Empirically estimated impacts of climate change on global crop production via increasing precipitation-evapotranspiration extremes

Raitzer, David A.; Drouard, Joeffrey Working Paper Empirically estimated impacts of climate change on global crop production via increasing precipitationevapotranspiration extremes ADB Economics Working Paper Series, No. 759 Provided in Cooperation with: Asian Development Bank (ADB), Manila Suggested Citation: Raitzer, David A.; Drouard, Joeffrey (2024) : Empirically estimated impacts of climate change on global crop production via increasing precipitation-evapotranspiration extremes, ADB Economics Working Paper Series, No. 759, Asian Development Bank (ADB), Manila, https://doi.org/10.22617/WPS240589-2 This Version is available at: https://hdl.handle.net/10419/310398 Standard-Nutzungsbedingungen: Die Dokumente auf EconStor dürfen zu eigenen wissenschaftlichen Zwecken und zum Privatgebrauch gespeichert und kopiert werden. Sie dürfen die Dokumente nicht für öffentliche oder kommerzielle Zwecke vervielfältigen, öffentlich ausstellen, öffentlich zugänglich machen, vertreiben oder anderweitig nutzen. Sofern die Verfasser die Dokumente unter Open-Content-Lizenzen (insbesondere CC-Lizenzen) zur Verfügung gestellt haben sollten, gelten abweichend von diesen Nutzungsbedingungen die in der dort genannten Lizenz gewährten Nutzungsrechte. Terms of use: Documents in EconStor may be saved and copied for your personal and scholarly purposes. You are not to copy documents for public or commercial purposes, to exhibit the documents publicly, to make them publicly available on the internet, or to distribute or otherwise use the documents in public. If the documents have been made available under an Open Content Licence (especially Creative Commons Licences), you may exercise further usage rights as specified in the indicated licence. https://creativecommons.org/licenses/by/3.0/igo/ ASIAN DEVELOPMENT BANK ASIAN DEVELOPMENT BANK 6 ADB Avenue, Mandaluyong City 1550 Metro Manila, Philippines www.adb.org EMPIRICALLY ESTIMATED IMPACTS OF CLIMATE CHANGE ON GLOBAL CROP PRODUCTION VIA INCREASING RECIPITATION– EVAPOTRANSPIRATION EXTREMES David A. Raitzer and Joeffrey Drouard ADB ECONOMICS WORKING PAPER SERIES NO. 759 December 2024 Empirically Estimated Impacts of Climate Change on Global Crop Production via Increasing Precipitation–Evapotranspiration Extremes To assess climate change effects on crop yields, remote sensing-derived yield and agrometeorological reanalysis data are used to construct a panel at 0.1-degree resolution for 2003–2015. Regressions controlling for grid cell-specific intercepts and time trends, temperature, rainfall, and cloudiness estimate the subregional relationships between yields and precipitation-evapotranspiration extremes for rice, wheat, and maize. Results imply that climate change will cause global yield reductions for all crops, with losses highest for wheat and maize, especially in South Asia and Southern Africa. About the Asian Development Bank ADB is committed to achieving a prosperous, inclusive, resilient, and sustainable Asia and the Pacific, while sustaining its efforts to eradicate extreme poverty. Established in 1966, it is owned by 69 members —49 from the region. Its main instruments for helping its developing member countries are policy dialogue, loans, equity investments, guarantees, grants, and technical assistance. ASIAN DEVELOPMENT BANK The ADB Economics Working Paper Series presents research in progress to elicit comments and encourage debate on development issues in Asia and the Pacific. The views expressed are those of the authors and do not necessarily reflect the views and policies of ADB or its Board of Governors or the governments they represent. ADB Economics Working Paper Series Empirically Estimated Impacts of Climate Change on Global Crop Production via Increasing Precipitation–Evapotranspiration Extremes David A. Raitzer and Joeffrey Drouard No. 759 | December 2024 David A. Raitzer (draitz[email protected]g) is a senior economist at the Economic Research and Development Impact Department, Asian Development Bank. Joeffrey Drouard ([email protected]) is an associate professor at the Université Côte d’Azur. Creative Commons Attribution 3.0 IGO license (CC BY 3.0 IGO) © 2024 Asian Development Bank 6 ADB Avenue, Mandaluyong City, 1550 Metro Manila, Philippines Tel +63 2 8632 4444; Fax +63 2 8636 2444 www.adb.org Some rights reserved. Published in 2024. ISSN 2313-6537 (print), 2313-6545 (PDF) Publication Stock No. WPS240589-2 DOI: