate from Deviation in ï¬?nal year production from Existing Existing Variability scenario, 2005–50 (%) Variability scenario, 2050 (1000 tonnes) Aus Aman Boro All rice Aus Aman Boro All rice Average scenario –0.10 –0.16 –0.18 –0.17 –114 –2270 –2862 –5246 A2 average –0.11 –0.17 –0.20 –0.18 –122 –2431 –3174 –5728 GFDL 2.1 –0.02 –0.22 –0.34 –0.27 –25 –3198 –5205 –8428 MIROC3.2 MEDRES –0.26 –0.20 –0.09 –0.15 –293 –2936 –1419 –4649 MPI ECHAM5 –0.08 –0.14 –0.06 –0.10 –95 –2074 –1056 –3224 NCAR CCSM3 –0.13 –0.15 –0.20 –0.17 –143 –2148 –3166 –5457 NCAR CCSM3 –0.05 –0.12 –0.33 –0.22 –55 –1800 –5026 –6881 B1 average –0.09 –0.14 –0.16 –0.15 –107 –2109 –2549 –4765 GFDL 2.1 –0.02 –0.14 –0.30 –0.22 –29 –2069 –4616 –6714 MIROC3.2 MEDRES –0.13 –0.15 –0.09 –0.12 –152 –2248 –1386 –3786 MPI ECHAM5 –0.15 –0.22 –0.03 –0.12 –175 –3219 –483 –3878 NCAR CCSM3 –0.08 –0.07 –0.18 –0.12 –90 –1010 –2799 –3899 NCAR CCSM3 –0.08 –0.14 –0.22 –0.18 –87 –1999 –3463 –5549 production loss ranges from 4.3 per cent for the Production losses across sub-regions more severe A2 scenarios to 3.6 per cent for Figure 6.8 shows the change in rice production the less severe B1 scenarios. Moreover, allowing in each of the 16 agro-climatic sub-regions. Dif- GCM uncertainty widens the range of rice pro- ferences in predicted climate changes and initial duction losses to between 2.0 and 6.5 per cent production patterns result in varying growth- (i.e. A2 emissions scenarios for MPI ECHAM5 effects at the regional level. The southern agro- and GFDL 2.1 respectively). climatic regions of Patuakhali (15) and Khulna Figure 6.7 and Table 6.6 show produc- (16) experience the largest decline in total rice tion losses associated with the three rice varie- production due to climate change. This is for ties. Boro rice is the most severely affected by three reasons. First, these two regions already climate change. Annual boro rice production experience signiï¬?cant declines in aus and aman growth rates fall from 3.05 per cent under the rice production due to climate variability, which Variability Scenario to 2.87 per cent under the now worsens under the Climate Change Sce- Average Climate Change Scenario. Over the 45- nario. Secondly, boro yields are severely affected year period 2005–50, cumulative losses in boro by the effect of climate change on mean rainfall, rice production equal 52.507Mt or an average temperature and CO2 levels. Finally, these two 1.166Mt per year. This loss in boro production is regions are the worst affected by rising sea lev- driven primarily by declining crop yields due to els, which permanently reduce cultivable land. climate change, rather than by the increase in the Overall rice production losses are therefore most frequency of major floods. Aus and aman produc- pronounced in these southern coastal regions. tion is also negatively affected by climate, albeit Moreover, this ranking of regions according to less severely and more as a result of the increased their vulnerability to climate change is consistent frequency of major floods. Climate change there- across the two emissions scenarios considered in fore has adverse implications for boro production this analysis (see Figure 6.9). and undermines its compensating role in offset- ting the aus and aman production losses caused Agricultural GDP impacts from climate by existing climate variability. Climate change will therefore exacerbate existing climate-related change food insecurity as well as Bangladesh’s vulner- Figure 6.10 shows the decline in agricultural ability to extreme climate events. GDP caused by climate change. Agriculture Economy-wide Impacts of Climate Risks 75 3,000 Aus rice Average annual growth rates, 2005–50 Existing variability scenario: 2.46% National Aus rice production quantity (1000mt) 2,500 Average climate change scenario: 2.36% Cumulative production loss, 2005–50 Average climate change scenario: 1,267 mil. mt 2,000 1,500 1,000 Existing variability scenario 500 Average climate change scenario 0 2005 10 15 20 25 30 35 40 45 50 40,000 Aman rice Average annual growth rates, 2005–50 35,500 National Aman rice production quantity (1000mt) Existing variability scenario: 2.41% Average climate change scenario: 2.26% 30,000 Cumulative production loss, 2005–50 Average climate change scenario: 26,270 mil. mt 25,000 20,000 15,000 10,000 Existing variability scenario 5,000 Average climate change scenario 0 2005 10 15 20 25 30 35 40 45 50 40,000 Boro rice Average annual growth rates, 2005–50 35,500 National Boro rice production quantity (1000mt) Existing variability scenario: 3.05% Average climate change scenario: 2.87% 30,000 Cumulative production loss, 2005–50 Average climate change scenario: 52,507 mil. mt 25,000 20,000 15,000 10,000 Existing variability scenario 5,000 Average climate change scenario 0 2005 10 15 20 25 30 35 40 45 50 Figure 6.7 Losses in national rice production by crop due to climate change, 2005–50 76 Deviation in average annual production growth rate from Deviation in production from Existing Variability Scenario (%) Existing Variability Scenario (percentage point) –25 –20 –15 –10 –5 0 5 –0.7 –0.6 –0.5 –0.4 –0.3 –0.2 –0.1 0.0 0.1 0.2 0.3 Scenario, 2050 National scenarios, 2050 National Dinajpur (1) Dinajpur (1) Rangpur (2) Rangpur (2) Ishwardi (3) Ishwardi (3) Tangail (4) Tangail (4) Dhaka (5) Dhaka (5) Average all scenarios Mymensingh (6) Mymensingh (6) Climate Change Risks and Food Security in Bangladesh Sylhet (7) Sylhet (7) Srimangal (8) Srimangal (8) Aus rice Average A2 scenarios Comilla (9) Comilla (9) Chittagong (10) Chittagong (10) Aman rice Rangamati (11) Rangamati (11) Maijdee Court (12) Average B1 scenarios Maijdee Court (12) Boro rice Jessore (13) Jessore (13) Faridpur (14) Faridpur (14) All rice crops Figure 6.9 Deviation in average ï¬?nal year rice production from the Existing Variability Scenario under different emissions Patuakhali (15) Patuakhali (15) Figure 6.8 Deviation in average ï¬?nal year rice production from the Existing Variability Scenario under the Average Climate Change Khulna (16) Khulna (16) Economy-wide Impacts of Climate Risks 77 60 Agricultural GDP Average annual growth rates, 2005–50 National agricultural GDP (2005 US$ bil.) 50 Existing variability scenario: 3.13% Average climate change scenario: 3.02% Additional cumulative economic loss, 2005–50 All climate change scenarios: US$36.02 bil. 40 A2 climate change scenarios: US$39.17 bil. B1 climate change scenarios: US$32.86 bil. 30 20 Existing variability scenario 10 Average climate change scenario 0 2005 10 15 20 25 30 35 40 45 50 Figure 6.10 Losses in national agricultural GDP due to climate change, 2005–50 growth rate declines from 3.13 per cent per year near-term will therefore reduce larger long-term under the Existing Variability Scenario to 3.02 economic costs. per cent under the Average Climate Change Sce- Model uncertainty implies that discounted nario. Cumulating these losses over the 45-year agricultural GDP losses may range from around period means that climate change costs Bangla- US$0.1 billion under the MPI ECHAM5 GCM desh’s agricultural sector a total of US$36 billion to US$0.29 billion per year under the GFDL in lost value-added during 2005–50. Discounting 2.1 GCM (see Table 6.8). On average the GCMs future losses at 5 per cent produces a present value indicate that agricultural GDP losses in the A2 of foregone real agricultural GDP of US$7.7 bil- emissions scenario will be almost 20 per cent lion, which is an average annual reduction in higher than in the B1 scenario. discounted agricultural GDP of 0.17 per cent during 2005–50 (see Table 6.7). Uncertainty National GDP impacts from climate regarding future emissions scenarios means that change the cumulative loss in agricultural GDP ranges from US$7.11 billion under the less severe B1 sce- Figure 6.11 shows the losses in national total nario to US$8.29 billion under the more severe GDP caused by climate change. The annual A2 scenario. Comparing Tables 6.7 and 6.4, the GDP growth rate declines by 0.06 per cent average loss in agricultural GDP due to climate per year over the 45-year period. This results change is a third of the agricultural GDP losses in a cumulative loss in total value-added of associated with existing climate variability. Cli- US$128.55 billion over 2005–50 (measured in mate change will thus substantially reduce agri- 2005 prices), which is 21 per cent of the losses cultural GDP beyond the losses already caused by already caused by existing variability. Discounted existing climate variability. These average losses economic losses are lower at US$25.73 billion. in agricultural GDP compound themselves over This is equivalent to an average drop in national time, starting at US$10.86 per capita in 2005–25 GDP of US$570 million per year or 1.15 per and rising to US$22.95 per capita in 2040–50. cent of total GDP compared to the Existing Vari- Reducing the impacts of climate change in the ability Scenario (see Table 6.7). This is the aver- 78 Climate Change Risks and Food Security in Bangladesh Table 6.7 Average GDP losses due to climate change, 2005–50 Agricultural GDP Total GDP 2005–50 2005–25 2025–40 2040–50 2005–50 2005–25 2025–40 2040–50 Average annual growth rate (%) Variability Scenario 3.13 3.11 3.13 3.19 4.44 4.20 4.59 4.69 Average Scenario 3.02 2.99 3.02 3.06 4.38 4.14 4.54 4.62 A2 average 3.01 2.98 3.01 3.05 4.37 4.13 4.52 4.63 B1 average 3.03 3.00 3.04 3.07 4.39 4.15 4.55 4.62 Cumulative economic loss (2005 US$ billion) Average Scenario 36.02 4.51 13.36 18.15 128.55 12.96 44.38 71.21 A2 average 39.17 4.70 14.45 20.02 146.79 14.33 50.74 81.73 B1 average 32.86 4.31 12.27 16.28 110.31 11.60 38.02 60.69 Discounted cumulative economic loss (2005 US$ billion) Average Scenario 7.70 2.22 3.12 2.36 25.73 6.33 10.24 9.17 A2 average 8.29 2.32 3.37 2.60 29.21 6.99 11.66 10.56 B1 average 7.11 2.13 2.88 2.11 22.26 5.66 8.82 7.78 Average annual discounted economic loss (2005 US$ billion) Average Scenario 0.17 0.11 0.21 0.24 0.57 0.32 0.68 0.92 A2 average 0.18 0.12 0.22 0.26 0.65 0.35 0.78 1.06 B1 average 0.16 0.11 0.19 0.21 0.49 0.28 0.59 0.78 End of period per capita discounted economic loss (2005 US$) Average Scenario 22.95 10.86 19.42 22.95 76.69 30.94 60.17 76.69 A2 average 24.71 11.34 20.67 24.71 87.04 34.20 67.76 87.04 B1 average 21.20 10.39 18.17 21.20 66.34 27.68 52.59 66.34 Table 6.8 GDP losses under different climate change scenarios, 2005–50 Discounted average annual GDP losses, 2005–50 (2005 US$ bil.) Agricultural GDP Total GDP 2005–50 2005–25 2025–40 2040–50 2005–50 2005–25 2025–40 2040–50 Average Scenario 0.17 0.11 0.21 0.24 0.57 0.32 0.68 0.92 A2 average 0.18 0.12 0.22 0.26 0.65 0.35 0.78 1.06 GFDL 2.1 0.29 0.19 0.35 0.39 1.02 0.59 1.20 1.63 MIROC3.2 MEDRES 0.15 0.09 0.18 0.22 0.58 0.31 0.69 0.96 MPI ECHAM5 0.10 0.06 0.12 0.16 0.45 0.23 0.55 0.74 NCAR CCSM3 0.19 0.13 0.24 0.26 0.67 0.37 0.82 1.04 UKMO HADCM 0.19 0.11 0.23 0.27 0.52 0.25 0.62 0.91 B1 average 0.16 0.11 0.19 0.21 0.49 0.28 0.59 0.78 GFDL 2.1 0.19 0.12 0.23 0.25 0.50 0.28 0.62 0.77 MIROC3.2 MEDRES 0.13 0.09 0.16 0.18 0.41 0.24 0.46 0.68 MPI ECHAM5 0.10 0.06 0.12 0.17 0.44 0.19 0.54 0.79 NCAR CCSM3 0.18 0.13 0.21 0.22 0.54 0.36 0.63 0.79 UKMO HADCM 0.19 0.13 0.23 0.25 0.58 0.35 0.70 0.87 Economy-wide Impacts of Climate Risks 79 60 National total GDP Average annual growth rates, 2005–50 Existing variability scenario: 4.44% Average climate change scenario: 4.38% 50 Additional cumulative economic loss, 2005–50 All climate change scenarios: US$128.55bn National agricultural GDP (2005 US$bn) A2 climate change scenarios: US$146.79bn B1 climate change scenarios: US$110.321bn 40 30 20 10 Existing variability scenario Average climate change scenario 0 2005 2010 2015 2020 2025 2030 2035 2040 2045 2050 Figure 6.11 Losses in national total GDP due to climate change, 2005–50 age annual economy-wide cost of climate change Household consumption impacts from in Bangladesh during 2005–50, and is equal to climate change about 8 per cent of foreign aid transfers to Bang- ladesh in 2005. In per capita terms this is equiva- Table 6.9 shows the reduction in real house- lent to a discounted US$76.69 per capita during hold consumption spending as a result of climate the full 2005–50 period (i.e. taking population change. The total loss in consumption spending and income growth into account).The economic over the 2005–50 period is US$104.77 billion cost of climate change also rises over time from (measured in 2005 prices), which suggests that US$0.57 billion per year during 2005–25 to over 80 per cent of the total economic cost of US$0.92 billion per year during 2040–50. climate change will be passed onto households Economic costs are higher under the average (i.e. compared to the US$128.55 billion loss in A2 scenario (US$29.21 billion overall; US$0.65 total GDP). The remaining economic cost will billion per year) than under the average B2 sce- be borne by the public sector and private invest- nario (US$22.26 billion overall; US$0.49 billion ment. Two-thirds of the decline in private con- per year). This uncertainty over future emission sumption will be experienced by households scenarios causes a wide divergence in the esti- working in the agricultural sector, including lan- mated economic cost of climate change over dless farm workers. However, the largest declines the coming decades (see Figure 6.12). Model are for larger-scale farmers, who rely more heav- uncertainty implies that average total GDP losses ily on agricultural incomes. By contrast, marginal range from US$0.41 billion per year under the farm households and landless farm workers rely MIROC3.2 MEDRES GCM (B1 scenario) to more on labour incomes and non-farm employ- over US$1 billion per year under the GFDL 2.1 ment, and are thus less directly affected by the GCM (A2 scenario) (see Table 6.8). Despite this economic losses from climate change. However, uncertainty, however, climate change will impose these households’ initial incomes are much lower a substantial economic cost on future develop- than larger-scale farmers, and so their welfare will ment in Bangladesh, thus justifying signiï¬?cant be more vulnerable to even small changes in per investments to curb its long-term impacts. capita incomes (see Table 6.9). Similarly, non-farm households’ consumption spending also declines, 80 Climate Change Risks and Food Security in Bangladesh 1.4 A2 scenarios 1.2 Discounted economic losses from climate change as a share of discounted GDP under the existing variability scenario (%) All scenarios 1.0 B2 scenarios 0.8 0.6 0.4 0.2 0.0 2005 2010 2015 2020 2025 2030 2035 2040 2045 2050 Figure 6.12 Cumulative discounted losses due to climate change as a share of total GDP, 2005–50 Table 6.9 Losses in national households’ consumption spending due to climate change, 2005–50 Population Cumulative loss in consumption Average per capita loss in Share of lost consumption in total share in spending, 2005–50 consumption, 2005–50 (US$) household consumption (%) 2005 (%) (2005 US$ billion) All A2 B1 All A2 B1 All A2 B1 scenarios scenarios scenarios scenarios scenarios scenarios scenarios scenarios scenarios All Households 100.00 104.77 118.30 91.24 10.47 11.82 9.12 1.62 1.83 1.40 Agricultural 72.42 65.82 74.41 57.23 9.08 10.27 7.90 1.59 1.81 1.38 Households Farm Households 57.26 56.39 63.83 48.94 9.84 11.14 8.54 1.64 1.86 1.42 Marginal Farms 20.32 8.87 10.01 7.73 4.36 4.92 3.80 1.38 1.56 1.20 Small-scale Farms 28.74 29.56 33.42 25.70 10.27 11.61 8.93 1.68 1.91 1.46 Large-scale farms 8.19 17.95 20.40 15.51 21.90 24.89 18.92 1.72 1.95 1.48 Landless workers 15.16 9.43 10.58 8.29 6.22 6.97 5.46 1.38 1.55 1.21 Non-farm 27.58 38.95 43.89 34.01 14.11 15.90 12.32 1.65 1.87 1.44 Households Low Education 19.26 14.53 16.31 12.76 7.54 8.46 6.62 1.52 1.71 1.33 Some Education 5.67 13.67 15.36 11.99 24.11 27.08 21.14 1.83 2.06 1.60 High Education 2.65 10.74 12.22 9.26 40.48 46.07 34.90 1.64 1.87 1.41 Note: Marginal Farms are less than 0.5 acres; Small-scale Farms are between 0.5 and 2.5 acres; and Large-scale Farms are larger than 2.5 acres. Low Education households have household heads that are illiterate or have completed some primary schooling; Some Education households’ heads have completed primary schooling and some secondary schooling; and High Education households’ heads have completed secondary schooling. For more information see Annex 2. especially for higher-educated households, who Consumption impacts across sub-regions are affected negatively by the deceleration in Per capita consumption declines in all sub- economic growth and rising food prices. regions (see Table 6.10). However, the least Economy-wide Impacts of Climate Risks 81 Table 6.10 Losses in regional farm households’ consumption spending due to climate change, 2005–50 Population Cumulative loss in consumption Average per capita loss Share of lost consumption share in spending, 2005–50 in consumption, in total household 2005 (%) (2005 US$ billion) 2005–50 (US$) consumption (%) All A2 B1 All A2 B1 All A2 B1 scenarios scenarios scenarios scenarios scenarios scenarios scenarios scenarios scenarios Farm Households 57.26 56.39 63.83 48.94 9.84 11.14 8.54 1.64 1.86 1.42 Dinajpur (1) 5.71 5.01 5.70 4.32 8.77 9.98 7.56 3.25 3.71 2.80 Rangpur (2) 7.17 5.94 6.64 5.25 8.28 9.25 7.31 1.73 1.94 1.52 Ishwardi (3) 1.70 1.41 1.62 1.20 8.31 9.55 7.07 1.79 2.06 1.52 Tangail (4) 4.05 3.79 4.37 3.21 9.37 10.80 7.93 1.26 1.46 1.07 Dhaka (5) 2.01 1.66 1.89 1.44 8.26 9.38 7.13 1.38 1.58 1.19 Mymensingh (6) 4.90 4.86 5.58 4.13 9.89 11.37 8.42 1.46 1.68 1.24 Sylhet (7) 3.41 2.71 3.19 2.24 7.95 9.33 6.57 1.13 1.33 0.93 Srimangal (8) 1.81 2.17 2.50 1.85 12.01 13.81 10.22 1.74 2.01 1.48 Comilla (9) 5.60 6.19 7.02 5.37 11.04 12.52 9.56 1.73 1.97 1.49 Chittagong (10) 2.60 2.82 3.18 2.47 10.85 12.23 9.48 1.75 1.97 1.52 Rangamati (11) 0.81 1.15 1.31 0.98 14.11 16.14 12.08 1.60 1.83 1.37 Maijdee Court (12) 2.01 2.32 2.58 2.05 11.50 12.81 10.19 1.91 2.14 1.69 Jessore (13) 4.28 3.73 4.25 3.21 8.71 9.93 7.49 1.64 1.88 1.41 Faridpur (14) 3.00 3.58 4.00 3.16 11.92 13.31 10.53 1.73 1.94 1.53 Patuakhali (15) 5.72 6.48 7.08 5.87 11.31 12.36 10.25 2.07 2.26 1.87 Khulna (16) 2.47 2.55 2.91 2.19 10.31 11.76 8.86 1.52 1.74 1.30 affected regions are Sylhet (7) and Dhaka (5).The model (2005). The effect is to compress cli- latter is partially insulated from climate change in mate change effects for the period 1984-2035 our analysis since the share of agriculture in this into the shorter period 2005-2035. How- regions’ total GDP is below that of most other ever, this assumption will not greatly affect regions. Many of the southern coastal regions our conclusions since the effects of climate experience signiï¬?cant declines in per capita change during 1984-2004 are fairly small, consumption and bear a substantial share of the especially relative to future climate change total cumulative cost of climate change for farm projections for the 2030s and 2050s. households. This is because these regions expe- 2 The estimation procedure of the 2005 SAM rienced large declines in rice production due to is described in Annex 3. sea level rises. However, if the initial per capita 3 This is measured by the cumulative loss in consumption level is controlled, then the largest total or national GDP during 2005–50. percentage declines in per capita consumption 4 All dollar values reported in this chapter are are in Maijdee Court (12) and Patuakhali (15) in constant 2005 US$. in the south and Dinajpur (1) in the northwest. 5 A lower or higher discount rate will not quali- These regions are the most vulnerable from an tatively change the results presented. A higher economic perspective. rate will result in lower estimated losses and a lower rate will result in higher estimated losses. The relative losses across simulations and model Notes experiments will largely be unchanged. 1 By starting the transition from historical to 6 This is lower than total GDP losses since 2030s data in 2005 we assume that the mean private consumption is only part of national of the historical 1970-1999 period did not income, which also includes government change from the mid-point of the historical consumption, investment demand and net period (1984) to the base year of the CGE exports. 7 Adaptation Options in the Agriculture Sector Vulnerability to climate risks and overall eco- resulted in changes in cropping pattern, greater nomic development are intricately linked, as diversiï¬?cation of agriculture, promotion of high- is shown in the preceding sections. Therefore, yielding varieties and increased cropping inten- adaptation in the agriculture sector must be well sity. Embankments in flood-prone areas (both integrated with both the broad national devel- coastal and inland) have also played a major opment goals and livelihood priorities at the role in reducing flood risks and protecting key local level. Not surprisingly, though, farmers and household assets. Over the last three decades, rural households have long adapted to a variety the Bangladesh government has invested over of climate risks. These coping strategies vary by US$10 billion (at constant 2007 prices) for flood geographic region and depend on the range of management embankments, coastal polder and prevailing socio-economic conditions. As the cyclone shelters (BCAS, personal communica- climate changes, more and different adaptations tion). With this protection, substantial increases will be required. An approach to studying crop- in production have been made possible. These ping adaptations through the crop simulation collective investments have resulted in signiï¬?cant approach of Chapter 5 is given in detail in Annex improvements in meeting national objectives of 1. In this chapter, a series of adaptations for which food-grain self-sufï¬?ciency. Substantial invest- ï¬?eld trials exist and farmer feedback is reported ments in early warning and preparedness sys- are described. These descriptions provide tem- tems (primarily improvements in flood forecast- plates for the development of other adaptations ing and cyclone warnings) have also minimized for farm-level implementation. (though not entirely eliminated) the risk from The presence of both formal and informal natural disasters. The Bangladesh Disaster Man- sources of support can play a critical role in agement Bureau plays a critical role in respond- minimizing climate risks. For instance, substan- ing to droughts and floods. Lastly, in addition to tial public-sector investments in agriculture and this direct support from the government depart- water have been made to help protect farmers ments, non-government organizations and other from a variety of existing climate risks. These donors have played an important role in support- measures include investments in water infrastruc- ing alternative livelihood activities. ture (e.g. embankments in floodplain and coastal Agriculture research and technology devel- areas to protect against floods and storm surges) opment has been essential to achieving higher and irrigation. and more stable crop yields. An active network Groundwater irrigation has provided a of agriculture research institutes exists in Bang- means for farmers to adapt to soil moisture deï¬?- ladesh. These include the Bangladesh Agricul- cits, particularly in drought-prone areas. This has ture Research Institute (BARI), Bangladesh Adaptation Options in the Agriculture Sector 83 Rice Research Institute (BRRI), Bangladesh The current large gap between actual and poten- Institute of Nuclear Agriculture (BINA) and tial yields suggests substantial on-farm opportu- the Bangladesh Agriculture University (BAU). nities to increase incomes and production. For These groups, among other things, develop and many communities, adoption of new technolo- test new crop varieties to increase national total gies can represent high downside risks unless production and resilience against climate risks. options are well tested in the ï¬?eld. Efforts to Extensive testing and ï¬?eld trials are undertaken provide ï¬?nancial and technical support to ensure before new varieties are released to extension sustainable production systems are thus required. organizations for dissemination. The Department Government agriculture extension ofï¬?cers play of Agriculture Extension (DAE) plays a vital role an important role in these regards. At the house- in disseminating new technologies down to the hold and farm level, private-level adaptations to farmer level through demonstration plots, by climate risks have included, inter alia: crop adjust- providing critical inputs and through training. ments in terms of crop mix and planting dates; Typically, DAE undertakes 5–10 demonstrations supplementary irrigation from ponds; mois- on a new crop variety at each block each year ture conservation approaches; adoption of new depending on budget resources. The DAE is one seed varieties (e.g. drought and saline resistant), of the largest governmental departments work- diversifying to ï¬?sheries and shrimp production; ing with approximately 60 per cent of the popu- and flood protection and drainage works. Some lation directly involved with crop production. current and past adaptation programmes for the The Department of Agricultural Information agriculture sector in Bangladesh are given in Service (AIS) under the DAE is instrumental in Table 7.1. preparing materials on speciï¬?c technology. These institutions will continue to be active in helping Bangladesh achieve food security. 7.1 Identifying and Evaluating In Bangladesh, though there is no speciï¬?c drought-tolerant rice variety, DAE does pro- Adaptation Options mote particular paddy varieties that are short Adaptation options can address several different durational to avoid the effects of drought. These types of climate risks. Broadly speaking, adapta- include BR25, BRRI Dhan 33 and BRRI Dhan tions can focus on increasing crop productivity, 39. BARI has also promoted some vegetables and improving irrigation efï¬?ciency or expanding crops like chilli, tomato, okra, cucumber, auber- water supply, crop diversiï¬?cation and intensiï¬?