d i s c u s s i o n pa p e r n u m B e r 4 august 2010 d e v e l o p m e n t a n d c l i m at e c h a n g e d i s c u s s i o n pa p e r s 1 56662 d e v e l o p m e n t a n d c l i m a t e c h a n g e The Costs of Agricultural Adaptation to Climate Change D I S C u S S I O N PA P E R N u M B E R 4 AuGuST 2010 D E V E L O P M E N T A N D C L I M A T E C H A N G E The Costs of Agricultural Adaptation to Climate Change Gerald C. Nelson, Mark W. Rosegrant, Jawoo Koo, Richard Robertson, Timothy Sulser, Tingju Zhu, Claudia Ringler, Siwa Msangi, Amanda Palazzo, Miroslav Batka, Marilia Magalhaes, Rowena Valmonte-Santos, Mandy Ewing, and David Lee Papers in this series are not formal publications of the World Bank. They are circulated to encourage thought and discussion. The use and cita- tion of this paper should take this into account. The views expressed are those of the authors and should not be attributed to the World Bank. Copies are available from the Environment Department of the World Bank by calling 202-473-3641. © 2010 The International Bank for Reconstruction and Development / THE WORLD BANK 1818 H Street, NW Washington, DC 20433, U.S.A. Telephone: 202-473-1000 Internet: www.worldbank.org/climatechange E-mail: feedback@worldbank.org All rights reserved. August 2010 This volume is a product of the staff of the International Bank for Reconstruction and Development / The World Bank. The findings, interpretations, and conclusions expressed in this volume do not necessarily reflect the views of the Executive Directors of The World Bank or the governments they represent. The World Bank does not guarantee the accuracy of the data included in this work. The boundaries, colors, denominations, and other information shown on any map in this work do not imply any judgement on the part of the World Bank concerning the legal status of any territory or the endorsement or acceptance of such boundaries. RIGHTS AND PERMISSIONS The material in this publication is copyrighted. Copying and/or transmitting portions or all of this work without permission may be a violation of applica- ble law. The International Bank for Reconstruction and Development / The World Bank encourages dissemination of its work and will normally grant permission to reproduce portions of the work promptly. For permission to photocopy or reprint any part of this work, please send a request with complete information to the Copyright Clearance Center Inc., 222 Rosewood Drive, Danvers, MA 01923, USA; telephone 978-750-8400; fax 978-750-4470; Internet: www.copyright.com. Cover image: Republic of Yemen. © Bill Lyons / World Bank Photo Library. All dollars are U.S. dollars unless otherwise indicated. iii Table OF CONTeNTS Adjustments to the IFPRI EACC Estimates for Agriculture vii 1. Introduction 1 2. Overview of the Modeling Methodology 2 2.1 ClimateData 2 2.2 Cropmodeling 3 2.3 TheIMPACT2009Model 4 2.4 ModelingClimateChangeinIMPACT 5 3. Modeling Results 7 3.1 TheEffectsofClimateChangeonYields 7 3.1.1 Directclimatechangeeffectsonrainfedandirrigatedyields 7 3.1.2 Indirecteffectsfromclimatechange:Waterstressforirrigatedcrops 10 3.2 ClimateChangeImpactsonAgricultureandHumanWell-being 15 3.2.1 Pricesandproduction 15 3.2.2 Tradeinagriculturalcommodities 19 3.2.3 Fooddemand 21 3.2.4 Welfareeffects 21 3.3 TheCostsofAdaptation 23 3.4 SensitivityAnalysis 26 3.5 Limitations 28 3.6 Conclusions 28 4. Annex. IFPRI's Climate Change Modeling Methodology 30 4.1 CropModeling 30 4.2 ClimateData 30 4.3 OtherAgronomicInputs 32 4.3.1 Soilcharacteristics 32 4.3.2 Cropvariety 32 iv THE COSTS OF AGRICu LT uRAL ADAPTATION TO CLIMATE CH A N G E 4.3.3 Croppingcalendar 32 4.3.4 CO2fertilizationeffects 33 4.3.5 Wateravailability 34 4.3.6 Nutrientlevel 34 4.4 FromDSSATtotheIMPACTmodel 34 4.5 TheIMPACT2009Model 34 4.6 ModelingClimateChangeinIMPACT 34 4.7 ModelingtheCostsofAdaptationtoClimateChange 36 4.8 EstimatingChildMalnutrition 37 4.9 AgriculturalResearchInvestments 37 4.9.1 Agriculturalresearchinvestmentssensitivityanalysis 38 4.10RuralRoads 39 4.10.1 Areaeffect 39 4.10.2 Yieldeffect 40 4.10.3 Scenarioresultsandadditionalroadcosts 40 4.11Irrigation 41 4.11.1 Areaexpansion 41 4.11.2 Irrigationefficiencyimprovements 41 4.11.3 Irrigationinvestmentssensitivityanalysis 42 4.12Population,incomeandclimatefuturescenarioassumptions 44 5. References 47 Tables 1 Yield changes by crop and management system under current climate and two climate change scenarios with and without CO2 fertilization effects (% change from yields with 2000 climate) 8 2 Water availability and use under current climate, and percent changes under two climate change scenarios in 2050 13 3 Yield changes for irrigated crops due to water stress under current climate and two climate change scenarios (% change from 2000 yields) 14 4 Population and income growth assumptions 15 5 World prices of selected crops and livestock products (uS$/metric ton) 15 6 Combined biophysical and economic yield effects from climate change, no CO2 fertilization 17 7 Climate change effects on crop production, no CO2 fertilization 18 8 Net cereal (rice, wheat, maize, millet, sorghum, and other grains) exports by region in 2000 and 2050 under scenarios with and without climate change (000 mt) 20 9 Value of net cereal trade by region (million uS$) 20 10 Per capita food consumption (kg per year) of cereals and meats with and without climate change 22 11 Daily per capita calorie availability with and without climate change 22 12 Total number of malnourished children in 2000 and 2050 (million children under 5 yrs) 23 D E V E L O P M E N T A N D C L I M AT E C H A N G E D I S C u S S I O N PA P E R S v Tables (ConTinued) 13 Investment and productivity scenarios for climate change adaptation 24 14 Daily calorie per capita consumption with adaptive investments (Kcals/person/day) 25 15 Number of malnourished children with adaptive investments (million children, under 5 years) 25 16 Additional annual investment expenditure needed to counteract the effects of climate change on nutrition (million 2000 uS$) 27 17 Percentage change in malnourished children with 10 percent increases in GDP, productivity growth, and population 28 18 Assumed multipliers of historic growth rates of agricultural research expenditures 38 19 Research investment sensitivity analysis 39 20 Road construction costs (2005 uS$ per km) 40 21 Percent yield increase with respect to road length, regional averages 40 22 Irrigation investment cost (uS 2000$ per hectare) 41 23 Results from alternate estimation of irrigation efficiency improvement costs (uS 2000 million per year) 43 24 Precipitation and temperature regional average changes, 2000 to 2050 44 Figures 1 The IMPACT 2009 modeling framework 3 2 Change in average maximum temperature, 2000­50, CSIRO 4 3 Change in average maximum temperature, 2000­50, NCAR 4 4 Change in precipitation, 2000­50, CSIRO 4 5 Change in precipitation, 2000­50, NCAR 4 6 IMPACT model units of analysis, the Food Production unit (FPu) 5 7 Yield changes by crop and management system under current climate and two climate change scenarios with and without CO2 fertilization effects (% change from yields with 2000 climate) 11 8 World prices of major grains (2000 uS$) 16 9 Net cereal (rice, wheat, maize, millet, sorghum, and other grains) trade by region in year 2000 and 2050 under scenarios with and without climate change (mmt) 21 10 Daily per capita calorie availability with and without climate change 23 11 Daily calorie availability, South Asia and Sub-Saharan Africa 24 12 Child malnutrition effects, South Asia and Sub-Saharan Africa (millions of children) 26 13 Daily calorie availability, East Asia and the Pacific, Europe and Central Asia, Latin America and the Caribbean, and Middle East and North Africa 26 vi THE COSTS OF AGRICu LT uRAL ADAPTATION TO CLIMATE CH A N G E Figures (ConTinued) 14 Child malnutrition effects, East Asia and the Pacific, Europe and Central Asia, Latin America and the Caribbean, and Middle East and North Africa (millions of children) 26 15 The SPAM data set development process 31 16 Rainfed crop planting month, 2000 climate 32 17 Rainfed planting month, 2500 climate, CSIRO GCM A2 scenario (AR4) 33 18 Rainfed planting month, 2500 climate, NCAR GCM A2 scenario (AR4) 33 19 Irrigated planting month, 2000 climate 33 20 Irrigated planting month, 2500 climate, CSIRO GCM A2 scenario (AR4) 33 21 Irrigated planting month, 2500 climate, NCAR GCM A2 scenario (AR4) 33 22 IMPACT model units of analysis, the Food Production unit (FPu) 35 23 Exogenous productivity growth rates (% per year) for selected crops and management type 45 vii Please note that the estimates of costs of adaptation for the agricultural sector presented in this report differ from the cost estimates used in the synthesis report of the "Economics of Adaptation to Climate Change." The note below provides an explanation for this difference. adjuSTmeNTS TO The IFPRI eaCC and life expectancy for health, infrastructure services, etc. eSTImaTeS FOR agRICulTuRe (d) As far as possible the analysis takes account of link- ages across sectors so as to avoid duplication of b a Ckg RO uN d estimates and the likelihood that changes in the development baseline in one sector will affect the The adjustments made to IFPRI's original estimates of cost of adaptation for other sectors. the cost of adaptation for agriculture have to be under- stood within the framework of analysis that was estab- Many economic scenarios have been used for discus- lished for the EACC Global Analysis. The key sions of climate change and the differences between elements are as follows: them are a significant source of difficulty in making comparisons. Rather than reinvent the wheel, the devel- (a) The costs are restricted to those incurred by the opment baseline for the EACC study was linked to the public sector and exclude what is often referred to UN's Medium Fertility population projection and aver- as "autonomous" adaptation--i.e. investment or age regional rates of economic growth derived from the expenditures by private individuals or companies in main economy-environment models used for economic response to changes in market prices or other analyses of climate change. These assumptions define a signals linked to climate change. consistent framework for analysing the costs of (b) Adaptation is measured relative to an "efficient" adaptation. baseline scenario economic and social development without climate change. This scenario allows for They do not represent an attempt to generate a definitive the impact of economic growth, population set of economic projections up to 2050. Since many of increase, urbanisation, etc up to 2050 on the level the component costs of adaptation are affected directly or and composition of public spending and indirectly by economic growth, monetary estimates of the investment. cost of adaptation depend upon these projections. (c) The cost of adaptation is defined as the additional Expressing the costs of adaptation as percentages of public expenditures necessary either to ensure that either GDP or sector expenditures under the baseline is sector welfare with climate change is not worse likely to provide a more robust way of assessing the over- than the level of sector welfare on the baseline all burden of adapting to climate change. without climate change. Sector welfare is given different interpretations according to the context, aN IlluSTRaTION ­ COaSTal PROTeCTION but in general the idea is to maintain the output of sector-specific services or outcomes--i.e. the level The implementation of this approach can be illustrated of malnutrition for agriculture, infant mortality by reference to coastal protection. viii THE COSTS OF AGRICu LT uRAL ADAPTATION TO CLIMATE CH A N G E Step 1 ­ A baseline scenario for development with- computable general equilibrium model of the world out climate change is defined. This includes a set of agricultural market which incorporates detailed agro- rules that govern the construction of coastal economic information to underpin the projections of defences that take account of growth in urban crop and other agricultural production. The baseline population and GDP in coastal zones. The time projections of the EACC model are used to determine profile of public investment and associated expen- demand for agricultural products while climate vari- diture on O&M in the sector is estimated for the ables, specified in considerable detail, influence patterns baseline scenario. Even without climate change of land use and agricultural production. Agricultural there will be significant expenditures on coast markets clear through adjustments in the prices of agri- defences because (a) the increase in the value of cultural products and volumes of trade, taking account income & urban assets in coastal zones due to of distribution and transport margins.1 economic growth justifies higher levels of protec- tion against flooding and storms, and (b) long run The key measure of welfare is the level of child malnu- changes means that some coastal zones are sinking trition in each country and for developing countries in and would require greater protection even without total. This is estimated using an equation that includes climate change--this is partly due to extraction of demographic variables and food availability in kilocalo- groundwater (Bangkok, Jakarta) or shifts in conti- ries per capita per day. Higher food prices lead to lower nental plates. consumption (availability) of food and thus higher malnutrition holding other variables constant. Thus, the Step 2 - The time profile of public investment and key impact of climate change on agricultural welfare is associated expenditure on O&M in the sector is mediated through food prices and their effect on food estimated for a specific climate scenario, on the consumption, since the other influences--life expec- assumptions that (i) baseline rates of population tancy, female education and access to safe water--are growth, GDP growth, urbanisation, etc remain the held constant across climate scenarios. same, and (ii) expenditures are adjusted to hold the level of services or of welfare achieved equivalent to In the baseline scenario without climate change the levels projected under the baseline scenario. For (NoCC) the IPRI results show that calorie availability coastal protection, this is interpreted as applying in developing countries would increase from an average the same rules for when coast defences would be of about 2700 kcal per person per day in 2000 to about built (in terms of the risks of flooding or storm 2890 in 2050, whereas it would fall to about 2420 under damage) and then calculating any changes in the both the NCAR and the CSIRO scenarios. Given other damage caused to assets and people who are not changes in the baseline scenario, the total number of protected. malnourished children falls from about 147 million in 2000 to 113 million in 2050 without climate change. Step 3 - The cost of adaptation is then calculated However, with climate change the reduction from the as the difference between the total of investment 2000 figure is much lower--only down to 137­138 and O&M expenditures under the climate scenario million depending upon the climate scenario used.2 and the equivalent total under the baseline scenario without climate change. This corresponds to identi- In the IMPACT model certain types of public spend- fying the costs associated with climate change ing--primarily investments in irrigation expansion and while holding constant the baseline assumptions for rural roads--are partly driven by changes in agricultural economic development, etc. aPP lIC aTION TO agRICulTuR e ­ 1 The IMPACT model is described in detail in a number of technical ma INTa ININ g wel FaRe papers produced by IFPRI. The discussion here relies upon Rosengrant et al (2002) and Rosengrant et al (2008). 