http://dx.doi.org/10.22617/WPS240589-2 The views expressed in this publication are those of the authors and do not necessarily reflect the views and policies ofthe Asian Development Bank (ADB) or its Board of Governors or the governments they represent. ADB does not guarantee the accuracy of the data included in this publication and accepts no responsibility for any consequence of their use. The mention of specific companies or products of manufacturers does not imply that they are endorsed or recommended by ADB in preference to others of a similar nature that are not mentioned. By making any designation of or reference to a particular territory or geographic area inthis document, ADB does not intend to make any judgments as to the legal or other status of any territory or area. This publication is available under the Creative Commons Attribution 3.0 IGO license (CC BY 3.0 IGO) https://creativecommons.org/licenses/by/3.0/igo/. By using the content of this publication, you agree to be bound bytheterms of this license. For attribution, translations, adaptations, and permissions, please read the provisions andterms of use at https://www.adb.org/terms-use#openaccess. This CC license does not apply to non-ADB copyright materials in this publication. If the material is attributed toanother source, please contact the copyright owner or publisher of that source for permission to reproduce it. ADB cannot be held liable for any claims that arise as a result of your use of the material. Please contact [email protected] if you have questions or comments with respect to content, or if you wish toobtain copyright permission for your intended use that does not fall within these terms, or for permission to use theADB logo. Corrigenda to ADB publications may be found at http://www.adb.org/publications/corrigenda. ABSTRACT Although agriculture is often considered vulnerable to climate change, recent gridded crop growth modelling intercomparison exercises have found that staple crop yields will be modestly affected by global warming. However, those crop growth models also do not fully reflect impacts of increasing climate extremes. This paper uses global remote sensing-derived yield and agrometeorological reanalysis data to construct a grid cell panel at 0.1-degree resolution for 2003–2015. Regressions that control for grid cell-specific intercepts and time trends, temperature, rainfall, and cloudiness empirically estimate the relationship between yields and precipitationevapotranspiration extremes for each growing season of rice, wheat, and maize by subregion. Estimated coefficients are applied to projections from an ensemble of global circulation models to project potential losses from climate change. All crops are found as having substantial potential future global yield reductions, but reductions are highest for wheat and maize, with losses most pronounced in Southern Asia and Southern Africa. Keywords: agriculture, climate change, drought, yield loss JEL codes: Q12, Q15, Q54, D22 2 I. BACKGROUND Agriculture is often considered one of the most vulnerable sectors to the impacts of climate change. Crop production is directly dependent on environmental conditions, with water provided by rainfall and solar radiation through cloud cover as essential inputs, temperatures shaping heat and cold stress, as well as the pace of plant geochemistry and water use, and growing periods defined by seasonal transitions. Early attempts to model the impacts of climate change on agricultural production confirmed the vulnerability of staple crop production to global warming, especially in tropical and subtropical regions. For example, Hasegawa et al. (2022) compiled a database of 8,703 simulations from 202 studies published between 1984 and 2020 of crop growth modelling. Mean results imply yield losses of 14% for rice, 44% for wheat, and 24% for maize in India; and losses of 4% for rice and 10% for maize in the People’s Republic of China (PRC) under a high-end emissions scenario by 2050, relative to 2000. More recent modelling studies, however, find much more limited impact of pronounced climate change. The latest 3b round of the Intersectoral Impact Modelling Intercomparison Project (ISIMIP), which involved 13 crop growth models applied to five downscaled global circulation models, forcing datasets, and tens of thousands of simulations, finds mean yield gains under a high-end emissions scenario for rice and wheat in India and the PRC through late in the century and more modest losses for maize, as well as broader patterns of rice and wheat gains (Orlov et