- gine (brinjal/eggplant), potato, cowpea, barley, cation, generating alternative enterprises (either maize, chickpea, linseed and sesame as drought farm or non-farm sector) to diversify household tolerant. BRRI has developed some flood-toler- income sources, and expanding access to train- ant varieties of paddy including BR11, 20, 21, 22, ing and credit. Existing strategies to deal speciï¬?- 23 and 24, and BRRI Dhan 31, 32, 33 and 34. cally with drought risks include: full irrigation Moreover, as was mentioned earlier, the boro crop for dry season boro and supplementary irriga- plays an important role in offsetting flood losses. tion for t. aman from groundwater and surface Finally, BRRI has developed some saline-resist- water sources, crop adjustments (e.g. replanting), ant paddy varieties including BR10 and 23, and moisture conservation practices and promotion BRRI Dhan 32, 41 and 47. Some vegetables and of horticultural crops. Existing strategies to deal other crops like chilli, tomato, okra, cucumber, speciï¬?cally with flood risks include: construction potato, cowpea, soybean and barley are promoted of embankments and drainage canals, harvesting as salt tolerant. These evolving new varieties will of crops from under water, changing the crop cal- continue to play a major role in helping farmers endar (e.g. late or early planting), raising seedlings adapt to changing and uncertain conditions. in a safe and dry place, double transplanting of Despite these innovations, poor adoption of seedlings and floating vegetable gardens. Existing technologies and innovations can be common. strategies to deal with coastal zone risks (e.g. tidal 84 Climate Change Risks and Food Security in Bangladesh Table 7.1 Sample of past and present programmes on adaptation in the agriculture sector Project Name Agency Location Sample Activities Reducing Vulnerability to CARE Satkhira, Gopalganj, Drought-tolerant crop cultivation; Climate Change Rajshahi Tree and plant nursery activities; Floating gardens and homestead vegetable gardens Livelihood Adaptation to FAO, DAE, Chapai, Nawabganj, Natore, Homestead gardening; Climate Change (LACC) Naogaon, Pirojpur, Khulna Drought-tolerant fruit tree gardening; Phase I Rainwater harvesting in mini ponds for supplementary irrigation for t. aman Livelihood Adaptation to FAO, DAE Rajshahi, Chapai Adaptation options have been identiï¬?ed but not implemented yet Climate Change (LACC) Nawabganj, Natore, Phase II Naogaon, Pirojpur, Khulna Disappearing Lands: Practical Action Gaibandha Sand bar vegetable cultivation in char lands; Supporting Communities Bangladesh Floating bed vegetable cultivation Affected by River Erosion Assistance to Local Community Action Aid Naogaon, Sirajganj, Homestead vegetable gardening; on Climate Change Adaptation Bangladesh Patuakhali Rice demonstrations; and DRR in Bangladesh Chickpea cultivation; Community based pond management for supplementary irrigation Barind Integrated Area BMDA Northern part of Bangladesh Groundwater irrigation; Development Project Management of surface water for crop production; Excavation of mini ponds Asia-Paciï¬?c Forum for BCAS Rajshai, Naogaon, Sirajganj, Zero-tillage maize cultivation; Environment and Development Gaibandha, Kurigram Chickpea cultivation; Shunamganj, Faridpur, Relay cropping of sweet gourd; Pirojpur, Cox’s Bazar, Floating bed vegetables cultivation Satkhira, Patuakhali, Barisal inundation and salinity intrusion) include: coastal Table 7.2 Sample adaptation options in the agriculture sector embankments and introduction of saline resist- Adaptation Option ant crops and bio-saline aquaculture (e.g. shrimp 1 Zero or minimum tillage to cultivate potato, aroid and groundnut cultivation). with water hyacinth and straw mulch The following 14 sample adaptation options 2 Zero-tillage cultivation of mashkalai, khesari, lentil and mustard (Table 7.2) were identiï¬?ed through a series of workshops with participation from a variety 3 Modiï¬?ed sorjan system (zuzubi garden) with vegetable cultivation in char land of research institutes, government agriculture 4 Floating bed vegetable cultivation extension ofï¬?cers, donor community representa- tives and practitioners in the ï¬?eld. These adapta- 5 Cultivating foxtail millet (kaon) in char land tions represent promising adaptation approaches 6 Parenga practice of t. aman cultivation system to increasing production (both existing and new 7 Relay cropping of sprouted seeds of aman rice in jute ï¬?elds crops) under constrained and changing environ- 8 Raising vegetables seedlings in polythene bags homestead ments. For all of these options, ï¬?eld trials exist and trellises farmer feedback is reported. Thus, these options 9 Zero-tillage maize cultivation have the potential for replication and scalability. 10 Chickpea cultivation using a priming technique The effectiveness and suitability of each of 11 Supplementary irrigation of t. aman from mini ponds these options will be dependent on a wide range 12 Year-round homestead vegetable cultivation of location-speciï¬?c factors. These may cover a 13 Pond-water harvesting for irrigation to cultivate rabi vegetables range of institutional, socio-economic, ï¬?nancial 14 Sorjan system for cultivating seasonal vegetables, fruits and ï¬?sh and environmental issues. Sustainability will in Adaptation Options in the Agriculture Sector 85 Table 7.3 Estimated costs and beneï¬?ts of selected adaptation part be dependent on the local capacity and the options capacity of the implementing support agency Adaptation Cost per Beneï¬?t per Proï¬?t per (both government and non-government) at the Option No. hectare (Tk) hectare (Tk) hectare (Tk) national and sub-national levels to provide both technical and material assistance. Potential for 1 potato 196,270 342,000 145,730 economic return will be a critical determinant 1 aroid 97,700 250,000 152,300 1 groundnut 79,095 90,000 10,905 of overall adoption. Detailed indicative cost and 2 mashkalai 29,950 52,500 22,550 ï¬?nancial beneï¬?t estimates were prepared (see 2 khesari 26,010 56,000 29,990 Table 7.3) for each of the identiï¬?ed options. The 2 lentil 30,531 75,000 44,469 unit costs are limited to those costs that would 2 mustard 37,540 67,500 29,960 be borne by the farmer to implement the adap- 3 262,500 535,000 272,500 tation option. That is, the cost to the govern- 4 34,025 59,750 25,725 ment agency to implement such options more 5 233,469 487,500 254,031 widely is not included here. These 14 adaptation 6 49,880 79,000 29,120 options represent potential no-regret strategies 7 49,830 78,000 28,170 for increasing incomes and building resilience to 8 165,575 432,000 266,425 climate risks. 9 98,315 129,000 30,685 Detailed factsheets follow, describing each 10 57,325 90,000 32,675 adaptation option (numbered above), including 11 76,705 100,875 24,170 information about the production package, geo- 12 125,000 372,000 247,000 graphical suitability, major advantages and disad- 13 151,575 294,000 142,425 vantages and the costs and ï¬?nancial beneï¬?ts of 14 253,084 573,052 319,968 implementation. More detailed information can Note: US$1 = Tk69 be found either through the Department of Agri- culture Extension or the FAO Livelihood Adap- tation to Climate Change (LACC) programme. 86 Climate Change Risks and Food Security in Bangladesh 1 Zero or minimum tillage to cultivate flooding and tidal surge, and areas where mulch potato, aroid and groundnut with water materials are readily available are the most suit- able.This is currently practised in Rangpur, Kuri- hyacinth and straw mulch gram, Gaibandha, Bogra, Sirajgang, Rajshahi, Nawabganj, Natore, Pabna, Kushtia, Faridpur, Summary Munsiganj, Madariganj and Barisal. This practice can produce several different types of crops (e.g. potato, aroid, groundnut, chickpea, Major advantages onion, garlic) with minimum tillage required. This practice is done on mostly medium high This option provides an additional crop and land in flood-prone areas during the rabi season. income for farmers. Farmers can also generate Farmers are already practising this approach when mulch materials as a byproduct of this option land is unfavourable to normal practices. Farmers which can generate material for household fuel. sow seeds on moist soil just after the recession This byproduct can be sold in the local market. of flood water. The land is then mulched with Mulch materials also have the added beneï¬?t of water hyacinth of about 30cm thick. The mulch- protecting soils from high temperatures and high ing conserves soil moisture and decreases evapo- evaporation which increases both microbial activ- ration from the soil. The process is as follows for ity and soil productivity. Finally, this approach can potato, and similar for the others: help to control the population of weeds. • Clear grasses and debris from the ï¬?eld; Major disadvantages • Sowing/planting time is November–Decem- This option will not be feasible without mulch ber; materials such as water hyacinth or straw. Moreo- • Apply fertilizers at the rate of 165kg–100kg– ver, thin application of mulching materials may 130kg–40kg/ha of urea-TSP-MOP-gypsum; not fully protect the tubers from sunlight, result- • Place germinated seed tubers in rows at 60cm ing in decreased quality of potato. apart and 25cm intervals within a row; • Cover the potato seeds with 30cm thick mass of water hyacinth; Approximate beneï¬?ts • Depending on the market price the crop can This option only requires land, potato seeds, be harvested partially or fully 70 days after water hyacinth or straw, fungicide and insecti- sowing. cides, fertilizers and labour. Approximately a total of Tk196,270 (US$2785) (including land-lease Potato is a photo-sensitive, succulent crop that cost) is required to cultivate potato in 1 hectare of needs more soil moisture during the vegetative land by using this option and farmers can harvest period. The water hyacinth is used as a mulch 19.0t of potato at a market price of approximately layer to preserve soil moisture as well as increase Tk18 per kg (total Tk342,000 or US$4854). production. During periods of dense fog and Moreover, farmers can harvest green potatoes as moist weather there is the potential for fungal cattle fodder and mulch material as well as fuel diseases (like late blight and early blight) that can for family consumption. Farmers can thus earn a severely affect potato yield. If this infection hap- net proï¬?t of Tk145,730 (US$2068) from 1 hec- pens during the early stages of potato cultivation, tare of land which would normally remain fallow production could decrease tremendously. during the rabi season. Per unit hectare cost of producing aroid and groundnut is Tk97,700 and Most suitable geographic area Tk79,095 (US$1386–1122) respectively. Per hec- tare proï¬?ts would be approximately Tk152,300 Coastal areas (saline and non-saline) and the and Tk10,905 (US$2161–154) respectively. central floodplains, depending on the degree of Adaptation Options in the Agriculture Sector 87 2 Zero-tillage cultivation of mashkalai, Manikganj, Nawabganj, Rajshahi, Pabna, Kushtia, khesari, lentil and mustard Meherpur, Jessore, Chuadunga, Jhenaida, Farid- pur, Barisal and Narail. Summary Major advantages In some cases, after the harvesting of the t. aman This option provides an additional crop and crop, the delay in the recession of flood water income for farmers. Farmers get byproducts that results in excessive soil moisture and unsuitable can be used as fodder and for family fuel con- conditions for planting. Given this situation, this sumption. Pulse crops are also leguminous fam- adaptation option broadcasts mustard, mashkalai ily crops which improve soil nutrients through or khesari in the t. aman ï¬?eld 10–15 days before the release of nitrogen. This helps to increase soil harvest using zero tillage approaches to generate productivity. Extension ofï¬?cers, in fact, often an extra crop. advise farmers to harvest the crop from the stem, careful not to uproot the plant. These pulses can Production package add 40–80kg per hectare of nitrogen, i.e. provide Zero-tillage cultivation of mustard, khesari and about 87–174kg of urea which will decrease fer- mashkalai is currently being practised by farm- tilizer costs. ers in flood-prone areas. Farmers sow seeds in the aman paddy ï¬?elds 10–15 days before harvest Major disadvantages time. This is also being grown in previously fal- Early harvest during the vegetative stages may low ï¬?elds in mid-October to November after reduce crop yields. During the production period, recession of flood waters.The process is as follows most of the land remains fallow and may disrupt for mustard, and similar for the others: cattle-grazing practices. Finally, the degree of soil moisture is an important determinant of seed • Method of seed sowing: broadcast; germination. Excess soil moisture could damage • If possible, supplemental irrigation may pro- seedlings. duce better yields (one irrigation during flowering stage and another one during fruit- Approximate beneï¬?ts ing stage); Land, seeds, fertilizers and labour are required. • For mustard cultivation the following doses of Approximately a total of Tk37,540 (US$532) is fertilizer might be used for better yield. Urea- needed to cultivate one hectare of mustard. Farm- TSP-MOP-gypsum-zinc sulphate-boric acid ers can harvest 1.5t of mustard at a market price = 225kg–160kg –75kg–140kg–4kg–12kg per of approximately Tk45 per kg (total Tk67,500 hectare; or US$958). Farmers also get the mustard plant • Seed rate: Mustard 8–10kg/ha; as cattle fodder and fuel for family consump- • Time of harvest: January–February; tion. Farmers can earn a net proï¬?t of Tk29,960 • Yield: Mustard 1.0–1.5t/ha. (US$425) from one hectare of land which nor- mally would have been fallow. The per hectare Most suitable geographic area cost of cultivating khesari, lentil and mashkalai Coastal areas (saline and non-saline) and central is Tk26,010, Tk30,531 and Tk29,950 (US$369, floodplains, depending on the degree of flood- US$433, US$425) respectively. The net proï¬?t is ing and tidal surge, are the most suitable. This is Tk29,990, Tk44,469 and Tk22,550 (US$425, currently practised in Kurigram, Sirajganj, Bogra, US$631, US$320) for khesari, lentil, and mashka- Joypurhat, Noagaon, Rajshahi, Jamalpur, Tangail, lai respectively. 88 Climate Change Risks and Food Security in Bangladesh 3 Modiï¬?ed sorjan system (zuzubi garden) season by using a modiï¬?ed sorjan system. A dedi- with vegetable cultivation in char land cated area of land (33 decimal) of loamy type soil is required for the modiï¬?ed sorjan system. This soil is best for making these beds and ditches. Jan- Summary uary to February is the best time for preparation. In many places, flood waters remain on crop Speciï¬?cally, the provisions are as follows: ï¬?elds and char lands for an extended period of time. In the absence of uplands, vegetables and Raised bed fruits cannot be grown. Most of the char lands 3m breadth x 0.5m height x (10m length or con- remain fallow after the recession of flood water sidering length of plot size). during the rabi and kharif 1 seasons. Normally vegetables and fruits must come from outside of Ditches the char lands to meet local demands. Moreo- ver, the communities in these areas typically are 2m breadth x 0.5m depth x (10m length or con- unable to afford the high price of vegetables and sidering length of plot size). fruits and thus cannot incorporate these items into their regular diets. The result is malnutrition Crops on beds from lack of minerals and vitamins. A modiï¬?ed zuzubi (variety Apel kul/BAU kul), recom- sorjan system with vegetable cultivation can help mended spacing (plant to plant and row to row) to increase production in these places. should be followed. Production package Crops on ditches Farmers in char land areas can produce vegetables Seasonal vegetables (cabbage, cauliflower, tomato, and fruit (zuzubi) during the rabi and kharif l aubergine, amaranth, Indian spinach, kang kong, Figure 7.1 Layout of modiï¬?ed sorjan system Adaptation Options in the Agriculture Sector 89 chilli, red amaranth, cucumber, bitter gourd, Most suitable geographic area snake gourd , bottle gourd, sweet gourd, etc.) can The char lands or coastal areas are most suitable be cultivated round the year if trellises are made depending on the degree of flooding and tidal on the ditches. surges.This is currently being practised in Sirajganj. Fruits Major advantages Papaya, lemon, etc. may be planted near the edge This option helps to diversify the crop mix and of the beds. increases the production of vegetables and fruits (both for household consumption and for sale in Design and layout local markets which increases income). This may • Top soil of middle 2m ditch is removed and also have positive nutritional impacts on com- kept aside; munities in the char lands. The zuzubi plant is a • Soil from the 2m plot (ditch) is dug at a depth flood-tolerant crop. of 0.5m and placed on the 3m plots; • The 3m plots are raised to 0.5m high beds to Major disadvantages measure three beds at the top; • The 2m plots at the ends are also dug at a depth Depending on the flood frequency and intensity, of 0.5m and the soil is placed on the beds; the ditches may silt with clay and sand. Moreover, • Slope of the bed is made uniform and com- large floods may damage the layout of the modi- pact by pressing; ï¬?ed sorjan system. Though the zuzubi plant is a • Top soil kept aside is spread uniformly on the flood-tolerant crop, prolonged exposure to flood raised bed. Thus the sorjan beds and furrows waters may still damage the crop and result in are made. disease. • Trellises are made with bamboo and other local materials on the furrows to support Approximate beneï¬?ts creeper vegetables. The modiï¬?ed sorjan system requires vegetables seeds, fruits saplings, fertilizers, bamboo, jute Cropping patterns sticks, spades and other resources. Approximately • The ï¬?ve beds (or more based on the size of a total of Tk233,469 (US$3313) is required to the plot) are numbered from 1 to 5 or more. cultivate fruits and vegetables using this approach. Two beds at either the east or south are ear- A farmer can typically harvest 16,500kg of veg- marked for vegetables cultivation and the rest etables and 3150kg of fruits (zuzubi) where for fruits and vegetables. the potential revenue generated is Tk487,500 • The edge of the beds is planted with creeper (US$6920) (assuming Tk20 per kg vegetables and vegetables, which is supported by the trellis Tk50 per kg of zuzubi). In addition, household over the furrows. consumption of these crops can help to supple- • Zuzubi should be planted on raised beds fol- ment family nutritional requirements. Farmers lowing standard spacing. Seasonal creeper can also use pruned materials for home fuel. In vegetables can be cultivated on the edges to summary, farmers can earn a potential net proï¬?t make trellises on the ditches. of Tk254,031 (US$3605) from one hectare of char land that would have otherwise remained fallow during the rabi season. 90 Climate Change Risks and Food Security in Bangladesh 4 Floating bed vegetable cultivation • In a few days the seeds will germinate and grow; Summary • Apply a little urea depending on the growth During high floods, land is water-logged and of the vegetables; seedlings are easily damaged. In these areas, pro- • Continue to harvest the vegetables by thinning duction of crops is difï¬?cult. Moreover, delayed to allow the remaining seedlings to grow. recession of flood waters can create a scarcity • When the water recedes the bed will touch of necessary cereals and vegetables. Considering the ground and the gourd plants will take this, floating vegetable beds (baira) can help to root on the ï¬?eld and start fruiting. meet the daily requirements for vegetables such • When the bed is about to touch the ground as lalshak, data, kang kong, okra, spinach, Indian other winter vegetables like cabbage, cauli- spinach, cucumber, bitter gourd, bottle gourd, flower, etc. can also be planted on the bed sweet gourd, radish, aubergine, onion and garlic ahead of the scheduled planting date and can and spices such as chilli and turmeric etc. be grown as a ï¬?eld crop with some fertilizer as needed. Suitable crops for floating beds Most suitable geographic area Amaranth (both leaf and stem), okra, aubergine, The coastal areas (both saline and non-saline kang kong, etc. can be grown under wet con- areas) and central floodplains are the most suita- ditions on floating substrata made of water hya- ble.The northeast region of the country may also cinth. Water hyacinth is abundantly available in introduce this practice as well. This is currently flood-prone and submerged areas. This is already being practised in Faridpur, Barisal, Gopalganj, being practised in certain locations. Bottle and Khulna and Gaibandha. sweet gourds also can be grown on floating sub- strata, but require that the floating beds touch the Major advantages ground after the recession of flood water for con- tinued rooting. Floating beds (baira) are a low-cost farmer inno- vation that can play a vital role in generating veg- Production packages etables for the family. During the dry season, baira can also result in compost to increase soil fertil- • Make a bamboo frame of 10m long and 1m ity as well as crop productivity. Farmers can also wide on water near the land and ï¬?x the loca- practise growing tree saplings on baira to generate tion with a bamboo pole; additional income. • Add floating piles of water hyacinth within the bamboo frame and repeat several times at Major disadvantages ï¬?ve to seven day intervals on the same piles until a heavy floating bed (about 60cm thick) Heavy rainfall and strong winds may damage is made; bairas. Sometimes damaged bairas may result in • Apply a small dose of TSP and MP on the the loss of harvest vegetables. If heavy rains come bed mixed with previously made compost of before germination, seeds may wash away. water hyacinth; • Mix seeds of red amaranth, stem amaranth, Approximate beneï¬?ts kang kong, okra and a few bottle gourd and Bamboo for making the baira frame, water hya- sweet gourd seeds in a proportionate quantity, cinth, vegetables seeds and some fertilizers are mix with soil and broadcast on the floating needed for this option. It is estimated that farm- bed after ï¬?ve to seven days of application of ers can harvest 175kg of vegetables per baira and fertilizers; that about 150 baira can be built on one hec- • Single crop vegetable seeds with recommended tare of land. Total harvested vegetables would be spacing may be sown on the floating bed; approximately 26,250kg. At a market price of Adaptation Options in the Agriculture Sector 91 about Tk20 (US$0.30) per kg, this results in a generate Tk272,500 (US$3867) net on land that total sale value of Tk525,000 (US$7451). Moreo- otherwise would remain fallow. Furthermore, ver, assuming the generation of 20t of water hya- farmers will consume some vegetables to meet cinth which can be sold in the market for about the basic family requirements. Tk10,000 (US$142), the farmer in total will 92 Climate Change Risks and Food Security in Bangladesh 5 Cultivating foxtail millet (koan) in currently being practised in Nilphamari, Rang- char land pur, Lalmonirhat, Kurigram, Gaibandha, Sirajganj and Jamalpur. Summary Major advantages Most of the char lands remain fallow after reces- sion of flood water in rabi and kharif 1 seasons. Kaon is a low-cost cereal crop that farmers can Foxtail millet (kaon) is a drought-tolerant, short easily grow on char lands that otherwise would duration crop that can be grown with minimum remain fallow during the rabi season. Kaon is also tillage during the rabi season immediately after a shallow-rooted crop which can decrease soil the recession of flood water. erosion and increase organic matter in the soil if, at the time of harvesting, cuttings are made at Production technology 20–30cm from the ground. Currently, farmers plough their char land two or Major disadvantages three times and then sow kaon seeds (10kg per hectare) following a broadcasting or line system. During seed germination and seedling stages, After two to three weeks, seeds germinate and prolonged droughts during the rabi season may the removal of weeds is needed for better growth severely affect the crop. Sometimes, early floods and increased yields. Farmers are using urea-TSP- and heavy water-logging may also impact the MP fertilizer at the rate of 100kg–75kg–40kg per production. hectare. If irrigation facilities are available then farmers use half the urea and the total amounts Approximate beneï¬?ts of TSP and MP fertilizer during ploughing of the This is a low-cost activity that needs only mini- land; the remainder of the urea may be applied mum tillage to prepare the land, kaon seeds and 35–40 days after germination of seed. When some fertilizer. Farmers can harvest 2.25t of mil- irrigation facilities are not available then farm- let from one hectare of char land and generate ers use all the fertilizer during ploughing of the Tk59,750 (US$848) in income. The production land. Irrigation is needed if drought conditions cost is Tk34,025 (US$482). Thus, the farmer gets are prolonged. Normally 2–2.5t of kaon may be Tk25,725 (US$365) as net proï¬?t from one hec- produced per hectare of char land. tare of land by using this option. Farmers can also use straw from the millet as fodder for cattle and Most suitable geographic area fuel. This material can be sold in the local market. This is most suitable on char lands depending on the occurrence and intensity of the floods.This is Adaptation Options in the Agriculture Sector 93 6 Parenga practice of t. aman cultivation Most suitable geographic area system This option is most suitable in floodplain areas where flood water recession is late. This option is Summary currently being practised in Kurigram, Gaibandha Prolonged flooding can damage aman seedlings and Sirajganj. and also the existing transplanted aman. Moreo- ver, if aman seedlings are damaged, there is often Major advantages not enough time to re-raise seedlings for trans- This option helps to increase the likelihood of planting. Farmers in some places are currently harvesting an additional cereal crop. Farmers also addressing this issue by sowing sprouted aman get straw as a byproduct which can be used as seeds in moist soil on medium-high or medium- fodder for household cattle with the remainder low land after recession of flood waters. sold in local markets for cash. Finally, because of the increased cropping intensity, the population Production package of weeds is reduced and thus weeding costs for the following crop cycle are reduced. In this approach, farmers ï¬?rst clear weeds and other debris from the land. Farmers then soak aman seeds for 24 hours till they sprout and then Major disadvantages broadcast (or directly sow) in moist soil during This option is vulnerable to heavy rainfall and the month of August. prolonged flooding (especially just after seeds are Under this option farmers use urea-TSP- sown in the ï¬?eld). MP-gypsum fertilizers at the rate of 200kg– 125kg–85kg–65kg per hectare of land. Farmers Approximate beneï¬?ts apply all of the TSP and gypsum and half of the The main requirements are medium-to-high land, MP fertilizer after ï¬?eld clearing and then the quality seeds of aman rice, fertilizers, insecticides remainder of the MP and one-third of the urea and pesticides. A total of Tk49,880 (US$708) is three to four weeks after sowing. During the till- needed to cultivate aman rice using the parenga ing stage, the second dose of urea (one-third of system on one hectare of land. Farmers can har- the total dose) may be applied with the remainder vest 3.5t of aman rice which at a market value of the urea and MP applied immediately before of Tk20 (US$0.30) per kg will gross Tk70,000 the panicle initiation stage. During the produc- (US$993). Furthermore, farmers can get straw tion period, one to two weedings are needed to for cattle fodder and fuel for family consumption. decrease infestation of insects and pests. Farmers The remainder can be sold in the local market for can harvest 3.5t of paddy from one hectare of cash (approximately Tk9000 or US$128). Farm- land and also get straw that can be used for cattle ers thus can earn Tk29,120 (US$413) from one fodder. hectare of land which would normally remain fallow. 