2 All of the figures are taken from the December 2009 version of the paper The IMPACT model used as the basis for estimating but they are the same in earlier and later versions. The climate scenarios do the costs of adaptation for agriculture is, in effect, a not take account of the impact of carbon fertilisation. D E V E L O P M E N T A N D C L I M AT E C H A N G E D I S C u S S I O N PA P E R S ix production. Thus, the NoCC scenario assumes that the and roads for each climate scenario with and without increase in food availability is underpinned by expendi- adaptation. This includes public spending that is built tures on irrigation and rural roads. Since overall food into the baseline scenario as it is required to support production is lower in the NCAR and CSIRO scenar- the level of food production that yields the reduction ios, public expenditures in these categories is also lower. in child malnutrition projected under the baseline So, without any adaptation to climate change public scenario. Hence, the cost on adaptation generated by spending would fall but the level of welfare--measured the IMPACT model that is consistent with the by child malnutrition--would also be lower. EACC definition is the difference between public spending on agricultural research, irrigation and roads The reference point for adaptation is to restore welfare for NCAR with adaptation scenario and public spend- to its baseline level, i.e. to reduce child malnutrition ing on the same categories for the NoCC scenario. As country by country to the levels that would have explained, it is necessary to adjust IFPRI's reported prevailed without climate change. In principle, this costs of adaptation to obtain these estimates for each could be achieved by public spending outside agriculture country. and linked sectors, for example by measures to increase life expectancy, female education or access to clean R uR a l R Oa dS water. A full optimisation model would look for the cheapest way of reducing child malnutrition and A second adjustment is required to ensure that the compute the cost of adaptation in that way. However, treatment of rural roads is consistent across different the EACC study focuses on sector-specific adaptation, sectors covered by the EACC study. The issue may be so the IMPACT model is used to estimate what public explained as follows. spending upon agricultural research, irrigation and roads is required to restore child malnutrition to its baseline The development baseline for infrastructure without values country by country. climate change assumes that public spending on paved and unpaved roads grows in accordance with equations It should be noted that a part of the additional spend- which explain the total length of roads and the propor- ing is driven directly by adaptation measures while the tion of roads that are paved as functions of income per remainder is a consequence of the complementary person, total population, urbanisation and a range of spending required to support the increased level of food country characteristics. Estimates of the cost of adapta- production. This distinction is important. Setting aside tion for roads are based on the additional costs of changes in the location and composition of food ensuring that these roads are built and maintained to production, adaptation will restore food availability design standards that reflect the changes in climate under each of the climate scenarios to what it would conditions projected under each climate scenario. have been without climate change. To the extent that However, the baseline scenario implies that there may food production is the same under the NoCC and, say, be some substantial increase in the length of rural roads NCAR with adaptation scenarios, public spending on purely as a consequence of economic development with- complementary inputs (irrigation, roads, etc) will be the out taking any account of the requirement to get food same in these two runs. Thus, the cost of adaptation will products to market. be the direct expenditures required to restore food production. In practice, the calculations are rather more On the other hand, the IMPACT model treats public complex because trade and changes in comparative spending on rural roads partly as complementary to advantage mean that food production is not exactly food production and partly as a source of adaptation. It restored on a country-by-country basis. does not attempt to calculate whether the spending on rural roads that it identifies as being necessary for either This is where the first of the adjustments to the the baseline or the climate change with adaptation IFPRI estimates is required. What IFPRI reports as scenario would be covered by the expansion in roads the cost of adaptation is the difference between the built into the development baseline for the infrastruc- overall levels of public spending on research, irrigation ture sector. x THE COSTS OF AGRICu LT uRAL ADAPTATION TO CLIMATE CH A N G E To address this, consider a hypothetical country A. In malnutrition. Comparison of the IFPRI scenarios with 2000 it has 100,000 km of roads of which 40,000 km and without climate change means that the cost of are intra-urban and inter-urban highways and 60,000 adaptation for rural roads is equivalent to the cost of km are rural roads. By 2050 the baseline development building and maintaining 20,000 km [ = 110,000 km projection for roads implies that it will have 300,000 (IMPACT NCAR with adaptation) ­ 90,000 km km of roads. Allowing for increases in income and (IMPACT NoCC)] of rural roads. urban population suggests that the length of intra- and inter-urban roads will increase to 200,000 km, so that However, taking account of the expansion in the provi- the length of rural roads will grow from 60,000 km to sion of rural roads built into the infrastructure baseline 100,000 km. means that the adjusted cost of adaptation should only include the cost of building and maintaining 10,000 km Quite separately, suppose that the IFPRI analysis based [ = 110,000 km (IMPACT NCAR with adaptation) ­ on the IMPACT model indicates that the length of 100,000 km (Infrastructure NoCC baseline)] of rural rural roads required in the NoCC scenario as to support roads. This is the consequence of ensuring that the same the projected increase in food production up to 2050 baseline scenario without climate change is applied will be 90,000 km. In that case, the apparent require- across all sectors. ment for additional public spending on rural roads in the NoCC identified by the IFPRI is covered by the The calculations required to make this adjustment are a increase in the provision of rural roads from 60,000 km little more complicated than the initial adjustment to to 100,000 km under the development scenario. the IFPRI costs of adaptation described above, but they are relatively straightforward once the full set of possi- Next, the IFPRI analysis concludes that the total length ble inequalities has been enumerated. The adjustment of rural roads under the NCAR with adaptation that has been applied uses the baseline development scenario would have to be 110,000 km in order to scenario for roads together with information on intra- support the restoration of food production to a level and inter-urban roads plus rural roads obtained from which is consistent with the NoCC estimate of child World Road Statistics and IFPRI. 1 1. INTROduCTION assessment report of the Intergovernmental Panel on Climate Change (IPCC et al., 2007) provide the climate inputs to the modeling work. Climate change will have large, but still uncertain, effects on agriculture. In this report we provide esti- The challenge of modeling climate change impacts in mates of the impacts on human well-being through agriculture arises in the wide ranging nature of processes effects on agricultural production, prices, and trade. Two that underlie the working of markets, ecosystems, and indicators provide the metrics to assess the impacts on human behavior. Our analytical framework integrates human well-being--per capita calorie consumption and modeling components that range from the macro to the child malnutrition count. We use these metrics to assess micro and from processes that are driven by economics the costs of adaptation with three types of investment-- to those that are essentially biological in nature. agricultural research, rural roads, and irrigation infra- structure and efficiency improvement. To provide some We begin this report with a discussion of the modeling idea of the uncertainties inherent in the climate change methodology and data used. The second part of the simulations, two general circulation models (GCMs) report provides the results of the analysis. An Annex using the A2 SRES scenario from the fourth provides additional technical details. 2 2. OveRvIew OF The mOdelINg Future climate data are assumed to be for the year 2050. To provide some idea of the uncertainties inherent in meThOdOlOgy the climate change simulations, results from two general circulation models (GCMs)--from NCAR (NCAR- An illustrative schematic of the links between the CCSM3) and CSIRO (CSIRO-Mk3.0)--using the A2 partial agriculture equilibrium model that emphasizes SRES scenario from the fourth assessment report of the policy and trade simulations and the biophysical Intergovernmental Panel on Climate Change (IPCC, modeling that emphasizes hydrology and agronomic M.L. Parry, O.F. Canziani, J.P. Palutikof, P.J. van der potential is shown in Figure 1. The modeling method- Linden and C.E. Hanson, 2007) are used.1 At one time ology reconciles the limited spatial resolution of macro- the A2 scenario was considered extreme although recent level economic models that operate through findings suggest it may not be. All scenarios have higher equilibrium-driven relationships at a national level with temperature in 2050, which results in greater evapora- detailed models of dynamic biophysical processes. The tion. When this water eventually returns to the earth as biophysical modeling combines crop modeling results precipitation, it can fall either on land or the oceans. from the Decision Support System for Agrotechnology The NCAR scenario is "wet" in the sense that average Transfer (DSSAT) crop modeling suite ( J. W. Jones et precipitation on land increases by about 10 percent al., 2003), which simulates responses of five important between 2000 and 2050. The CSIRO scenario is "dry", crops (rice, wheat, maize, soybeans, and groundnuts) to with land-based precipitation totals increasing only climate, soil, and nutrients and the SPAM data set of about 2 percent. crop location and management techniques (Liang You and Stanley Wood, 2006). This analysis is done at a All climate variables are assumed to change linearly spatial resolution of one half degree. The results are fed between their values in 2000 and 2050. This assumption into IFPRI's global agricultural supply and demand eliminates any random extreme events such as droughts projection model, IMPACT. An overview of the or high rainfall periods and also assumes that the forc- modeling process is presented here with more details in ing effects of GHG emissions proceed linearly; that is, the Annex. 1 NCAR and CSIRO AR4 data were downscaled by Kenneth Strzepek 2.1 Cl I maTe d aTa and colleagues at the MIT's Center for Global Change Science. We acknowledge the international modeling groups for providing their data for analysis, the Program for Climate Model Diagnosis and Intercomparison (PCMDI) for collecting and archiving the model To simulate today's climate the Worldclim current data, the JSC/CLIVAR Working Group on Coupled Modeling conditions data set (www.worldclim.org) is used which (WGCM) and their Coupled Model Intercomparison Project is representative of 1950­2000 and reports monthly (CMIP) and Climate Simulation Panel for organizing the model data analysis activity, and the IPCC WG1 TSU for technical support. The average minimum and maximum temperatures and IPCC Data Archive at Lawrence Livermore National Laboratory is monthly average precipitation. supported by the Office of Science, U.S. Department of Energy. D E V E L O P M E N T A N D C L I M AT E C H A N G E D I S C u S S I O N PA P E R S 3 FIguRe 1. The ImPaCT 2009 mOdelINg FRamewORk Model Inputs and Scenario Definitions Area elasticities w.r.t. Urban growth & changes in crop prices food habits (demand elasticities) Supply, demand, and Yield elasticities w.r.t. trade data from Income growth crop, labor, and capital prices FAOSTAT, IFPRI, UN, projections World Bank, and others Area and yield annual Population growth growth rates projections Water Simulation Water Supply Domestic Prices ­ Renewable H2O (world price, trade wedge, marketing margin) ­ Effective H2O for Irrigated and Rainfed Crops Water Demand Supply Projection Demand Projection ­ Irrigation ­ Livestock ­ Domestic ­ Industry ­ Environment Iteration for Net Trade World Market exports ­ imports Clearing Climate Scenarios Malnutrition Rainfall, Runoff, Potential ET Results World Update Adjust Trade Balance Inputs World Price Imports = Exports Go to NO YES Next Year Model Calculations (Food) Source: Rosegrant et al. 2001. with no gradual speedup in climate change. The effect average maximum temperatures than does CSIRO. of this assumption is to underestimate negative effects The CSIRO scenario has substantial precipitation from climate variability. declines in the western Amazon while NCAR shows declines in the eastern Amazon. These figures illus- Figure 2 and Figure 3 show the change in average trate qualitatively the range of potential climate maximum temperature between 2000 and 2050 for outcomes with current climate modeling capabilities the CSIRO and NCAR scenarios. Figure 4 and and thus an indication of the uncertainty in climate Figure 5 show changes in average precipitation. In change impacts. each set of figures the legend colors are identical; i.e., a specific color represents the same change in temper- 2 . 2 C R O P m O d e lI N g ature or precipitation across the two scenarios. A quick glance at these figures shows the substantial differ- DSSAT provides a common data interfaces to several ences that exist across these two climate scenarios. For extremely detailed process models of the daily develop- example the NCAR scenario has substantially higher ment of different crop varieties from planting to 4 THE COSTS OF AGRICu LT uRAL ADAPTATION TO CLIMATE CH A N G E FIguRe 2. ChaNge IN aveRage maxImum FIguRe 3. ChaNge IN aveRage maxImum TemPeRaTuRe, 2000­50, CSIRO TemPeRaTuRe, 2000­50, NCaR FIguRe 4. ChaNge IN PReCIPITaTION, FIguRe 5. ChaNge IN PReCIPITaTION, 2000­50, CSIRO 2000­50, NCaR Source: Authors' data. 2 harvest-ready. The system requires daily weather data, 2 . 3 T h e ImPa C T 2 0 0 9 m O d e l including maximum and minimum temperature, solar radiation, and precipitation, a description of the soil The IMPACT model was initially developed at the physical and chemical characteristics of the field, and International Food Policy Research Institute (IFPRI) to crop management, including crop, variety, planting project global food supply, food demand, and food secu- date, plant spacing, and inputs such as fertilizer and rity to 2020 and beyond (M.W. Rosegrant, S. Msangi, irrigation. For this report, five crops--rice, wheat, C. Ringler, T.B. Sulser, T. Zhu and S.A. Cline, 2008). It maize, soybeans, and groundnuts--are directly is a partial equilibrium agricultural model with 32 crop modeled with DSSAT. All other crops in the and livestock commodities, including cereals, soybeans, IMPACT model are mapped to one or more of these roots and tubers, meats, milk, eggs, oilseeds, oilcakes and crops based on similarity in photosynthetic metabolic meals, sugar, and fruits and vegetables. IMPACT has pathways. 115 country (or in a few cases country aggregate) regions, within each of which supply, demand, net trade, Not all of the data needed for DSSAT are readily avail- and prices for agricultural commodities are determined. able, so various approximation techniques were devel- Large countries are further divided into major river oped, as described in the Annex. DSSAT is run at 0.5 basins. The result, portrayed in Figure 6, is 281 discrete degree intervals for the locations where the SPAM data locations, called food production units (FPUs). The set says the crop is currently grown. The results from model links the various countries and regions through this analysis are then aggregated to the IMPACT FPU level as described below. 2 Rosegrant et al. (2008) provides technical details on the IMPACT model. D E V E L O P M E N T A N D C L I M AT E C H A N G E D I S C u S S I O N PA P E R S 5 FIguRe 6. ImPaCT mOdel uNITS OF aNalySIS, The FOOd PROduCTION uNIT (FPu) Source: Authors' data. international trade using a series of linear and nonlinear equation (1) as well as the water availability coefficient equations to approximate the underlying production (WAT) for irrigated crops. See Table 24 at the end of and demand relationships. World agricultural commod- the Annex for intrinsic growth coefficients for selected ity prices are determined annually at levels that clear crops. These rates depend on crop, management system, international markets. Growth in crop production in and location. For most crops, the average is about 1 each country is determined by crop and input prices, percent per year from effects that are not modeled. But exogenous rates of productivity growth and area expan- in some countries the growth is assumed to be negative, sion, investment in irrigation, and water availability. while in others it is as high as 5 percent per year for Demand is a function of prices, income, and population some years. growth and contains four categories of commodity demand--food, feed, biofuels feedstock, and other uses. YC tni = tni × ( PStni ) iin × ( PFtnk ) ikn × (1) k (1 + gy tni ) - YCtni (WATtni ) 2.4 mOdel INg ClImaT e ChaNge I N I mPa CT We generate relative climate change productivity effects by calculating location-specific yields based on DSSAT Climate change effects on crop productivity enter into results for 2000 and 2050 climate as described above the IMPACT model by affecting both crop area and and then construct a set of yield growth rate changes yield. For example, yields (YC) are altered through the cause by climate change. These rate changes are then intrinsic yield growth coefficient, gy tni , in the yield used to alter gy tni . Rainfed crops react to changes in 6 THE COSTS OF AGRICu LT uRAL ADAPTATION TO CLIMATE CH A N G E precipitation and temperature as modeled in DSSAT. artifacts, we place a 20 percent cap on yield increases For irrigated crops, the effect of temperature is derived over the no-climate-change amount at the pixel level. from the DSSAT results and water stress effects are captured in the hydrology model built into IMPACT, Harvested areas in the IMPACT model are also increasing the value of WAT in equation (1). affected by climate change. In any particular FPU, land may become more or less suitable for a crop and will One of the more significant challenges for this research impact the intrinsic area growth rate gaini in the area is spatial aggregation. FPUs are large areas. For example, growth calculation. Water availability from the hydrol- the Ganges FPU in India is the entire length of the ogy model will affect the WAT variable for irrigated Ganges River in India. Within an FPU, there can be crops as it does with yields. large variation in climate and agronomic characteristics. A major challenge was to come up with an aggregation AC tni = tni × ( PStni ) iin × ( PStnj ) ijn × (1 + gatni ) - (2) j i scheme to take outputs from the crop modeling process ACtni (WATtni ) to the IMPACT FPUs. The process used starts with the SPAM data set, with a spatial resolution of 5 arc- minutes (approximately 10 km at the equator) that Crop calendar changes due to climate change cause two corresponds to the crop/management combination. The distinct issues. When the crop calendar in an area physical area in the SPAM data set is then used as the changes so that a crop that was grown in 2000 can no weight to find the area-weighted average yield across longer be grown in 2050, we implement an adjustment each FPU. This is done for each climate scenario to gatni that will bring the harvested area to zero by including the no-climate-change baseline. The ratio of 2050. However, when it becomes possible to grow a the area-weighted average yield in 2050 to the crop in 2050 where it could not be grown in 2000, we no-climate-change yield is used to adjust the yield do not add this new area. For example, the growing growth rate in equation (1) to reflect the effects of season in parts of Ontario, Canada is too short for climate change. maize in 2000 but adequate in 2050. Because we do not include this added area in 2050 our estimates of future In some cases the simulated changes in yields from production are biased downward somewhat. The effect climate change are large and positive. This usually has is likely to be small, however, as new areas have other one of two causes; starting from a low base (which can constraints on crop productivity, in particular soil be common in marginal production areas) and unrealis- characteristics. tically large effects of CO2 fertilization. To avoid these 7 3. mOdelINg ReSulTS the field are approximately 50 percent less than in experiments in enclosed containers. And another report ( Jorge A. Zavala et al., 2008) finds that higher levels of The results of our analysis are reported in three parts-- atmospheric CO2 increase the susceptibility of soybean the biological effects of climate change on crop yields, plants to the Japanese beetle and maize to the western the resulting impacts on economic outcomes including corn rootworm. So the actual, field benefits of CO2 prices, production, consumption, trade, calorie availabil- fertilization remain uncertain. ity and child malnutrition, and finally the costs of adap- tation to climate change. DSSAT has an option to include CO2 fertilization effects at different levels of atmospheric concentration. 