al. 2024). The positive and less damaging effects compared with previous exercises are driven, in part, by increased estimates of the effects of carbon dioxide (CO2) fertilization, or the increase in photosynthesis that results as CO2 concentrations rise in high end emissions scenarios. At the same time, the process-based crop growth models used in these exercises have limited ability to replicate the effects of extreme climatic events, such as drought or excessive rainfall. The models have typically been initially developed to ascertain yield potential, or the maximum yield possible under temperature and solar radiation constraints, without water or nutrient constraints imposed. They are typically further calibrated to replicate crop performance in yield trials under favorable conditions. For example, models used in ISIMIP3b were unable to explain more than 80% of yield variance in major Asian producers’ yields of rice, wheat, or maize and consistently overestimated production during drought events (Heinicke et al. 2022). Under climate change, there is general agreement among global circulation models that drought will increase in frequency and intensity (IPCC 2021). Even though in many regions rainfall will increase, as 3 warmer air can contain more water vapor, higher temperatures increase evaporation and evapotranspiration rates even more, so that conditions are often drier. In addition, climate change makes seasonal variations stronger and amplifies weather variability in many regions, which compounds the effects of higher temperatures. Higher rainfall is often through increased precipitation intensity in short periods, creating additional losses through water stagnation in fields. While many studies have shown that drought and inundation are damaging to crop yields, few studies have quantified the damage in manner that can be reconciled with climate projections to show global and regional yield loss under climate change. A number of studies have used the standardized precipitation index (e.g., Hendrawan, Komori, and Kim 2023; Vogel at al. 2019), but this index only captures precipitation rather than other determinants of drought related to evapotranspiration. Other prior studies either use country-level data (Matiu, Ankerst, and Menzel 2017), which give few and coarse observations, or data for a single country (e.g., Mohammed et al. 2022; Leng 2021), which reveal only country-specific relationships. Santini et al. (2022) perform somewhat similar analysis to this study by using global spatial yield data, but at 0.5degree resolution, which is 25 times coarser compared with the data used in the study, and with only dummy representation of extreme events. None of these studies use estimated coefficients with climate projections to show implications of future climate change. With the availability of global remote sensing-derived crop yield data at 0.1-degree resolution over successive years, along with climatic data for the same units and periods, it becomes possible to empirically estimate the effect of drought and inundation on yield using hundreds of thousands of observations that are much more broadly representative. The empirical estimates can then be applied to geospatial projections of drought and inundation under a high emissions climatic scenario to determine implications for crop production in specific regions. These effects of extreme climatic events can be considered as additional to the effects typically captured in global grided crop modelling exercises, which do not replicate extreme climatic events well and can potentially alter conclusions that climate change impacts on agriculture are modest or beneficial. 4 II. OBJECTIVES The purpose of this research is to quantify how the production of rice, wheat, and maize, globally and in specific regions, will be impacted by the increased frequency and intensity of climatic extremes under climate change using large, globally representative, spatial data spanning more than a decade. Such an approach has become possible because of the increasing availability of earth-observation based data for both agricultural production and meteorological conditions. The empirical approach to estimate the effects of extreme events on yields relies on panel data regression, with spatial grid cells as the panel unit. The approach employed uses fixed effects to control for time invariant differences in grid cells and variables that account for common trends over time. Moreover, the preferred specification accounts for grid cell-specific trends over time to eliminate