94 Climate Change Risks and Food Security in Bangladesh 7 Relay cropping of sprouted seeds of Most suitable geographic area aman rice in jute ï¬?elds This option is most suitable for medium-high land where sufï¬?cient drainage exists. This can Summary also be extended to saline, non-saline and cen- Farmers normally cultivate jute on medium-high tral floodplain areas where proper infrastructure land from mid-April to mid-September.Typically exists. This option is currently being practised in after harvest, sowing or transplantation of aman is Faridpur and Barisal. not possible and so kharif 2 crops are typically not planted and the land remains fallow. To address Major advantages this, farmers in the southeastern part of Bang- This option results in an additional crop which ladesh are practising relay cropping of sprouted helps to increase total cereal production and gen- aman seeds in jute ï¬?elds. This approach results erate income for households. Farmers also get in an additional short duration crop on the same straw which can be used as fodder and family ï¬?eld where cultivation of transplanted aman rice fuel consumption. would not have been possible. Major disadvantages Production package Under this option it is important to control the Farmers sow jute seeds on medium-high land water depth during the sowing of seeds in the from mid-April to mid-May and normally har- jute ï¬?eld. If the water depth increases (more than vest mid-August to mid-September. Sprouted 2.5–5cm of water) and no measures are taken, aman seeds are sown on the jute ï¬?eld 15–20 days seeds may get damaged. Also, during the jute cut- before harvest/cutting of jute. Farmers select ting period, caution must be taken to not damage medium-high land where drainage facilities are the aman seedlings. For instance, in some places, sufï¬?cient and control of water application depth the jute plant is left on the ï¬?elds after cutting is possible. Before the sowing of aman seeds in which damages the seedlings and reduces overall the jute ï¬?elds, the land is cleared of weeds and rice production. others debris. Farmers will soak aman seeds for 24 hours for sprouting and then broadcast in the Approximate beneï¬?ts jute ï¬?eld when standing water depth is not more Resources required include medium-high land, than 2.5–5cm. It is very important to control quality seeds of aman rice, fertilizers, insecticides the water depth of the jute ï¬?eld while the seeds and pesticides. A total of Tk49,830 (US$707) is sprout over the next seven days. If water depth needed to cultivate aman rice on one hectare of cannot be controlled then seeds could be damaged. jute ï¬?eld. Farmers can harvest 3.5t of rice which Farmers use urea-TSP-MP-gypsum fertiliz- at market price of Tk20 (US$0.30) per kg results ers at the rate of 200 kg–125kg–85kg–65kg. All in Tk70,000 (US$993) gross. Farmers also get the TSP and gypsum and half of the MP fertilizer straw byproducts that can be used as cattle fod- dose are applied after the jute ï¬?eld is cleared.The der and fuel for family consumption. Remain- remainder of the MP and one-third of the urea der material can be sold in the local market for is applied three to four weeks after sowing. Dur- about Tk8000 (US$113). Farmers thus can earn ing the tillering stage, the second dose of urea Tk28,170 (US$400) from one hectare of land (one-third of total dose) may be applied and the that would normally remain fallow after harvest rest of the urea and MP may be applied imme- of the jute. diately before the panicle initiation stage. During the production period, one to two weedings are required to decrease the susceptibility to insects and pests. Adaptation Options in the Agriculture Sector 95 8 Raising vegetable seedlings in During land preparation, farmers apply half the polythene bags on homestead trellises cow dung and the total amount of TSP fertiliz- ers. After preparation, beds are raised by about 1m width, 25–30cm height and 25–30cm length Summary depending on land size; 30cm to 35cm of space The floodplain region can sometimes be char- is kept as a trench between the two beds to use acterized by early floods, prolonged water-log- for irrigation water and to drain excess rain- ging and even late recession of flood water.These water. Beds are levelled and then seedlings are problems affect vegetable production during the transplanted during mid-August to September rabi season. Communities typically need vegeta- following the spacing for speciï¬?c vegetables ble seedlings for early plantation to harvest before (tomato = 60cm x 40cm; aubergine = 75cm x a regular season. Farmers in these areas cannot 60cm; cabbage = 60cm x 45cm). The remain- cultivate winter vegetables due to the unavail- der of the cow dung is used during plantation of ability of vegetables seedlings. To overcome this, the seedlings and the rest of the fertilizer doses farmers therefore raise vegetable seedlings in pol- may be applied in the ï¬?eld three to ï¬?ve weeks ythene bags on trellises near the homestead for after plantation. Farmers can harvest on average early rabi and kharif 1 vegetable production on 36–50t of tomato, aubergine and cabbage on one medium-high land. hectare of land. In some floodplain regions, farmers are also Production system producing other vegetables (e.g. cucumber, sweet Farmers make covered trellises to overcome the gourd, ash gourd and bitter gourd) in poly bags impacts of heavy rainfall and late floods.These are during the early kharif l season and transplanting typically located near the homestead. Soil from on to raised pits of water hyacinth and soil. high lands is mixed with cow dung following the recommended ratio (60:40) and then kept for Most suitable geographic area seven to ten days for decomposition. After that, The most suitable area for this practice is in the mixture is placed in polythene bags (7.5 x flood-prone areas and coastal areas where irriga- 12.5cm) and two to three seeds are sown per tion is available. This is currently being practised poly bag in June to mid-July. These poly bags are in Faridpur, Gopalganj, Barisal and Habiganj. then set up on the trellis.Within a few days, seeds are germinated and kept in the trellis for 30–45 Major advantages days before transplanting on the main land. This coincides with the flood water receding and land This option helps to increase the overall produc- thus becomes favourable for ploughing. Farm- tion of vegetables. This technique helps to over- ers prepare the ï¬?eld and transplant the seedlings come the typical water-logging situation which in August to September or as early as possible. constrains early vegetable production during the Farmers typically cultivate tomato, aubergine, rabi season. Farmers also get byproduct materi- cabbage, etc. using this approach. als that can be used as cattle fodder. Farmers can Fertilizer doses for some common vegeta- sell the vegetable seedlings themselves in the local bles: markets at a high price. • Tomato: Urea-TSP-MP and cow dung = Major disadvantages 550kg–450kg–250kg–10000kg; Signiï¬?cant delay in water recession on medium- • Aubergine: Urea-TSP-MP and cow dung = high land may delay the transplant time which 375kg–150kg–250kg–10000kg; can decrease yields. Long-lasting drought condi- • Cabbage: Urea-TSP-MP and cow dung = tions during the rabi season might increase irri- 250kg–150kg–200kg–15000 kg. gation costs and decrease yields. 96 Climate Change Risks and Food Security in Bangladesh Approximate beneï¬?ts age can harvest 36–50t of vegetables at a market The resources required include quality vegetable price of Tk12 (US$0.17) per kg, grossing a total seeds, fertilizers, insecticides, pesticides and suf- of Tk432,000 (US$6131). Thus, farmers can earn ï¬?cient medium-high land. A total of Tk165,575 a net proï¬?t of Tk266,425 (US$3781) from one (US$2350) (including land-lease value) is needed hectare of land which normally would not be cul- to cultivate early rabi vegetables on one hec- tivated due to the late recession of flood waters. tare of land using this option. Farmers on aver- Adaptation Options in the Agriculture Sector 97 9 Zero-tillage maize cultivation (NPKSZnB) for high yields and 175kg– 40kg–100kg–20kg–2kg–1kg/ha for moder- Summary ate yields; • Apply all of the PKS and 50 per cent N as Maize is a crop that grows all year. The yield of the base dose and the remaining N in two maize is comparatively higher than rice, wheat or equal instalments at 30 and 55 days after sow- any other cereal crops. In the Barind tract areas, ing along the maize rows; farmers often harvest and keep t. aman rice on • Keep single maize seeding per hill at 20 days the ï¬?eld to dry for two to three weeks. This is after sowing and use thin plants as fodder; then collected for threshing to separate the rice • Green maize cobs can be harvested depend- from the straw. During this time, the soil loses ing on need. Mature cobs are harvested at 127 its moisture due to the high rate of evaporation days after sowing; and the land becomes hard. This situation makes • Yields of about 4t/ha can be harvested with it difï¬?cult for farmers to plough the land and local composite. Hybrid yields are almost therefore in many cases the land remains fallow. double. Farmers can easily cultivate maize by using the existing soil moisture in the fallow land during It is to be noted that the land should have the the rabi season following a zero-tillage system. necessary residual moisture to germinate maize seeds. Otherwise, a minimum irrigation may be Production package needed before sowing the maize seeds or after For cultivating maize using the existing soil germination. After germination of maize seeds, moisture on fallow lands, a crop calendar needs farmers plough the land by using a spade/hoe and to be identiï¬?ed. The exact period of transplan- then apply organic and chemical fertilizer mixed tation of t. aman seedlings and harvesting dur- with soil. Farmers may irrigate the ï¬?eld accord- ing the kharif 2 season must be determined and ing to the degree of soil moisture and availability coordinated with the maize cultivation. Maize is of irrigation water to increase the production of typically cultivated during the rabi season (seed- maize. sowing time in mid-November to early Decem- ber). Harvested t. aman rice should not be kept Most suitable geographic area on the ï¬?eld. Maize seeds are sown immediately In addition to drought-prone areas, the zero-till- after harvest using a ‘dribbling’ approach whereby age maize cultivation option is also suitable in the a few seeds are sown using a sharp stick or ï¬?n- coastal saline and non-saline areas and the central ger. The following steps need to be followed for floodplains. This is currently being practised in zero-tillage maize cultivation: Rajshahi and Nawabganj. • Weeds and other debris are cleared from the ï¬?eld before sowing seeds; Major advantages • Two to ï¬?ve maize seeds are dribbled per hill This option helps to increase the cropping inten- on no-till muddy soil in 25cm intervals in sity and decrease the area of fallow land. This rows 70cm apart; system also helps to increase family income and • Fertilizers are applied at the following doses: consumption. Farmers also get byproducts (e.g. 250kg–200 kg–185 kg–105kg nitrogen- stem and cob) which can be sold in the market phosphorus-potassium-sulphur (NPKS) per and/or used to meet family fuel demands. Some ha land in bands along the maize rows. Fer- farmers also use the stem to make fences to pro- tilizers rate for hybrid varieties is 250kg– tect homestead-based vegetable gardens. Finally, 50kg–140kg–40kg–4kg–2kg/ha of nitrogen- during the vegetative stage of maize, the leaves phosphorus-potassium-sulphur-zinc-boron can be used as cattle fodder. 98 Climate Change Risks and Food Security in Bangladesh Major disadvantages Tk98,315 (US$1395) (including land-lease cost) Maize can uptake many nutrients, which decreases is required to cultivate maize per hectare of land. overall soil fertility and productivity. Therefore, Farmers can harvest 7.5t of maize. Thus, farmers it should not be used as a mono crop. Presently can earn net Tk30,685 (US$435) from one hec- maize is used only in the poultry sector for poul- tare of land that normally would remain fallow try feed. during the rabi season. Approximate beneï¬?ts Quality seeds, hoe/spade, labour, fertilizers and pesticides are required. A total of approximately a Adaptation Options in the Agriculture Sector 99 10 Chickpea cultivation using a priming and fuel demand and can potentially be used to technique generate extra income. The chickpea is also a leguminous family plant which absorbs nitrogen and can help increase overall soil productivity. Summary For this reason, the chickpea plant should only Farmers typically sow chickpea seeds directly in be cut after harvest and not entirely uprooted. the ï¬?eld. However, this is not optimal under low Moreover, continuous cultivation of chickpeas soil-moisture conditions. Priming is one approach helps to decrease demand for urea for the next to address this. In the priming technique, seeds crop. Finally, up to 10–15 crop days are saved by are soaked in water for a period of time based on priming. the thickness of the seed coating. In drier areas, chickpea seeds are soaked at night for a period of six to eight hours. Then the soaked chickpea Major disadvantages seeds are spread in a shaded place where there is In the Barind tract areas, farmers are spreading enough air movement for air drying before sow- cut t. aman paddy on the ï¬?elds for prolonged ing in the ï¬?eld. After sowing/seeding, the land is periods thereby increasing the loss of soil mois- ploughed and levelled to preserve the moisture. ture. This affects the germination of seeds. More- Good tillage is essential for moisture preserva- over, in some cases, land cannot be ploughed due tion. Primed seeds will be germinated after four to heavy soil and, as a result, most of the land to ï¬?ve days. remains fallow. Most suitable geographic area Approximate beneï¬?ts Priming is practised throughout the country Chickpea seeds, fertilizers, bio-fertilizer and depending on the antecedent soil-moisture con- insecticides are the major inputs for this practice. ditions, recession of flood water and seed-sow- Approximately, a total of Tk57,325 (US$813) ing time. In addition to drought-prone areas, this (included land-lease cost) is needed to cultivate option is most suitable in the central floodplain, chickpeas on one hectare of land. A farmer can coastal saline and non-saline areas. This is cur- typically harvest 1.8t of chickpeas at a sale value rently being practised in Rajshahi, Nawabganj of Tk50 (US$0.70) per kg. Thus, farmers can and Naogaon. generate about Tk90,000 (US$1280) in revenues. Farmers also get dry plant and husk byproducts that can be used as cattle fodder and fuel for fam- Major advantages ily needs. The remainder can be sold in the mar- This option helps to increase pulse production ket for cash. Farmers, thus, can earn net Tk32,675 and increase the cropping intensity. In addition (US$463) from one hectare of land during the to the crop, farmers also get plant and pulse husk rabi season. byproducts which can be used to meet fodder 100 Climate Change Risks and Food Security in Bangladesh 11 Supplementary irrigation of t. aman in these ponds from July to November for both from mini ponds household consumption and market sales. This is an important source of protein for rural com- munities. Summary In the absence of water for irrigation, rainwa- Major disadvantages ter harvesting in mini ponds for supplementary irrigation of t. aman during the dry period is an Excavation of a pond within the crop land option. These mini ponds are typically excavated reduces the area for crop cultivation. Moreover, within the crop land. Farmers use comparatively due to the soil erosion in the Barind tract, these lower areas to dig these ponds. ponds may become silted in a few years, requir- ing re-excavation. Most suitable geographic area Approximate beneï¬?ts In addition to the Barind tract areas, this prac- Rice seeds, fertilizers, insecticides, pesticides tice can be implemented in char land and coastal and land are the major inputs to implement areas depending on the availability of water for this option. Approximately a total of Tk76,705 irrigation from other sources. This is currently (US$1088) (included land-lease cost) is needed being practised in Rajshahi. to cultivate t. aman rice on one hectare of land using this option. Farmers can harvest 4.5t of Major advantages t. aman and straw (14,500 bundles) for a sale This option helps to decrease the loss of t. aman value of Tk90,000 and Tk10,875 (US$1277, production by providing supplemental irrigation. US$154) respectively when the unit prices are This decreases the dependency on groundwater Tk20 (US$0.30) per kg and Tk0.75 (US$0.01) for irrigation. It is reported that having access to per bundle. Farmers can also harvest ï¬?sh. Thus, supplemental irrigation can improve rice yields farmers can earn net Tk24,170 (US$343) from by up to 23 per cent and net economic proï¬?t one hectare of land which would normally not by 75 per cent (FAO LACC Project). Farmers be possible. can also cultivate rapid-growth varieties of ï¬?sh Adaptation Options in the Agriculture Sector 101 12 Year-round homestead vegetable Table 7.4 Common vegetable cultivation patterns cultivation Bed Rabi Kharif 1 Kharif 2 No. (Mid Oct–Mid Mar) (Mid Mar–Mid July) (Mid July–Mid Oct Summary 1 Tomato Okra Data (katua data) 2 Lalshak, aubergine Indian spinach Indian spinach Each farm family typically has about 30 decimals 3 Lalshak, aubergine Kang kong Jute vegetable of land around the homestead. Farmers can eas- 4 Radish Okra, lalshak Onion, lalshak ily use this fallow land for cultivating vegetables 5 Batishak, country Chilli, lalshak Chilli around the year to fulï¬?l family requirements. Any bean surplus can be sold in the market to increase fam- Note: No. of beds no. of crops may vary depending on space availability ily income. Homestead vegetable cultivation is an employment-generation activity for women and Cultivation of runner-type vegetables on the roof children, a source of additional income and also increases vegetable and fruit consumption. Sweet gourd, bottle gourd, ash gourd, country bean and other runner-type vegetables can be Production package cultivated on the roof of the house. A farmer can earn about Tk1000 (US$14) per year from the Cultivation of vegetables in open space of the sale of excess vegetables. Only Tk200 (US$2.80) homestead (bed method) is required for seeds and materials. Five beds, each 3m in length, 1m wide and 20cm in height, should be prepared in a sunny and open Cultivation of runner-type vegetables at edge/bank of space near the homestead. Three kg of decom- pond posed cow dung or recommended fertilizers of Bamboo poles can also be planted at a distance speciï¬?c vegetables should be applied thoroughly of about 1.5–2m from the banks/edge of a to each bed before the sowing of seeds or plant- homestead pond to make a trellis for the culti- ing of seedlings. The cultivation pattern of year- vation of runner-type vegetables. This method round vegetables is shown in Table 7.4. has some advantages of utilizing unused spaces The farmer needs only Tk1500 (US$21) per near a pond as well as providing shade for ï¬?sh. year to cover all cost related to vegetable cultiva- The farmer typically needs about Tk500–700 tion in these ï¬?ve beds.Typically,Tk2500 (US$35) (US$7.10–9.90) for the total cost of materials. extra can be generated after the family require- About Tk2500 (US$35) can be earned from the ments are fulï¬?lled. sale of extra vegetables in local market after the family requirements are fulï¬?lled. Cultivation of runner-type vegetables on a platform (trellis) Cultivation of papaya in land surrounding the homestead A trellis (5m x 4m) is made with bamboo, jute and edges/banks of pond sticks and string and located in a sunny space near Papaya can be cultivated on the edges/banks of a the homestead. Country bean, long yard bean, ash pond or surrounding the homestead by digging gourd, bottle gourd, bitter gourd, snake gourd, holes with the dimension 60 x 60 x 60cm, 2m cucumber and other runner-type vegetables can apart from each other. 10kg cow dung, 500g TSP, be cultivated on this trellis all year round. The 250g MP, 50g boron fertilizer and 20g zinc sul- farmer needs approximately Tk600 (US$8.50) phate mixing are used with the soils. After plan- for the bamboo, necessary fertilizers and seeds to tation of saplings, 50g urea and MOP fertilizer cultivate vegetables. The farmer can earn about should be applied to each plant per month, with Tk1800 (US$25) per year by selling extra vegeta- this doubled during the flowering stages. Farm- bles after the family requirements are fulï¬?lled. ers can earn about Tk5000 (US$70) from 100 plants. Initial production costs are about Tk2200 (US$31). 102 Climate Change Risks and Food Security in Bangladesh Most suitable geographic area may result in decreases in the availability of cow This adaptation option is suitable in many dif- dung. ferent areas. In particular, flood-prone, drought- prone, haor (wetland system) and beel (floodplain Approximate beneï¬?ts pond) areas are most suitable. This is currently Vegetable seeds, fertilizers, bio-fertilizer, insecti- being practised to varying degrees throughout cides, bamboo, jute sticks and rope are the major the country, but limited in scale. materials needed to implement this option. Approximately a total of Tk10,000 (US$142) Major advantages (considering a 20 decimal homestead area for each family) is needed to cultivate different sea- This option increases family consumption of veg- sonal runner-type vegetables, leafy vegetables and etables and fruits and improves family income. fruits. The farmer will typically get Tk25,000 to Tk35,000 (US$354–496) from the sale of extra Major disadvantages vegetables. Farmers may also get plant material Due to the intensive use of land for cultivating that can be used as cattle fodder. Farmers can vegetables and fruits, this option may reduce earn Tk15,000 to Tk20,000 (US$212–283) net the overall area devoted to cattle grazing. This from a single farm homestead. Adaptation Options in the Agriculture Sector 103 13 Pond-water harvesting for irrigation to applied three weeks and ï¬?ve weeks after planting cultivate rabi vegetables of the vegetable seedlings. Given that these are high value crops, farmers must pay attention to Summary weeding, irrigation, insecticide and pesticide use. About 36–40t of tomato can be harvested from From October to May, rainfall is very low. The one hectare of land. demand for vegetables during the rabi season is high. Winter vegetables require irrigation Most suitable geographic area for maintaining soil moisture at different stages (especially during the vegetative, flowering and The flood-prone and coastal areas are the most fruit-bearing stages). Supplementary irrigation suitable. This is currently being practised in can play a vital role in production of vegetables. Rajshahi, Nawabganj, Naogaon and Natore. In the Barind areas, some farmers are cultivating rabi vegetables with irrigation from mini ponds. Major advantages Typically, rainwater is harvested or surface water This option helps to increase vegetable produc- is diverted into mini ponds during the monsoon tion and consumption for the family. In drought- and then used to grow vegetables during the rabi prone areas, it is very difï¬?cult to cultivate early season. This practice helps to harvest vegetables rabi vegetables without following this option. earlier in the season. This reduces the area of fallow land. After the harvest of vegetables, farmers can grow boro rice Production package or high value horticultural crops (e.g. onion, gar- The fertilizer dose for tomato is lic). Farmers can also culture ï¬?sh in these ponds for both household consumption and for sale in Urea-TSP-MP and Cow dung = 550kg– local markets. 