3.1 The eFF e CTS OF ClImaTe ChaN g e To capture the uncertainty in actual field effects, we ON yI eld S simulate two levels of atmospheric CO2 in 2050--369 ppm (the level in 2000) and 532 ppm, the levels in 2050 Climate change alters temperature and precipitation used in the A2 scenario GCM runs. The results with patterns as shown in Figure 2 to Figure 5. These 369 ppm are called the no-CO2 fertilization; with 532, changes have both a direct effect on crop production the results are called with-CO2 fertilization. For most and indirect effects through changes in irrigation water tables, we report only no-CO2 fertilization results under availability and evapotranspiration potential. In this the assumption that this is the most likely outcome in section, we report on the direct effects on rainfed yields farmers' fields. of changing temperature and precipitation, irrigation yields through temperature effects alone, and the indi- D 3.1.1 irectclimatechangeeffectsonrainfedand rect effects of water availability through irrigation- irrigatedyields related changes in water availability. Table 1 reports the statistics and Figure 7 presents A particular challenge is how to include the effects of figures that show the direct biological effects of the two CO2 fertilization. Plants produce more vegetative climate change scenarios on yields with and without matter as atmospheric concentrations of CO2 increase. CO2 fertilization on the five crops we model with The effect depends on the nature of the photosynthetic DSSAT. Rainfed crops are modeled with both water process used by the plant species. So-called C3 plants and temperature stress effects. For irrigated crops, only use CO2 less efficiently than C4 plants so C3 plants are temperature stress is included here. Water scarcity more sensitive to higher concentrations of CO2. It effects on irrigation are dealt with in a later section. remains an open question whether these laboratory results translate to actual field conditions. A recent For most crops, yield declines predominate when no report on field experiments on CO2 fertilization CO2 fertilization is allowed. Irrigated and rainfed wheat (Stephen P. Long et al., 2006), finds that the effects in and irrigated rice are especially hard hit. The East Asia 8 THE COSTS OF AGRICu LT uRAL ADAPTATION TO CLIMATE CH A N G E and the Pacific region combines both China, which is fertilization effect allowed, yields decline less and in temperate for the most part, and Southeast Asia, which many locations some yield increases occur relative to is tropical, so the differential effects of climate change 2000 climate. However, rainfed maize and irrigated and in these two climate zones are masked. In China, some rainfed wheat still see substantial areas of reduced crops fare reasonably well, because higher future yields. Sub-Saharan Africa sees mixed results with small temperatures are favorable in locations where current declines or increases in maize yields and large negative temperatures are at the low end of the crop's optimal effects on rainfed wheat. The Latin America and temperature. India and other parts of South Asia are Caribbean region has mixed yield effects, with some particularly hard hit by climate change. With the CO2 crops up slightly and some down. Table 1. yIeld ChaNgeS by CROP aNd maNagemeNT SySTem uNdeR CuRReNT ClImaTe aNd TwO ClImaTe ChaNge SCeNaRIOS wITh aNd wIThOuT CO 2 FeRTIlIzaTION eFFeCTS (% ChaNge FROm yIeldS wITh 2000 ClImaTe) Region CSIRONoCF NCARNoCF CSIROCF NCARCF Maize, irrigated East Asia and the Pacific ­1.3 ­2.6 ­0.8 ­1.9 Europe and Central Asia 0.0 ­1.3 0.1 ­1.2 Latin America and the Caribbean ­2.8 ­3.0 ­2.3 ­2.5 Middle East and North Africa 0.1 ­1.0 ­0.4 ­1.1 South Asia ­6.4 ­5.5 ­4.4 ­3.6 Sub­Saharan Africa 0.3 0.6 0.5 0.8 Developing Countries ­2.0 ­2.8 ­1.4 ­2.1 Developed Countries ­1.2 ­8.7 ­1.2 ­8.6 World ­0.8 ­5.6 ­0.6 ­5.2 Maize, rainfed East Asia and the Pacific 1.5 ­3.9 3.7 ­2.0 Europe and Central Asia 25.0 3.7 32.8 12.4 Latin America and the Caribbean ­0.4 ­1.9 2.2 0.4 Middle East and North Africa 58.6 ­46.7 61.8 ­46.3 South Asia ­2.9 ­7.8 0.2 ­4.9 Sub­Saharan Africa ­2.4 ­4.6 ­0.8 ­2.7 Developing Countries 0.2 ­2.9 2.6 ­0.8 Developed Countries 0.6 ­5.7 9.5 2.5 World 1.0 ­3.4 5.3 0.5 rice, irrigated East Asia and the Pacific ­13.0 ­19.8 4.4 ­1.1 Europe and Central Asia ­4.1 ­15.1 15.0 5.7 Latin America and the Caribbean ­6.4 ­0.8 ­1.2 7.0 Middle East and North Africa ­13.3 ­29.5 1.7 ­14.4 South Asia ­15.5 ­17.5 2.5 1.4 Sub­Saharan Africa ­11.4 ­14.1 5.7 2.4 Developing Countries ­14.4 ­18.5 2.4 ­0.5 (Continuedonnextpage) D E V E L O P M E N T A N D C L I M AT E C H A N G E D I S C u S S I O N PA P E R S 9 Table 1. (continued) Region CSIRONoCF NCARNoCF CSIROCF NCARCF Developed Countries ­3.5 ­5.5 10.5 9.0 World ­13.8 ­17.8 2.8 ­0.0 rice, rainfed East Asia and the Pacific ­4.5 ­5.8 2.5 1.8 Europe and Central Asia 49.8 ­1.0 61.3 ­6.1 Latin America and the Caribbean 5.3 ­1.8 12.7 6.7 Middle East and North Africa 0 0 0 0.0 South Asia 0.1 2.6 8.5 10.2 Sub­Saharan Africa 0.1 ­0.5 8.1 7.3 Developing Countries ­1.3 ­1.4 6.5 6.4 Developed Countries 17.3 10.3 23.4 17.8 World ­1.3 ­1.4 6.5 6.4 soybean, irrigated East Asia and the Pacific ­8.2 ­13.4 9.1 3.6 Europe and Central Asia 31.9 30.1 32.9 30.5 Latin America and the Caribbean ­1.2 ­2.5 19.5 18.2 Middle East and North Africa ­4.2 ­14.0 5.6 ­5.0 South Asia ­9.5 ­11.5 12.0 10.3 Sub­Saharan Africa 4.6 5.0 17.8 17.8 Developing Countries ­8.0 ­12.3 10.3 5.8 Developed Countries 2.5 ­2.7 15.0 9.0 World ­0.4 ­5.4 13.7 8.0 soybean, rainfed East Asia and the Pacific ­3.6 ­8.6 17.0 11.5 Europe and Central Asia 25.5 5.9 37.0 5.9 Latin America and the Caribbean ­2.6 4.2 19.1 19.1 Middle East and North Africa 17.5 ­84.2 26.0 ­76.4 South Asia ­13.8 ­13.6 4.4 7.9 Sub­Saharan Africa ­3.5 ­5.8 19.1 17.8 Developing Countries ­2.3 1.7 19.5 18.0 Developed Countries 14.1 6.6 19.5 15.1 World 1.1 2.3 18.0 16.3 Wheat, irrigated East Asia and the Pacific ­2.7 ­7.1 3.7 ­0.6 Europe and Central Asia ­9.4 ­19.8 ­3.3 ­14.7 Latin America and the Caribbean 0.3 ­5.6 6.5 0.9 Middle East and North Africa ­12.8 ­19.7 ­5.8 ­13.4 South Asia ­47.1 ­53.9 ­38.3 ­45.8 Sub­Saharan Africa 0.7 1.4 7.3 9.7 Developing Countries ­28.3 ­34.3 ­20.8 ­27.2 Developed Countries ­5.7 ­4.9 ­1.3 ­0.1 World ­25.6 ­31.1 ­18.5 ­24.4 (Continuedonnextpage) 10 THE COSTS OF AGRICu LT uRAL ADAPTATION TO CLIMATE CH A N G E Table 1. (continued) Region CSIRONoCF NCARNoCF CSIROCF NCARCF Wheat, rainfed East Asia and the Pacific ­14.8 ­16.1 ­5.4 ­9.2 Europe and Central Asia ­0.3 ­1.8 8.5 8.0 Latin America and the Caribbean 2.3 4.2 12.2 11.8 Middle East and North Africa ­2.6 ­8.1 8.8 2.0 South Asia ­44.4 ­43.7 ­28.9 ­28.0 Sub­Saharan Africa ­19.3 ­21.9 ­11.2 ­15.9 Developing Countries ­1.4 ­1.1 9.3 8.5 Developed Countries 3.1 2.4 9.7 9.5 World 1.0 0.8 9.7 9.1 groundnut, irrigated East Asia and the Pacific ­11.1 ­13.7 3.6 1.2 Europe and Central Asia ­34.4 ­50.3 ­22.6 ­41.5 Latin America and the Caribbean 0.0 0.0 0.0 0.0 Middle East and North Africa ­11.6 ­28.5 4.3 ­15.6 South Asia ­6.7 ­10.6 9.4 5.0 Sub­Saharan Africa ­11.5 ­11.3 3.9 4.2 Developing Countries ­10.0 ­13.1 5.2 2.0 Developed Countries ­4.6 ­10.7 12.1 5.0 World ­9.2 ­12.7 6.2 2.5 groundnut, rainfed East Asia and the Pacific ­5.1 ­6.5 11.3 9.7 Europe and Central Asia 0.0 0.0 0.0 0.0 Latin America and the Caribbean 0.9 7.1 18.1 17.9 Middle East and North Africa ­20.5 23.6 ­11.8 23.6 South Asia ­8.1 ­8.9 9.1 6.7 Sub­Saharan Africa ­4.1 ­8.6 14.2 8.8 Developing Countries ­4.7 ­7.9 12.9 8.6 Developed Countries ­18.3 ­5.0 2.7 11.6 World ­4.9 ­7.9 12.7 8.7 Source: Authors' estimates. The results in this table are derived by growing a crop in DSSAT at 0.5 degree intervals around the world. At each location, the yield is calculated with 2000 climate, existing soil conditions, and rates of nitrogen application assumed relevant for that country. Then 2050 climate data replace the 2000 climate data and the crop is grown again. The values reported in this table are the area-weighted averages of these two figures. The first two columns report results without CO2 fertilization; the last two columns with CO2 fertilization. I 3.1.2 ndirecteffectsfromclimatechange:Water though water availability for irrigated crops. In addition, stressforirrigatedcrops higher temperatures under climate change will for the most part increase evapotranspiration requirements of Climate change will have a direct impact on regional crops. The impacts of climate change on effective rain- hydrology and therefore affect agricultural production fall, potential and actual evapotranspiration and runoff 11 (Continuedonnextpage) NCAR CF NCAR CF NCAR CF FIguRe 7. yIeld ChaNgeS by CROP aNd maNagemeNT SySTem uNdeR CuRReNT ClImaTe Sub-Saharan Sub-Saharan Sub-Saharan aNd TwO ClImaTe ChaNge SCeNaRIOS wITh aNd wIThOuT CO 2 FeRTIlIzaTION eFFeCTS Africa Africa Africa South Asia South Asia South Asia NCAR NoCF NCAR NoCF NCAR NoCF Rainfed Soybean Middle East and Middle East and Middle East and Rainfed Maize Rainfed Rice North Africa North Africa North Africa CSIRO CF CSIRO CF CSIRO CF Latin America and Latin America and Latin America and the Caribbean the Caribbean the Caribbean Europe and Europe and Europe and Central Asia Central Asia Central Asia CSIRO NoCF CSIRO NoCF CSIRO NoCF East Asia and East Asia and East Asia and the Pacific the Pacific the Pacific D E V E L O P M E N T A N D C L I M AT E C H A N G E D I S C u S S I O N PA P E R S 30.0 20.0 10.0 0.0 ­10.0 ­20.0 ­30.0 30.0 20.0 10.0 0.0 ­10.0 ­20.0 ­30.0 30.0 20.0 10.0 0.0 ­10.0 ­20.0 ­30.0 (% ChaNge FROm yIeldS wITh 2000 ClImaTe) NCAR CF NCAR CF NCAR CF Sub-Saharan Sub-Saharan Sub-Saharan Africa Africa Africa South Asia South Asia South Asia NCAR NoCF NCAR NoCF NCAR NoCF Irrigated Soybean Middle East and Middle East and Middle East and Irrigated Maize Irrigated Rice North Africa North Africa North Africa Latin America and Latin America and Latin America and CSIRO CF CSIRO CF CSIRO CF the Caribbean the Caribbean the Caribbean Europe and Europe and Europe and Central Asia Central Asia Central Asia CSIRO NoCF CSIRO NoCF CSIRO NoCF East Asia and East Asia and East Asia and the Pacific the Pacific the Pacific 30.0 20.0 10.0 0.0 ­10.0 ­20.0 ­30.0 30.0 20.0 10.0 0.0 ­10.0 ­20.0 ­30.0 30.0 20.0 10.0 0.0 ­10.0 ­20.0 ­30.0 12 THE COSTS OF AGRICu LT uRAL ADAPTATION TO CLIMATE CH A N G E FIguRe 7. (continued) Irrigated Wheat Rainfed Wheat 30.0 30.0 20.0 20.0 10.0 10.0 0.0 0.0 ­10.0 ­10.0 ­20.0 ­20.0 ­30.0 ­30.0 the Caribbean North Africa the Pacific East Asia and Central Asia Europe and Latin America and Middle East and South Asia Africa Sub-Saharan North Africa the Pacific East Asia and Central Asia Europe and the Caribbean Latin America and Middle East and South Asia Africa Sub-Saharan CSIRO NoCF CSIRO CF NCAR NoCF NCAR CF CSIRO NoCF CSIRO CF NCAR NoCF NCAR CF Irrigated Groundnut Rainfed Groundnut 30.0 30.0 20.0 20.0 10.0 10.0 0.0 0.0 ­10.0 ­10.0 ­20.0 ­20.0 ­30.0 ­30.0 North Africa South Asia Africa Sub-Saharan North Africa the Pacific East Asia and Central Asia Europe and the Caribbean Latin America and Middle East and the Pacific East Asia and Central Asia Europe and the Caribbean Latin America and Middle East and South Asia Africa Sub-Saharan CSIRO NoCF CSIRO CF NCAR NoCF NCAR CF CSIRO NoCF CSIRO CF NCAR NoCF NCAR CF Source: Authors' data. (or internal renewable water) were analyzed for the two CSIRO, IRW increase is less than with NCAR; the climate change scenarios using the global hydrological Middle East & North Africa and Sub-Saharan Africa model linked with IMPACT. regions both see reductions of about 4 percent. Table 3 shows water availability and use results. For Irrigation water requirement is the amount of water each region we report internal renewable water (IRW), needed to grow irrigated crops without water stress. irrigation water requirements, actual consumption and Table 2 reports irrigation water requirements in 2000 the ratio of consumption to requirements, called the and 2050 under current climate, and the percent irrigation water supply reliability index (IWSR). changes in 2050 under the two climate change scenarios relative to 2050 requirements under current climate. IRW is the water (surface runoff plus net groundwater Changes in irrigation water requirements from 2000 to recharge) available from precipitation falling on a study 2050 reflect the increased demand for food, changes in area such as a river basin or a country. In the NCAR irrigated area, and changes in irrigation water use effi- results, all regions see increased IRW in 2050. With ciency. Changes in 2050 irrigation water requirements D E V E L O P M E N T A N D C L I M AT E C H A N G E D I S C u S S I O N PA P E R S 13 Table 2. waTeR avaIlabIlITy aNd uSe uNdeR CuRReNT ClImaTe, aNd PeRCeNT ChaNgeS uNdeR TwO ClImaTe ChaNge SCeNaRIOS IN 205 Water Water Wateravailability Wateravailability Wateravailability availabilityand availabilityand andusechange, andusechange, andusechange, usein2000, usein2050,no 2000­2050,no 2000­2050,NCAR 2000­2050,CSIRO currentclimate climatechange climatechange relativetono relativetonoclimate (km3/yr) (km3/yr) (%) climatechange(%) change(%) east asia & Pacific Internal renewable water 9,248.0 9,248.0 0 8.2 4.3 Irrigation water requirement 345.2 277.4 ­19.7 5.3 1.2 Irrigation water consumption 238.6 219.0 ­8.2 5.1 1.5 IWSR (%) 69.1 78.9 15.1 ­12.7 europe & Central asia Internal renewable water 4,916.0 4,916.0 0 18.0 8.8 Irrigation water requirement 77.9 70.1 ­9.9 ­11.0 0.6 Irrigation water consumption 72.6 65.7 ­9.6 ­6.7 ­3.8 IWSR (%) 93.3 93.6 ­4.1 ­5.7 latin america & Caribbean Internal renewable water 13,232.0 13,232.0 0 10.7 0.6 Irrigation water requirement 103.0 108.9 5.8 ­13.3 8.9 Irrigation water consumption 96.1 103.4 7.6 ­0.3 ­4.9 IWSR (%) 93.3 94.9 1.8 0.3 Middle East & North Africa Internal renewable water 179.0 179.0 0 11.5 ­3.6 Irrigation water requirement 89.3 101.3 13.4 2.8 ­6.5 Irrigation water consumption 85.9 97.4 13.4 ­1.4 ­11.8 IWSR (%) 96.2 96.1 0.0 1.2 south asia Internal renewable water 1,788.0 1,788.0 0 14.0 2.0 Irrigation water requirement 489.1 515.3 5.4 ­2.6 0.9 Irrigation water consumption 367.1 386.5 5.3 ­0.9 1.3 IWSR (%) 75.1 75.0 2.8 ­1.2 sub­saharan africa Internal renewable water 3,762.0 3,762.0 0 6.5 ­3.9 Irrigation water requirement 38.3 51.0 33.2 ­8.5 ­9.7 Irrigation water consumption 37.9 50.3 32.6 ­8.5 ­8.5 IWSR (%) 99.0 98.5 ­0.5 ­0.4 developed Internal renewable water 7,479.0 7,479.0 0 10.9 7.3 Irrigation water requirement 103.8 107.6 3.8 5.3 1.2 Irrigation water consumption 102.4 106.0 3.5 5.1 1.5 IWSR (%) 98.7 98.5 ­0.2 0.3 developing Internal renewable water 33,101.0 33,101.0 0 10.8 2.4 Irrigation water requirement 1,142.8 1,124.1 ­1.6 ­11.0 0.6 Irrigation water consumption 898.3 922.1 2.7 ­6.7 ­3.8 IWSR (%) 78.6 82.0 4.9 ­4.4 Source: Authors' estimates. Note: The values in the last two columns are the percent change in the row variable relative to an outcome with no climate change. For example, in the East Asia and Pacific region, IWSR is 78.9 percent in 2050 without climate change. With the NCAR scenario, the ISWR increases to 90.8 percent, an increase of 15.1 percent. 14 THE COSTS OF AGRICu LT uRAL ADAPTATION TO CLIMATE CH A N G E under climate change scenarios are due to changes in CSIRO effects are especially large. Latin America and crop evapotranspiration potential from higher tempera- the Caribbean yields are relatively unaffected, although tures, changes in effective rainfall, and changes in crop this is in part due to the small amount of irrigated irrigated harvested areas as a result of supply effects production in that region. from changes in agricultural commodity prices. Irrigation water consumption is the water actually used Table 3. yIeld ChaNgeS FOR IRRIgaTed by irrigated crops. The consumption value is always CROPS due TO waTeR STReSS uNdeR CuR- smaller than the requirements value because it is impos- ReNT ClImaTe aNd TwO ClImaTe ChaNge sible in practice to deliver precisely the correct amount SCeNaRIOS (% ChaNge FROm 2000 yIeldS) of water. The ratio of consumption to requirements is called irrigation water supply reliability (IWSR). The 2050 smaller the ratio, the greater the water stress on irri- Noclimate gated crop yields. Region change NCAR CSIRO Rice Across the group of developing countries, IWSR East Asia & Pacific ­4.8 ­1.2 ­6.7 improves under the NCAR GCM and worsens under Europe & Central Asia ­1.9 ­3.2 ­3.3 the CSIRO GCM. However regional effects of climate Latin America & ­0.1 ­0.1 ­0.1 change vary. Reliability improves slightly for Latin Caribbean America and the Caribbean and for the Middle East Middle East & North ­8.3 ­3.3 ­3.2 Africa and North Africa. For Sub-Saharan Africa reliability South Asia ­8.9 ­6.3 ­8.1 worsens slightly under both scenarios. In East Asia and Sub-Saharan Africa ­0.3 ­0.4 ­0.3 the Pacific and South Asia, reliability increases under the Developed 0.0 0.0 0.0 NCAR scenario but declines under the CSIRO scenario. Developing ­6.3 ­3.5 ­7.0 Wheat Yield reductions of irrigated crops due to water stress East Asia & Pacific ­21.9 ­3.1 ­32.6 are directly estimated in IMPACT using empirical rela- Europe & Central Asia ­0.9 ­1.1 ­0.5 tionships developed by FAO (Doorenbos and Kassam, Latin America & ­0.2 ­2.1 ­0.2 1979), taking into account the growing demand for Caribbean water outside agriculture as well as agricultural Middle East & North ­1.4 ­5.6 ­0.5 demands. The results are shown in Table 3. Both Africa NCAR and CSIRO scenarios result in more precipita- South Asia ­14.4 ­17.4 ­14.8 tion over land than with no climate change in most Sub-Saharan Africa ­0.3 ­0.8 ­0.6 parts of the world, but the CSIRO scenario has rela- Developed ­1.7 0.0 ­1.7 tively small increases. Combined with growing demand Developing ­11.6 ­4.1 ­15.3 for water outside of agriculture the consequence is often Maize substantial yield decline. For example, in East Asia and East Asia & Pacific ­9.0 ­8.7 ­19.9 the Pacific, with no climate change, the combined Europe & Central Asia ­0.8 ­0.4 ­0.8 effects of non-agricultural demand growth and Latin America & ­4.1 ­0.1 ­3.0 increased irrigated area result in an average 4.8 percent Caribbean decline in irrigated rice yields. With the NCAR Middle East & North ­7.2 ­1.5 ­5.5 Africa scenario, that decline is only 1.2 percent. However, with South Asia ­20.0 ­13.9 ­21.1 the drier CSIRO scenario the irrigated yield loss is 6.7 Sub-Saharan Africa ­9.0 ­8.7 ­19.9 percent. Irrigated rice, wheat, and maize yield losses are Developed ­0.1 ­1.4 0.0 all large with CSIRO for East Asia and the Pacific. Developing ­8.0 ­9.1 ­14.0 South Asia yields for all crops see large yield declines under both scenarios. In Sub-Saharan Africa, irrigated Source:Authors' estimates. maize yields decline under both scenarios but the D E V E L O P M E N T A N D C L I M AT E C H A N G E D I S C u S S I O N PA P E R S 15 Table 4. POPulaTION aNd INCOme gROwTh aSSumPTIONS Average Average annualgrowth 2000(constant 2050(constant annualgrowth 2000(million) 2050(million) rate(%) 2000US$) 2000US$) rate(%) Population Percapitaincome South Asia 1,361 2,306 1.05 462 3,490 4.04 East Asia and Pacific 1,825 2,218 0.39 906 10,344 4.87 Europe and Central Asia 488 456 ­0.14 2,600 17,269 3.79 Latin America and the 513 754 0.77 3,999 16,091 2.78 Caribbean Middle East and North 259 453 1.12 1,597 5,908 2.62 Africa Sub-Saharan Africa 666 1,732 1.91 563 1,247 1.59 Developed 948 1,163 0.41 28,629 79,951 2.05 Developing 5,136 7,961 0.88 1,333 7,362 3.42 World 6,084 9,124 0.81 5,588 16,612 2.18 Source:EACC Study estimates. 3.2 ClImaTe ChaNge ImPaCTS ON annum. Per capita income growth is assumed to be high- agRICulTuRe aNd humaN well-beINg est in East Asia and the Pacific at 4.87 percent per annum. Sub-Saharan Africa has the lowest growth per The direct and indirect effects of climate change on capita income growth at 1.59 percent per annum. agriculture play out through the economic system, alter- ing prices, production, productivity investments, food 3.2.1 Pricesandproduction demand, food consumption and ultimately human well-being. World prices are a useful single indicator of the effects of climate change on agriculture. Table 5 shows the For this study, a common set of income and population prices effects of various permutations of climate change, growth assumptions were used. Table 4 provides an over- with and without the CO2 fertilization effect. Figure 8 view of these assumptions. Population growth is assumed show world price effects for the major grains respec- to be highest in Sub-Saharan Africa at 1.91 percent per tively, assuming no CO2 fertilization effect. Table 5. wORld PRICeS OF SeleCTed CROPS aNd lIveSTOCk PROduCTS (uS$/meTRIC TON) 2050 Noclimatechange NCARNoCF CSIRONoCF NCARCFeffect CSIROCFeffect Agriculturalproducts 2000 US$/metricton(%increaseover2000) %changefrom2050NoCFresults Rice 190 307 (61.6) 421 (121.6) 406 (113.7) ­17.0 ­15.1 Wheat 113 158 (39.8) 334 (195.6) 307 (171.7) ­11.4 ­12.5 Maize 95 155 (63.2) 235 (147.4) 240 (152.6) ­11.2 ­12.6 Soybeans 206 354 (71.8) 394 (91.3) 404 (96.1) ­60.6 ­62.2 (Continuedonnextpage) 16 THE COSTS OF AGRICu LT uRAL ADAPTATION TO CLIMATE CH A N G E Table 5. (continued) 2050 Noclimatechange NCARNoCF CSIRONoCF NCARCFeffect CSIROCFeffect Agriculturalproducts 2000 US$/metricton(%increaseover2000) %changefrom2050NoCFresults Beef 1,925 2,556 (32.8) 3,078 (59.9) 3,073 (59.6) ­1.3 ­1.5 Pork 911 1,240 (36.1) 1,457 (59.9) 1,458 (60.0) ­1.3 ­1.5 Poultry 1,203 1,621 (34.7) 1,968 (63.6) 1,969 (63.7) ­1.9 ­2.1 Source:Authors' estimates. Notes: Prices are in 2000 uS$. Numbers in parantheses are percent increases over 2000. The last two columns in this table report the percentage difference between the price in 2050 with and without the CO2 fertilization effect. For example, with the NCAR scenar- io, assuming CO2 fertilization is effective in the field results in a 17.0 percent reduction in the 2050 world rice price relative to the level reached with no CO2 fertilization. The decline in prices of livestock products with CO2 fertilization reflects the reduced cost of feed. percent for maize and 11 to 14 percent for soybeans. If FIguRe 8. wORld PRICeS OF majOR CO2 fertilization is effective in farmers' fields, these gRaINS (2000 uS$) price increases are 11 percent to 17 percent smaller for 450 rice, wheat, and maize and over 60 percent smaller for 400 soybeans. 350 Livestock are not directly affected by climate change in Dollars per Metric Ton 300 the IMPACT model but the effects of higher feed 250 prices caused by climate change pass through to live- 200 stock, resulting in higher meat prices. For example, 150 beef prices are 33 percent higher by 2050 with no 100 climate change and 60 percent higher with climate 50 change and no CO2 fertilization of crops. With CO2 0 fertilization, crop-price increases are less so the beef price increase is about 1.5 percent less than with no Rice Wheat Maize Soybeans Other grains CO2 fertilization. 2000 2050 No climate change 2050 CSIRO NoCF 2050 NCAR NoCF Table 6 combines the biophysical effects of climate change on yields with the indirect effects from water Source:Authors' estimates. stress in irrigated crops and autonomous adjustments to yield due to price effects directly on yields and on productivity growth. With no climate change, world prices for the most important agricultural crops--rice, wheat, maize, and Table 7 reports crop production effects of climate soybeans will increase between 2000 and 2050, driven change, accounting for both the changes in yield shown by population and income growth and biofuels demand. in Table 6, and changes in crop area induced by climate Even with no climate change, the price of rice would change. For each crop the first row is 2000 production rise by 62 percent, maize by 63 percent, soybeans by 72 and the second is 2050 production with no climate percent and wheat by 39 percent. Climate change change. The third to fifth rows are the difference results in additional price increases--a total of 32 to 37 between the scenario with climate change production percent for rice, 94 to 111 percent for wheat, 52 to 55 and no-climate-change production in 2050. For D E V E L O P M E N T A N D C L I M AT E C H A N G E D I S C u S S I O N PA P E R S 17 Table 6. COmbINed bIOPhySICal aNd eCONOmIC yIeld eFFeCTS FROm ClImaTe ChaNge, NO CO 2 FeRTIlIzaTION Europe Latin Middle EastAsia and America Eastand Sub- South andthe Central andthe North Saharan Developed Developing Asia Pacific Asia Caribbean Africa Africa Countries Countries World rice 2000 (kg/ha) 2,068 3,054 2,077 2,438 4,076 1,089 4,437 2,549 2,606 2050 No CC 3,175 3,859 4,272 3,568 6,246 2,269 6,226 3,486 3,556 (kg/ha) NCAR (%) ­11.1 ­5.2 2.0 0.9 ­6.0 ­0.5 2.7 ­7.1 ­6.9 CSIRO (%) ­10.9 ­8.1 ­0.1 2.7 ­15.5 ­3.1 2.8 ­8.8 ­8.4 Wheat 2000 (kg/ha) 2,503 3,782 2,075 2,463 1,680 1,827 3,375 2,468 2,726 2050 No CC 5,559 5,476 4,186 3,941 3,753 3,353 5,329 4,596 4,778 (kg/ha) NCAR (%) ­44.9 10.1 ­6.5 2.3 3.9 ­26.5 1.3 ­14.1 ­9.3 CSIRO (%) ­48.5 10.9 ­15.1 4.8 ­2.4 ­30.5 ­2.8 ­18.0 ­13.1 Maize 2000 (kg/ha) 1,868 4,214 3,706 2,957 5,696 1,483 8,625 3,029 4,404 2050 No CC 2,464 7,292 6,676 4,927 7,268 2,206 12,799 5,124 7,170 (kg/ha) NCAR (%) 4.4 7.9 18.8 5.0 4.1 ­1.7 5.4 4.7 9.8 CSIRO (%) 1.1 10.2 13.4 2.4 ­1.5 0.1 ­1.9 6.7 5.1 Millet 2000 (kg/ha) 800 1,528 844 1,512 1,017 655 1,436 753 759 2050 No CC 1,689 3,009 2,368 3,585 1,812 1,772 2,142 1,811 1,814 (kg/ha) NCAR (%) ­0.7 10.9 3.5 7.5 ­0.4 8.0 ­0.1 6.8 6.8 CSIRO (%) ­2.7 8.7 4.1 6.6 1.6 6.6 ­1.4 5.1 5.0 sorghum 2000 (kg/ha) 799 3,089 1,237 2,891 4,978 843 3,596 1,124 1,395 2050 No CC 1,438 5,665 4,706 5,440 5,708 1,663 5,142 2,101 2,335 (kg/ha) NCAR (%) 1.5 9.3 10.7 2.4 0.8 8.6 2.5 8.7 8.2 CSIRO (%) ­0.9 6.9 6.6 2.9 ­0.4 5.8 ­0.9 6.6 5.4 Source: Authors' estimates. Note:The rows labeled "NCAR (% change)" and "CSIRO (% change)" indicate the percent change in yield due to climate change in 2050 relative to yields in 2050 without climate change. For example, South Asia rice yields were 2,068 kg/ha in 2000. With no cli- mate change, South Asia rice yields are predicted to increase to 3,054 kg/ha in 2050. With the CSIRO scenario, South Asia rice yields predictions are 10.9 percent lower than with no climate change in 2050. example Sub-Saharan agriculture maize production The negative effects of climate change are especially would increase by 45 percent with no climate change pronounced in Sub-Saharan Africa and South Asia; all (from 37.1 mmt to 53.9 mmt). Relative to no climate of the major crops have production declines (relative to change, the 2050 CSIRO climate results in a 9.6 the no climate change scenario) under the two GCMs. percent decline in production. For East Asia and the Pacific, the results are mixed, and 18 THE COSTS OF AGRICu LT uRAL ADAPTATION TO CLIMATE CH A N G E depend on both crop and GCM. Rice production higher biological yields. For other major producing effects are uniformly negative, while wheat and maize areas, lower precipitation and higher temperatures mean are mixed. Comparing the all- developed-country aver- lower yields. With increasing incomes and population, age to the all-developing country average, developing growth in demand for meat means higher consumption countries fare worse for almost all crops under both of maize for feed. The resulting higher maize prices scenarios. One striking result is that maize production induce further yield growth and area expansion in the in developed countries increases substantially with US. Note that this result is driven by the choice of climate change. This result is due entirely to increases in GCM. Other GCMs report lower precipitation in the the U.S. Both GCMs show substantial precipitation US Midwest and would result in dramatically different increases in the U.S. Midwest allowing substantially scenario outcomes. Table 7. ClImaTe ChaNge eFFeCTS ON CROP PROduCTION, NO CO 2 FeRTIlIzaTION East Europe Latin Middle Asiaand and America Eastand Sub- South the Central andthe North Saharan Developed Developing World Asia Pacific Asia Caribbean Africa Africa Countries Countries Rice 2000 (mmt) 119.8 221.7 1.1 14.9 5.5 7.5 20.4 370.3 390.7 2050 No 168.9 217.0 2.6 17.8 10.3 18.3 20.3 434.9 455.2 CC (mmt) 2050 No 41.0 ­2.1 143.0 19.9 88.0 145.6 ­0.2 17.4 16.5 CC (%) CSIRO (% ­14.3 ­8.1 ­0.2 ­21.7 ­32.9 ­14.5 ­11.8 ­11.9 ­11.9 change) NCAR (% ­14.5 ­11.3 ­0.8 ­19.2 ­39.7 ­15.2 ­10.6 ­13.6 ­13.5 change) Wheat 2000 (mmt) 96.7 102.1 127.5 23.5 23.6 4.5 205.2 377.9 583.1 2050 No 191.3 104.3 252.6 42.1 62.0 11.4 253.7 663.6 917.4 CC (mmt) 2050 No 97.8 2.2 98.1 79.1 162.7 153.3 23.6 75.6 57.3 CC (%) CSIRO (% ­43.7 1.8 ­43.4 11.4 ­5.1 ­33.5 ­7.6 ­29.2 ­23.2 change) NCAR (% ­48.8 1.8 ­51.0 17.4 ­8.7 ­35.8 ­11.2 ­33.5 ­27.4 change) Maize 2000 (mmt) 16.2 141.9 38.0 80.1 8.2 37.1 297.9 321.3 619.2 2050 No 18.7 264.7 62.7 143.1 13.1 53.9 505.1 556.2 1,061.3 CC (mmt) 2050 No 15.4 86.5 65.0 78.7 59.8 45.3 69.6 73.1 71.4 CC (%) CSIRO (% ­18.5 ­12.7 ­19.0 ­0.3 ­6.8 ­9.6 11.5 ­10.0 0.2 change) NCAR (% ­8.9 8.9 ­38.3 ­4.0 ­9.8 ­7.1 1.8 ­2.3 ­0.4 change) Millet 2000 (mmt) 10.6 2.3 1.2 0.0 0.0 13.1 0.5 27.3 27.8 (Continuedonnextpage) D E V E L O P M E N T A N D C L I M AT E C H A N G E D I S C u S S I O N PA P E R S 19 Table 7. (continued) East Europe Latin Middle Asiaand and America Eastand Sub- South the Central andthe North Saharan Developed Developing World Asia Pacific Asia Caribbean Africa Africa Countries Countries 2050 No 12.3 3.5 2.14 0.1 0.1 48.1 0.8 66.2 67.0 CC (mmt) 2050 No 16.0 52.2 78.3 267.2 60.0 142.5 141.0 CC (%) CSIRO (% ­19.0 4.2 ­4.3 8.8 ­5.5 ­6.9 ­3.0 ­8.5 ­8.4 change) NCAR (% ­9.5 8.3 ­5.2 7.2 ­2.7 ­7.6 ­5.6 ­7.0 ­7.0 change) Sorghum 2000 (mmt) 8.4 3.1 0.1 11.4 1.0 19.0 16.9 43.0 59.9 2050 No 9.6 3.4 0.4 28.0 1.1 60.1 20.9 102.6 123.5 CC (mmt) 2050 No 14.3 9.7 300.0 145.6 10.0 216.3 23.7 138.6 106.2 CC (%) CSIRO (% ­19.6 1.4 ­2.7 2.3 0.3 ­2.3 ­3.1 ­2.5 ­2.6 change) NCAR (% ­12.2 6.7 ­10.4 4.3 0.7 ­3.0 ­7.3 ­1.5 ­2.5 change) Source:Authors' estimates. Note:The rows labeled "2050 No CC (% change)" indicate the percent change between production in 2000 and 2050 with no cli- mate change. The rows labeled "CSIRO (% change)" and "NCAR (% change)" indicate the additional percent change in production in 2050 due to climate change relative to 2050 with no climate change. For example, South Asia sorghum production was 8.4 mmt in 2000. With no climate change, South Asia sorghum production is predicted to increase to 9.6 mmt in 2050, an increase of 13.9 per- cent. With the CSIRO scenario, South Asia sorghum production in 2050 is 19.6 percent lower than with no climate change in 2050 (7.72 mmt instead of 9.6 mmt); mmt = million metric tons. T 3.2.2 radeinagriculturalcommodities a small net exporter in 2000 and becomes a net importer of cereals in 2050 with no climate change. As with the earlier studies, our simulations result in Both climate change scenarios result in substantial trade flow adjustments with climate change. Table 8 and increases in South Asian net imports relative to no Figure 9 report net cereal flows. With no climate climate change. The East Asia and Pacific region is a change, developed-country net exports increase from net importing region in 2000 and imports grow 83.4 mmt to 105.8 mmt between 2000 and 2050, an substantially with no climate change. Depending on increase of 27 percent. Developing-country net imports climate change scenario, this region either has slightly mirror this change. With the NCAR results and no less net imports than with the no-climate-change CO2 fertilization, developed-country net exports scenario or becomes a net exporter. In Latin America increase slightly (0.9 mmt) over no climate change. and the Caribbean, the 2050 no-climate-change With the drier CSIRO scenario, on the other hand, scenario is increased imports relative to 2000 but the 3 developed-country net exports increase by 39.9 mmt. CSIRO and NCAR climate scenarios result in smaller net imports in 2050 than in 2000. Regional results show important differences in the effects of climate change on trade and the differential 3 The results with CO2 fertilization increase developed-country exports effects of the three scenarios. For example, South Asia is by an additional 12 to 18 percent relative to no climate change. 20 THE COSTS OF AGRICu LT uRAL ADAPTATION TO CLIMATE CH A N G E The effects of climate change on trade flow values are For example, without climate change, 2050 developed even more dramatic than on production because of country production of maize increases by 207.2 mmt climate change effects on prices. Without climate (an increase of 70 percent); in developing countries, change, the value of developing country net imports of maize production increases by 234.9 mmt (73 percent). cereals in 2050 is 214 percent greater than in 2000. With both CSIRO and NCAR scenarios, developed With the wetter NCAR scenario, 2050 net imports country production increases more than developing value is 262 percent greater than in 2000; with the drier country production, but the magnitudes of these CSIRO scenario it is 361 percent greater. changes are much greater with CSIRO than with NCAR. The result is much greater net exports of maize The climate scenario differences in trade flows are (and other major rainfed crops) from developed coun- driven by geographical differences in production effects. tries with CSIRO than with NCAR. Similar differences Table 8. NeT CeReal (RICe, wheaT, maIze, mIlleT, SORghum, aNd OTheR gRaINS) exPORTS by RegION IN 2000 aNd 2050 uNdeR SCeNaRIOS wITh aNd wIThOuT ClImaTe ChaNge (000 mT 2050 NoClimate CSIRONo CSIROCF NCARCF Region 2000 Change CF NCARNoCF effects(%) effects(%) South Asia 15,013 ­19,791 ­53,823 ­51,663 ­15.0 ­8.1 East Asia and the Pacific ­19,734 ­72,530 ­55,086 8,158 9.1 ­58.5 Europe and Central Asia 8,691 178,097 64,916 34,760 4.4 6.5 Latin America and the Caribbean ­11,358 ­38,063 ­3,114 ­2,848 251.7 239.5 Middle East and North Africa ­51,753 ­84,592 ­66,708 ­64,459 ­0.0 0.6 Sub-Saharan Africa ­22,573 ­65,122 ­29,236 ­28,011 53.1 49.5 Developed Countries 83,352 105,809 145,740 106,672 12.1 18.4 Developing Countries ­83,352 ­105,809 ­145,740 ­106,672 12.1 18.4 Source:Authors' estimates. Note:The last two columns in this table report the percentage difference between the net imports in 2050 with climate change and with the CO2 fertilization effect. For example, Sub-Saharan countries import 28.0 mmt under the NCAR climate scenario and no CO2 fertilization effects. CO2 fertilization adds 49.5 percent. Table 9. value OF NeT CeReal TRade by RegION (mIllION uS$) 2000 2050NoClimateChange 2050CSIRONoCF 2050NCARNoCF South Asia 2,589 ­2,238 ­14,927 ­14,727 East Asia and the Pacific ­1,795 ­7,980 ­8,879 6,530 Europe and Central Asia 750 24,276 14,377 6,662 Latin America and the Caribbean ­1,246 ­6,027 ­342 480 Middle East and North Africa ­5,600 ­12,654 ­17,723 ­17,703 Sub-Saharan Africa ­2,995 ­12,870 ­10,914 ­11,153 Developed Countries 8,500 18,184 39,219 30,733 Developing Countries ­8,500 ­18,184 ­39,219 ­30,733 Source:Authors' estimates. D E V E L O P M E N T A N D C L I M AT E C H A N G E D I S C u S S I O N PA P E R S 21 FIguRe 9. NeT CeReal (RICe, wheaT, maIze, mIlleT, SORghum, aNd OTheR gRaINS) TRade by RegION IN