450kg–250kg–10,000kg; Major disadvantages for aubergine it is urea-TSP-MP and Cow In the absence of rainfall and the unavailability dung = 375kg–150kg–250kg–10,000kg. of surface water, farmers may fail to harvest rain- water in the mini pond. Prolonged droughts may Farmers select land nearest to the pond and pre- dry out the pond. pare the land by ploughing and applying half of the cow dung and the total amount of TSP ferti- Approximate beneï¬?ts lizers.After preparation of the land, beds are raised The resources required for this option include with a width of 1m, a height of 25–30cm and a quality vegetables seeds or seedlings, fertilizers, length depending on land size; 30–40cm of space insecticides, pesticides and land with a mini pond. is kept between beds for irrigation and drain- Tk151,575 (US$2151) is needed to cultivate early age of excess rainwater. After levelling the beds, vegetables during the rabi season on one hectare farmers transplant the seedlings, usually from of land. A farmer can harvest 24.5t of vegetables mid-September to November, following the sug- which, at a market price of about Tk12 (US$0.17) gested spacing for speciï¬?c vegetables (tomato = per kg, will gross Tk294,000 (US$4173). Thus, 60cm x 40cm; aubergine = 75cm x 60cm). The farmers can earn net Tk142,425 (US$2021) from remainder of the cow dung is used during trans- one hectare of land. planting of seedlings. Urea and MP fertilizers are 104 Climate Change Risks and Food Security in Bangladesh 14 Sorjan system for cultivating seasonal Table 7.5 Common vegetable cropping patterns for sorjan system vegetables, fruits and ï¬?sh Bed No. Cropping patterns on Crops on bed edge bed tops Summary 1 Amaranth, okra, red Bitter gourd, hyacinth bean amaranth, tomato Tidal surges, water-logging, saline water intru- sion and soil salinity increases due to sea level 2 Indian spinach, vegetables Bitter gourd, hyacinth bean rise make the production of vegetables and fruits seedlings for cabbage, cauliflower difï¬?cult. Demand for these crops is typically met from outside supply. For this option, farmers can 3 Papaya, chilli, red amaranth Ribbed gourd, hyacinth bean grow vegetables and fruits on raised beds and 4 Banana, kang kong, red Ribbed gourd, marma creeper vegetables on the edges to meet day-to- amaranth, aubergine day demands. In addition, farmers can earn extra 5 Banana, amaranth, red Snake gourd, bitter gourd cash from the sale of remainder vegetables and amaranth, aubergine fruits at the local market. Major advantages Production techniques This option increases the cropping intensity and • A model sorjan has ï¬?ve raised beds (3m wide decreases the area of fallow land. Family con- each) and six ditches (2m wide and 1.5m sumption of fruits and vegetables will increase, deep). It may be increased or decreased based as will income. on land size and shape. Normally a 28m x 11m piece of land and clayey soil is most suitable Major disadvantages for making these raised beds and ditches. The dry months (January–March) are best for pre- Various extreme events may increase the risk of paring the sorjan (see Figure 7.1, Plate 7.3). inundation and affect the layout of the sorjan sys- • Slope of the bed is made uniform and com- tem. Bed heights consequently may need to be pact; raised, which increases the production costs. • Trellises are made with bamboo and other local materials on the furrows to support Approximate beneï¬?ts creeper vegetables. In the sorjan system the following resources may • Fish may be cultured in the ditches during be required: vegetable seeds, fruit saplings, ferti- wet months. lizers, bamboo, jute sticks, a spade and others. A total of Tk7795 (US$110) is needed to cultivate Cropping patterns vegetables, fruits and ï¬?sh in the coastal zone using this system. Farmers can typically harvest 400– Two beds can be earmarked for vegetable cultiva- 20kg, 130kg and 18kg of vegetables, fruits and tion and the rest for fruits and vegetables. The ï¬?sh respectively. This results in a gross value of edge of the beds can be planted with creeper vege- Tk17,650 (US$250) when the market price per tables, which can be supported by the trellis over kg is Tk20 for vegetables, Tk125 (US$1.80) for the furrows. The cropping patterns in Table 7.5 ï¬?sh and Tk55 (US$0.78) for fruit (zuzubi). Each could be used for the ï¬?ve beds (as an illustration). year, farmers can also get some fuel each year from pruning the zuzubi stem. Thus, farmers can Most suitable geographic area earn net Tk9855 (US$139) from one model sor- The coastal areas (both saline and non-saline), jan (11m x 28m) in the coastal zone which would depending on the degree of flooding and tidal normally remain fallow. Over a hectare of land, surges, are the most suitable. This is currently the production costs are Tk253,084 (US$3592), being practised in Barisal, Gopalganj, Pirojpur gross income is Tk573,052 (US$8133), and the and Jhalokathi. net income is Tk319,968 (US$4541). 8 The Way Forward – Turning Ideas into Action The year 2007 was indicative of the challenges agricultural research and development; promote that Bangladesh faces to achieving food security. education and skills development; increase access Severe flooding from July to September 2007 to ï¬?nancial services; enhance irrigation efï¬?ciency affected over 13 million people in 46 districts and overall water and land productivity; strengthen and caused extensive damage to agricultural pro- climate risk management; and develop protective duction and physical assets (e.g. housing, embank- infrastructure. Continued developmental plan- ments). With hardly any time to recover, on 15 ning and investment is needed to build resilience November 2007 Cyclone Sidr made landfall at both national and local scales. across the southern coast of the country, causing This study is largely focused on the impacts of over 3000 deaths.The total economic damages of climate risks (both climate variability and climate these two events amounted to over US$1 billion change) on food security in Bangladesh.The risks US (World Bank, 2008). Moreover almost 2 mil- from climate change include higher temperatures lion tonnes of rice were lost, putting government and changing precipitation patterns, increased cereal stocks in a precarious situation. Finally, that flood intensity and frequency, droughts and sea same year the unabated increase in the interna- level rise effects on agriculture production. The tional prices of oil and food, of which Bangladesh future is also expected to bring elevated CO2 con- is a net importer, put further strains on both gov- centrations, which have a beneï¬?cial effect on crop ernment budgets and household livelihoods. growth. A suite of models is used to approach What these events demonstrated was the the complex questions of climate variability and inherent vulnerability of Bangladesh to climate change and food security in Bangladesh. In this risks. It also showed the degree to which food study, future climate is projected using the most security remains a challenge for the country. Cli- recent climate science available. Detailed hydro- mate change has the potential to signiï¬?cantly logic models project changes in future flooding. affect Bangladesh’s efforts to provide food to a Country speciï¬?c data is used to derive more growing nation. The challenges that the agricul- realistic and accurate agricultural impact func- ture sector will face as it adapts to climate change, tions and simulations. Finally, a dynamic CGE is however, coincide well with the needs required used to better understand the degree to which to address the climate variability risks of today. economic effects will buffer against the physical Both processes of adapting to climate change and losses predicted from climate variability and cli- stimulating the agriculture sector to achieve rural mate change. Signiï¬?cant impacts on growth and growth and support livelihoods require efforts to, household consumption are projected. among other things: diversify household income The models used here are among the best sources; improve crop productivity; support greater mathematical representations available of the 106 Climate Change Risks and Food Security in Bangladesh physical and economic responses to these exog- uncertain future changes. This, however, is not a enous climate changes. However, like all modell- cause for inaction. Rather, a no-regrets strategy ing approaches, uncertainty exists as parameters is to promote activities and policies that help the may not be known with precision and functional national government and households build resil- forms may not be fully accurate. Thus, careful ience in the agriculture sector to existing climate sensitivity analysis and an understanding and risks today. This aligns well with existing devel- appreciation of the limitations of these models opment strategies and plans. By doing so, the are required. Further collection and analysis of country and households will be better prepared critical input and output observations (e.g. cli- for whichever future outcome materializes. The mate data, farm-level practices and irrigation adaptation options identiï¬?ed are only a small constraints) will enhance this integrated frame- sub-set of what can be done today. work methodology and future climate impact assessments. The southern and northwestern regions are the most Some key messages that emerge from this vulnerable. This is due to the confluence of several study include: different climate risks and existing poor baseline conditions The impacts of existing climate variability are enormous. From a socio-economic perspective, the south A no-regret strategy is to focus ï¬?rst on the near-term and northwest regions have long been areas of climate risks to build future resilience extensive poverty. As these communities cur- Bangladesh is clearly one of the most vulnerable rently in many cases exist on the margins, both countries to climate risks today. The agriculture climate variability and climate change threaten to sector is impacted by annual flooding, water increase these vulnerabilities. The sub-regions in shortages during the dry season and frequent the south sit at the confluence of multiple climate coastal cyclones and storm surges. Though the risks. These areas experience the largest decline relative severity of these disasters has decreased in rice production due to climate change. This substantially since the 1970s, these remain criti- is for three reasons. First, these regions already cal challenges to rural poverty and growth. experience signiï¬?cant declines in aus and aman Thus, continued substantial public investment rice production due to climate variability, which in protective infrastructure (e.g. cyclone shelters, is expected to worsen under climate change. embankments), early warning and preparedness Second, boro yields are severely affected by the systems and programmes to build resilience at effects of changes in mean rainfall and tempera- the household level (e.g. income diversiï¬?cation, ture. Thus, reductions in boro production limit identiï¬?ed adaptation options in this study) can the ability for these regions to compensate for play a critical role in minimizing these impacts. lost aus and aman rice production during extreme The impacts of future changes in temperatures events. Projections, moreover, are conservative as and precipitation, increased CO2 levels, flooding, access to irrigation is assumed limitless.Third, the droughts and sea level rise are highly uncertain. In south is also affected the most by rising sea lev- some cases not only is the magnitude of change els, which permanently reduces cultivable land. not known with precision, but also the direction The largest percentage declines in per capita of change. Despite this, the simulation estimates consumption will be in these regions. Similarly, in this study suggest that the simulated cumula- the northwest region is particularly vulnerable as tive economy-wide impacts of existing climate impacts have a disproportionate share relative to variability alone (US$594 billion) are almost the low existing household consumption. These ï¬?ve times that of climate change (US$129 bil- two areas are where priority must be given and lion). That is, the existing inter- and intra-annual where substantial opportunities for adaptation variation is signiï¬?cantly larger than the projected are possible. The Way Forward – Turning Ideas to Action 107 Increased investments in adaptation in the agriculture livelihoods and develop sustainably. As popula- sector are critical to ensuring continued growth and tions grow, the ability for many countries to poverty alleviation meet basic food requirements and effectively Bangladesh will continue to depend on the manage future disasters will be critical for sus- agriculture sector for economic growth. Rural taining long-term economic growth. These are households will continue to depend on the agri- challenges above and beyond those that many culture sector for income and livelihoods. Floods, countries are already currently facing. Moreover, droughts and cyclones will continue to affect the most developing countries lack the ï¬?nancial and performance of the agriculture sector.Though the technical capacities to manage these increasing government has made substantial investments to risks. Thus, strategic prioritization and improved increase the resilience of the poor (e.g. new high- planning and management of existing assets and yielding crop varieties, protective infrastructure, budget resources are critical. Largely, these strate- disaster management), as has been shown these gic choices will be dependent on the economics variability impacts may be exacerbated by long- of these impacts. term effects of climate change. The integrated framework used in this analy- Households have for a long time adapted to sis provides a broad and unique approach to esti- these dynamic conditions to maintain their liveli- mating the hydrologic and biophysical impacts hoods. The nature of these adaptations and the of climate change, the macro-economic and determinants of success depend on the availabil- household-level impacts and an effective method ity of assets, resources, labour, skills, education for assessing a variety of adaptation practices and social capital. The adaptation options iden- and policies. The framework presented here can tiï¬?ed, in fact, are currently being implemented serve as a useful guide to other countries and in many locations with assistance from both the regions faced with similar development chal- government and donor community. However, lenges and objectives of achieving food security. the scale of these efforts remains limited and is In assessing the impacts, several different mod- not commensurate with the probable impacts. elling environments must be integrated to pro- Moreover, the current large gap between actual vide a more nuanced and complete picture of and potential yields suggests substantial on-farm how food security may be impacted by climate opportunities for growth and poverty reduction. change. This approach is needed to better under- Expanded availability of modern rice varieties, stand the relative impacts from multiple climate irrigation facilities, fertilizer use and labour could risks (e.g. floods, droughts, climate change) and increase average yields at rates that could poten- how these relate in the context of an evolving tially more than offset the climate change impacts. socio-economic baseline (e.g. population, prices, Signiï¬?cant additional planning and investments international trade). Moreover, such a framework in promoting these types of adaptations are still allows for extensive scenario analysis to identify needed. and understand key sensitivities. This is critical to making decisions in a highly uncertain future. 8.1 A Framework for Assessing Finally, through this integration of multiple disci- plines, a richer and more robust set of adaptation the Economic Impacts of Climate options and policies for the agriculture sector Change can be identiï¬?ed and tested. Continued reï¬?ne- ments to the assessment approach developed in The precise impact of climate change on coun- this volume will further help to sharpen criti- tries in the developing world remains to be seen. cal policies and interventions by the Bangladesh This much is known, however: climate change government. poses additional risks to many developing coun- tries in their efforts to reduce poverty, promote Annex 1 – Using DSSAT to Model Adaptation Options The following sections detail adaptive responses Table A1.1 Cultivars available in the DSSAT v4.5 CERES-Rice that may be tested according to the capabilities of model the DSSAT models (Hoogenboom et al, 2003). 1 IRRI Originals 24 RD 23 (cal.) Arranged to approximate the planning process for 2 IRRI Recent 25 CICA8 any given year, farm-level practices are described 3 Japanese 26 Low Temp. Sen that may affect the selection of the crop and 4 N. American 27 Low Temp. Tol variety to grow, the sequence and timing of the 5 IR 8 28 17 BR11, t. aman 6 IR 20 29 18 BR22, t. aman cropping calendar, decisions relating to how each 7 IR 36 30 19 BR3, t. aman ï¬?eld is planted and what types of applications are 8 IR 43 31 20 BR3, boro made to adjust for mid-season deï¬?ciencies. The 9 Labelle 32 CPIC8 beneï¬?ts of regional-scale programmes to reduce 10 Mars 33 Lemont external hazards from damaging crops are also 11 Nova 66 34 RN12 explored and additional factors are identiï¬?ed that 12 Peta 35 TW will affect the interpretation of DSSAT simula- 13 Starbonnett 36 IR 64 14 UPLRI5 37 Heat Sensitive tion results. 15 UPLRI7 38 BR14 16 IR 58 39 IR 72 Cultivar Selection 17 18 SenTaNi IR 54 40 41 BR11 Pant-4 19 IR 64 42 Jaya Biology 20 IR 60 (Est) 43 BPRI10 21 IR 66 44 Zheng Dao 9380 In addition to deciding which species (rice, wheat, 22 IR 72x 45 CL-448 etc.) to grow, speciï¬?c cultivars within that species 23 RD 7 (cal.) 46 PR114 may be more or less adapted to future climate Note: Cultivars known to be grown in Bangladesh are in bold. The Dhan-29 in a particular location. The DSSAT CERES- was added to DSSAT for this study based upon calibrations at the Bangladesh Rice model contains 46 different cultivars (Table Agricultural Research Council. A1.1), including some known to be currently used in Bangladesh. (Table A1.4). In addition to testing the range of Each of these cultivars is represented as a known cultivars for a particular location, the sen- collection of 8 genetic coefï¬?cients affecting sitivity of any particular genetic coefï¬?cient may different aspects of growth and environmental be assessed using hypothetical cultivars that may resilience (Table A1.2). Similarly, the DSSAT serve as models in the engineering of new breed- CERES-Wheat model contains 10 cultivars ing programmes. Additional biological resilience (Table A1.3) described by 40 genetic coefï¬?cients may be simulated by adjusting salinity levels or Annex 1 – Using DSSAT to Model Adaptation Impacts 109 Table A1.2 Genetic coefï¬?cients in the DSSAT v4.5 CERES-Rice Table A1.4 Genetic coefï¬?cients in the DSSAT v4.5 CERES- model Wheat model 1 P1 Time period for basic vegetative phase 1 P1V Days at optimum vernalizing temperature required 2 P20 Longest day length at which the development occurs at to complete vernalization a maximum rate 2 P1D Percentage reduction in development rate in a 3 P2R Extent to which phasic development leading to panicle photoperiod 10 hours shorter than the threshold initiation is delayed for each hour increase in photoperiod relative to that at the threshold above P20 3 P5 Grain ï¬?lling (excluding lag) phase duration 4 P5 Time period from beginning of grain ï¬?lling to 4 G1 Kernel number per unit canopy weight at anthesis physiological maturity 5 G2 Standard kernel size under optimum conditions 5 G1 Potential spikelet number coefï¬?cient 6 G3 Standard, non-stressed dry weight of a single tiller 6 G2 Single grain weight under ideal conditions at maturity 7 G3 Tillering coefï¬?cient 8 G4 Temperature tolerance coefï¬?cient 7 PHINT Time interval between successive leaf tip appearances 8 AWNS Awn score Table A1.3 Cultivars available in the DSSAT v4.5 CERES-Wheat 9 ECONO Code for the ecotype model 10 GRNMN Minimum grain N 11 GRNS Standard grain N 1 Spring – High Lat 6 Spring – Low Lat 12 HTSTD Standard canopy height 2 Winter – Europe 7 Maris Fundin 13 KCAN PAR extinction coefï¬?cient 3 Winter – USA 8 Newton 4 Winter – Ukraine 9 Manitou 14 LAWRS Lamina area to weight ratio of standard ï¬?rst leaf 5 Facultative 10 Chelsea SRW-US 15 LAWR2 Lamina area to weight ratio, phase 2 16 LA1S Area of standard ï¬?rst leaf Note: The Kanchan and Sowgat cultivars were added to DSSAT for this study 17 LAVS Area of standard vegetative phase leaf based upon calibrations at the Bangladesh Agricultural Research Council. 18 LARS Area of standard reproductive phase leaf 19 LLIFE Life of leaves during vegetative phase the amount of water available for root uptake. 20 LT50H Cold tolerance when fully hardened The simulated direct response function of each 21 NFGL N stress factor, growth, lower species’ growth to carbon dioxide fertilization is 22 NFGU N stress factor, growth, upper independent of their particular cultivars in the 23 NFPL N stress factor, photosynthesis, lower models, although growth rates may be handled 24 NFPU N stress factor, photosynthesis, upper differently by each cultivar. The DSSAT CERES 25 PARUV PAR conversion to dm ratio, before last leaf stage 26 PARUR PAR conversion to dm ratio, after last leaf stage models do not directly simulate pests and diseases. 27 P1 Duration of phase end juvenile to double ridges 28 P2 Duration of phase double ridges to end leaf Location growth The same agricultural practices may also be tested 29 P3 Duration of phase end leaf growth to end spike growth in numerous divisions with varying soil types and 30 P4 Duration of phase end spike growth to end grain water regimes to determine where a speciï¬?c crop ï¬?ll lag can be grown productively. Areas whose yield 31 P4SGE Stem growth end stage underperforms may be tested with alternative 32 RDGS1 Root depth growth rate, early phase cultivars or crops in order to maximize utility for 33 RDGS2 Root depth growth rate, later phases local residents. 