yeaR 2000 aNd 2050 uNdeR SCeNaRIOS wITh aNd wIThOuT ClImaTe ChaNge (mmT) 200 150 100 Millions of Metric Tons 50 0 ­50 ­100 ­150 South Asia East Asia and Europe and Latin America and Middle East Sub-Saharan Africa Developed Developing the Pacific Central Asia the Caribbean and North Africa Countries Countries 2000 2050 No Climate Change 2050 CSIRO NoCF 2050 NCAR NoCF Source: Authors' estimates. exist for wheat, where the climate change effects on change declines about 10 percent in developing coun- yield are much more dramatic in developing countries tries and 9 percent in developed countries. Cereal than in developed countries. consumption decrease from climate change is 25 percent in developed countries and 21 percent in devel- 3.2.3 Fooddemand oping countries. The level of food available for consumption is deter- 3.2.4 Welfareeffects mined by the interaction of supply, demand, and the resulting prices with individual preferences and income. Our measures of the welfare effects of climate change Table 10 shows average per capita consumption of cere- are the change in calorie availability and in the number als and meat products in 2000 and in 2050 under vari- of malnourished children brought about by climate ous climate change scenarios. In the developing country change. group per capita cereal consumption declines and per capita meat consumption increases between 2000 and The declining consumption for cereals in particular 2050 with no climate change. Climate change reduces translates into similarly large declines in calorie avail- meat consumption growth slightly and causes a more ability as the result of climate change. Results are substantial fall in cereals consumption. These results are presented in Table 11 and Figure 10. Without climate the first evidence of the negative welfare effects of change, calorie availability increases throughout the climate change. Both climate change scenarios have world between 2000 and 2050. The largest increase, of similar consequences. Meat consumption with climate 13.8 percent, is in East Asia and the Pacific, but the 22 THE COSTS OF AGRICu LT uRAL ADAPTATION TO CLIMATE CH A N G E Table 10. PeR CaPITa FOOd CONSumPTION (kg PeR yeaR) OF CeRealS aNd meaTS wITh aNd wIThOuT ClImaTe ChaNge 2050 NoClimate CSIRONo NCARNo CSIROCFeffect NCARCFeffect 2000 Change CF CF (%) (%) Meat South Asia 6 16 14 14 0.9 0.8 East Asia and the Pacific 40 71 66 66 0.7 0.6 Europe and Central Asia 42 56 51 51 0.8 0.7 Latin America and the Caribbean 57 71 64 64 1.0 0.9 Middle East and North Africa 23 39 36 36 0.7 0.6 Sub-Saharan Africa 11 18 16 16 1.0 0.8 Developed Countries 88 100 92 92 0.8 0.7 Developing Countries 28 41 37 37 0.8 0.7 Cereals South Asia 164 157 124 121 7.0 7.1 East Asia and the Pacific 184 158 124 120 8.1 8.3 Europe and Central Asia 162 169 132 128 5.3 4.9 Latin America and the Caribbean 123 109 89 87 6.1 5.9 Middle East and North Africa 216 217 172 167 5.5 5.1 Sub-Saharan Africa 117 115 89 89 7.4 7.1 Developed Countries 118 130 97 94 6.8 6.3 Developing Countries 164 148 116 114 7.1 7.1 Source:Authors'estimates. Table 11. daIly PeR CaPITa CalORIe avaIlabIlITy wITh aNd wIThOuT ClImaTe ChaNge 2050 NoClimate NCARNo CSIRONo Change CF CF NCARCF CSIROCF 2000 kcal/day kcal/day kcal/day effects(%) effects(%) South Asia 2,424 2,660 2,226 2,255 4.3 4.3 East Asia and the Pacific 2,879 3,277 2,789 2,814 4.3 4.3 Europe and Central Asia 3,017 3,382 2,852 2,885 2.7 2.9 Latin America and the Caribbean 2,879 2,985 2,615 2,628 2.7 2.8 Middle East and North Africa 2,846 3,119 2,561 2,596 3.6 3.7 Sub-Saharan Africa 2,316 2,452 1,924 1,931 6.5 6.9 Developed Countries 3,450 3,645 3,190 3,215 2.3 2.5 Developing Countries 2,696 2,886 2,410 2,432 4.4 4.4 Source:Authors'estimates. average consumer in all countries gains--by 3.7 percent With climate change, however, calorie availability in in Latin America, 5.9 percent in Sub-Saharan Africa, 2050 is not only lower than the no-climate-change and 9.7 percent in South Asia. scenario in 2050; calorie availability actually declines D E V E L O P M E N T A N D C L I M AT E C H A N G E D I S C u S S I O N PA P E R S 23 large relative to the no climate change scenario. There is FIguRe 10. daIly PeR CaPITa CalORIe almost no difference in calorie outcome between the avaIlabIlITy wITh aNd wIThOuT ClImaTe two climate scenarios. ChaNge Table 12 reports summary statistics for the child 4,000 malnourishment indicator. With no climate change only 3,500 Sub-Saharan Africa sees an increase in the number of 3,000 malnourished children between 2000 and 2050. All Calories per Capita 2,500 other parts of the developing world see declines in the 2,000 number of malnourished children, driven by rapid 1,500 income and agricultural productivity growth. Climate 1,000 change eliminates much of that improvement. In East Asia and the Pacific, instead of 10 million malnourished 500 children in 2050, we find 14.2 to 14.5 million. In South 0 Asia, instead of 52.3 million malnourished children in South Asia and thePacific East Asia Central Asia Europe and the Caribbean Latin America and North Africa Middle East and Africa Sub-Saharan 2050, we find 58.5 to 59.1 million. In Sub-Saharan Africa, the effect of climate change is to increase the no-climate-change result by some 11 million children. 2000 No Climate Change 2050 NCAR NoCF 2050 CSIRO NoCF If CO2 fertilization is in fact effective in the field, the negative effect of climate change on child malnutrition Source: Authors' estimates. is reduced somewhat but not enough to offset the nega- tive effects of climate change on child malnutrition. relative to 2000 levels throughout the world. For the 3 . 3 T h e CO S T S O F a d aP TaT I O N average consumer in a developing country the decline is 10 percent relative to 2000. With CO2 fertilization, the The challenge of estimating the costs of adaptation declines are 3 percent to 7 percent less severe but still includes both choosing a baseline and then determining Table 12. TOTal NumbeR OF malNOuRIShed ChIldReN IN 2000 aNd 2050 (mIllION ChIldReN uNdeR 5 yRS) 2050 NoClimate NCARNo CSIRONo NCARCF CSIROCF 2000 Change CF CF effect(%) effect(%) South Asia 75.62 52.29 59.06 58.56 ­2.7 ­2.7 East Asia and Pacific 23.81 10.09 14.52 14.25 ­9.0 ­9.0 Europe and Central Asia 4.11 2.70 3.73 3.66 ­4.4 ­4.9 Latin America and Caribbean 7.69 4.98 6.43 6.37 ­4.7 ­4.8 Middle East and North Africa 3.46 1.10 2.09 2.01 ­10.3 ­11.3 Sub-Saharan Africa 32.67 41.72 52.21 52.06 ­5.4 ­5.6 All Developing Countries 147.36 112.88 138.04 136.91 ­4.6 ­4.8 Source:Authors' estimates. Note: The last three columns in this table report the percentage difference between the number of malnourished children in 2050 with and without the CO2 fertilization effect. For example, with the NCAR GCM, assuming CO2 fertilization is effective in the field results in a 2.7 percent decline in the number of malnourished children in South Asia relative to the climate change outcome in 2050 without CO2 fertilization. 24 THE COSTS OF AGRICu LT uRAL ADAPTATION TO CLIMATE CH A N G E what to include in the adaptation costs. The metric used to determine costs is the number of malnourished chil- Table 13. INveSTmeNT aNd PROduCTIvITy dren. The assumed public sector goal is to invest in SCeNaRIOS FOR ClImaTe ChaNge agricultural productivity enhancements that return the adaPTaTION number of malnourished children with climate change to the number that occur in the baseline. For the Developing country agricultural productivity investments, increase in intrinsic productivity growth rates over baseline EACC study of which this report is a part, the baseline growth rates is a world without climate change. Yield growth rate for all crops ­ 60% Animal numbers growth rate ­ 30% Production growth rate for oils and meals ­ 40% The choice of baseline affects some of the investments Irrigated area growth rate ­ 25% Rainfed area growth rate ­ 15% decrease considered in the costs of adaptation. For example, Increase in basin water efficiency ­ 0.15 by 2050 climate change reduces the productivity of agricultural Developed country agricultural productivity investments investments so investments made in a no-climate Yield growth rate for all crops ­ 30% change scenario become less productive when climate Animal numbers growth rate ­ 15% Production growth rate for oils and meals ­ 30% change occurs. So investments must compensate both for the reduced productivity of existing investments and Source: Authors' data. the need for additional productivity to deal with climate change effects. malnutrition counts for the various developing country A second issue is how to account for costs undertaken regions. Finally, Table 16 reports the annualized addi- by the private sector as it adjusts to climate change. An tional investment costs needed to meet the malnutrition important example of this is the change in trade flows numbers in Table 15. indicated in Table 9. As farmers cope with changes in productivity brought about because of climate change, The additional investments needed to reduce child production levels are altered and trade flows are malnutrition numbers to the no-climate-change results changed. There are clearly costs associated with these adjustments. Farmers must alter production practices, buy new seeds, and perhaps change capital equipment. However, we have no mechanism to estimate these FIguRe 11. daIly CalORIe avaIlabIlITy, private sector expenditures so these costs are not SOuTh aSIa aNd Sub-SahaRaN aFRICa included. 2,700 Three categories of productivity-enhancing investments 2,500 are considered--agricultural research, irrigation expan- 2,300 sion and efficiency improvements, and rural roads. We do two experiments. The first experiment, reported in 2,100 the first part of Table 13, is investments in the develop- ing world only needed to reduce childhood malnutrition 1,900 near the levels of the without-climate-change scenario. The second experiment is to include additional produc- 1,700 tivity enhancements in developed countries to assess the potential for spillovers. 1,500 South Asia Sub-Saharan Africa Table 14 reports the effects on daily per capita calorie 2000 No climate change NCAR NCAR + availability for these two experiments. Table 15 reports NCAR + + CSIRO the results for child malnutrition for the two climate CSIRO + CSIRO ++ scenarios relative to the no-climate-change scenario. Source: Authors' data. Figure 11 and Figure 14 are graphs of the calorie and D E V E L O P M E N T A N D C L I M AT E C H A N G E D I S C u S S I O N PA P E R S 25 Table 14. daIly CalORIe PeR CaPITa CONSumPTION wITh adaPTIve INveSTmeNTS (kCalS/ PeRSON/day LatinAmerica MiddleEast Europeand EastAsiaand andthe andNorth Sub-Saharan Developing SouthAsia CentralAsia thePacific Caribbean Africa Africa Countries 2000 2,424 2,879 3,017 2,879 2,846 2,316 2,696 2050 no climate 2,660 3,277 3,382 2,985 3,119 2,452 2,886 change NCAR 2,226 2,789 2,852 2,615 2,561 1,924 2,410 NCAR + 2,531 3,161 3,197 2,994 2,905 2,331 2,768 NCAR + + 2,564 3,198 3,235 3,027 2,941 2,367 2,803 CSIRO 2,255 2,814 2,885 2,628 2,596 1,931 2,432 CSIRO + 2,574 3,200 3,243 3,011 2,954 2,344 2,801 CSIRO ++ 2,612 3,241 3,285 3,048 2,996 2,384 2,840 Source: Authors' estimates Note:NCAR + and CSIRO + include only agricultural productivity investments in the developing world. NCAR ++ and CSIRO ++ include all productivity improvements in developed countries. The climate change results presented in this table assume no CO2 fer- tilization effects. Table 15. NumbeR OF malNOuRIShed ChIldReN wITh adaPTIve INveSTmeNTS (mIllION ChIldReN, uNdeR 5 yeaRS) Europeand LatinAmerica MiddleEast Sub-Saharan EastAsiaand Central andthe andNorth Developing SouthAsia Africa thePacific Asia Caribbean Africa World 2000 75.62 32.67 23.81 4.11 7.69 3.46 147.36 2050 No climate 52.37 38.78 12.02 2.96 5.43 1.15 112.71 change NCAR 58.16 48.72 16.55 3.91 6.73 2.02 136.10 NCAR + 53.19 40.03 12.40 3.15 5.29 1.23 115.28 NCAR + + 52.73 39.26 12.05 3.08 5.16 1.18 113.47 CSIRO 58.17 49.02 16.52 3.91 6.72 2.02 136.37 CSIRO + 53.04 40.32 12.25 3.12 5.26 1.21 115.20 CSIRO ++ 52.53 39.44 11.86 3.04 5.12 1.16 113.16 Source:Authors' estimates. Note:NCAR + and CSIRO + include only agricultural productivity investments in developing countries. NCAR ++ and CSIRO ++ include all productivity improvements in developed countries. The climate change results presented in this table assume no CO2 fer- tilization effects. are shown in Table 16. The additional annual invest- investment needs dominate, making up about 40 ments vary somewhat by climate scenario. With the percent of the total. Of that amount, the vast majority is wetter NCAR scenario the additional annual costs are for rural roads. South Asia investments are about $1.5 $7.1 billion. With the drier CSIRO scenario the costs billion per year with Latin America and Caribbean increase to $7.3 billion. Sub-Saharan African close behind with $1.2 billion per year. East Asia and 26 THE COSTS OF AGRICu LT uRAL ADAPTATION TO CLIMATE CH A N G E FIguRe 12. ChIld malNuTRITION eFFeCTS, FIguRe 14. ChIld malNuTRITION eFFeCTS, SOuTh aSIa aNd Sub-SahaRaN aFRICa eaST aSIa aNd The PaCIFIC, euROPe aNd (mIllIONS OF ChIldReN) CeNTRal aSIa, laTIN ameRICa aNd The CaRIbbeaN, aNd mIddle eaST aNd NORTh 2,700 aFRICa (mIllIONS OF ChIldReN) 2,500 25 2,300 20 2,100 15 1,900 1,700 10 1,500 South Asia Sub-Saharan Africa 5 2000 No climate change NCAR NCAR + 0 NCAR + + CSIRO East Asia and Europe and Latin America Middle East CSIRO + CSIRO ++ the Pacific Central Asia and the Caribbean and North Africa Source: Authors' data. Source: Authors' data. FIguRe 13. daIly CalORIe avaIlabIlITy, eaST aSIa aNd The PaCIFIC, euROPe aNd CeNTRal regions with irrigation efficiency investments substantial aSIa, laTIN ameRICa aNd The CaRIbbeaN, as well. Unlike Sub-Saharan Africa, road investments in aNd mIddle eaST aNd NORTh aFRICa these regions are relatively small. 3,400 3,300 With additional investments in the developed countries, spillover effects to the developing world reduce the need 3,200 for adaptation investments slightly. For example, with 3,100 the NCAR scenario, the annual investment need is $7.1 3,000 billion if productivity expenditures are only in the developing world. With developed country productivity 2,900 investments, that amount drops to $6.8 billion. 2,800 2,700 The key message embodied in these results is the importance of improving agricultural productivity as a 2,600 means of meeting the challenges that climate change 2,500 represents. The path to the needed agricultural produc- East Asia and Europe and Latin America Middle East the Pacific Central Asia and the Caribbean and North Africa tivity gains varies by region and to some extent by Source: Authors' data. climate scenario. 3 . 4 S eN S I T IvI T y a N a lyS I S the Pacific needs are just under $1 billion per year. To assess the sensitivity of our results, we did three Agricultural research is important in all three of these types of sensitivity analysis--a 10 percent increase in D E V E L O P M E N T A N D C L I M AT E C H A N G E D I S C u S S I O N PA P E R S 27 Table 16. addITIONal aNNual INveSTmeNT exPeNdITuRe Needed TO COuNTeRaCT The eFFeCTS OF ClImaTe ChaNge ON NuTRITION (mIllION 2000 uS$) Latin Europeand America MiddleEast Sub- EastAsia Central andthe andNorth Saharan Developing SouthAsia andPacific Asia Caribbean Africa Africa World nCar with developing country investments Ag. Research 172 151 84 426 169 314 1,316 Irrig. Expansion 344 15 6 31 ­26 537 907 Irrig. Efficiency 999 686 99 129 59 187 2,158 Rural Roads (Area Exp.) 8 73 0 573 37 1,980 2,671 Rural Roads (Yield Incr.) 9 9 10 3 1 35 66 Total 1,531 934 198 1,162 241 3,053 7,118 nCar with developing country and developed country investments Ag. Research 158 141 46 385 146 297 1,174 Irrig. Expansion 331 12 5 29 ­31 528 874 Irrig. Efficiency 995 684 98 128 59 186 2,151 Rural Roads (Area Exp.) 6 61 0 528 31 1,911 2,536 Rural Roads (Yield Incr.) 8 8 9 3 1 33 62 Total 1,499 905 159 1,072 206 2,956 6,797 Csiro with developing country investments Ag. Research 185 172 110 392 190 326 1,373 Irrig. Expansion 344 1 1 30 ­22 529 882 Irrig. Efficiency 1,006 648 101 128 58 186 2,128 Rural Roads (Area Exp.) 16 147 0 763 44 1,911 2,881 Rural Roads (Yield Incr.) 13 9 11 3 1 36 74 Total 1,565 977 222 1,315 271 2,987 7,338 Csiro with developing country and developed country investments Ag. Research 168 157 66 335 162 302 1,191 Irrig. Expansion 330 1 ­1 28 ­27 519 850 Irrig. Efficiency 1,002 645 100 127 58 185 2,119 Rural Roads (Area Exp.) 14 129 0 686 36 1,822 2,687 Rural Roads (Yield Incr.) 13 8 9 3 1 34 68 Total 1,528 941 174 1,179 230 2,863 6,915 Source:Authors' estimates. Note:These results are based on yield changes that do not include the CO2 fertilization effect. per capita income everywhere, a 10 percent increase in Population growth has the largest effect in absolute intrinsic productivity growth everywhere, and a 10 terms with elasticities ranging from 0.5 to 1.3 and the percent increase in population. The results can be inter- effect is negative. With all other exogenous variables preted as elasticities. For example, the elasticity of held constant, more population growth means more malnourished children with respect to GDP is ­0.14 malnourished children. Productivity growth is more and with intrinsic productivity growth from ­0.28 to effective than income growth in reducing the number of ­0.32. malnourished children. For example, a 10 percent 28 THE COSTS OF AGRICu LT uRAL ADAPTATION TO CLIMATE CH A N G E increase in intrinsic productivity reduces the number of marine fisheries made less productive as mangrove malnourished children in the developing world by 2.9 swamps are affected. Rice production in river deltas will to 3.2 percent. A 10 percent increase in income reduces be particularly hard hit. For example, over 30 percent of the number of malnourished children by only 1.4 the rice growing area in Vietnam would be lost to a 1 percent. meter sea level rise. Fourth, some parts of the world, in particular the rivers that derive from glaciers in the 3.5 lI mITaTIONS mountains of Asia, might see more varied flows with effects on irrigated agriculture and fisheries based on There are seven categories of climate change impacts water sourced from rivers. Fifth, we do not include that cannot currently be modeled due to data limita- autonomous adjustment costs such as those incurred by tions. Incorporation of most of these effects would farmers and traders as they adjust to changes in trade almost certainly make the effects of climate change flows. Sixth, we do not include the potential production significantly worse than the already negative picture in FPUs where production is not possible under current shown here. First, direct effects on livestock are not climate conditions but might be with 2050 cliamte. included. These effects range from less productive Finally, we do not include the effect of climate variabil- pastures for ruminants because of heat and precipitation ity and extreme events as current GCM scenarios do changes to increased stress in livestock confinement not account for them. systems. Second, pests and diseases, from traditional weeds that are more robust to larger insect populations 3 . 6 C O N Cl uS I O N S to more infectious diseases, might be a more serious problem with higher temperatures and locations with This analysis brings together, for the first time, detailed more precipitation. Third, the analysis does not take modeling of crop growth under climate change with into account the effect of sea level rise on coastal agri- insights from a detailed global partial agriculture equi- cultural resources. Coastal rice paddies might see saline librium trade model. Several important conclusions can intrusion, coastal seafood pens might be lost, and be drawn. Table 17. PeRCeNTage ChaNge IN malNOuRIShed ChIldReN wITh 10 PeRCeNT INCReaSeS IN gdP, PROduCTIvITy gROwTh, aNd POPulaTION Latin EastAsia Europeand America MiddleEast Sub- andthe Central andthe andNorth Saharan Developing SouthAsia Pacific Asia Caribbean Africa Africa Countries 10 percent increase in GDP everywhere No CC (% change) ­0.9 ­1.0 ­0.3 ­0.2 ­2.6 ­2.3 ­1.4 NCAR (% change) ­0.8 ­3.5 ­0.3 ­0.2 ­3.5 ­1.7 ­1.4 CSIRO (% change) ­0.8 ­3.5 ­0.3 ­0.2 ­3.6 ­1.7 ­1.4 10 percent increase in intrinsic productivity growth everywhere No CC (% change) ­2.1 ­2.8 ­5.6 ­5.6 ­6.4 ­4.2 ­3.2 NCAR (% change) ­1.7 ­4.8 ­3.9 ­4.5 ­7.7 ­3.0 ­2.8 CSIRO (% change) ­1.8 ­5.1 ­4.2 ­4.6 ­8.5 ­3.2 ­2.9 10 percent increase in population everywhere No CC (% change) 5.4 5.0 5.8 6.4 8.1 13.1 8.3 NCAR (% change) 5.2 5.9 5.0 5.7 10.0 11.9 7.9 CSIRO (% change) 5.2 6.0 5.1 5.7 10.2 11.9 7.9 Source: Authors' estimates. D E V E L O P M E N T A N D C L I M AT E C H A N G E D I S C u S S I O N PA P E R S 29 Regardless of climate change scenario, agriculture will consumption and trade flow effects exhibit considerable be negatively affected by climate change. When differences depending on the climate scenario. biophysical impacts of climate change are integrated into the IMPACT economic modeling framework, food Increases in investments to increase agricultural produc- prices increase sharply for key crops with adverse conse- tivity, including agricultural research, improvements in quences for the poor. Even without climate change, irrigation efficiency and expansion of irrigated area, and world prices are forecast to increase, from 39 to 72 rural road construction can compensate for much of the percent for the most important crops, driven by popula- effects of climate change on calorie availability and tion and income growth in the developing world child malnutrition. We estimate these costs to be in the outstripping expected productivity enhancements. range of US$7.1­7.3 billion annually (constant 2000 Climate change from unmitigated emissions of green- dollars) for direct agriculture and related investments house gases will cause even greater price increases. Rice (public agricultural research and development, irrigation prices are projected to be 13 percent higher in 2050 efficiency and expansion, and rural roads. compared to a no-climate change case, wheat prices 70 to 87 percent higher, and maize prices rise 34 percent. Changes in the volume and direction of international Price increases due to climate change are lower if CO2 trade in agricultural commodities are another impor- fertilization is considered, but the recent insights from tant avenue to compensate for the differential impacts field experiments suggest that benefits from carbon of climate change, which is also taken into account in fertilization are less than previously estimated. Higher our modeling framework. Thus, more open interna- food prices as a result of lower crop yields mean reduced tional trade should continue to be promoted to partially food availability and more malnourished children. offset adverse effects and uncertainty from climate change. There remains great uncertainty about where the partic- ular impacts will occur and the resulting production, 30 4. aNNex. IFPRI'S ClImaTe ChaNge crop from planting to harvest-ready. It requires daily weather data, including maximum and minimum mOdelINg meThOdOlOgy temperature, solar radiation, and precipitation, a description of the soil physical and chemical character- The challenge of modeling climate change impacts istics of the field, and crop management, including crop, arises in the wide ranging nature of processes that variety, planting date, plant spacing, and inputs such as underlie the working of markets, ecosystems, and fertilizer and irrigation. human behavior. Our analytical framework integrates modeling components that range from the macro to the For maize, wheat, rice, groundnuts, and soybeans, we micro and from processes that are driven by economics use the DSSAT crop model suite, version 4.0 ( J. W. to those that are essentially biological in nature. Jones, G. Hoogenboom, C. H. Porter, K. J. Boote, W. D. Batchelor, L. A. Hunt, P. W. Wilkens, U. Singh, A. J. An illustrative schematic of the linkage in IFPRI's Gijsman and J. T. Ritchie, 2003). For mapping these IMPACT model between the global agricultural policy results to other crops in IMPACT, the primary assump- and trade modeling of the partial agriculture equilib- tion is that plants with similar photosynthetic metabolic rium model with the hydrology and agronomic poten- pathways will react similarly to any given climate tial modeling is shown in Figure 1. change effect in a particular geographic region. Millet, sorghum, sugarcane, and maize all use the C4 pathway The modeling methodology used here reconciles the and are assumed to follow the DSSAT results for maize limited spatial resolution of macro-level economic in the same geographic regions. The remainder of the models that operate through equilibrium-driven rela- crops use the C3 pathway. The climate effects for the tionships at a national or even more aggregate regional C3 crops not directly modeled in DSSAT follow the level with detailed models of dynamic biophysical average from wheat, rice, soy, and groundnut from the processes. The climate change modeling system same geographic region, with two exceptions. The combines a biophysical model (the DSSAT crop model- IMPACT commodities of "other grains" and dryland ing suite) of responses of selected crops to climate, soil legumes are directly mapped to the DSSAT results for and nutrients with the SPAM data set of crop location wheat and groundnuts, respectively. and management techniques (Liang You and Stanley Wood, 2006), illustrated in Figure 15. These results are 4 . 2 C lI m aTe d aTa then aggregated and fed into the IMPACT model. DSSAT requires detailed daily climate data, not all of 4.1 C ROP mOdel INg which are readily available, so various approximation techniques were developed. To simulate today's climate The DSSAT crop simulation model is an extremely we use the Worldclim current conditions data set detailed process model of the daily development of a (www.worldclim.org) which is representative of D E V E L O P M E N T A N D C L I M AT E C H A N G E D I S C u S S I O N PA P E R S 31 FIguRe 15. The SPam daTa SeT develOPmeNT PROCeSS Forest Percentage Ag . Shrublands, Savanna, Grasslands >60% Croplands 40-60% Cropland/Natural 30-40% Vegetation <30% Water bodies (a) Crop Production Statistics (b) Land Cover (c) Agricultural Land Cover Pre-Processing Production shares (High/Low inputs) Harvested to physical area (CI) Potential gross revenue per pixel per cropper input level Optimisation (MAX: Gross Revenue) MIN: Cross Entropy Simultaneous allocation across all cropsinto agricultural share of each pixel (d) Crop * Input Level Specific (e) Crop Area Allocation Biophysical Suitability Any other mapped crop distribution evidence Source: Authors' data. 1950­2000 and reports monthly average minimum and Rainfall rates were obtained at three-hourly intervals for maximum temperatures and monthly average precipita- the years 1981, 1985, 1991, and 1995. A day was deter- tion. Site-specific daily weather data are generated mined to have experienced a precipitation event if the stochastically using the SIMMETEO software. At each average rainfall rate for the day exceeded a small thresh- location, 30 iterations were run and the mean of the yield old. The number of days experiencing a rainfall event values used to represent the effect of the climate variables. within each month was then counted up and averaged over the four years. Precipitation rates and solar radiation data were obtained from NASA's LDAS website (http://ldas.gsfc. The monthly values were regressed nonlinearly using nasa.gov/). We used the results from the Variable the Worldclim monthly temperature and climate data, Infiltration Capacity (VIC) land surface model. For elevation from the GLOBE dataset (http://www.ngdc. shortwave radiation (the sunlight plants make use of ), noaa.gov/mgg/topo/globe.html) and latitude. These monthly averages at 10 arc-minute resolution were regressions were used to estimate monthly solar radia- obtained for the years 1979­2000. Overall averages for tion data and the number of rainy days for both today each month were computed between all the years (for and the future. These projections were then used by example, the January average was computed as [ January SIMETEO to generate the daily values used in 1979 + January 1980 + ... + January 2000 ] / 22). DSSAT. 32 THE COSTS OF AGRICu LT uRAL ADAPTATION TO CLIMATE CH A N G E For future climate, we use the fourth assessment report locations crops can be grown in 2000 but not in 2050, 4 A2 runs using the CSIRO and NCAR models. At one or vice versa. For rainfed crops, we assume that a crop is time the A2 scenario was considered an extreme planted in the first month of a four month contiguous scenario although recent findings suggest it may not be. block of months where monthly average maximum We assume that all climate variables change linearly temperature does not exceed 37 degrees Celsius (about between their values in 2000 and 2050. This assumption 99 degrees F), monthly average minimum temperature eliminates any random extreme events such as droughts does not drop below 5 degrees Celsius (about 41 or high rainfall periods and also assumes that the forc- degrees F) and monthly total precipitation is not less ing effects of GHG emissions proceed linearly; that is, than 60 mm. See Figure 16 to Figure 18. we do not see a gradual speedup in climate change. The effect of this assumption is to underestimate negative For irrigated crops we assume that precipitation is not a effects from climate variability. constraint and only temperature matters, avoiding freez- ing periods. The starting month of the irrigated grow- 4.3 OT he R agRONO mIC INP uTS ing season is identified by 4 contiguous months where the monthly average maximum temperature does not Six other agronomic inputs are needed--soil character- exceed 45 degrees Celsius (about 113 degrees F) and istics, crop variety, cropping calendar, CO2 fertilization the monthly average minimum temperature does not effects, irrigation, and nutrient levels. drop below 8.5 degrees Celsius (about 47 degrees F). See Figure 19 to Figure 21. 4.3.1 Soilcharacteristics DSSAT uses many different soil characteristics in determining crop progress through the growing season. John Dimes of ICRISAT and Jawoo Koo of IFPRI FIguRe 16. RaINFed CROP PlaNTINg collaborated to classify the FAO soil types into 27 mONTh, 2050 ClImaTe meta-soil types. Each soil type is defined by a triple of soil organic carbon content (high/medium/low), soil rooting depth as a proxy for available water content (deep/medium/shallow), and major constituent (sand/ loam/clay). The dominant soil type in a pixel is used to represent the soil type for the entire pixel. 4.3.2 Cropvariety DSSAT includes many different varieties of each crop. For the results reported here, we use the maize variety Garst 8808, a winter wheat variety, a large-seeded Source: Compiled by Authors. Virginia runner type groundnut variety, a maturity group 5 soybean variety, and for rice, a recent IRRI indica rice variety and a Japonica variety. The rice varieties are assigned by geographic area according to whichever is 4 NCAR and CSIRO AR4 data downscaled by Kenneth Strzepek and more commonly cultivated within the region. colleagues at the MIT's Center for Global Change Science. We acknowledge the international modeling groups for providing their data for analysis, the Program for Climate Model Diagnosis and 4.3.3 Croppingcalendar Intercomparison (PCMDI) for collecting and archiving the model data, the JSC/CLIVAR Working Group on Coupled Modeling (WGCM) and their Coupled Model Intercomparison Project Climate change will alter the cropping calendar in some (CMIP) and Climate Simulation Panel for organizing the model data analysis activity, and the IPCC WG1 TSU for technical support. The locations, shifting the month in which a crop can be IPCC Data Archive at Lawrence Livermore National Laboratory is safely planted forward or back. Furthermore, in some supported by the Office of Science, U.S. Department of Energy. D E V E L O P M E N T A N D C L I M AT E C H A N G E D I S C u S S I O N PA P E R S 33 FIguRe 17. RaINFed PlaNTINg mONTh, 2050 FIguRe 19. IRRIgaTed PlaNTINg mONTh, ClImaTe, CSIRO gCm a2 SCeNaRIO (aR4) 2050 ClImaTe Source: Compiled by Authors. Source: Compiled by Authors. FIguRe 18. RaINFed PlaNTINg mONTh, 2050 FIguRe 20. IRRIgaTed PlaNTINg mONTh, ClImaTe, NCaR gCm a2 SCeNaRIO (aR4) 2050 ClImaTe, CSIRO gCm a2 SCeNaRIO (aR4) Source: Compiled by Authors. Source: Compiled by Authors. FIguRe 21. IRRIgaTed PlaNTINg mONTh, Developing a climate based growing season algorithm 2050 ClImaTe, NCaR gCm a2 SCeNaRIO for winter wheat was challenging. Our solution was to (aR4) treat winter wheat differently than other crops. Rather than using a cropping calendar, we let DSSAT use planting dates throughout the year and choose the date that provides the best yield for each pixel. 4.3.4 CO 2 fertilizationeffects Plants produce more vegetative matter as atmospheric concentrations of CO2 increase. The effect depends on the nature of the photosynthetic process used by the plant species. So-called C3 plants use CO2 less effi- ciently than C4 plants so C3 plants are more sensitive Source: Compiled by Authors. to higher concentrations of CO2. It remains an open 34 THE COSTS OF AGRICu LT uRAL ADAPTATION TO CLIMATE CH A N G E question whether these laboratory results translate to 4 . 4 FR O m d S SaT TO Th e I mPa C T mO d e l actual field conditions. A recent report on field experi- ments on CO2 fertilization (Stephen P. Long, Elizabeth DSSAT is run for five crops--rice, wheat, maize, A. Ainsworth, Andrew D. B. Leakey, Josef Nosberger soybeans, and groundnuts--at 0.5 degree intervals for the and Donald R. Ort, 2006), finds that the effects in the locations that the SPAM data set says the crop is field are approximately 50 percent less than in experi- currently grown. Other crops are assumed to have ments in enclosed containers. And another report productivity effects similar to these five crops as described ( Jorge A. Zavala, Clare L. Casteel, Evan H. DeLucia above. The results from this analysis are then aggregated and May R. Berenbaum, 2008) finds that higher levels to the IMPACT FPU level as described below. of atmospheric CO2 increase the susceptibility of soybean plants to the Japanese beetle and maize to the 4 . 5 T h e ImPa C T 2 0 0 9 m O d e l 5 western corn rootworm. So the actual, field benefits of CO2 fertilization remain uncertain. The IMPACT model was initially developed at the International Food Policy Research Institute (IFPRI) to DSSAT has an option to include CO2 fertilization project global food supply, food demand and food secu- effects at different levels of CO2 atmospheric concen- rity to year 2020 and beyond (Rosegrant et al. (2008)). It tration. To capture the uncertainty in actual field effects, is a partial equilibrium agricultural model with 32 crop we simulate two levels of atmospheric CO2 in 2050-- and livestock commodities, including cereals, soybeans, 369 ppm (the level in 2000) and 532 ppm, the CO2 roots and tubers, meats, milk, eggs, oilseeds, oilcakes and levels in 2050 actually used in the A2 scenario. meals, sugar, and fruits and vegetables. IMPACT has 115 country (or in a few cases country aggregate) Our aggregation process from SPAM pixels and the crop regions, within each of which supply, demand, and prices model results to IMPACT FPUs results in some improb- for agricultural commodities are determined. Large able yield effects in a few locations. To deal with these, countries are further divided into major river basins. The we introduce the following caps. In the crop modeling result, portrayed in Figure 22, is 281 spatial units, called analysis we cap yield increases at 20 percent at the pixel food production units (FPUs). The model links the vari- level. In addition, we cap the FPU-level yield increase at ous countries and regions through international trade 30 percent. Finally, we limit the negative effect of climate using a series of linear and nonlinear equations to on yield growth in IMPACT to ­2 percent per year. approximate the underlying production and demand relationships. World agricultural commodity prices are 4.3.5 Wateravailability determined annually at levels that clear international markets. Growth in crop production in each country is Rainfed crops receive water either from precipitation at determined by crop and input prices, exogenous rates of the time it falls or from soil moisture. Soil characteris- productivity growth and area expansion, investment in tics influence the extent to which previous precipitation irrigation, and water availability. Demand is a function events provide water for growth in future periods. of prices, income, and population growth and contains Irrigated crops receive water automatically in DSSAT as four categories of commodity demand--food, feed, needed. Soil moisture is completely replenished at the biofuels feedstock, and other uses. beginning of each day in a model run. To assess the effects of water stress on irrigated crops, a separate 4 . 