34 RSFRS Reserves fraction of assimilates going to stem 35 TI1LF Tillering threshold (leaf number to start tillering) Calendar Adjustment 36 37 WFGU WFPU Water stress factor, growth, upper Water stress factor, photosynthesis, upper 38 WFPGF Water factor, genotype sensitivity to stress when Planting and harvesting dates grain ï¬?lling 39 TBGF Temperature base, grain ï¬?lling The cropping calendar used by farmers in Bang- 40 P1DPE Day length factor, pre-emergence ladesh has been developed to take advantage of 110 Climate Change Risks and Food Security in Bangladesh current climate. As climate shifts occur, how- Table A1.5 Planting method options in the DSSAT v4.5 models ever, so may the optimal planting and harvesting 1 Bedded 6 Nursery schedules. In order to determine the sensitivity of 2 Cutting 7 Pre-germinated seed the yield to the planting date, a series of DSSAT 3 Dry Seed 8 Ratoon model simulations may be run with incremental 4 Horizontally planted sticks 9 Transplants planting dates over the course of several weeks to 5 Inclined (45°) sticks 10 Vertically planted sticks months. Harvest dates may be determined auto- Note: Known common practices in Bangladesh is in bold. matically by the models or speciï¬?ed according to local requirements. Table A1.6 Tillage implements available in the DSSAT v4.5 models Sequence 1 Animal-drawn implement 18 Fertilizer applicator, anhydrous Bangladesh is one of the few countries in the 2 Bedder 19 Harrow, spike world that is able to have three planting sea- 3 Blade cultivator 20 Harrow, tine sons in a given year. In the future, however, the 4 Chisel plough, straight point 21 Lister 5 Chisel plough, sweeps 22 Manure injector sequence of crops and fallow periods may need 6 Chisel plough, twisted shovels 23 Matraca hand planter to be adjusted to maximize yield. The DSSAT 7 Cultivator, ï¬?eld 24 Moldboard plough models allow crop sequences to be tested over an 8 Cultivator, ridge till 25 Mulch treader annual period or even in multi-year cycles. Future 9 Cultivator, row 26 Plank stresses may require increased crop rotation with 10 Disk plough 27 Planter, no-till legumes to replenish nutrients in the soils, and 11 Disk, 1-way 28 Planter, row the treatment of crop residuals and fallow peri- 12 Disk, double disk 29 Planting stick (hand) 13 Disk, tandem 30 Rod weeder ods may have large effects on nitrogen and water- 14 Drill, deep furrow 31 Roller packer cycle processes. The potential for changing crop 15 Drill, double-disk 32 Rotary hoe sequence and rotations in response to the climate 16 Drill, no-till 33 Subsoiler change scenarios may therefore be tested using 17 Drill, no-till into sod 34 V-Ripper DSSAT models. spacing and planting depth may also be deter- Planting Systems mined. Together, there is a wide range of poten- tial planting methods that may be tested under Method climate change scenarios. In addition to the planting date discussed in the previous section, the DSSAT CERES-Rice Inputs model recognizes several planting options that may be adapted to future climates. Rice may be Irrigation planted according to ten different options (Table A1.5), ranging from dry seed to inclined sticks to If irrigation is available, simulated applications transplants (as is common for aman production may be made according to a set schedule or in Bangladesh). The environment in which the automated according to need – deï¬?ned accord- transplants are grown and their age and weight ing to the percentage of saturation in a soil layer may be adjusted. Thirty-four tillage options may extending down to a particular depth. An option also be tested for optimization (Table A1.6). exists to build a bund around rice plots to retain water, and an application may be made according to any of 11 methods; applied either as a given Density quantity or until the soil reaches a particular level Planting may be done at varying density in a uni- of saturation (Table A1.7). An efï¬?ciency factor form distribution, in rows or on hills. The row may be adjusted to represent lost runoff, and irri- Annex 1 – Using DSSAT to Model Adaptation Impacts 111 Table A1.7 Irrigation options in the DSSAT v4.5 models Table A1.8 Fertilizer types in the DSSAT v4.5 models 1 Alternating furrows 7 Furrow 1 Ammonium nitrate 13 Liquid phosphoric acid 2 Bund height 8 Percolation rate 2 Ammonium nitrate 14 Monoammonium 3 Constant flood depth 9 Puddling (for rice only) sulphate phosphate 4 Drip or trickle 10 Sprinkler 3 Ammonium 15 Potassium chloride 5 Flood depth 11 Water table depth polyphosphate 6 Flood 4 Ammonium sulphate 16 Potassium nitrate 5 Anhydrous ammonia 17 Potassium sulphate Note: Options that may be appealing for use in Bangladesh are in bold. 6 Aqua ammonia 18 Rhizobium 7 Calcitic limestone 19 Rock phosphate gation can be exclusively scheduled for particular 8 Calcium ammonium 20 Single super phosphate growth stages if necessary. Regardless of the types nitrate solution of irrigation provided, the largest difference will 9 Calcium hydroxide 21 Triple super phosphate be between irrigated and rainfed ï¬?elds. The gap 10 Calcium nitrate 22 Urea 11 Diammonium 23 Urea ammonium nitrate that exists between these two options will have phosphate solution profound implications on potential grain yield, 12 Dolomitic limestone 24 Urea super granules demand for surface water resources, and stresses Note: Fertilizers known to be in use in Bangladesh are in bold (Yearbook of on the water table. Agricultural Statistics of Bangladesh, 2005). Fertilizer Table A1.9 Fertilizer and organic amendment application options in the DSSAT v4.5 models In addition to incorporating crop residuals from a previous season’s harvest, the DSSAT models 1 Applied in irrigation water 2 Band on saturated soil, 2cm flood, 92% in soil allow fertilizer applications to be made according 3 Banded beneath surface to a schedule or in an automated manner.Twenty- 4 Banded on surface ï¬?ve different chemical fertilizers (Table A1.8) and 5 Bottom of hole, deep placement 19 application methods (Table A1.9) are available 6 Broadcast on flooded/saturated soil, 15% in soil in the DSSAT models and may be automated 7 Broadcast on flooded/saturated soil, 30% in soil depending on the nitrogen stress at a given soil 8 Broadcast on flooded/saturated soil, 45% in soil level. Fourteen organic amendments may also be 9 Broadcast on flooded/saturated soil, 60% in soil 10 Broadcast on flooded/saturated soil, 75% in soil simulated (Table A1.10). Like irrigation, the price 11 Broadcast on flooded/saturated soil, 90% in soil and availability of fertilizers will largely determine 12 Broadcast on flooded/saturated soil, none in soil their use in the future, so any gains in yield need 13 Broadcast, incorporated to be weighed against increases in cost. 14 Broadcast, not incorporated 15 Deeply placed urea super granules/pellets, 100% in soil Environmental modiï¬?cations 16 Deeply placed urea super granules/pellets, 95% in soil 17 Foliar spray The DSSAT models also allow farm-level envi- 18 Injected ronmental modiï¬?cations that may reduce plant 19 On the seed stresses. For example, water stresses may be amel- iorated by adjusting the rate of soil drainage (e.g. by simulating the addition of a semi-permeable Table A1.10 Organic amendments available in the DSSAT v4.5 material at the base of the soil column), and sched- models uled periods of shading may reduce heat stress. 1 Barnyard manure 8 Maize residue 2 Bush fallow residue 9 Macuna residue 3 Compost 10 Peanut residue Independent options 4 Cowpea residue 11 Pearl millet residue Baseline experiments that are calibrated to his- 5 Crop residue 12 Pigeon pea residue torical division level yields can be run with the 6 Green manure 13 Sorghum residue 7 Liquid manure 14 Soybean residue adjustment of a single practice, allowing the 112 Climate Change Risks and Food Security in Bangladesh sensitivity of yield to each particular approach effectiveness of each approach may be evalu- to be evaluated. Yield sensitivity for each of the ated along with the extent to which the new following management practices (as applicable) management departs from traditional practices. may then be assessed for any given location and Assumptions about water management may also climate scenario: be incorporated to determine irrigation avail- ability and develop strategies to maintain scarce • Crop selection resources. This allows a projected beneï¬?t to be • Known cultivar selection associated with the cost of each simulated adapta- • Genetic coefï¬?cient values tion option. Some potential combined practices • Soil proï¬?le include modiï¬?cations to: • Tillage schedule • Tillage implement • Cultivar selection and irrigation calendar • Tillage depth • Planting density and planting depth • Planting date • Tillage implement and fertilizer implementa- • Planting density tion option • Planting geometry • Cultivar selection, irrigation amount and ferti- • Planting depth lizer type • Planting method • Cultivar selection, planting date, and irriga- • Temperature of transplant environment (if tion amount applicable) • Genetic coefï¬?cient values, planting date, plant- • Transplant age (if applicable) ing method, planting geometry, planting • Transplant weight (if applicable) depth, irrigation calendar, irrigation amount, • Irrigation calendar irrigation type, fertilizer calendar, fertilizer • Irrigation type type, fertilizer amount and fertilizer imple- • Irrigation amount mentation option. • Fertilizer calendar • Fertilizer type Sequential adjustments • Fertilizer implementation option • Organic amendment calendar The DSSAT crop modelling system allows for • Organic amendment type simulations of crop cycling that allows evalua- • Organic amendment implementation option tion of multi-seasonal adaptation strategies. The • Harvest date inclusion of a legume cycle or a multi-seasonal • Environmental modiï¬?cations strategy of fertilizer application may be a cost- • External flood control effective way to maximize yield from a particular • External salinity control. plot of land. The effect of changing one season’s practices may be assessed for lingering impacts on the following seasons using this crop model- Combined practices ling framework. Thus, the cropping strategy of a Once sensitive practices are identiï¬?ed, strategies particular location may be analysed throughout may be developed to combine adapted practices an annual or multi-year cycle of cultivation and for maximum seasonal yield. If multiple strate- fallow periods to maximize sustainability and gies produce similar shifts in yield, the cost- yields. Annex 2 – Description of the CGE Model Tables A2.1 and A2.2 present the equations of Factor incomes are distributed to households a simple closed-economy computable general in each region using ï¬?xed income shares based equilibrium (CGE) model that is used at this on the households’ initial factor endowments stage to illustrate how climate change affects the – Equation 3. Total household incomes Y are economic outcomes examined in this analysis. then either saved (based on marginal propensities The model is recursive dynamic and can there- to save Ï…) or spent on consumption C (accord- fore be separated into a static ‘within-period’ ing to marginal budget shares β) – Equation 4. component wherein producers and consum- Consumption spending includes a ‘subsistence’ ers maximize proï¬?ts and utility, and a dynamic component λ that is independent of income and ‘between-period’ component wherein the model determined by household population H. Savings is updated based on the demographic model and are collected in a national savings pool and used previous period results, thereby reflecting changes to ï¬?nance investment demand I (i.e., savings- in population, labour supply, and the accumula- driven investment closure) – Equation 5. Finally, tion of capital and technology. a single price P equilibrates national product In the static component of the model, pro- markets, thus avoiding the necessity of modelling ducers in each sector s and agro-climatic region inter-regional trade flows – Equation 9. r produce a level of output Q in time period t The model variables and parameters are cali- by employing the factors of production F under brated to observed data from a regional social constant returns to scale (exogenous productivity accounting matrix (SAM) (described in Annex α) and ï¬?xed production technologies (ï¬?xed fac- 3) that captures the initial equilibrium structure tor shares δ) – Equation 1. Proï¬?t maximization of the economy in 2005. Parameters are then implies that factor payments W are equal to aver- adjusted over time to reflect demographic and age production revenues – Equation 2. Labour economic changes and the model is resolved for a supply L, land supply N and capital supply K are series of new equilibriums for the 45-year period ï¬?xed within a given time period, implying full 2005–50. Three dynamic adjustments occur employment of factor resources. Land and labour between periods: changes in land and labour sup- market equilibrium is deï¬?ned at the regional ply; capital accumulation; and technical change. level, so land and labour is mobile across sec- Between periods the model is updated to tors but wages and rental rates vary by region reflect long-term growth rates in land supply N and – Equation 6. National capital market equilib- labour supply L. These are imposed through the rium implies that capital is mobile across both parameters σ and φ – Equations 10 and 11 – which sectors and regions, and earns a national rental remain unchanged across simulations. For capital rate (i.e. regional capital returns are equalized) supply K, the model endogenously determines – Equation 8. the national rate of accumulation – Equation 12. 114 Climate Change Risks and Food Security in Bangladesh Table A2.1 Simple CGE model equations Static model equations ΠF Production function δ fsr 1 Qsrt = αsrt . fsrt Factor payments Wfrt Σ Σ . F = s fsrt s δfsr . Pst . Qsrt 2 Household income Yhrt = Σ θ .W fs hf frt . Ffsrt 3 Consumption demand Pst . Dhsrt = βhsr . (1 – Ï…hr ).Yhrt 4 Investment demand Pst . Ist = Ï?s . Συ hr hr .Y hrt 5 Labour market equilibrium ΣF s fsrt = Lfrt f is labour 6 Land market equilibrium Ffsrt = Nfsrt.(1 + εfsrt).λfrt f is land 7 Capital market equilibrium ΣF rs fsrt = Kft and Wfrt = Wfr 't f is capital 8 Product market equilibrium ΣD hr hsrt = ΣQ r srt + Ist 9 Recursive dynamic equations Labour supply Lfrt = Lfrt–1.(1 + σfr ) f is labour 10 Land expansion Nfsrt = Nfsrt–1 . (1 + Ï•fsr – ηfsrt) f is land 11 Pst–1 . Ist – 1 Σ Capital accumulation f is capital 12 Kft = Kft–1 . (1 + Ï€ – Ï„t) + s K Technical change αsrt = αsrt–1.(1 + γ – µ).(ω . ν) f is labour 13 The level of investment I from the previous period ital returns. Finally, the model captures total factor is converted into new capital stocks using a ï¬?xed productivity through the production function’s capital price κ. This is added to previous capital shift parameter α. The rate of technical change γ stocks after applying a ï¬?xed long-term rate of is exogenously determined based on long-term depreciation Ï€. New capital is allocated to regions trends and is applied to all simulations. and sectors endogenously in order to equalize cap- Annex 2 – Description of the CGE Model 115 Table A2.2 Simple CGE model variables and parameters Subscripts Static model exogenous parameters f Factor groups (land, labour and capital) α Production shift parameter (factor productivity) h Household groups β Household average budget share m GCMs and emission scenarios δ Factor input share parameter r Regions (agro-climatic) θ Household share of factor income s Economic sectors Ï? Investment commodity expenditure share t Time periods Ï… Household marginal propensity to save Endogenous variables Dynamic updating exogenous parameters D Household consumption demand quantity γ Long-run unbiased productivity growth rate F Factor demand quantity κ Base price per unit of capital stock I Investment demand quantity Ï€ Long-run capital depreciation rate K National capital supply σ Long-run labour supply growth rate L Regional labour supply φ Long-run land expansion rate P Commodity price Climate-related exogenous parameters Q Output quantity w Yields: climate-affected deviation from base W Average factor return v Yields: flood-affected deviation from base Y Total household income Ï„ Major flood: additional capital depreciation ε Major flood: land loss from inundation η Major flood: decelerating land expansion µ Major flood: decelerating productivity growth λ Sea level rise: long-run land deviation from base Climate Impact Channels in the yield loss. In the presence of changes, the param- eter falls below one and the yield in a particular Economy-wide Model year is below its long-term trend. The climate- Climate variability and change is imposed on and flood-affect yield deviations can compound the simple economy-wide model via various each other, causing yields in a particular year to climate-related exogenous parameters (see Table fall below long-term trends. However, there are A2.3). The ï¬?rst impact channel is changes in rice no permanent yield losses caused by climate vari- crop yields (wheat is not considered here). The ability during a typical year (i.e. yields can return hydro-crop models produce two parameters that to long-term trend rates in subsequent years). reflect deviations in annual crop yields from its The second impact channel is the additional exogenously-determined long-term trend.1 The economic losses associated with extreme climate ï¬?rst parameter w is the yield deviations caused events. When a major flood year is randomly by changes in climate conditions, including tem- drawn from the climate datasets then there are perature, rainfall and CO2 levels (see Equation four additional ‘extreme event’ impacts that take 13). This climate-affected yield parameter w var- place over and above the climate- and flood- ies around a base year value of one depending affected yield changes described above. First, on the year selected randomly from the climate there is a deceleration in long-term rate of land data series (i.e. values greater than one represent expansion, which is governed by the parameter φ better-than-base-year yields). The second param- – Equation 11. This is achieved by assigning the eter v taken from the hydro-crop models is the offsetting parameter η with the negative value of crop yield deviation caused by changes in mean the long-term land expansion rate (i.e. -φ), thus flooding (see Equation 13). When there are no reducing land expansion during a major flood changes in mean flooding the flood-affected year to zero. Second, there are land losses from yield parameter is equal to one and there is no severe water inundation during major floods. 116 Climate Change Risks and Food Security in Bangladesh Table A2.3 Summary of climate impact channels in economy-wide model simulations Impact channel Affected sectors Description of impact For each year in the Monte Carlo experiment (during the 45-year repeated draws in each simulation) Climate-affected yield impacts ω Aus, aman and boro rice Deviation in rice crop yields from base value due to changes in rainfall, temperature and CO2 levels (0 < ω < ∞). Climate-affected yield impacts ν Aus and aman rice Decline in rice crop yields caused by water-logging (0 < ν < 1). Optimal year has no water logging (i.e. ν = 1). For major flood years when drawn in the Monte Carlo experiment (reflecting conditions in 1970–71, 1974–75, 1984–85, 1987–88, 1988–89, 1998–99) Decelerating crop land expansion η All crops and shrimp ï¬?sheries, Long-term rate of annual land expansion is reduced to zero except for irrigated boro rice,during major flood year (η = -φ). Land expansion continues in wheat, and pulses subsequent period (i.e. η returns to zero). Land inundation from flooding ε Aus and Aman rice Flooding reduces land supply according to observed crop land losses from historical crop production data (ε < 0). Crop land returns to cultivation in subsequent year (i.e. ε returns to zero). Capital stock losses Ï„ All non-heavy-industry sectors Capital depreciation rates increase during major flood year reducing the physical stock of capital (Ï„ < 0). Normal depreciation rates resume in subsequent period (i.e. Ï„ returns to zero). Decelerating productivity growth µ All agricultural sectors Long-term rate of annual land expansion is reduced to zero during major flood year (µ = -γ). Productivity growth continues in subsequent period (i.e. µ returns to zero). Shocks only taking place in the climate change simulations Rising sea levels λ All crops and shrimp ï¬?sheries Crop land gradually and permanently declines due to rising sea levels in affected regions (0 < λ < 1). Historical data suggests that the land cultivated by assigning the offsetting parameter µ with the for aus and aman are severely affected during negative value of the exogenous rate of technical major flood years (see Figure 2.3). Accordingly, change (i.e. -γ), thus reducing the growth rate of the land inundation parameter ε is set equal to total factor productivity to zero during a major the land declines observed in each region from flood year to zero. ofï¬?cial agricultural production data – Equation The third climate impact channel captured in 7. Third, productive capital stocks are lost during the CGE model is the rise in sea levels caused by major flood years, which limit both agricultural climate change. Independent analysis estimates the and non-agricultural production.This is captured amount of land lost in each of the 16 sub-regions in the model by doubling the exogenous econ- resulting from increases in sea levels by the 2030s omy-wide rate of depreciation Ï€ during major and the 2050s.These are imposed on the total sup- flood years (1988 and 1998) and increasing it by ply of agricultural land through the parameter λ 50 per cent during the other less severe flood – equation 7.The base year reflects current condi- years. In other words, the additional depreciation tions and λ has a value of one. As sea levels gradu- parameter Ï„ – equation 12 – rises from zero in ally rise, the land area in particular regions decline typical climate years to 0.5 when the 1988 and and λ has a value less than one. Since the rising sea 1998 flood years are drawn from the climate level is only associated with climate change it is datasets. Finally, there is a deceleration in long- only imposed on the climate change simulations. term technical progress, which is governed by The CGE therefore captures three climate- the parameter γ – Equation 13. This is achieved related impacts: Annex 2 – Description of the CGE Model 117 1 Yield losses each year resulting from potential Farm land is divided into: sub-optimal climate conditions (including temperature, rainfall, CO2 and mean flood 1 Marginal farms with less than 0.5 acres; changes) as estimated by the hydro-crop 2 Small-scale farms with between 0.5 and 2.5 models; acres; 2 Lost land, capital and productivity growth 3 Large-scale farms with more than 2.5 acres. during major flood years as observed in his- torical production data; All factors are assumed to be fully employed, and 3 Lost cultivable land resulting from rising sea capital is immobile across sectors. New capital levels caused by climate change. from past investment is allocated to regions/sec- tors according to proï¬?t rate differentials under a The economy-wide model also accounts for ‘putty-clay’ speciï¬?cation. the predicted increase in the frequency of major The full model still assumes national product flooding resulting from climate change. The markets. However, international trade is captured return period for the 1988 and 1998 flood years by allowing production and consumption to shift are reduced by one-third. In other words, 1988 imperfectly between domestic and foreign mar- and 1998 are characterized as the 1/33 and 1/50 kets, depending on the relative prices of imports, year floods respectively (in relation to water dis- exports and domestic goods. Since Bangladesh charges). The frequency of these floods in the is a relatively small economy, world prices are sample for the random selection of years for the assumed to be ï¬?xed and the current account bal- future climate sequences is increased to 1/25 and ance is maintained by a flexible real exchange 1/33 for the 1988 and 1998 floods respectively. rate (i.e. the price index of tradable-to-non-trad- able goods). Production and trade elasticities are Extensions in the Full Bangladesh econometrically estimated. A linear expenditure system determines Model household consumption levels and permits non- The simpliï¬?ed model illustrates how climate var- unitary income elasticities. Households are disag- iability and change affects economic outcomes in gregated across agricultural and nonagricultural our analysis. However, the full Bangladesh model groups. Agricultural households are separated drops certain assumptions.2 Constant elasticity of into landowners and landless agricultural work- substitution (CES) production functions allows ers. Landowning farm households in each of the factor substitution based on relative factor prices 16 agro-climatic regions are separated into mar- (i.e. δ is no longer ï¬?xed).The model identiï¬?es 36 ginal, small-scale and large-scale farm households. sectors in each of the 16 agro-climatic regions Non-agricultural households are disaggregated (i.e. 17 in agriculture, 14 in industry and 5 in according to the education level of their house- services). Intermediate demand in each sector hold head (i.e. uneducated or illiterate, primary- (excluded from the simple model) is determined school educated and secondary-school edu- by ï¬?xed technology coefï¬?cients. Based on the cated or higher). There are a total of 52 distinct 2005 household income and expenditure survey, household groups in the full CGE model. These labour markets are further segmented across four household groups pay taxes to the government skill groups: based on ï¬?xed direct and indirect tax rates. Tax revenues ï¬?nance exogenous recurrent spending, 1 Illiterate or uneducated workers; resulting in an endogenous ï¬?scal deï¬?cit. 2 Workers with primary education; 3 Workers with some secondary schooling; Notes 4 Workers with secondary or higher schooling. 1 The base year crop yields in the CGE model (i.e. 2005) are set as the average yield achieved 118 Climate Change Risks and Food Security in Bangladesh for a given agro-climatic region and crop dur- sented in Thurlow (2004). The CGE model ing the 1970–99 baseline. These then grow at used in the current study falls within the an exogenous rate reflecting the long-term broader class of structural neo-classical rate of technical change without the effects models described in Dervis et al (1982) and of climate variability or change. Robinson (1989). 2 A mathematical speciï¬?cation of the under- lying recursive dynamic CGE model is pre- Annex 3 – Constructing the Social Accounting Matrix for Bangladesh Key Features of the 2005 General structure of SAMs Bangladesh SAM A SAM is an economy-wide data framework that usually represents the real economy of a single A social accounting matrix (SAM) is a consist- country.1 More technically, a SAM is a square ent data framework that captures the information matrix in which each account is represented by contained in the national income and product a row and column. Each cell shows the payment accounts and the input-output table, as well as from the account of its column to the account of the monetary flows between institutions. A SAM its row – the incomes of an account appear along is an ex-post accounting framework because, its row, its expenditures along its column. The within its square matrix, total receipts must equal underlying principle of double-entry accounting total payments for each account contained within requires that, for each account in the SAM, total the SAM. Since the required data is not drawn revenue (row total) equals total expenditure (col- from a single source, information from various umn total). Table A3.1 shows an aggregate SAM sources must be compiled and made consistent. (with verbal explanations in place of numbers). This process is valuable since it identiï¬?es incon- The SAM distinguishes between ‘activi- sistencies among Bangladesh’s statistical sources ties’ (the entities that carry out production) and and highlights areas where data reliability is weak- ‘commodities’ (representing markets for goods est. SAMs are economy-wide databases which are and non-factor services). SAM flows are valued used in conjunction with analytical techniques to at producers’ prices in the activity accounts and strengthen the evidence underlying policy deci- at market prices (including indirect commodity sions. The 2005 SAM extends previous SAMs taxes and transactions costs) in the commodity constructed by the International Food Policy accounts. The commodities are activity outputs, Research Institute (IFPRI) by including more either exported or sold domestically, and imports. agricultural sectors, and disaggregating agri- In the activity columns, payments are made to cultural production across 64 districts/zilas and commodities (intermediate demand) and factors non-agricultural production across the 6 major of production (value-added comprising operat- regional divisions/states of the country. ing surplus and compensation of employees). In the commodity columns, payments are made to domestic activities, the rest of the world and Table A3.1 Basic structure of a SAM Activities Commodities Factors Households Government Investment Rest of the World Total Activities Marketed output Activity income Commodities Intermediate inputs Private consumption Government Investment, change Exports Total demand consumption in stocks Factors Value-added Factor income Households Factor income to Transfers to Foreign remittances Household income households households received Government Sales taxes, import Factor and corporate Direct household Government transfers Government income tariffs taxes taxes from rest of world Savings Household savings Government savings Foreign savings Savings Rest of the World Imports Repatriated earnings Government transfers Foreign exchange to rest of world outflow Total Activity expenditures Total Factor expenditures Household Government Investment Foreign exchange supply expenditures expenditures inflow Annex 3 – Constructing the Social Accounting Matrix for Bangladesh 121 various tax accounts (for domestic and import sub-sectors), livestock (2 sub-sectors), ï¬?sheries (2 taxes).This treatment provides the data needed to sub-sectors) and forestry. With the exception of model imports as perfect or imperfect substitutes forestry and ‘other ï¬?sheries’, which are disaggre- vis-à-vis domestic production. gated across divisions, all agricultural sub-sectors The government is disaggregated into a core are disaggregated across Bangladesh’s 64 districts government account and different tax collection or zilas. Given severe land constraints in Bang- accounts, one for each tax type. This disaggrega- ladesh, agricultural land is disaggregated across tion is necessary since otherwise the economic three categories: (1) marginal farmers (farm interpretation of some payments is often ambig- households with less than 0.5 acres of cultivated uous. In the SAM, direct payments between the land); (2) small-scale farmers (households with government and households are reserved for between 0.5 and 2.5 acres of cultivated land); transfers. Finally, payments from the government and (3) medium- and large-scale farmers (house- to factors (for the labour services provided by hold with more than 2.5 acres of cultivated land public-sector employees) are captured in the gov- – equivalent to 1 hectare of land). Land alloca- ernment services activity. Government consump- tion across crops varies across different parts of tion demand is a purchase of the output from the the country and across farm households with dif- government services activity, which in turn pays ferent land endowments (see Table A3.3). Farm labour. households’ land endowments are typically small The SAM contains a number of factors of in Bangladesh, with the average farm household production, which earn incomes from their use cultivating just over 1 acre of land.This varies sig- in the production process and then pay their niï¬?cantly across divisions, with the largest average incomes to households, government and the cultivated land sizes in Rajshahi and the smallest rest of the world. Capital earnings or proï¬?ts are in Chittagong and Dhaka. This aggregation hides taxed according to average corporate tax rates even greater variation across districts. All farm and some proï¬?ts may be repatriated abroad. The households devote a majority of their land to rice remaining capital earnings, together with other production, although most households produce a factors’ earnings (e.g. land and labour) are paid to diverse range of crops. This regional and sectoral households. Households use their incomes to pay heterogeneity justiï¬?es the detailed spatial disag- taxes, save, and consume domestically produced gregation of crops in the new Bangladesh SAM. and imported commodities. The 2005 SAM retains non-agriculture sec- toral detail from the 2002/03 IFPRI SAM, but these non-agricultural sectors are now disaggre- Structure of the 2005 Bangladesh SAM gated across six regional divisions (see Table A3.4). The new SAM extends previous Bangladesh The Dhaka division is the largest in terms of its SAMs produced by IFPRI by: (1) updating the contribution to gross domestic product (GDP), previous 2002/03 SAM to 2004/05; (2) disag- accounting for two-ï¬?fths of the overall economy. gregating the agricultural sector across a greater Agriculture is least important in the Dhaka divi- number of crops; (3) disaggregating agricul- sion, accounting for 10 per cent of its economy, tural production and land and livestock markets which is substantially below the national aver- across 64 districts or zilas; and (4), disaggregat- age contribution of agriculture of 20 per cent of ing non-agricultural production and labour and GDP. Agriculture is especially important for the capital markets across 6 divisions. The next sec- smaller divisions of Barisal, Khulna and Sylhet, tion describes the various data sources used to where the sector accounts for around a third to a produce the new 2005 SAM, while this section half of divisional GDP. describes its overall structure. Factors markets are deï¬?ned at various levels The SAM identiï¬?es 62 sectors, of which 23 of spatial aggregation. Land and livestock capital are in agriculture (see Table A3.2). Agricultural are speciï¬?c to each zila and, as mentioned above, production is divided into crop agriculture (18 agricultural land in each zila is further disag- 122 Climate Change Risks and Food Security in Bangladesh Table A3.2 Sectors in the 2005 Bangladesh SAM gregated into marginal, small-scale and large- Sector Activity Commodity Description Disaggregation scale farms. Capital and labour are deï¬?ned at the no. code code divisional level. Labour is further disaggregated Agriculture across four education-based categories: (1) illiter- 1 arausl cauric Rice Aus (local) Zila 2 araush Rice Aus (hyv) Zila ate landless workers whose households still derive 3 aramnl camric Rice Aman (local & trans) Zila incomes from agriculture (i.e. farm labourer 4 aramnh Rice Aman (hyv & hybrid) Zila 5 arborl cboric Rice Boro (local) Zila families); (2) low-skilled workers (i.e. primary 6 arborh Rice Boro (hyv & hybrid) Zila schooling or less) and illiterate workers whose 7 awheat cwheat Wheat Zila 8 aocere cocere Other cereals Zila households derive incomes from wage employ- 9 ajutef cjutef Jute Zila ment and/or non-farm enterprises; (3) semi- 10 asugar csugar Sugarcane Zila 11 aocash cocash Other cash crops Zila skilled workers (i.e. some secondary schooling); 12 apulse cpulse Pulses Zila and (4) high-skilled workers (i.e. completed sec- 13 arapes crapes Rapeseed Zila 14 aooilc cooilc Other oil crops Zila ondary schooling and/or tertiary qualiï¬?cations). 15 aspice cspice Spices Zila These factor incomes are paid to households and 16 apotat cpotat Potatoes Zila 17 aveges cveges Vegetables Zila are supplemented by social security payments 18 afruit cfruit Fruits Zila from the government and remittances received 19 alives clives Livestock Zila 20 apoult cpoult Poultry Zila from abroad. 21 ashrmp cshrmp Shrimp farming Zila The model identiï¬?es ‘agricultural’ and ‘non- 22 aoï¬?sh coï¬?sh Other ï¬?shing Division 23 afores cfores Forestry Division agricultural’ households depending on whether Industry the household receives any income from work- 24 amines cmines Mining & quarrying Division 25 aaumll caumll Rice milling (Aus) Division ing in the agricultural sector. However, even agri- 26 aammll cammll Rice milling (Aman) Division cultural households derive at least some of their 27 abrmll cbrmll Rice milling (Boro) Division 28 aocmll cocmll Other cereal milling Division incomes from non-farm enterprises and off-farm 29 aedoil cedoil Edible oils Division wage employment. Agricultural households are 30 asugrp csugrp Sugar processing Division 31 aofood cofood Other food processing Division separated into the three land endowment cat- 32 abevtb cbevtb Beverages and tobacco Division egories discussed above (i.e. marginal, small and 33 aleath cleath Leather and footwear Division 34 ajtext cjtext Jute textiles Division medium/large) within each district. The SAM 35 ayarns cyarns Yarn Division also identiï¬?es households who are landless but 36 amclth cmclth Mill cloth Division 37 aoclth coclth Other cloth Division derive some of their income from working in 38 agarms cgarms Ready-made garments Division the agricultural sector. These landless households 39 aknitw cknitw Knitwear Division 40 aotext cotext Other textiles Division are only disaggregated across divisions because 41 awoodp cwoodp Wood and paper Division labour markets are identiï¬?ed at the division level. 42 achems cchems Chemicals Division 43 aferts cferts Fertilizers Division Finally, non-agricultural households are disag- 44 apetrl cpetrl Petroleum products Division gregated across divisions and according to the 45 anmetl cnmetl Non-metallic minerals Division 46 ametal cmetal Metal products Division education level of the head of the household (i.e. 47 amachs cmachs Machinery Division low-skilled, semi-skilled and high-skilled).2 49 aconst cconst Construction Division 50 antgas cntgas Natural gas Division Table A3.5 shows the share of different 51 aelect celect Electricity Division households’ incomes derived from factor and 52 awater cwater Water Division Services non-factor sources as reflected in the 2005 53 atrade ctrade Retail and wholesale trade Division SAM. Table A3.6 shows the income shares taken 54 ahotel chotel Hotels and catering Division 55 atrans ctrans Transport Division directly from the 2005 Household Income and 56 acomms ccomms Communications Division Expenditure Survey (HIES) (Bangladesh Bureau 57 abusre cbusre Business and real estate Division 58 afsrvs cfsrvs Financial services Division of Statistics, 2005a). A comparison shows that the 59 acsrvs ccsrvs Community and social Division SAM captures the relative importance of factors services 60 apadmn cpadmn Public administration Division in generating different households’ incomes.3 61 aeduca ceduca Education Division From Table A3.5, labour income is most 62 aheals cheals Health and social works Division important for non-agricultural households, Source: 2005 Bangladesh SAM. which are in turn most dependent on the type of Annex 3 – Constructing the Social Accounting Matrix for Bangladesh 123 Table A3.3 Average cultivated crop land allocation across divisions and scale of production National Divisions Farm size Barisal Chitta- Dhaka Khulna Rajshahi Sylhet Marginal Small Large gong farms farms farms Population 137,649 9164 27,019 43,727 15,879 33,026 8835 27,971 39,567 11,274 Households 28,166 1760 4816 9200 3479 7406 1505 5829 7523 1738 Average cultivated land (ac) All crops and shrimp 1.06 1.37 0.82 0.82 1.10 1.40 1.19 0.38 1.98 7.36 Rice 0.80 1.09 0.65 0.60 0.74 1.02 1.12 0.26 1.48 5.60 Aus (local) 0.05 0.23 0.09 0.02 0.05 0.02 0.11 0.02 0.10 0.42 Aus (high yield) 0.04 0.05 0.06 0.03 0.04 0.03 0.08 0.02 0.08 0.25 Aman (local)vv 0.18 0.62 0.15 0.13 0.19 0.16 0.23 0.05 0.31 1.45 Aman (high yield) 0.19 0.06 0.13 0.11 0.20 0.35 0.18 0.07 0.36 1.28 Boro (local) 0.04 0.02 0.02 0.03 0.02 0.03 0.22 0.01 0.06 0.30 Boro (high yield) 0.29 0.11 0.21 0.28 0.24 0.43 0.31 0.11 0.58 1.89 Wheat 0.03 0.01 0.01 0.02 0.05 0.06 0.01 0.01 0.06 0.21 Other cereals 0.01 0.00 0.00 0.00 0.01 0.02 0.00 0.00 0.02 0.07 Jute 0.04 0.01 0.01 0.05 0.07 0.05 0.00 0.02 0.08 0.23 Sugar cane 0.01 0.00 0.00 0.01 0.01 0.02 0.00 0.00 0.02 0.09 Other cash crops 0.00 0.00 0.00 0.00 0.01 0.01 0.00 0.00 0.01 0.02 Pulses 0.03 0.10 0.01 0.02 0.04 0.03 0.00 0.01 0.04 0.22 Rapeseed 0.03 0.01 0.01 0.04 0.02 0.04 0.00 0.01 0.06 0.17 Other oil crops 0.01 0.03 0.03 0.01 0.01 0.01 0.00 0.00 0.02 0.12 Spices 0.04 0.06 0.04 0.03 0.03 0.04 0.01 0.02 0.07 0.22 Potatoes 0.03 0.01 0.02 0.01 0.01 0.07 0.02 0.01 0.06 0.17 Vegetables 0.02 0.02 0.03 0.01 0.02 0.01 0.02 0.01 0.03 0.10 Fruits 0.01 0.02 0.01 0.01 0.01 0.01 0.00 0.00 0.02 0.07 Shrimp farming 0.01 0.00 0.00 0.00 0.06 0.00 0.00 0.01 0.01 0.07 Source: Authors’ calculations using the 2005 Agricultural Census and 2005 Bangladesh SAM labour category in which their household head of Bangladesh, with most households relying on falls. Marginal farmers and landless agricultural small-scale agriculture and allocating more than households are more dependent on illiterate two-ï¬?fths of the total consumption spending to and lower-skilled labour incomes. By contrast, food products. large-scale farmers derive a greater share of their In summary, the 2005 Bangladesh SAM income from capital earnings and land earnings makes full use of available data sets to produce and less from labour. a SAM with a stronger focus on agriculture but Household expenditure patterns are shown with the retained non-agricultural detail of pre- in Table A3.7 and are mainly determined by vious SAMs. Moreover, the SAM reflects the spa- income levels. Per capita consumption is low- tial heterogeneity of the country, both in terms of est for marginal and landless agricultural house- agro-ecological conditions and cropping patterns, holds, and almost three times as high on aver- as well as the varying concentrations of non- age for high-skilled non-agricultural households. agricultural production. Since the SAM captures Food consumption as a share of total consump- in detail Bangladesh’s sub-national and sectoral tion spending is lowest for higher-income large- characteristics, it is an ideal tool for examining scale farm and high-skilled non-agricultural agricultural investment policies, rural-urban link- households. However, these household groups ages and transformation and environmental and form only a small share of the total population climate-related scenarios. 124 Climate Change Risks and Food Security in Bangladesh Table A3.4 National and divisional per cent of gross domestic product (GDP) National Barisal Chittagong Dhaka Khulna Rajshahi Sylhet GDP shares across divisions 100.0 6.9 21.7 39.3 9.7 13.7 8.7 GDP shares within divisions 100.0 100.0 100.0 100.0 100.0 100.0 100.0 Agriculture 20.2 40.8 16.7 10.4 31.7 24.1 37.8 Mining & quarrying 1.2 0.1 1.4 1.1 2.0 0.2 2.6 Manufacturing 15.8 8.6 19.7 18.1 6.6 18.2 7.3 Grain milling 2.1 1.6 1.5 2.3 1.1 4.2 0.7 Edible oils 0.3 0.3 0.3 0.4 0.2 0.7 0.1 Sugar processing 0.2 0.1 0.1 0.2 0.1 0.3 0.1 Other food processing 1.0 0.8 0.7 1.1 0.5 2.0 0.3 Beverages /tobacco 0.3 0.2 0.2 0.3 0.1 0.5 0.1 Leather/footwear 0.2 0.0 0.2 0.4 0.0 0.1 0.1 Jute textiles 0.1 0.0 0.1 0.2 0.0 0.2 0.0 Yarn 1.0 0.4 1.7 1.2 0.3 0.4 0.2 Mill cloth 0.7 0.3 1.3 0.9 0.2 0.3 0.1 Other cloth 0.7 0.3 1.2 0.8 0.2 0.3 0.1 Ready-made garments 3.3 1.3 5.7 4.2 1.1 1.3 0.7 Knitwear 1.7 0.7 2.9 2.1 0.6 0.7 0.3 Other textiles 0.2 0.1 0.3 0.2 0.1 0.1 0.0 Wood & paper 1.0 1.1 1.0 1.2 0.8 0.7 0.4 Chemicals 0.6 0.2 0.1 0.1 0.0 3.5 0.0 Fertilizers 0.1 0.0 0.0 0.0 0.0 0.8 0.0 Petroleum products 0.0 0.0 0.0 0.1 0.0 0.0 0.0 Non-metallic minerals 0.7 0.5 0.3 0.4 0.1 1.0 3.2 Metals products 0.9 0.4 1.2 1.3 0.1 0.4 0.1 Machinery 0.2 0.2 0.2 0.1 0.2 0.3 0.0 Other manufacturing 0.6 0.1 0.8 0.5 0.7 0.5 0.6 Construction 10.6 15.8 8.5 7.6 12.8 13.0 19.3 Natural gas 0.1 0.1 0.3 0.1 0.0 0.0 0.0 Electricity 1.5 0.1 3.5 1.5 0.4 0.5 0.6 Water 0.1 0.0 0.2 0.1 0.0 0.0 0.0 Private services 42.8 29.3 42.1 52.1 36.7 38.3 27.4 Retail & wholesale trade 12.7 7.7 12.2 18.3 11.4 6.9 3.2 Hotels & catering 0.5 0.2 1.7 0.1 0.2 0.4 0.1 Transport 9.4 8.1 8.6 8.7 12.1 12.1 8.2 Communications 1.2 1.4 1.9 0.9 0.4 1.4 0.8 Business & real estate 7.5 4.7 7.6 7.6 5.6 11.2 5.9 Financial services 1.9 1.7 2.6 1.9 1.6 1.6 1.7 Community services 9.6 5.5 7.4 14.6 5.5 4.7 7.5 Public and related services 7.7 5.2 7.5 8.9 9.8 5.7 5.0 Public administration 2.8 1.7 4.0 3.3 2.0 1.5 1.6 Education 2.6 2.0 2.3 2.3 5.8 2.8 1.5 Health and social work 2.2 1.4 1.2 3.2 1.9 1.5 1.9 Source: Authors’ calculations using the 2005 Bangladesh SAM. The initial task in building a SAM involves publications. This information often uses: (1) dif- compiling data from various sources into the ferent disaggregation of sectors, production fac- SAM framework. This information is drawn tors, and socio-economic household groups; (2) from national accounts, household and agricul- different years and/or base-year prices; and (3) tural surveys, foreign trade statistics, government different data collection and compilation tech- budgets, balance of payments and various other niques. Consequently, the initial or prior SAM Annex 3 – Constructing the Social Accounting Matrix for Bangladesh 125 Table A3.5 Household factor income shares from the 2005 Bangladesh SAM Labour Capital Agriculture All Per capita factors income Illiterate Low- Semi- High- Physical Cattle Land Adjusted (US$) skilled skilled skilled capital revenues* All households 14.8 7.8 14.4 10.0 38.6 1.8 12.6 17.4 100.0 366 Agricultural 16.6 6.9 11.3 7.5 37.9 2.5 17.3 23.8 100.0 317 Farm households 8.8 5.8 11.6 8.7 41.5 3.0 20.6 28.3 100.0 337 Marginal 24.1 11.4 14.2 4.1 30.6 4.0 11.6 18.3 100.0 160 Small-scale 9.0 6.7 11.6 7.6 41.5 3.3 20.3 28.3 100.0 331 Large-scale 0.9 1.7 10.2 12.6 46.9 2.1 25.5 33.4 100.0 796 Landless 57.5 12.8 9.5 1.0 19.2 0.0 0.0 0.0 100.0 244 Non-agricultural 15.8 10.5 21.2 14.9 37.7 0.0 0.0 0.0 100.0 526 Low-skilled 41.4 24.4 2.5 0.3 31.3 0.0 0.0 0.0 100.0 280 Semi-skilled 1.3 4.2 59.1 0.7 34.6 0.0 0.0 0.0 100.0 826 High-skilled 0.0 0.0 3.8 47.5 48.7 0.0 0.0 0.0 100.0 1672 *Adjusted agricultural revenues include labour and capital earnings from the agricultural sector as well as land returns. Source: 2005 Bangladesh SAM. Social security payments, foreign remittance earnings and other non-factor incomes are excluded from this table. Table A3.6 Household factor income shares from the 2005 Household Income and Expenditure Survey Labour wages and in-kind receipts Non-farm Livestock Agriculture All Per capita enterprise product factors income Illiterate Low- Semi- High- Land Revenues revenues revenues (US$) skilled skilled skilled All households 14.6 4.6 8.9 12.2 37.8 1.1 – 20.7 100.0 250 Agricultural 15.3 3.9 7.6 10.2 36.0 1.4 – 25.5 100.0 246 Farm households 6.8 3.1 7.7 11.9 38.2 1.7 – 30.7 100.0 281 Marginal 19.3 6.6 10.6 5.9 31.2 1.8 – 24.5 100.0 185 Small–scale 5.4 3.0 8.3 11.1 40.0 1.8 – 30.3 100.0 269 Large–scale 0.5 0.6 4.7 17.2 40.1 1.4 – 35.6 100.0 485 Landless 57.4 8.3 7.1 2.2 25.0 0.0 – 0.0 100.0 153 Non-agricultural 20.4 7.8 13.2 16.9 41.7 0.0 – 0.0 100.0 234 Low–skilled 40.2 14.4 2.8 0.8 41.8 0.0 – 0.0 100.0 166 Semi-skilled 1.2 2.4 46.7 1.8 48.0 0.0 – 0.0 100.0 286 High-skilled 0.0 0.0 1.1 63.5 35.4 0.0 – 0.0 100.0 644 Source: Authors’ calculations using the 2005 Household Income and Expenditure Survey (Bangladesh Bureau of Statistics, 2005a). Social security payments, foreign remittance earnings and other non-factor incomes are excluded from this table. Note that household population weights differ between SAM (census-based) and the HIES. inevitably includes imbalances between row and in the SAM is discussed below. The notation for column account totals. SAM entries is (row, column) and the values are The prior macro SAM is based on national in millions of 2005 Bangladesh taka. and government accounts and balance of pay- ments. The disaggregated SAM is built so that Factors, Activities: 3,388,539 the totals from the macro SAM are preserved This is the value of gross domestic product (GDP) (i.e. shares are used from other sources not actual at factor cost or alternatively, total value-added numbers). This section explains how each macro generated by labour, capital and land. Sectoral SAM entry is derived and disaggregated to arrive GDP is drawn from national accounts and con- at the prior micro SAM. Table A3.8 shows the tains information on 20 aggregate sectors (Bang- 2005 macro SAM for Bangladesh. Each entry ladesh Bureau of Statistics, 2008b). GDP is then 126 Climate Change Risks and Food Security in Bangladesh Table A3.7 Household consumption Consumption share (%) Population (1000s) Per capita consumption Food Non-food Taka per year US$ per year All households 41.6 56.0 137,649 21,543 335 Agricultural 40.9 57.1 99,685 20,069 312 Farm households 39.5 58.5 78,811 21,129 328 Marginal 42.4 52.8 27,971 16,876 262 Small-scale 40.6 58.8 39,567 20,708 322 Large-scale 33.7 65.1 11,274 33,158 515 Landless 47.8 49.8 20,873 16,067 250 Non-agricultural 42.9 53.9 37,965 25,414 395 Low-skilled 45.3 49.5 26,518 19,074 297 Semi-skilled 49.4 50.1 ,7799 35,262 548 High-skilled 26.7 71.5 ,3648 50,447 784 Source: Authors’ calculations using the 2005 Bangladesh SAM. Per capita consumption differs from Table A3.5 due to additional income sources missing from that table (i.e. social security and foreign remittances) and additional expenditure items missing from this table (i.e. direct taxes and savings). further disaggregated across the full 62 sectors crops. This cultivated land area information was using shares from the 2001/02 Bangladesh SAM then combined with ofï¬?cial regional crop yield (Arndt et al, 2002).Value-added is further divided estimates for 1999/2000 (Bangladesh Bureau of into the returns to labour; capital and land using Statistics, 2002) – the year for which these esti- the 2001/02 SAM. mates were most available – to derive an estimate Labour income is split across four educa- of total production in each crop and zila. This tional groups: ‘illiterate’ refers to workers without was then scaled to match the level of produc- any education and living in landless agricultural tion and land area observed at the national level households; ‘low-skilled’ includes workers with in ofï¬?cial agricultural production data for the primary schooling or less (Class I to V); ‘semi- 2004/05 season (Bangladesh Bureau of Statistics, skilled’ includes workers with some secondary 2008c). Thus, in estimating agricultural produc- schooling (Class VI to IX); and ‘high-skilled’ tion in each region, the full range of available data includes workers who have completed second- was employed. The land used in each agricultural ary school or higher education (SSC/HSC activity was further disaggregated according to and above). Workers’ incomes from wage and the land size of the farm household. The three non-farm enterprises are drawn from the 2005 land sizes include: (1) marginal landholders (less Household Income and Expenditure Survey than 0.5 acres); (2) small-scale farmers (between (HIES) (Bangladesh Bureau of Statistics, 2005a). 0.5 and 2.5 acres); and (3) medium- and large- Capital is split into non-livestock physical capital scale farmers (more than 2.5 acres). The same and livestock capital. production technology was assumed for the same Each activity is then disaggregated across zilas crops in different zilas. (for agricultural sectors) or divisions (for non- Non-agricultural GDP was disaggregated agricultural sectors). Agricultural land alloca- across six regional divisions based on labour and tion by crop and zila was taken from the 2005 non-farm enterprise earnings reported in HIES. Agriculture Sample Survey (ASS) (Bangladesh This assumes that the same sector in each region Bureau of Statistics, 2005b).The ASS interviewed employs the same production technology. 