6 m O d e lI N g C lI m aTe C h aNg e I N hydrology model is used, as described below. I m Pa C T 4.3.6 Nutrientlevel Climate change effects on crop production enter into the IMPACT model by altering both crop area and DSSAT allows a choice of nitrogen application amounts yield. Yields are altered through the intrinsic yield and timing. We vary the amount of elemental N from 15 to 200 kg per hectare depending on crop, manage- ment system (irrigated or rainfed) and country. 5 See Rosegrant et al. 2008 for technical details. D E V E L O P M E N T A N D C L I M AT E C H A N G E D I S C u S S I O N PA P E R S 35 FIguRe 22. ImPaCT mOdel uNITS OF aNalySIS, The FOOd PROduCTION uNIT (FPu) Source: Authors' data. growth coefficient, gy tni , in the yield equation (1) as as described above and converted to a growth rate well as the water availability coefficient (WAT) for irri- which is then used to alter gy tni . Rainfed crops react to gated crops. See Table 24 for figures showing these rates changes in precipitation as modeled in DSSAT. for the most important crops. These growth rates range depend on crop, management system, and location. For For irrigated crops, water stress from climate change is most crops, the average of this rate is about 1 percent captured as part of the hydrology model built into per year from effects that are not modeled. But in some IMPACT, a semi-distributed macro-scale hydrology countries the growth is assumed to be negative while in module that covers the global land mass (except others it is as high as 5 percent per year for some years. Antarctica and Greenland). It conducts continuous hydrological simulations at monthly or daily time steps at YC tni = tni × ( PStni ) iin × ( PFtnk ) ikn × (1 + gy tni ) - (1) a spatial resolution of 30 arc-minutes. The hydrological k 6 module simulates the rainfall-runoff process, partitioning YCtni (WATtni ) incoming precipitation into evapotranspiration and runoff Climate change productivity effects are produced by 6 tni - yield intercept for year t, determined by yield in previous year; calculating location-specific yields for each of the five PStni - output price in year t; PFtni - input prices in year t. - input and crops modeled with DSSAT for 2000 and 2050 climate output price elasticities. 36 THE COSTS OF AGRICu LT uRAL ADAPTATION TO CLIMATE CH A N G E that are modulated by soil moisture content. A unique may become more or less suitable for any crop and will feature of the module is that it uses a probability distribu- impact the intrinsic area growth rate, in the area growth tion function of soil water holding capacity within a grid calculation. Water availability will affect the WAT factor cell to represent spatial heterogeneity of soil properties, for irrigated crop area. enabling the module to deal with sub-grid variability of soil. A temperature-reference method is used to judge AC tni = tni × ( PStni ) iin × ( PStnj ) ijn × (1 + gatni ) - (2) j i whether precipitation comes as rain or snow and deter- AC tni (WATtni ) mines the accumulation or melting of snow accumulated in conceptual snow storage. Model parameterization was done to minimize the differences between simulated and Crop calendar changes due to climate change cause two observed runoff processes, using a genetic algorithm. The distinct issues. When the crop calendar in an FPU model is spun up for five years at the beginning for each changes so that a crop that was grown in 2000 can no simulation run to minimize any arbitrary assumption of longer be grown in 2050, we implement an adjustment initial conditions. Finally, simulated runoff and evapo- to gatni that will bring the harvested area to zero--or transpiration at 30 arc-minute grid cells are aggregated to nearly so--by 2050. However, when it becomes possible the 281 FPUs of the IMPACT model. to grow a crop in 2050 where it could not be grown in 2000, we do not add this new area. An example is that One of the more challenging aspects of this research has parts of Ontario, Canada that have too short a growing been to deal with spatial aggregation issues. FPUs are season in 2000 will be able to grow maize in 2050, large areas. For example, the India Ganges FPU is the according to the climate scenarios used. As a result our entire length of the Ganges River in India. Within an estimates of future production are biased downward FPU, there can be large variations in climate and agro- somewhat. The effect is likely to be small, however, as nomic characteristics. A major challenge was to come up new areas have other constraints on crop productivity, in with an aggregation scheme to take outputs from the crop particular soil characteristics. modeling process to the IMPACT FPUs. The process we used proceeds as follows. First, within an FPU, choose the 4 . 7 m O d e lI N g T h e C O S T S O F appropriate SPAM data set, with a spatial resolution of 5 a d aP TaT I O N TO ClI m aT e Ch aN ge arc-minutes (approximately 10 km at the equator) that corresponds to the crop/management combination. The This section describes the methodology used to provide physical area in the SPAM data set is then used as the estimates of the costs of adapting to climate. weight to find the weighted-average-yield across the FPU. This is done for each climate scenario (including A key issue is the metric for adaptation. We use average the no-climate-change scenario). The ratio of the per capita calorie consumption and an associated measure weighted-average-yield in 2050 to the no-climate-change of human well-being, the number of malnourished chil- yield is used to adjust the yield growth rate in equation dren under 5. We use the underweight definition of (1) to reflect the effects of climate change. malnutrition, the proportion of children under five falling below minus 2 standard deviations from the median In some cases the simulated changes in yields from weight-for-age standard set by the U.S. National Center 7 climate change are large and positive. This usually arises for Health Statistics and the World Health Organization . from two major causes; (1) starting from a low base (which can occur in marginal production areas) and (2) unrealistically large effects of carbon dioxide fertiliza- 7 We use the underweight definition of malnutrition, which is low weight for age or weight for age; more than a standard deviation of 2 tion. To avoid these artifacts, we place a cap on the below the median value of the reference (healthy) population. Two changes in yields at 20 percent gains over the alternate definitions are · Stunting. Low height for age or height for age more than a stan- no-climate-change outcome at the pixel level. dard deviation of 2 below the median value of the reference (healthy) population · Wasting. Low weight for height or weight for height more than a Harvested areas in the IMPACT model are also standard deviation of 2 below the median value of the reference affected by climate change. In any particular FPU, land (healthy) population. D E V E L O P M E N T A N D C L I M AT E C H A N G E D I S C u S S I O N PA P E R S 37 4.8 eSTI maTINg Ch Ild mal NuTRIT I O N NMALt = MALt × POP 5t The IMPACT model provides data on the per capita where NMAL = number of malnourished children, calorie availability by country. Child malnutrition has and many determinants of which calorie intake is one. The percentage of malnourished children under the age of POP5 = number of children 0-5 years old five is estimated from the average per capita calorie in the population consumption, female access to secondary education, the quality of maternal and child care, and health and sanita- For this report, we assume that life expectancy, maternal tion (Rosegrant et al. (2008)). The precise relationship education and clean water access are held constant in all used to project the percentage of malnourished children future scenarios and limit investments to three areas: is based on a cross-country regression relationship of agricultural research and development spending, rural Smith and Haddad (2000), and can be written as follows: roads, and irrigation area expansion and efficiency improvements that alter calorie availability and child KCALt malnutrition estimates. The approach is to estimate the MAL = -25.24 × ln - 71.76 × t , 2000 KCAL2000 productivity growth needed to meet a malnutrition or calorie availability target and then estimate the invest- LFEXPRAT - 0.22 × SCH - t , 2000 t , 2000 ment expenditures needed in research, irrigation, and 0.08 × WATER t , 2000 road to generate that productivity growth. The basic process is as follows. where · Run the NoCC scenario and estimate the number MAL = percentage of malnourished of malnourished children in 2050 children · Run a CC scenario and estimate the number of malnourished children in 2050 KCAL = per capita kilocalorie availability · Find a blend of agricultural productivity growth rate increases (crop, animal numbers and oils and LFEXPRAT = ratio of female to male life meals) that produces a scenario with climate change expectancy at birth where number of the malnourished children in 2050 is roughly equal to the number of malnour- SCH = total female enrollment in ished children in 2050 for the NoCC scenario and secondary education (any age estimate the implied investment costs. group) as a percentage of the female age-group corresponding 4 . 9 a g R I Cu lTuR a l R eSe a R Ch to national regulations for I N v eS Tm e N T S secondary education, and The process of estimating agricultural research invest- WATER = percentage of population with ments uses published research and expert opinion to access to safe water estimate yield responsiveness to research expenditures and estimation of future expenditures on the basis of t,t 2000 = the difference between the vari- historical expenditure growth rates. Most of the data on able values at time t and the base public agricultural research are from the Agricultural year t2000 Science and Technology Indicators (ASTI) data set available at http://www.asti.cgiar.org/ and converted Data on the percentage of malnourished children (MAL into 2000 US$ values by the GDP deflator obtained are taken from the World Development Indicators. from the IMF's International Financial Statistics. For a Other data sources include the FAO FAOSTAT data- few countries, OECD Science and Technology base, and the UNESCO UNESCOSTAT database. Indicators data and Eurostat data on gross domestic 38 THE COSTS OF AGRICu LT uRAL ADAPTATION TO CLIMATE CH A N G E expenditure on R&D for agricultural sciences are used 2050 after being converted to 2000 US $ values.8 For China, ARbaseline = ARn (4) the Ministry of Science and Technology (MOST) data n = 2010 for public agricultural research spending is used. For For a given scenario, we determine the change in invest- some countries where public agricultural research data ment implied in changes in agricultural performance are not available, ASTI estimates of public agricultural relative to the baseline. The scenario agricultural research are used.9 For these countries, ASTI uses agri- research costs (ARscenario ) are computed as follows: cultural GDP of the country and the average intensity ratio of the region that the country is located to gener- Yld 2050 - Yld 2050 e Scenario Baselin ate an estimate. Baseline Yld 2050 ARScenario = 1+ Yield ARBaseline (5) Research Baseline research expenditures in 2050 are estimated by applying the multipliers, g a , in Table 18 to the historical growth rates, g h , obtained from data on agricultural and y research spending discussed above. The historical with Yld 2050 being the average of cereal yields. growth rate for most countries is computed as an aver- age of the annual historical growth rates for a recent ten ARScenario represents the change needed to achieve the year period (or less when data are not available). For the new level of productivity to achieve the target. remaining countries, regional average historical growth rates are computed from the data set and used for indi- A 4.9.1 griculturalResearchInvestmentsSensitivity vidual countries. Analysis Since the initial analysis was done, an improved meth- Table 18. aSSumed mulTIPlIeRS OF odology to estimate agricultural research investments hISTORIC gROwTh RaTeS OF agRICulTuRal has been developed. This new methodology uses more ReSeaRCh exPeNdITuR detailed estimates of yield responsiveness to research expenditures. It also introduces lags between research Period Multiplierofhistoricgrowthrate(%) expenditure and the resulting increase in productivity. 2000­2010 9 The estimation of baseline agricultural research spend- 2011­2020 8 ing remains the same. 2021­2030 7 2031­2040 6 The first change is to differentiate yield elasticity with 2040­2050 5 respect to research expenditures ( Research ) by the follow- Yield Source: Compiled by authors. ing regions--Asia, Sub-Saharan Africa (SSA), Latin America and the Caribbean (LAC), Western Asia and North Africa (WANA), and North America and For the main results, it is assumed that the yield elastic- Europe (NAE), based on the following references-- ity with respect to research expenditures is ( Research ) Yield A.D. Alene and O. Coulibaly (2009), A. K. Kiani et al. 0.296 for all countries and regions. 8 There are no data or estimates for North Korea, Singapore, Agricultural research investment (ARn) for every year Afghanistan, Equatorial Guinea, Somalia, Djibouti, Kazakhstan, after 2010 is calculated as follows: Kyrgyzstan, Tajikistan, Turkmenistan, Ukraine, Uzbekistan, Armenia, Azerbaijan, Belarus, and Georgia. 9 These countries are Angola, Cameroon, Cape Verde, Central African Republic, Chad, Comoros, Democratic Republic of Congo, Guinea Bissau, gh ga ARn = + 1 ARn-1 (3) Lesotho, Liberia, Mozambique, Namibia, Rwanda, Sao Tome and Principe, 100 Sierra Leone, Swaziland, Zimbabwe, Bolivia, Ecuador, Peru, Venezuela, Antigua and Barbuda, Guyana, Jamaica, Surinam, Trinidad and Tobago, Algeria, Bahrain, Iraq, Israel, Lebanon, Kuwait, Libya, Qatar, Saudi Arabia, Turkey, United Arab Emirates, Bhutan, Cambodia, Mongolia, and Total investments over the period are: Luxembourg. D E V E L O P M E N T A N D C L I M AT E C H A N G E D I S C u S S I O N PA P E R S 39 (2008), C. Thirtle et al.(2003), and (D. Schimmelpfenig increase market access and reduce transaction costs. We and C. Thirtle (1999). The regional elasticities are 0.344 consider two relationships between roads and agricul- for Asia, 0.363 for SSA, 0.197 for LAC, and 0.171 for tural production--the effects of area expansion and WANA, and 0.063 for NAE. yield growth. We also take into consideration the time lag between 4.10.1 Areaeffect investments and impacts on productivity. The initial effects are small, grow over time until the maximum Expanded crop area requires roads to deliver inputs and effect on yield is reached and then taper off. We assume move goods from fields to market. We assume that any that the elasticities described above are the cumulative growth in cropped area requires a similar growth in effect of the initial expenditure. The scenario agricul- rural roads and that it is a one to one relationship. Rural tural research costs for each region ( ARScenario ) are road length data were taken from World Road Statistics computed as follows: 2002. We use information from the latest available year, typically 2000, to calculate rural road length (r2000) as ( yld Scenario - yld Baseline ) ARScenario ,t ARScenario = 1 + t2050 = 2010 Yield total roads minus highways minus motorways. yld Baseline eResearch ,t Rural road investment costs are calculated by multiplying The resulting level of spending ( ARScenario ) computed for the extra road length between 2000 and 2050 by the road each region represents the change needed to achieve the construction cost per km (C r ) values in Table 20, derived new level of productivity to achieve the target. The from various World Bank road construction project docu- effect of this new methodology is to raise the estimate ments. The values in the table are in 2005 US dollars; of developing country research investment costs by they are deflated to 2000 US dollars for the analysis. about $900 million annually. We calculate the extra road length required due to area 4.10 Ru R al RO adS increase (ra) as follows: a 2050 - a2000 Higher yields and more cropped area require maintain- ra = r2000 × (6) a2000 ing and increasing the density of rural road networks to Table 19. ReSeaRCh INveSTmeNT SeNSITIvITy aNalySIS LatinAmerica MiddleEast EastAsia Europeand andthe andNorth Sub-Saharan Developing SouthAsia andPacific CentralAsia Caribbean Africa Africa World nCar with developing country investments Original 172 151 84 426 169 314 1,316 Revised 195 145 688 653 312 259 2,252 nCar with developing country and developed country investments Original 158 141 46 385 146 297 1,174 Revised 181 137 587 596 278 245 2,024 Csiro with developing country investments Original 185 172 110 392 190 326 1,373 Revised 223 154 744 594 338 268 2,322 Csiro with developing country and developed country investments Original 168 157 66 335 162 302 1,191 Revised 206 143 615 517 297 249 2,027 Source:Authors'estimates. 