2.8 million households and asked them to indi- cate whether they cultivated any agricultural Commodities, Activities: 3,558,916 land during the 2004/05 season and to identify This is the value of intermediate inputs used in how much of the land was allocated to different the production process. The aggregate value is Table A3.8 2005 macro SAM for Bangladesh (millions of Taka) Activities Commodities Factors Households Government Taxes Investment Rest of World Total Activities 6,947,454 Commodities 3,558,916 2,892,513 206,985 777,864 613,880 3,558,916 Factors 3,388,539 3,388,539 Households 3,300,422 109,625 226,043 Government 364,169 25,837 Taxes 243,196 49,772 71,201 Savings 672,375 73,396 32,094 Rest of World 859,508 38,345 Total 6,947,454 128 Climate Change Risks and Food Security in Bangladesh derived at the sector-level using GDP estimates taken from national accounts (Bangladesh Bureau described above. The technical coefï¬?cients used of Statistics, 2008a). Total private consumption in the SAM are derived from the 2001/02 SAM was distributed across commodities and different (Arndt et al, 2002). household groups using information from HIES (Bangladesh Bureau of Statistics, 2005a). How- Activities, Commodities: 6,947,454 ever, HIES only sampled 10,080 households out This is the value of total marketed output. Since of a total population of 137 million people. The all output is assumed to be supplied to markets, survey is thus strictly representative at the divi- this value is equivalent to gross output, where sional level and its estimates of consumption are gross output is the sum of intermediate demand unreliable at the zila level. Accordingly, per capita and GDP at factor cost. The SAM distinguishes expenditures were estimated for different house- between regional activities and national com- hold groups at the divisional level and then mul- modities. Regional producers therefore supply tiplied by zila level population estimates from the their output into a national commodity (i.e. there Agricultural Sample Survey (Bangladesh Bureau is no explicit treatment of inter-divisional trade). of Statistics, 2005b). These estimates were then scaled to match national consumption aggregates for each commodity from HIES. Taxes, Commodities: 243,196 While the macro SAM in Table A3.8 shows only a single row and column for taxes, this account Commodities, Government: 206,985 actually consists of a number of distinct tax The total value of government consumption accounts. These include speciï¬?c accounts for spending is taken from government accounts direct, indirect and trade taxes as reported in (International Monetary Fund, 2007) and dis- government accounts (International Monetary aggregated across commodities using informa- Fund, 2007). The commodity tax entry can tion from the 2001/02 SAM (Arndt et al, 2002), therefore be disaggregated to include indirect adjusted for observed changes in public adminis- sales taxes (92,752) and import tariffs (150,446). tration, education and health in national accounts These aggregate values were taken from govern- (Bangladesh Bureau of Statistics, 2008b). ment accounts for 2005 (International Monetary Fund, 2007). Aggregate tax revenues were disag- Commodities, Investment: 777,864 gregated across commodities using information The aggregate value of investment demand on value-added tax and import tariff rates from is taken from national accounts (Bangladesh the 2001/02 SAM (Arndt et al, 2002). Bureau of Statistics, 2008b) and disaggregated across commodities using information from the Rest of World, Commodities: 859,508 2001/02 SAM (Arndt et al, 2002). Note that this The value of total imports of goods and services aggregate value includes both public and private was initially taken from national accounts (Bang- investment. ladesh Bureau of Statistics, 2008b). Goods imports were disaggregated using 2007 foreign trade data Commodities, Rest of World: 613,880 (Bangladesh Bureau of Statistics, 2008a) and serv- The value of total exports of goods and serv- ices trade from the balance of payments (Interna- ices was taken from national accounts (Bangla- tional Monetary Fund, 2007). desh Bureau of Statistics, 2008b). Goods exports were disaggregated using 2007 foreign trade data Commodities, Households: 2,892,513 (Bangladesh Bureau of Statistics, 2008a) and serv- The payment from households to commodities is ices exports from the balance of payments (Inter- equal to household consumption of marketed pro- national Monetary Fund, 2007). duction.The total level of private consumption is Annex 3 – Constructing the Social Accounting Matrix for Bangladesh 129 Households, Factors 3,300,422 Households, Government: 109,625 This is the total labour value-added generated This is social security and other transfers paid by during production as well as livestock and land the government to households. The total level returns. The distribution of labour income across of social transfers was taken from government households is determined using household labour accounts (International Monetary Fund, 2007). income shares as reported in HIES (Bangladesh This was disaggregated across households using Bureau of Statistics, 2005a). Land and livestock social security incomes reported by households in returns were based on land and stock holdings HIES (Bangladesh Bureau of Statistics, 2005a). reported in the 2005 Agricultural Sample Survey (Bangladesh Bureau of Statistics, 2005b). Capital Households, Rest of World: 226,043 returns were distributed using non-farm enter- This is foreign workers’ remittances to domestic prise earnings and returns to assets (e.g. imputed households as reported in the balance of pay- rent, interest earnings and property rents). ments (International Monetary Fund, 2007). This was disaggregated across households using Taxes, Factors: 49,772 reported foreign remittance incomes in HIES These are corporate taxes paid on the proï¬?ts (Bangladesh Bureau of Statistics, 2005a). earned by capital. It is paid to the government and is derived from government accounts (Inter- Government, Taxes: 364,169 national Monetary Fund, 2007). The same cor- The tax accounts in the micro SAM are separated porate tax rate is assumed across all divisions. into import tariffs, export taxes, sales taxes and direct taxes. Each account adds up tax revenue Rest of World, Factors: 38,345 from all sources and then transfers these funds These are remitted proï¬?ts by the capital factor to the government. The entries correspond to and are equal to the value of foreign factor pay- government accounts (International Monetary ments in the balance of payments (International Fund, 2007). Monetary Fund, 2007). Government, Rest of World: 25,837 Taxes, Households: 71,201 Government income from the rest of the world is The value of direct taxes on households is equiva- equivalent to the value of foreign grants and ofï¬?- lent to PAYE taxes and is taken from government cial transfers in the balance of payments (Interna- accounts (International Monetary Fund, 2007). tional Monetary Fund, 2007). Tax payments are distributed across households using information on tax and deduction pay- Savings, Government: 73,396 ments from HIES (Bangladesh Bureau of Statis- This is value of public savings. It is the sum of tics, 2005a). the ï¬?scal surplus (after receiving foreign grants) and the value of public investment or capital Savings, Households: 672,375 expenditure. It is equal to the ï¬?scal surplus in This is value of domestic private savings and government accounts (International Monetary is calculated as a residual to balance aggregate Fund, 2007). household income and expenditure accounts when constructing the macro SAM. Household Savings, Rest of World: 32,094 groups in the SAM are assumed to have savings This is the current account deï¬?cit or the total rates in proportion to the share of capital earn- value of foreign savings. It is derived from the ings in total household earnings (scaled to match balance of payments (International Monetary the macro SAM control total). Fund, 2007). 130 Climate Change Risks and Food Security in Bangladesh Balancing the Prior SAM tors and households. Since the aggregate national SAM is balanced, this results in imbalances for The range of datasets used to construct the prior the household accounts only. These house- micro SAM implies that there will inevitably hold accounts were again balanced using cross- be imbalances (i.e. row and column totals are entropy, but holding all other non-household- unequal). Cross-entropy econometrics is used related entries of the national SAM constant. to reconcile SAM accounts (see Robinson et al, Given the imbalances in the household survey 2001). This approach begins with the construc- between incomes and expenditures, and then the tion of the prior SAM which, as explained in the additional imbalances caused by different house- previous section, used a variety of data from a hold factor income shares in the macro SAM, the number of sources of varying quality. This prior target household income/expenditure total for SAM provided the initial ‘best guess’ for the the ï¬?nal balanced SAM was an average of the estimation procedure. Additional information is income and expenditure totals in the unbalanced then brought to bear, including knowledge about prior SAM. aggregate values from national accounts and tech- nology coefï¬?cients. A balanced Bangladesh SAM Cross-entropy estimation of the balanced was then estimated by minimizing the entropy ‘distance’ measure between the ï¬?nal SAM and SAM the initial unbalanced prior SAM, taking into Table A3.9 summarizes the equations deï¬?ning account of all additional information. the SAM estimation procedure. Starting from an initial estimate of the SAM, additional informa- Balancing procedure for the Bangladesh tion is imposed in the form of constraints on the estimation. Equation 1 speciï¬?es that row sums SAM and corresponding column sums must be equal, The balancing procedure takes places in two which is the deï¬?ning characteristic for a consist- stages. First, a national SAM and supply-use table ent set of SAM accounts. Equation 2 speciï¬?es is constructed using primarily national accounts, that sub-accounts of the SAM must equal con- government budgets and balance of payments. trol totals, and that these totals are assumed to This was disaggregated across activities and com- be measured with error (Equation 3). An exam- modities using sectoral GDP estimates from the ple would be the estimate of GDP provided by agricultural census, HIES and previous SAMs national accounts, which is the total value of the for Bangladesh. The SAM contains aggregate Factor-Activity matrix in the prior SAM. The entries for factors and households. This aggregate matrix G is an aggregator matrix, with entries national SAM was then balanced using cross- equal to 0 or 1. The index k is general and can entropy. Larger standard errors were applied to include individual cells, column/row sums and non-agricultural production estimates, since this any combination of cells such as macro aggre- is less recent data, and on household demand gates. Equation 4 allows for the imposition of because total consumption from the household information about column coefï¬?cients in the survey is 25 per cent below the aggregate ï¬?gure SAM rather than cell values, also allowing for reported in national accounts (making a shares error (Equation 5). approach to estimating commodity consump- The error speciï¬?cation in Equations 2 and tion less accurate). Smaller standard errors were 3 describes the errors as a weighted sum of a imposed on agricultural production because a speciï¬?ed ‘support set’ (the V parameters). The number of data sources, including the large sam- weights (W ) are probabilities to be estimated, ple agricultural survey, reported similar land area starting from a prior on the standard error of and production levels. measurement of aggregates of flows (Equation After balancing the aggregate national SAM, 8) or coefï¬?cients (Equation 9). The number of the SAM was disaggregated across regions, fac- elements in the error support set (w) determines Annex 3 – Constructing the Social Accounting Matrix for Bangladesh 131 Table A3.9 Cross-entropy SAM estimation equations Index Deï¬?nition i, j row (i) and column (j) entries k set of constraints w set of weights Symbol Deï¬?nition Ti,j SAM in values Ai , j and Ai , j SAM in column coefï¬?cients Gk ,i , j aggregator matrix for each constraint k γ k and γ k and aggregate value for constraint k ek error on each constraint k eiAj , error on each cell coefï¬?cient W and W and weights and prior on error term for each constraint k or cell coefï¬?cient i,j V error support set indexed over w for each constraint k or cell coefï¬?cient i,j Equations ∑T i i, j = ∑ Ti , j j (1) ∑∑ G i j k ,i , j â‹… T i, j = γ k (2) γ k = γ k + ek (3) T i, j Ai, j = with ∑A =1∀ i (4) ∑T i, j i, j i i Ai , j = Ai , j + eiAj for some i, j , (5) e k = ∑ W k, w â‹… V k, w (6) w eiAj = ∑ Wi ,Aj , w â‹… Vi ,Aj , w , (7) w ∑W w k ,w = 1 with 0 ≤ Wk , w ≤ 1 (8) ∑W w A i, j ,w = 1 with 0 ≤ Wi ,Aj , w ≤ 1 (9)   min  ∑ ∑Wk ,w â‹… (ln Wk ,w − ln Wk ,w )+ ∑∑∑Wi,Aj ,w (ln Wi,Aj ,w − ln Wi,Aj ,w ) (10)  k w i j w  how many moments of the error distribution are this minimand is uniquely appropriate and that estimated. The probability weights must be non- using any other minimand introduces unwar- negative and sum to one (Equations 8 and 9).The ranted assumptions (or information) about the objective function is the cross-entropy distance errors. between the estimated probability weights and Various constraints were imposed on the their prior for the errors in both coefï¬?cients and model according to the perceived reliability of aggregates of SAM flows. It can be shown that the Bangladesh data. Certain values that appeared 132 Climate Change Risks and Food Security in Bangladesh in national accounts were maintained in order to remain consistent with the overall macro structure Notes of the economy.The macro-economic aggregates 1 For general discussions of SAMs see Pyatt that were maintained in the micro-SAM include: and Round (1985) and Reinert and Roland- total labour value-added; total capital value- Holst (1997); for perspectives on SAM-based added; household ï¬?nal demand; government modelling see Pyatt (1988) and Robinson spending; investment demand; exports; imports; and Roland-Holst (1988). government borrowing/saving; current account 2 Note that ‘low-skilled non-agricultural’ balance; sales taxes; import tariffs; direct taxes on households include both household heads enterprises; government transfers to enterprises; who are illiterate and those who completed enterprise transfers to the rest of the world; enter- some level of primary schooling prise transfers to government; household trans- 3 There are differences in the interpretation of fers to government; government transfers to the factor income sources in the SAM and survey. rest of the world; and household foreign transfers The SAM separates land returns from agri- received. Since the household survey (HIES) and cultural labour and capital earnings, while the agricultural production and GDP estimates the HIES reports agricultural revenues after from national accounts were taken from data for subtracting production costs (i.e. returns to the same year, the standard errors applied to the all agricultural factors). Survey agricultural various components were uniform. income shares are therefore larger than land returns in the SAM. Similarly, non-farm enterprise earnings are a form of ‘mixed income’. In other words, they include both labour and capital earnings, and are therefore typically higher than capital earnings alone. References Adger, W. N., Agrawala, S., Mirza, M.M.Q., Conde, C., ling studies of elevated [CO2] impacts on crop yield O’Brien, K., Pulhin, J., Pulwarty, R., Smit, B. and and food supply’, New Phytologist, 5pp Takahashi, K. (2007) ‘Assessment of adaptation prac- Alley, R. B., Clark, P. U., Huybrechts, P. and Joughin, I. tices, options, constraints and capacity’ in Parry, M. (2005) ‘Ice-sheet and sea level changes’, Science, vol L., Canziani, O. F., Palutikof, J. P., van der Linden, 310, pp456–460 P. J. and Hanson, C. E. (eds) Climate Change 2007: Ali, A. (1999) ‘Climate change impacts and adaptation Impacts, Adaptation and Vulnerability. Contribution of assessment in Bangladesh’, Climate Research, vol 12, Working Group II to the Fourth Assessment Report of the pp109–116 Intergovernmental Panel on Climate Change, Cambridge Ali, A. (2006) ‘Mathematically predicting the impacts University Press, Cambridge, UK, pp717–743 of climate change and sea level rise on storm surges Adler, R. F., Huffman, G. J. and Keehn, P. R. (1994) in the Bangladesh coastal region’, presented at the ‘Global rain estimates from microwave adjusted Workshop on Climate Change Impact Modelling, geosynchronous IR data’, Remote Sensing Review, Climate Change Cell, Department of Environment, vol 11, pp125–52 Government of the People’s Republic of Bangla- Aerts, J. C. J. H., De Vente, J., Hassan, A. and Martin T. desh, Dhaka, 26–27 February C. (2000) Spatial Integration of Hydrological Monitoring Arndt, C., Dorosh, P., Fontana, M. and Zohir, S. (2002) and Remote Sensing Applications (SPIHRAL), Final Opportunities and Challenges in Agriculture and Gar- Report, Resource Analysis, Delft, Netherlands ments: A General Equilibrium Analysis of the Bangladesh Agrawala, S., Ota, T., Ahmed, A. U., Smith, J. and van Economy, TMD Discussion Paper no 107, IFPRI, Aalst, M. (2003) Development and Climate Change in Washington D.C. Bangladesh: Focus on Coastal Flooding and the Sundar- Bangladesh Bureau of Statistics (BBS) (1998) Agricul- bans, Environment Directorate and Developmental tural Statistical Year Book of Bangladesh, Dhaka, Bang- Co-operation Directorate, Organization for Eco- ladesh nomic Co-operation and Development, Paris BBS (2002) Regional Estimates of Agricultural Crop Produc- Ahmad, A. S. M. S., Munim, A. A., Begum, Q. N. and tion: 1985/86-1999/2000, Dhaka, Bangladesh Choudhury, A. M. (1996) ‘El Nino-southern oscilla- BBS (2005a) Household Income and Expenditure Survey tion and rainfall variation over Bangladesh’, Mausam, 2004, Dhaka, Bangladesh vol 47, no 2, pp157–62 BBS (2005b) Agriculture Sample Survey 2005, Dhaka, Ahmed, A. U. and Mirza, M. M. Q. (2000) ‘Review of Bangladesh causes and dimensions of floods with particular ref- BBS (2006a) Agriculture Sample Survey of Bangladesh, erence to flood ’98: national perspectives’ in Ahmad, Dhaka, Bangladesh Q. K., Chowdhury, A. K. A., Imam, S. H. and Sarker, BBS (2006b) Report of the Household Income and Expendi- M. (eds) Perspectives on Flood 1998, University Press ture Survey, Dakha, Bangladesh Ltd, Dhaka, pp67–84 BBS (2008a) Foreign Trade Statistics, Dhaka, Bangladesh, Ahmed, A. U. (2006) Bangladesh: Climate Change Impacts available at www.bbs.gov.bd and Vulnerability, Climate Change Cell, Bangladesh, BBS (2008b) National Accounts 2005, Dhaka, Bangla- p42 desh, available at www.bbs.gov.bd Ainsworth, E.A., Leakey, A.D.B., Ort, D.R. and Long, BBS (2008c) Handbook of Agricultural Statistics 2007, S.P. (2008) ‘FACE-ing the facts: Inconsistencies and Dhaka, Bangladesh, available at www.bbs.gov.bd interdependence among ï¬?eld, chamber and mode- 134 Climate Change Risks and Food Security in Bangladesh BBS (2009) National Accounts 2009, Dhaka, Bangladesh Easterling, W. E., Aggarwal, P. K., Batima, P., Brander, BCA (2002) Bangladesh Country Almanac, available at K. M., Erda, L., Howden, S. M., Kirilenko, A., Mor- www.cimmyt.org/bangladesh/Programs/Program4. ton, J., Soussana, J.-F., Schmidhuber, J. and Tubiello, htm F. N. (2007) ‘Food, ï¬?bre and forest products’, in M. Bangladesh University of Engineering and Technology L. Parry, O. F. Canziani, J. P. Palutikof, P. J. van der (1993) Multipurpose Cyclone Shelter Programme, Final Linden and C. E. Hanson (Eds) Climate Change Report, UNDP/World Bank, Dhaka, Bangladesh 2007: Impacts, Adaptation and Vulnerability, Contribu- Brammer, H. (1996) The Geography of the Soils of Bangla- tion of Working Group II to the Fourth Assessment desh, University Press, Dhaka, Bangladesh, 287pp Report of the Intergovernmental Panel on Climate BRRI (Bangladesh Rice Research Institute) (2007) Change, Cambridge University Press, Cambridge, Modern Rice Cultivation, Bangladesh Rice Research UK, pp273–313 Institute, Joydebpur, Bangladesh, 66pp Faisal, I. M. and Parveen, S. (2003) ‘Food security in CEGIS (2006) Impacts of Sea Level Rise on Land-use the face of climate change, population growth, and Suitability and Adaptation Options, Draft Final Report, resource constraints: Implications for Bangladesh’, Submitted to the Ministry of Environment and For- Environmental Management, vol 34, no 4, pp487–98 est, Government of Bangladesh and United Nations Farr, T. G. et al (2007), ‘The Shuttle Radar Topography Development Programme (UNDP), Dhaka, Bang- Mission’, Review of Geophysics, vol 45 ladesh Fischer, G., Shah, M. and van Velthuizen, H. (2002) CEGIS (2009), image courtesy of Maminul Haque Climate Change and Agricultural Vulnerability, Inter- Sarker at CEGIS, appears in ‘Where warming hits national Institute for Applied Systems Analysis, Lax- hard’, Nature Reports – Climate Change, 15 January enburg, Austria, p152 2009 Food and Agriculture Organization of the United Chen, J., Lin, H. and Pei, Z. (2007) ‘Application of Nations (2008). FAOSTAT, Rome, available at ENVISAT ASAR data in mapping rice crop growth www.fao.org in Southern China’ Geoscience and Remote Sensing Gomosta, A. R., Quayyum, H. A. and Mahbub, A. A. Letters, IEEE, vol 4, no 3, pp431–435, Doi: 10.1109/ (2001) ‘Tillering duration and yielding ability of LGRS.2007.896996 rice varieties in the winter rice season of Bangla- Chowdhury, M. R. (2000) ‘An assessment of flood desh’, in Peng, S. and Hardy, B. (eds) Rice Research forecasting in Bangladesh: The experience of the for Food Security and Poverty Alleviation, International 1998 flood’, Natural Hazards, vol 22, pp139–163 Rice Research Institute, Philippines, 692pp Chowdhury, M. R. (2003) ‘The El Nino-Southern Goodbred, S. L. and Kuehl, S. A. (2000) ‘Enormous Oscillation (ENSO) and seasonal flooding – Bangla- Ganges-Brahmaputra sediment discharge during desh’, Theoretical Applied Climatology, vol 76, pp105–24 strengthened early Holocene monsoon’, Geology, Chowdhury, J. A. and co-editors (2005c) Yearbook of vol 28, no 12, December, pp1083–86 Agricultural Statistics of Bangladesh, Bangladesh Bureau Habibullah, M., Ahmed, A.U. and Karim, Z. (1998) of Statistics, Dhaka, Bangladesh, p344 ‘Assessment of Foodgrain Production Loss Due to Cline,W. (2007) Global Warming and Agriculture – Impact Climate Induced Enhanced Soil Salinity’ in Huq, Estimates by Country, Center for Global Develop- S., Karim, Z., Asaduzzaman, M. and Mahtab, F. (eds) ment, Peterson Institute, Washington D.C, p186 Vulnerability and Adaptation to Climate Change for Copenhagen Diagnosis (2009) Updating the World on Bangladesh,Kluwer Academic Publishers, Dordrecht, the Latest Climate Science, University of New South pp55–70 Wales Climate Change Research Centre (CCRC), Hassan, A. (2006). ‘Recent trends in Bangladesh (tech- Sydney, Australia, p60 nical notes)’, Centre for Environmental and Geo- Cuny, F. C. (1983) Disasters and Development, Oxford graphical Information Services (CEGIS), Dhaka, University Press, New York Bangladesh Dervis, K., de Melo, J. and Robinson, S. (1982) General Hassan, A. and Shah, M. A. R. (2006) ‘Impacts of sea Equilibrium Models for Development Policy, Cambridge level rise on suitability of agriculture and ï¬?sheries: A University Press, New York case study on southwest region of Bangladesh’, pre- Dyurgerov, M. D. and Meier, M. F. (2005) Glaciers and sented in the Workshop on Climate Change Impact Changing Earth System: A 2004 Snapshot, Institute of Modelling, Climate Change Cell, Department of Arctic and Alpine Research, University of Colorado, Environment, Government of Bangladesh, Dhaka, Boulder, CO, p117 26–27 February References 135 Hassan, A., Chowdhury, E. H. and Khatun, M. F. (2007) IPCC (2001) Climate Change 2001: The Scientiï¬?c Basis, Classiï¬?cation of Flood for Seasonal Flood Management, Contribution of Working Group I to the Third paper presented on International Conference on Assessment Report of the Intergovernmental Panel Water & Flood Management – 2007, Dhaka, Bang- on Climate Change, Cambridge University Press, ladesh Cambridge, UK Hatï¬?eld, J., Boote, K., Fay, P., Hahn, L., Izaurralde, C., IPCC (2007a) Climate Change 2007:The Scientiï¬?c Basis, Kimball, B. A., Mader, T., Morgan, J., Ort, D., Pol- Contribution of Working Group I to the Fourth ley, W., Thomson, A., Wolfe, D., (2008) ‘Agriculture’, Assessment Report of the Intergovernmental Panel Chapter 2 in The effects of climate change on agricul- on Climate Change, Solomon, S., et al (eds), Cam- ture, land resources, water resources, and biodiversity in the bridge University Press, Cambridge, UK United States, a report by the U.S. Climate Change IPCC (2007b) Climate Change 2007: Impacts, Adap- Science Program and the Subcommittee on Global tation and Vulnerability, Contribution of Working Change Research, Washington, DC, 362pp Group II to the Fourth Assessment Report of the Haque, C. E. (1997) Hazards in a Fickle Environment: Intergovernmental Panel on Climate Change, Parry, Bangladesh, Kluwer Academic Publishers, Dordrecht M., et al (eds), Cambridge University Press, Cam- Hendry, G. R., and Kimball, B. A. (1994) ‘The FACE bridge, UK Programme’, Agriculture and Forest Meteorology, vol Islam, M. N., Mannan, M. A., Devkota, L. P. and Nessa, 70, pp3–14 M. (2005) Validation of PRECIS Regional Climate Hofer, T. and Messerli, B. (2007) Floods in Bangladesh Model in Bangladesh, Climate Change Cell, Bangla- – History, Dynamics and Rethinking the Role of the desh, Dhaka, p46 Himalayas, United Nations University Press, Tokyo, IWM and CEGIS (2007), Investigating the Impact of Rel- New York, Paris, p468 ative Sea level Rise on Coastal Communities and their Hoogenboom G., Jones, J. W., Porter, C. H., Wilkens, Livelihoods in Bangladesh, Final Report, June 2007, P. W., Boote, K. J., Batchelor, W. D., Hunt, L. A. and IWM /CEGIS/Government of Bangladesh, Dhaka Tsuji, G. Y. (2003) DSSAT v4 vol. 1, University of Jianchu, X., Shrestha, A., Vaidya, R., Eriksson, M., Hawaii, Honolulu, HI Hewitt, K. (2007) ‘The Melting Himalayas: Regional Hopson, T. and Webster, P. (2007) ‘A 1–10 day ensem- Challenges and Local Impacts of Climate Change ble forecasting scheme for the major river basins of on Mountain Ecosystems and Livelihoods’, ICI- Bangladesh: forecasting severe floods of 2003–2007’, MOD Technical Paper, June, 24pp Journal of Hydrometeorology, in press. Jones, J., Hoogenboom, W. G., Porter, C. H., Boote, K. Horton, R., Herweijer, C., Rosenzweig, C., Liu, J. J., Batchelor,W. D., Hunt, L.A.,Wilkens, P.W., Singh, P., Gornitz, V. and Ruane, A. C. (2008) ‘Sea level U., Gijsman, A. J. and Ritchie, J. T. (2003) ‘The rise projections for current generation CGCMs DSSAT cropping system model’, European Journal of based on the semi-empirical method’, Geophysi- Agronomy, vol 18, nos 3–4, pp235–65 cal Research Letters, vol 35, L02715, doi:10.1029/ Joyce, R. J., Janowiak, J. E., Arkin, P. A. and Xie, P. 2007GL032486 (2004) ‘CMORPH: A method that produces global Huq, S., Karim, Z., Asaduzzaman, M. and Mahtab, F. precipitation estimates from passive microwave and (eds) (1999) Vulnerability and Adaptation to Climate infrared data at high spatial and temporal resolution’, Change in Bangladesh, Kluwer Academic Publishers, Journal of Hydrometeorology, vol 5, pp487–503 Dordrecht, Netherlands Kanamitsu, M., Ebisuzaki, W., Woollen, J., Yang, S.