40 THE COSTS OF AGRICu LT uRAL ADAPTATION TO CLIMATE CH A N G E Table 20. ROad CONSTRuCTION COSTS Table 21. PeRCeNT yIeld INCReaSe wITh (2005 uS$ PeR km) ReSPeCT TO ROad leNgTh ( yldinc Roads ), RegIONal aveRageS South Asia 575,000 Sub-Saharan Africa 600,000 Latin America 0.043 Middle East and North Africa 585,000 Sub-Saharan Africa 0.240 Latin America and Caribbean 580,000 Western Asia and North Africa 0.085 East Asia and Pacific 555,000 South Asia 0.170 ECA 590,000 East Asia and the Pacific 0.158 Developed 621,000 Eastern Europe and Central Asia 0.141 Source: Various World Bank road construction project documents. Source: Compiled by authors. if a2050 - a2000 < 0 then ra = 0 The yield values used in this calculation ( yld xxxx ) are an Finally we multiply ra by road unit cost to get the cost average for all cereals modeled--rice, wheat maize, of new roads needed to support crop area expansion sorghum, millet and an `other grains' category. We ( RRa ). calculate the increase in road investment due to a yield increase ( RR y ) as follows: RRa = raC r (7) yld 2050 4.10.2 Yieldeffect - 1 × yldinc Roads yld 2000 RR y = Yield × r2000 × C r (8) Rural road density has been shown to be among the e Roads most important contributors to productivity growth in agriculture. This is due to the impact that better roads The total investment in rural roads ( RRbaseline ) for the have in reducing the transport component of input costs baseline run is calculated as follows: and transaction costs of marketing products. In addi- tion, roads improve the flow of information on market RRbaseline = RRa + RR y (9) conditions, new technologies, and reduce the potential risks to their enterprises. 4.10.3 Scenario Results and additional Road Costs The investments in rural roads needed to achieve a given yield effect includes two components. The first, To calculate the effect of a particular scenario on called says how much of a given yield increase is driven road costs, we use the cereal yield in 2050 from the Yield by road expansion. Table 21 reports regional averages baseline and the respective scenario model run, e Roads for this variable. For example, in Latin America 4.3 and yldinc Roads to calculate the target costs of rural roads percent of any yield increase is driven by road ( RRScenario ) as follows: expansion. Scenario yld 2050 The second component is the elasticity of yields with Baseline - 1 × yldinc Roads yld 2050 respect to road expansion. Table 6 in Fan, P. Hazell and RRScenario = 1 + Yield RRBaseline (10) e Roads S. Thorat (1998) reports the elasticity of total factor productivity to road investments as 0.072 in India using data from the 1970s through the early 1990s. We use this value for all countries. D E V E L O P M E N T A N D C L I M AT E C H A N G E D I S C u S S I O N PA P E R S 41 4.11 IRRI gaTION Table 22. IRRIgaTION INveSTmeNT COST Irrigation investments to meet a productivity target (uS 2000$ PeR heCTaRe) include two components. Costs for expanding irrigated Region Irrigationcost area and costs related to the increase of irrigation water South Asia 6,023 use efficiency. East Asia and Pacific 9,916 Eastern Europe and Central Asia 4,997 4.11.1 Areaexpansion Latin America and Caribbean 15,929 The total investments in irrigation area are calculated by Middle East and North Africa 9,581 multiplying the estimated net irrigated area increase Sub-Saharan Africa 18,252 between 2000 and 2050 by the cost of irrigation per Source:Literature review of World Bank, Food and Agriculture hectare. Total irrigated area data generated by IMPACT Organization (FAO) and International Water Management Institute have to be adjusted for cropping intensity ( rn ) because (IWMI) documents, project reports, and meta-evaluations directly the IMPACT results include multiple cropping seasons related to completed and on-going irrigation projects. and therefore overstates the physical area. Net irrigated area ( an ) for each year n is calculated as Net follows: IMPACT, the concept of basin efficiency (BE) is used an to account for changes in irrigation efficiency at all Net an = 1000 × 100 (11) levels within a river basin (N. Haie and A.A. Keller, rn 2008, A. Keller and J. Keller, 1995). It fully accounts for The annual changes in net irrigated area for each year the portion of diverted irrigation water that returns to are given by rivers or aquifer systems and can be reused repeatedly by downstream users. This approach avoids the limita- Net Net Net a n = a n+1 - an (12) tion of the classical irrigation efficiency concept that treats return flow as "losses." if Net Net an < 0 then a n = 0 (13) BE is defined as the ratio of beneficial irrigation water consumption (BC) to total irrigation water consumption The year-to-year changes are summed for the entire (TC); that is, changes in precipitation are excluded from period between 2000 and 2050 to get aggregate net irri- this calculation: Net gated area change a 2050- 2000 . The aggregate year-to- BC year change between 2000 and 2050 is multiplied by BE = (15) ( ) irrigation unit cost c Irrig to get the total costs of TC increased irrigation between 2000 and 2050 ( IR ). BE in the base year is calculated as the ratio of the net IR = a Net 2050 - 2000 × cirrig (14) irrigation water demand (NIRWD) to the total irriga- tion water consumption based on Shiklomanov (1999). Irrigation unit costs vary by region, as indicated in Table NIRWD is defined as. 23. In a few countries where better information is avail- ( ) cp able, it is used instead. NIRWD = kc cp ,st ET 0st - PE cp ,st AI (16) cp st 4.11.2 Irrigationefficiencyimprovements · cp--index for the IMPACT crop. Includes all Improvements in irrigation efficiency are another source IMPACT crops that receive irrigation of agricultural productivity improvements, especially as · st--index for the crop growth stages. FAO has water scarcity becomes a world-wide problem. In divided the crop growing period into four stages, 42 THE COSTS OF AGRICu LT uRAL ADAPTATION TO CLIMATE CH A N G E each with separate crop coefficient (kc) values. See report) and calculated associated investment costs. Let R.G. Allen et al. (1998) for details. subscript "0" denote the base scenario and "1" denote an · kc --crop coefficient. Each crop growth stage is alternate irrigation investment scenario, and assume that associated with a corresponding crop coefficient area with more efficient irrigation (for example, through (R.G. Allen, et al., 1998) that adjusts reference ET adoption of enhanced management or advanced tech- for the characteristics of a particular crop. nologies such as sprinklers) accounts for a share of X of · ET 0 --reference evapotranspiration. total irrigated area in 2050, we have: Evapotranspiration describes the sum of evapora- TC1 = BC 0 × (1 - X ) + BC ×X tion and plant transpiration from the Earth's land 0 (17) E0 surface to atmosphere. Evaporation accounts for the movement of water to the air from sources such as = TC 0 + BC 0 × X the soil, canopy interception, and water bodies. Transpiration accounts for the movement of water We assume all consumption in high efficiency irrigation within a plant and the subsequent loss of water as is beneficial consumption. Assuming that beneficial vapor through stomata in its leaves. Reference consumption is the same in the base scenario as in the evapotranspiration is defined as the ET that occurs alternate scenario, from a standardized "reference" crop such as clipped BC 0 grass or alfalfa. E1 = (18) · PE--effective rainfall (rainfall that is actually avail- TC1 able for plant growth) Bringing (17) into (18) and simplifying results in: · AI cp --irrigated area for crop cp in the basin E0 X = 1- E1 (1 - E0 ) (19) This calculation generates globally consistent estimates for BE for the base year. 4.11.3 Irrigationinvestmentssensitivityanalysis For the future, we project small enhancements in BE, with levels increasing to 0.5­0.8 by 2050 under the An alternate methodology uses beneficial consumption baseline. An upper level of BE is set at 0.85 as a practi- in the base and alternative investment scenarios from cal maximum. IMPACT model simulations, rather than assuming that beneficial consumption is the same in both scenarios. To account for the investment costs associated with This leads to: increasing irrigation efficiency, we used one-third of the BC 0 × (1 - X ) cost of recent irrigation modernization projects using TC1 = + BC1 × X (20) BE0 sprinklers as a proxy. Based on a literature review of World Bank, FAO, and International Water Management Institute (IWMI) documents, project In the above equation we still assume that all water reports, and meta-evaluations directly related to consumption in the high irrigation efficiency areas is completed and on-going irrigation projects focusing on beneficial consumption. irrigation modernization only, we identified per-hectare investment cost of US$2,144 for East, South, Southeast, Bring BE0 = BC 0 TC 0 and BE1 = BC1 TC1 into the and Central Asia; US$4,311 for Sub-Saharan Africa above equation and simplify to get: and Latin America; and US$953 for the Middle East BE0 and North Africa. We used one third of these values to X = 1- BE1 (1 - BE0 ) (21) proxy investment costs for irrigation efficiency enhance- ment under alternative climate change scenarios. in which is the ratio of beneficial consumption of For an increased agricultural investment cost scenario, 2050 in the alternate investment scenario to that in the we increase BE values by a given amount (0.15 for this baseline scenario, namely = BC1 BC 0 . The values of D E V E L O P M E N T A N D C L I M AT E C H A N G E D I S C u S S I O N PA P E R S 43 can be calculated from beneficial consumption results basin efficiency values range from 0.41 (Brazil) to 0.82 of IMPACT simulations for the baseline and alternate (Colorado Basin, United States). A reviewer comment irrigation investment scenarios. TC1 is bounded by the suggested setting the minimum basin efficiency value to available renewable water supply for irrigation. 0.55. We implemented this sensitivity analysis using the alternate estimate methodology and results are also We then multiply X by the investment costs and irri- presented in Table 23 (alternate estimate 2). As can be gated area in 2050. seen, total costs under estimate 2 are slightly higher than under estimate 1. The reason for this is that bene- IEinv = X × IEcost × AI (22) ficial consumption is now already higher in the base year while the denominator is smaller (see Eq. 21), and Table 23 provides a comparison between old and new this increase continues out into the future. In none of results for irrigation efficiency investments (method 1 the cases is the maximum achievable BE value of 0.85 and alternate estimate 1, respectively). Our base year reached. Table 23. ReSulTS FROm alTeRNaTe eSTImaTION OF IRRIgaTION eFFICIeNCy ImPROve- meNT COSTS (uS 2000 mIllION PeR yeaR) NCARw.DevelopingCountryInvestments Method1Alternateestimate1 Alternateestimate2 South Asia 999 343 351 East Asia and Pacific 686 522 533 Europe and Central Asia 99 107 110 Latin America and Caribbean 129 190 208 Middle East and North Africa 59 71 71 Sub-Saharan Africa 187 203 249 Developing 2,158 1,436 1,522 nCar w. developing Country investments + developed Country Productivity increases South Asia 995 342 351 East Asia and Pacific 684 521 531 Europe and Central Asia 98 107 110 Latin America and Caribbean 128 190 207 Middle East and North Africa 59 70 71 Sub-Saharan Africa 186 202 249 Developing 2,151 1,433 1,519 Csiro w. developing Country investments South Asia 1,006 347 356 East Asia and Pacific 648 499 510 Europe and Central Asia 101 110 113 Latin America and Caribbean 128 189 206 Middle East and North Africa 58 70 70 Sub-Saharan Africa 186 202 248 Developing 2,128 1,417 1,503 Csiro w. developing Country investments + developed Country Productivity increases South Asia 1,002 347 356 East Asia and Pacific 645 498 509 (Continuedonnextpage) 44 THE COSTS OF AGRICu LT uRAL ADAPTATION TO CLIMATE CH A N G E Table 23. (continued) NCARw.DevelopingCountryInvestments Method1Alternateestimate1 Alternateestimate2 Europe and Central Asia 100 110 112 Latin America and Caribbean 127 189 206 Middle East and North Africa 58 70 70 Sub-Saharan Africa 185 201 247 Developing 2,119 1,413 1,500 Source:Author calculations. 4.12 P OP ulaTION, INCOme a Nd ClI m aTe We report results for two climate scenarios--the F uT u Re SC eN aRIO aSS umPT I O N S NCAR and CSIRO GCMs with the A2 scenario from AR4. For each of the two 2050 scenarios we use crop All simulations use standard IMPACT model assump- model results with 369 ppm CO2 to be the no-CO2 tions for elasticities and intrinsic productivity and area fertilization results and with 532 ppm CO2 to represent growth changes. Income elasticities decline with income CO2 fertilization results. growth. For population growth, we use the 2006 UN medium variant projections. For income growth, we rely Then we simulate agricultural productivity increases in on the estimates provided by the World Bank for this the developing world needed to address the malnour- study. All income and price values are in constant 2000 ished children goals. US dollars. Table 24. PReCIPITaTION aNd TemPeRaTuRe RegIONal aveRage ChaNgeS, 2000 TO 2050 GCM prec(mm) prec(%) tmin(C) tmax(C) East Asia and Pacific CSIRO 21.9 2.1 1.66 1.56 East Asia and Pacific NCAR 76.21 7.6 2.61 2.08 Europe and Central Asia CSIRO 26.21 6.1 1.82 1.67 Europe and Central Asia NCAR 56.14 13.2 4.35 3.65 Latin America and the Caribbean CSIRO ­8.36 ­0.6 1.57 1.62 Latin America and the Caribbean NCAR 28.39 1.9 2.03 1.91 Middle East and North Africa CSIRO ­2.36 ­2.0 1.65 1.56 Middle East and North Africa NCAR 26.96 22.1 2.8 2.54 South Asia CSIRO 14.51 1.6 1.79 1.64 South Asia NCAR 100.95 11.2 2.37 1.76 Sub-Saharan Africa CSIRO ­27.75 ­3.5 1.69 1.79 Sub-Saharan Africa NCAR 69.58 8.6 2.29 1.77 All Developing CSIRO 6.44 0.8 1.71 1.66 All Developing NCAR 56.85 7.5 3.08 2.58 World CSIRO 9.09 1.8 1.3 1.22 World NCAR 45.55 9.1 2.28 1.91 Source: Compiled by authors. D E V E L O P M E N T A N D C L I M AT E C H A N G E D I S C u S S I O N PA P E R S 45 FIguRe 23. exOgeNOuS PROduCTIvITy gROwTh RaTeS (% PeR yeaR) FOR SeleCTed CROPS aNd maNagemeNT TyPe Rainfed Wheat Irrigated Wheat 5 6 4 5 3 4 2 3 1 2 0 1 ­1 0 ­2 ­1 yr1015 yr1520 yr2025 yr2530 yr3035 yr3540 yr4045 yr4550 yi1015 yi1520 yi2025 yi2530 yi3035 yi3540 yi4045 yi4550 average std-ave std+ave min max average std-ave std+ave min max Rainfed Soybeans Irrigated Soybeans 4 5 3.5 4 3 2.5 3 2 1.5 2 1 1 0.5 0 0 ­0.5 ­1 ­1 ­1.5 ­2 yr1015 yr1520 yr2025 yr2530 yr3035 yr3540 yr4045 yr4550 yi1015 yi1520 yi2025 yi2530 yi3035 yi3540 yi4045 yi4550 average std-ave std+ave min max average std-ave std+ave min max Rainfed Maize Irrigated Maize 5 5 4 4 3 3 2 2 1 1 0 0 ­1 ­1 ­2 ­2 yr1015 yr1520 yr2025 yr2530 yr3035 yr3540 yr4045 yr4550 yi1015 yi1520 yi2025 yi2530 yi3035 yi3540 yi4045 yi4550 average std-ave std+ave min max average std-ave std+ave min max (Continuedonnextpage) 46 THE COSTS OF AGRICu LT uRAL ADAPTATION TO CLIMATE CH A N G E FIguRe 23. (continued) Rainfed Rice Irrigated Rice 4 4 3.5 3 3 2.5 2 2 1.5 1 1 0 0.5 0 ­1 ­0.5 ­1 ­2 ­1.5 yr1015 yr1520 yr2025 yr2530 yr3035 yr3540 yr4045 yr4550 yi1015 yi1520 yi2025 yi2530 yi3035 yi3540 yi4045 yi4550 average std-ave std+ave min max average std-ave std+ave min max Source: Food and Agriculture Organization (FAO). 47 5. ReFeReNCeS Resources. Water," In Resources and Irrigation Division Discussion Paper Arlington, VA: Winrock International. Alene, A.D. and O. Coulibaly. 2009. "The Impact of Kiani, A. K. ; M. Iqbal and T. Javed. 2008. "Total Factor Agricultural Research on Productivity and Poverty in Productivity and Agricultural Research Relationship: Sub-Saharan Africa." Food Policy, 34, pp. 198­209. Evidence from Crops Sub-Sector of Pakistan's Punjab." European Journal of Scientific Research, 23(1), pp. 87­97. Allen, R.G.; L.S. Pereira; D. Raes and M. Smith. 1998. "Crop Evapotranspiration Guidelines for Computing Long, Stephen P.; Elizabeth A. Ainsworth; Andrew D. Crop Water Requirements," In. Rome: FAO. B. Leakey; Josef Nosberger and Donald R. Ort. 2006. "Food for Thought: Lower-Than-Expected Crop Yield Fan, S.; P. Hazell and S. Thorat. 1998. "Government Stimulation with Rising Co2 Concentrations." Science, Spending, Growth and Poverty: An Analysis of 312(5782), pp. 1918­21. Interlinkages in Rural India." Environment and Production Technology Division Discussion Paper, 33. Rosegrant, M.W.; S. Msangi; C. Ringler; T.B. Sulser; T. Zhu; and S.A. Cline. 2008. "International Model for Haie, N. and A.A. Keller. 2008. "Effective Efficiency as Policy Analysis of Agricultural Commodities and Trade a Tool for Sustainable Water Resources Management." (Impact): Model Description." Washington, D.C.: JAWRA, 10, pp. 1752­688. International Food Policy Research Institute. IPCC; M.L. Parry; O.F. Canziani; J.P. Palutikof; P.J. van Schimmelpfenig, D. and C. Thirtle. 1999. "The der Linden and C.E. Hanson eds. Climate Change 2007: Internationalization of Agricultural Technology: Impacts, Adaptation and Vulnerability. Contribution of Patents, R&D Spillovers, and Their Effects on Working Group Ii to the Fourth Assessment Report of the Productivity in the European Union and United States." Intergovernmental Panel on Climate Change. Cambridge, Contemporary Economic Policy, 17(4), pp. 457­68. UK: Cambridge University Press, 2007. Shiklomanov, I. A. 1999. "Electronic Data Provided to Jones, J. W.; G. Hoogenboom; C. H. Porter; K. J. Boote; the Scenario Development Panel, World Commission W. D. Batchelor; L. A. Hunt; P. W. Wilkens; U. Singh; on Water for the 21st Century," In. A. J. Gijsman and J. T. Ritchie. 2003. "The Dssat Cropping System Model." European Journal of Smith, L. and L. Haddad. 2000. "Explaining Child Agronomy, 18(3-4), pp. 235­65. Malnutrition in Developing Countries: A Cross- Country Analysis," In IFPRI Research Report. Keller, A. and J. Keller. 1995. "Effective Efficiency: A Washington, D.C.: International Food Policy Research Water Use Concept for Allocating Freshwater Institute. 48 THE COSTS OF AGRICu LT uRAL ADAPTATION TO CLIMATE CH A N G E Thirtle, C. ; L. Lin and J. Piesse. 2003. "The Impact of Research-Led Agricultural Productivity Growth on Poverty Reduction in Africa, Asia and Latin America." World Development, 31(12), pp. 1959­75. You, Liang and Stanley Wood. 2006. "An Entropy Approach to Spatial Disaggregation of Agricultural Production." Agricultural Systems, 90(1-3), pp. 329­47. Zavala, Jorge A.; Clare L. Casteel; Evan H. DeLucia and May R. Berenbaum. 2008. "Anthropogenic Increase in Carbon Dioxide Compromises Plant Defense against Invasive Insects." Proceedings of the National Academy of Sciences, 105(13), pp. 5129­33. The World Bank Group 1818 H Street, NW Washington, D.C. 20433 USA Tel: 202-473-1000 Fax: 202-477-6391 Internet: www.worldbank.org/climatechange