-K., Hussain, S. G. (1995) ‘Decision support system for Hnilo, J., Fiorino, M. and Potter, J. (2002) ‘NCEP/ assessing rice yield losses from annual flooding in DOE AMIP-II Reanalysis (R-2)’, Bulletin of the Bangladesh’, dissertation submitted to the Graduate American Meteorology Society, vol 83, no 7, pp1019–37 Division of the University of Hawaii, p137 Karim, Z., Ahmed, M., Hussain, S.G. and Rashid, Kh. Hussain, S. G. (2006) Agriculture Water Demand and B. (1994) ‘Impact of climate change on the produc- Drought Modeling, Proceedings of Workshop on Cli- tion of modern rice in Bangladesh’, Implications of mate Change Impact Modeling, Climate Change Climate Change for International Agriculture: Crop Mod- Cell, Department of Environment, Government of eling Study, US Environmental Protection Agency, the People’s Republic of Bangladesh, Dhaka Washington D.C. International Monetary Fund (2007) Bangladesh: Statis- Karim, Z., Hussain, S.G. and Ahmed, M. (1996) ‘Assess- tical Appendix, Country Report No. 07/229, Wash- ing impacts of climate variation on foodgrain pro- ington D.C. duction in Bangladesh’, Journal of Water, Air and Soil Pollution, vol 92, pp53–62 136 Climate Change Risks and Food Security in Bangladesh Karim, Z., Hussain, S. G. and Ahmed, A. U. (1998), Mahmood, R., Meo, M., Legates, D.R. and Morris- ‘Climate change vulnerability of crop agriculture’, sey, M.L. (2003) ‘The CERES-Rice model-based in S. Huq, Z. Karim, M. Asaduzaman and F. Mahtab estimates of potential monsoon season rainfed rice (eds), Vulnerability and Adaptation to Climate Change productivity in Bangladesh’, Professional Geographer, for Bangladesh, Kluwer Academic Publishers, Dor- vol 55, no 2, pp269–73 drecht, , Netherlands, pp39–54 Meehl, G.A., Covey, C., Delworth,T., Latif, M., McAv- Khalil, G. M. (1990) ‘Flood in Bangladesh: a ques- aney, D., Mitchell, J.P.B., Stouffer, R.J. and Taylor, tion of disciplining the river’, Natural Hazards vol K.E. (2007) The WCRP CMIP3 multi-modal data- 3, pp379–401 set: A new era in climate change research, Bulletin of Kimball, B. A. and Bernacchi, C. J. (2006) ‘Evapo- the American Meteorological Society, vol 88, pp1383– transpiration, canopy temperature, and plant water 1394 relations’ in Managed Ecosystems and CO2: Case Stud- Mendelsohn, R. and Schlesinger, M. (1999) ‘Climate ies, Processes, and Perspectives. Springer-Verlag, Berlin, response functions’, Ambio, vol 28, no 4, pp362–66 Germany, pp311–324 Mirza, M. M. Q. (2002) ‘Global warming and changes Kotera, A., and Nawata, E. (2007) ‘Role of plant height in the probability of occurrence of floods in Bangla- in the submergence tolerance of rice: A simulation desh and implications’, Global Environmental Change, analysis using an empirical model’, Agricultural Water vol 12, pp127–38 Management, vol 89, pp49–58 Mirza, M. M. Q. (2003) ‘Three Recent Extreme Floods Lal, M., Meehl, G. A., and Arblaster, J. M. (1998a) ‘Sim- in Bangladesh: A Hydro-Meteorological Analysis’, ulation of Indian Summer Monsoon Rainfall and its Natural Hazards, vol 28, pp35–64 Intraseasonal Variability in NCAR’s Climate System Mirza, M. M. Q. and Dixit, A. (1997) ‘Climate change Model’, National Center for Atmospheric Research, and water resources in the GBM basins’, Water Nepal, draft, Boulder, CO vol 5, no 1, pp71–100 Lal, M., Singh, K. K., Raghore, L. S., Srinivasan, G., and Mitchell, T. D., Carter, T. R., Jones, P. D., Hulme, M. Saseendran, S. A. (1998b) ‘Vulnerability of Rice and and New, M. (2004) ‘A comprehensive set of high- Wheat Yields in NW India to Future Changes in resolution grids of monthly climate for Europe and Climate’, Agriculture and Forest Meteorology, vol 89, the globe: The observed record (1901–2000) and 16 pp101–114 scenarios (2001–2100)’, in Tyndall Centre Working Lal, M., Singh, K. K., Srinivasan, G., Rathore, L. S. and Paper 55 Mall, R. K. (1998c) ‘Growth and Yield Responses of MPO (1987), National Water Plan, Phase I, Master Plan Soybean in Madya Pradesh, India to ClimateVariability Organization Development, Dhaka, Bangladesh and Change’, submitted to Agriculture and Forest Mete- NAPA (National Adaptation Programme of Action) orology, Indian Institute for Technology, New Delhi (2005), Ministry of Environment and Forest, Bang- Latif, M. A., Islam, M. R., Ali, M.Y. and Saleque, M. A. ladesh, 46pp (2005) ‘Validation of the system of rice intensiï¬?ca- NCDC (2008) National Climatic Data Center Summary tion (SRI) in Bangladesh’, Field Crops Research, vol of the Day, available at http://iridl.ldeo.columbia. 93, pp281–292 edu/SOURCES/.NOAA/.NCDC/.DAILY/. Long, S. P., Ainsworth, E., Leakey, A. and Morgan, P. GLOBALSOD (2005) ‘Global food insecurity: Treatment of major New York City Panel on Climate Change (2009), Cli- food crops with elevated carbon dioxide or ozone mate Risk Information, Center for Climate Systems under large-scale fully open-air conditions suggests Research, Hartford, CT, release version, February 17 recent models may have overestimated future yields’, Nicholls, R.J. and Leatherman, S.P. (1995) ‘The poten- Philosophical Transactions of the Royal Society B, vol tial impact of accelerated sea level rise on devel- 360, pp2011–20 oping countries’, Journal of Coastal Research, vol 14 Long, S. P., Ainsworth, E. A., Leakey, A. D. B. (2006) (special issue) ‘Food for thought: lower-than-expected crop yield Nicholls, R. J., Wong, P. P., Burkett, V. R., Codignotto, stimulation with rising CO2 concentrations’, Science, J. O., Hay, J. E., McLean, R. F., Ragoonaden, S. and vol 312, pp1918–1921 Woodroffe, C. D., (2007) ‘Coastal systems and low- Mahmood, R. (1997) ‘Impacts of air temperature vari- lying areas’, in Parry, M. L., Canziani, O. F., Palutikof, ations on the boro rice phenology in Bangladesh: J. P., van der Linden, P. J. and C.E. Hanson, (eds) implications for irrigation requirements’, Agricultural Climate Change 2007: Impacts, Adaptation and Vulner- and Forest Meteorology, vol 84, pp233–247 ability. Contribution of Working Group II to the Fourth References 137 Assessment Report of the Intergovernmental Panel on Shiffer, R.A. and Rossow, W.B. (1983) ‘The Inter- Climate Change, Cambridge University Press, Cam- national Cloud Climatology Project (ISCCP) bridge, UK, pp315–356 – The 1st project of the World Climate Research POWER (2008) Prediction of World Energy Resource Programme’, Bulletin of the American Meteorological (POWER), NASA Climatology Resource for Agro- Society, vol 64, no 7, pp779–84 climatology: Daily Averaged Data (Evaluation Ver- Siddiqui, K. U. and Hossain, A. N. H. A. (eds) (2006) sion), available at http://power.larc.nasa.gov Options for Flood Risk and Damage Reduction in Bang- Pyatt, G. (1988) ‘A SAM approach to modeling’, Jour- ladesh, University Press Limited, Dhaka, Bangladesh nal of Policy Modeling, vol 10 SRES (2000) Special Report on Emissions Scenarios, A Pyatt, G. and Round, J. (1985) Social Accounting Matrices: Special Report of Working Group III of the Inter- A Basis for Planning, World Bank, Washington, D.C. governmental Panel on Climate Change, Cambridge Rahmstorf, S. (2007) ‘A semi-empirical approach to University Press, Cambridge, UK, p599 projecting future sea level rise’, Science, vol 315, Tanner, T. M., Hassan, A., Islam, K. M. N., Conway, pp368–70 D., Mechler, R., Ahmed, A.U. and Alam, M. (2007) Rashid, H. and Paul, B. K. (1987) ‘Flood problems in ORCHID: Piloting Climate Risk Screening in Bangla- Bangladesh: is there any indigenous solution?’, Envi- desh, Detailed Research Report, Institute of Develop- ronmental Management, vol 11, no 2, pp155–173 mental Studies, University of Sussex, Brighton, UK Reinert, A. and Francois, J. (eds) (1997) Applied Meth- Thomalla, F., Cannon, T., Huq, S., Klein, R. J. T. and ods for Trade Policy Analysis: A Handbook, Cambridge Schaerer, C. (2005) ‘Mainstreaming Adaptation to University Press, New York Climate Change in Coastal Bangladesh by building Reinert, K. A. and Roland-Holst, D. W. (1997). ‘Social Civil Society Alliances’ Proceedings of the Solu- accounting matrices’, in J. F. Francois and K.A. tions to Coastal Disasters Conference 2005, Ameri- Reinert, Applied Methods for Trade Policy Analysis: A can Society of Civil Engineers (ASCE), Charleston, Handbook, Cambridge University Press, Cambridge South Carolina, 8-11 May, pp668–684 Robinson, S. (1989) ‘Multisectoral models’, in Chen- Thurlow, J. (2004) ‘A Standard Recursive Dynamic ery, H. and Srinivasan, T. N. (eds) Handbook of Devel- Computable General Equilibrium Model for South opment Economics, vol II, Elsevier Science Publishers, Africa: Extending the Static IFPRI model’,Working Amsterdam paper 1-2004, Trade and Industrial Policy Strategies, Robinson, S. and Roland-Holst, D.W. (1988). ‘Macro- Pretoria economic structure and computable general equi- Timsina, J., and Humphreys, E. (2006a) ‘Applications librium models’, Journal of Policy Modeling, vol 10 of CERES-Rice and CERES-Wheat in Research, Robinson, S., Cattaneo, A. and El-Said, M. (2001) Policy and Climate Change Studies in Asia: A ‘Updating and estimating a social accounting matrix Review’, International Journal of Agricultural Research, using cross entropy methods’, Economic Systems vol 1, no 3, pp202–225 Research, vol 13 Timsina, J., and Humphreys, E. (2006b) ‘Performance Rosenzweig, C. and Iglesias, A. (2006) Potential Impacts of CERES-Rice and CERES-Wheat models in of Climate Change on World Food Supply: Data Sets rice–wheat systems: A review’, Agricultural Systems, from a Major Crop Modeling Study, Goddard Institute vol 90, pp5–31 for Space Studies, Columbia University, New York Timsina, J., Singh, V., Badaruddin, M. and Meisner, C. Sarwar, M. G. M. (2005) ‘Impacts of Sea Level Rise (1998) ‘Cultivar, Nitrogen, and moisture effects on a on the Coastal Zone of Bangladesh’, Masters Thesis, rice-wheat sequence: Experimentation and simula- Lund University, Sweden, 45pp tion’, Agronomy Journal, vol 90, no 2, pp119–130 Sattar, S. (2000) ‘Bridging the rice yield gap in Bangla- Tubiello, F.N., Amthor, J. S., Boote, K. J., Donatelli, M., desh’, in M. K. Papademetriou, F.J. Dent, E. M. Easterling, W., Fischer, G., Gifford, R. M., Howden, Herath (eds), Bridging the Rice Yield Gap in Asia and M., Reilly, J. and Rosenzweig, C. (2007a) ‘Crop the Paciï¬?c, FAO RAP Publication 2000/16, FAO, response to elevated CO2 and world food supply’ Rome (A comment on “Food for Thought…â€? by Long et Selvaraju, R., Subbiah, A. R., Baas, S. and Juergens, I. al in Science vol 312, pp1918–1921, 2006), European (2006) ‘Livelihood adaptation to climate variability Journal of Agronomy, vol 26, no 3, pp215–223, April and change in drought-prone areas of Bangladesh’, Tubiello, F. N., Soussana, J.-F. and Howden, S. M. FAO Institutions for Rural Development Series, (2007b) ‘Crop and pasture response to climate Case Study 5, 115pp change’, Proceedings of the National Academy of Sciences, 138 Climate Change Risks and Food Security in Bangladesh vol 104, no 50, pp19686–19690, doi:10.1073/pnas. World Bank (2000) Bangladesh: Climate Change and 0701728104 Sustainable Development, Report No. 21104-BD, UNDP-GOB (1989) Bangladesh Flood Policy Study-Final Rural Development Unit, South Asia Region, Report, United Nations Development Programme- World Bank, Dhaka, Bangladesh, p95 Governement of Bangladesh, Dhaka, Bangladesh World Bank (2007) Floods 2007 Damage and Needs University of Washington Climate Impacts Group and Assessment Report, Dhaka, Bangladesh Washington Department of Ecology (2008) Sea World Bank (2008) Cyclone Sidr in Bangladesh, Damage, Level Rise in the Coastal Waters of Washington State, Loss, and Needs Assessment for Disaster Recovery and Climate Impacts Group, Seattle, WA Reconstruction, Dhaka, Bangladesh Warrick, R. A. and Ahmad, Q. K. (1996) (eds) The World Bank (2009) World Development Indicators, avail- Implications of Climate and Sea level Change for Bangla- able at www.worldbank.org/data desh, Kluwer Academic Publishers, London Wunsch, C. (1999) ‘The interpretation of short climate Watson, R.T., Zinyowera, M. C., Moss, R. H., Dokken, records, with comments on the North Atlantic and D. J. (1997) Summary for Policy Makers; The Regional Southern oscillations’, Bulletin of the American Mete- Impacts of Climate Change, A Special Report of IPCC orological Society, vol 80, pp245–55 Working Group II, IPCC, Geneva, Switzerland Yoshida, S. (1981) Fundamentals of Rice Crop Science, Webster, P.J. and Hoyos, C. (2004) ‘Prediction of mon- International Rice Research Institute, Los Baños, soon rainfall and river discharge on 15–30 day time Laguna, Philippines, 269pp scales’, Bulletin of American Meteorological Society, vol 85, no 11, pp1745–65 Index activity accounts 119, 126–128 annual temperature changes 24, 25 actual yields 8–9, 42, 59, 83, 102, 106, 107 aquaculture 18–19 adaptation aroid 86 agriculture 2, 4, 56–59, 82–99 aus rice production DSSAT model 108–112 economy-wide impacts 65, 66, 67, 69, 74, future 105–107 75, 76 importance 1 flooding 5, 7 agriculture future 41, 50, 51–52, 53, 54, 55, 56, 57, 58 adaptation 2, 56–59, 82–99 regional vulnerability 106 economy-wide impacts 60, 62–63, 65, 68–70, spatial distribution 6 71–72, 74–77, 78 see also rice production flooding 2–3, 5, 7–8, 10, 14–15 Average Climate Change Scenarios 73, 74, 76 future impacts 21, 33, 41–59 GDP 1, 5, 7, 60, 65, 68–70, 74–77, 78, 124 baira 90–91 households 117 Bangladesh Meteorological Department (BMD) inputs 110–112 43 management practices 45–46 baselines research and development 82–83 adaptation 111–112 SAM 119–132 development of 42–46 success of 6–10 flood 34, 35, 36, 48 see also crop production; rice production precipitation 23 agro-climatic regions 31, 32 sea-level rise 49 agro-ecological zones 2, 32, 123 vulnerability 106 aman rice production water levels 28, 37, 38 adaptation 93, 94, 97, 100 bed systems 88–91, 95, 101, 103, 104 economy-wide impacts 65–66, 67, 69, 74, biophysical crop models 2, 4, 42–43 75, 76 BMD see Bangladesh Meteorological flooding 5 Department future 41, 50, 51, 52, 53, 54, 55, 56, 57, 58 boro rice production monsoons 15, 16 economy-wide impacts 60, 66, 67, 69, 73–74, regional vulnerability 106 75, 76 spatial distribution 6 flooding 5, 7 variability 7–8, 9 future 41, 50, 51, 52, 53, 54–56, 57, 58 see also rice production regional vulnerability 106 annual flooding 10–15 spatial distribution 6 annual peak flows 37 see also rice production 140 Climate Change Risks and Food Security in Bangladesh box-and-whisker diagrams 51 cropping calendar 14, 109–110 Brahmaputra river 12–13, 33–34 crop production 6, 8, 14, 41–59, 60–81, 115–117 see also Ganges–Brahmaputra–Meghna Basin see also agriculture; rice production cross-entropy econometrics 130–132 calendar cropping 14, 109–110 cultivars 42, 43–45, 108–109 calibration 29–30 see also varieties capital cultivatable land 61, 64, 74, 116, 117, 126 accumulation 71 113 cyclones 105 CGE model 63, 64, 117 market equilibrium 113, 114 DAE see Department of Agriculture Extension SAM 122 damage functions 46–48 supply 113–114 damages 2–3, 52–56, 105 carbon dioxide Decision Support System for Agrotechnology CERES crop models 42–43 Transfer (DSSAT) 41–42, 49, 108–112 crop performance 1, 41, 49–51, 52, 54, 56 delta approach 30, 49, 51 fertilization 109 density of planting 110 CERES see Crop Environment Resource Department of Agriculture Extension (DAE) 83 Synthesis developing countries 107 CES see constant elasticity of substitution Dhaka 5 CGE see computable general equilibrium model disaster management 1, 106 char land 88–89, 92 discharge 10, 11–13, 15, 16, 28, 29, 30, 33–34 chickpea cultivation 99 see also rivers classiï¬?cation of floods 10 ditches 89–89 climate change simulation double-entry accounting 119 adaptation 110 dribbling approach 97 crop performance 41, 51, 52 droughts 15–16, 23, 83 economy-wide impacts 64, 72–73, 74, 76, 78 dry seasons 15–16, 18, 23, 31–33 floods 28, 37, 39 see also lean seasons future 21–27 DSSAT see Decision Support System for coastal flooding 48–49, 53, 81, 83–84 Agrotechnology Transfer combined practices 112 dynamic components 113–114 commodity accounts 119–121, 128 dynamic modelling 3, 4 computable general equilibrium (CGE) model 3, 4, 60, 61–64, 73, 105, 113–118 economy-wide impacts 60–81, 106, 115–117, conservation of water 20 119–132 constant elasticity of substitution (CES) 117 education 63, 64, 72 consumption elasticities 117, 118 CGE model 113, 114, 117 embankments 82 economy-wide impacts 60, 63, 71–72, 79–81 emissions scenarios SAM 123, 126 crop performance 42, 49, 50–56 country-speciï¬?c surveys 2 economy-wide impacts 73, 76 Crop Environment Resource Synthesis floods 31, 33 (CERES) 42–46, 108–112 future 21, 22 crop models integrated modelling 3 CERES 42–46, 108–112 national GDP 77–79 economy-wide 3, 4, 60, 61–64, 73, 113–118 environmental modiï¬?cations 111 see also CGE model evaluation of adaptation 83–104 Index 141 evaporation 28, 29 government 82, 83, 121, 128, 129, 130 existing climate variability 61–62, 64–72, 73, gross domestic product (GDP) 74, 76, 77, 106 agriculture 1, 5, 7, 60, 65, 68–70, 74–77, 78, exports 119, 129 121 extreme events 2–3, 12, 61, 63–64, 105, 115 national impacts 70–71, 77–79 see also droughts; floods SAM 124, 126, 130 groundnuts 86 factor payments 113, 114, 129 groundwater 1, 5, 6, 15, 18–19, 52, 82 factors 121–122, 125, 126–128, 129 farm level harvesting dates 109–110 adaptation 82–104 Himalayas 19 flood damage 48 historical data 4, 7–10, 11–13, 23, 62, 116 practices 45, 108–112 homestead cultivation 93–94, 95, 101 size 71, 79, 121 household level fertilizers adaptation 83, 107 adaptation 90–104, 107 consumption 60, 63, 71–72, 79–81, 117 carbon dioxide 109 income 113, 114, 117, 122–123, 125, 126 future crop performance 45 welfare 4 60 ï¬?sh 104 hydro-crop models 61, 62, 63, 72, 115 floating agriculture 56–59, 90–91 hydrology 4, 14–15, 28–40, 105 floods adaptation 82, 83, 86, 87, 92, 93, 94, 95 identiï¬?cation of adaptation options 83–104 agriculture 2–3, 5, 7–8, 14–15 imports 121, 128 coastal areas 48–49, 53, 80, 83–4 income damages 46–48, 52–53, 105 economy-wide impacts 72, 79, 87 economy-wide model 115–117 household 113, 114, 117, 122–123, 125, 126 future trends 26, 28–40, 41, 52–53 infrastructure 1, 20, 35, 82, 106 hydrologic models 4, 105 inputs 110–112 living with 10–15 integrated approaches 3–4, 54–56, 107 protection 6, 10, 30, 34, 82, 83 Intergovernmental Panel on Climate Change rivers 28–34, 35, 37–39 (IPCC) 17, 21, 22, 26, 31, 49 see also inundation international adaptation 56 foxtail millet 84, 92 inundation 3, 17, 48–49, 53, 115–116 fruits 89, 104 see also floods; sea-level rise future impacts 3, 4, 21–59, 105 investment 1, 5, 82, 106–107, 113, 114, 128–129 Ganges–Brahmaputra–Meghna (GBM) basin 5, IPCC see Intergovernmental Panel on Climate 19, 20, 28–30, 31–33 Change Ganges river 12, 13, 19, 33 irrigation 5, 45, 82, 92, 95, 97, 100, 103, GBM see Ganges–Brahmaputra–Meghna basin 110–111 GCMs see global climate models GDP see gross domestic product jute ï¬?elds 94 genetic information 42, 43–45, 108–109 glacier retreat 19 Kanchan 44 global climate models (GCMs) 3, 21, 22, 30–31, kaon 84, 92 51, 77 kharif 15, 16, 88, 92, 94, 95, 97, 101 Gorai River 39 see also monsoons 142 Climate Change Risks and Food Security in Bangladesh labour 63, 64, 113, 114, 117, 122–123, 126 MSL see mean sea level land mulching 86 extreme events 63–64 multiple practices 112 flooding 14–15, 28, 34–35, 115–116, 117 mustard 87 labour 113 rice cultivation 8, 9 national level SAM 121, 122, 123, 126–128 CGE model 113, 117 sea-level rise 17, 26 crop performance 41, 50, 51, 53, 54–56, 73, 74 landless households 72, 79, 122 flood area 36 large-scale farm households 71, 79, 121 GDP impacts 70–71, 77–79 lean-seasons 5, 15–16 hydrologic super model 30 see also dry seasons non-agricultural sector 70–71, 117, 121, 122, lentils 87 123, 130 literature review 2–3 non-farm level livelihoods 10–15 consumption 79, 80 location 84–85, 109 enterprises 122, 125, 126, 128, 129 households 71, 72, 80 macro-micro models 62 no-regrets strategies 106 macro social accounting matrix 125–130 northwestern regions 106 maize 6–7, 97–98 management 9, 20, 45–46 onset times 37–39 marginal farmers 72, 79, 121 optimal climate simulation 60, 61, 62, 64–65, mashkalai 84, 85, 87 66, 68, 70, 71 mean sea level (MSL) 5 organic amendments 111 Meghna river 12, 13, 18, 33–34, 38 see also Ganges–Brahmaputra–Meghna basin papaya 101 MIKE BASIN model 28 parenga practice 93 minimum tillage 86–87 peak flows 5, 10, 12–13, 28, 37 mini ponds 100 phenological stage 43, 46–47 modelling plant height 46, 47 CGE 3, 4, 60, 61–64, 73, 105, 113–118 planting dates 109–110 DSSAT 108–112 polythene bags 95–96 hydro-crop models 61, 62, 63, 72, 115 ponds 101, 103 hydrology 4, 28–40 population 19, 64, 107 integrated 3–4, 107 potato 86 uncertainty 105–106 potential yields 8–9, 42, 59, 83, 107 modiï¬?ed sorjan system 88–89 poverty 1, 106–107 monsoons precipitation aman rice 15 crop performance 41, 51 crop performance 52 flood hydrology 31, 33 future trends 23, 28, 31, 33, 34 future trends 4, 21, 22–25, 26 GBM basin 6 historical trends 11, 12 onset and recession 37–39 see also monsoon; rainfall salinity 18 priming techniques 99 see also kharif prior social accounting matrix 125, 130–132 Monte Carlo process 35, 62, 72, 116 processing sector 60, 70 monthly temperature changes 24, 25 production see agriculture; rice production Index 143 production functions 2, 63, 114, 117, 118 sea-level rise productivity 1, 2, 5, 63, 114 coastal areas 1–2, 17–19 proï¬?t maximization 113 crop performance 41, 48–49, 54, 101 protected areas 35 cultivatable land 61, 74, 116, 117 pulses 7, 87, 99 future trends 21, 26 see also inundation rabi vegetables 103 seasonal level rainfall 29, 30, 31, 33, 52 adaptation options 104, 112 see also monsoons; precipitation Bangladesh 5–6 raised beds 88, 95, 104 future trends 22, 23 RCMs see regional climate models precipitation 12 recession of floods 37–39, 95 seedbed temperature 45 regional climate models (RCMs) 22 selection of global climate models 30–31 regional level self-sufï¬?ciency 6 adaptation 82 sensitivity analysis 3, 105–106 crop performance 57 sequence of crops 110, 112 flood types 11, 15 small-scale farmers 121 hydrology 19–20 social accounting matrix (SAM) 62, 118, studies 2 119–132 vulnerability 106 soil 9–10, 42, 43, 44, 86, 97 see also sub-regional level sorjan systems 88–89, 104 relay cropping 94 southern regions 106 research and development 82–83 spatial level 6, 8, 34–35, 121 reserves of grain 10–11 Special Report on Emissions Scenarios (SRES), rice production IPCC 21, 31 adaptation options 108–112 sprouted seeds 94 climate variability 7–10, 60–81 SRES see Special Report on Emissions Scenarios crop calendar 14 static components 113, 114 cropping intensity 6, 7 straw 86, 92 droughts 16 sub-optimal climate conditions 65 future performance 41–59 sub-regional level increases 1 crop performance 43, 44, 53, 54, 56, 58 losses 105 delineation 31, 32 regional vulnerability 106 economy-wide impacts 63, 68, 69, 74, 79–81 see also aman rice; aus rice; boro rice flooding 35, 46, 47 rivers vulnerability 106 discharge 10, 11–13, 15, 16, 28, 29, 30, 33–34 see also regional level GBM basin 5, 19, 20, 28–30, 31–33 sub-sets of models 30–31 glacial retreat 19 sugar cane 7 salinity 18 supplementary irrigation 100, 101 see also discharge supply management 20 roof vegetable cultivation 101 surface and groundwater irrigation see irrigation runner-type vegetables 101 t.aman see transplanted aman salinity 3, 5, 17, 18–19, 59, 83 taxes 121, 128, 129 SAM see social accounting matrix technical change 63, 64, 113, 114, 116 savings 129, 130 technology adoption 83 144 Climate Change Risks and Food Security in Bangladesh Teesta River 38 vegetable cultivation 88–91, 95–96, 101–102 temperature vulnerability 5–20, 72, 82, 105, 106 agriculture 1, 2 crop performance 1, 41, 51 warming trends 2, 21–22, 23, 31 flood hydrology 29, 30, 31, 32 see also temperature future predictions 4 water glaciers 19 conservation 20 sea-level rise 17 disasters 1–2 transplant 45 harvesting 98 warming trends 4, 21–22, 22–25 management 20 temporal level 35–39 water hyacinth 85–86, 90 TFP see total factor productivity water levels tidal fluctuations 49 flood damage 46, 47, 48 tillage 86–87, 97–98, 110 future trends 34 time slices 22, 42, 62 hydrology 28, 30 timing 5, 12, 13, 37–39 peak 37 total factor productivity (TFP) 63, 64 relay cropping 94 trade 117 rivers 12, 17–18, 30 transplanted aman (t.aman) 93, 97, 100 welfare 4, 60, 71 see also aman production wet seasons 23 transplant environments 45 see also kharif; monsoons trellises 95–96, 101 wheat production cultivar selection 108, 109 uncertainty future 44–45, 45, 50, 51, 52, 55, 56, 57, 58 CGE model 117–118 increases 1, 6 crop performance 42–43 Worst Case Scenarios 65, 66, 70, 71 emissions scenarios 73, 77, 79 flood damage 47–48 year-round homestead vegetable cultivation future trends 21 101–102 GCMs 60 yields modelling 105–106 climate variability 8–10 unflooded production 49–52 crop models 3 urea 92, 93, 94, 95, 99 economy-wide model 115, 117 future 41–59 validation 29–30 gap 8–9, 42, 59, 83, 107 variability 5, 7–10, 23, 30, 47, 60–81, 115–117 varieties 7–8, 83 zero tillage 86–87, 97–98 see also cultivars zuzubi garden 88–89