effects of local-level time trends in both independent and dependent variables, so that the analysis is derived from exogenous anomalies. In addition to drought indicators, climatic variables that reflect less extreme changes are included, so as to control for the effects of variables that typically determine yields in gridded crop growth modelling exercises. Additional control variables, such as local CO2 variations, local air pollution, and irrigation are also tested to determine if they enhance model performance. Once coefficients for climatic extremes are empirically estimated, they are applied to projections of climatic extremes under climate changes to determine overall implications for crop production. III. METHODS A. Data A range of different data sources have been obtained and processed for use in the empirical analysis and future projections.1 This section begins by presenting the raw datasets and then explaining how they were aggregated and processed. All datasets have been obtained or harmonized to a resolution of 0.1-degree cells, which are approximately 11 kilometers (km) by 11 km at the equator, unless already noted. 1 The following sections describe the various datasets accessed on 23 July 2024, unless otherwise noted. 5 1. Raw Data2 Crop yield. Yield data are obtained from the Copernicus Climate Change Project (Wit et al. 2022). This dataset provides insights into the spatio-temporal variations of crop yield at 0.1-degree spatial resolution for different crop types. The data are from a 1-km grid cell fraction of absorbed photosynthetic active radiation (FAPAR) derived from satellite observations on a 10-day basis, interpreted via a crop-specific light use efficiency and temperature function for the reduction assimilation rate of photosynthesis, with growth apportioned between stems, leaves, and grains depending on the growth stage. As FAPAR is the major time-varying observed variable (crop parameters are constant) driving the estimates, water stress is reflected through reductions in leaf area that determine observed FAPAR. These yield data have been confirmed to replicate interannual variability well for areas dominated by the observed crops (Chevuru et al. 2023). This research focuses on rice, wheat, and maize yields for the period from 2003 to 2015. The crops were chosen for the reason that they are the most important staple food sources globally and account for more than half of all calories consumed by people. This reference period is chosen because data for all variables in the preferred specification and robustness checks are consistently available for this period. More specifically, rice crops often have two cropping seasons within a single campaign year, which is defined as the calendar year of the harvest. The data separate these two growing periods. In the same way, winter and spring wheat are treated as separate within the same campaign year. For each of these crops, data are provided on (i) crop development stage, as a time-series with dekadal (10-day) intervals, indicating whether the crop is in the emergence phase, flowering phase, or at physiological maturity; and (ii) the total weight of storage organs (TWSO), which represents crop yield. Meteorological. The Copernicus Climate Change Project enables access to agrometeorological indicators derived from reanalysis (AgERA5) at daily and 0.1-degree resolution for use in agroecological studies (Boogaard et al. 2020).3 Based on the hourly European Centre for MediumRange Weather Forecasts (ECMWF) “fifth generation reanalysis for the global climate and weather” (ERA5; Hersbach et al. 2020) data at surface level, the AgERA5 dataset is aggregated to daily time steps and interpolated to the 0.1-degree resolution using regression equations trained on the ECMWF “high resolution atmospheric model.” From the AgERA5 dataset, daily 2 Additional data used in robustness checks are described in Appendix 2. 3 C3S (Copernicus Climate Change Service) Climate Data Store. 2020. Agrometeorological Indicators from 1979 to Present Derived from Reanalysis (accessed 10 May 2024). 12 Table 3: Estimates of Climate and Weather Effects on Log Global Crop Yields Statistics Maize Winter Wheat Spring Wheat Wet Rice 1 Wet Rice 2 spei_6_mean 0.027*** 0.032*** 0.080*** -0.007*** 0.033*** (0.0003) (0.001) (0.001) (0.0003) (0.001) spei_6_mean2 -0.016*** -0.024*** -0.059*** -0.004*** -0.006*** (0.0003) (0.0004) (0.001) (0.0002) (0.0004) temp_max_mean 2.307*** 3.229*** 0.1 5.300*** 4.053*** (0.027) (0.058) (0.070) (0.055) (0.086) temp_max_mean2 -0.004*** -0.006*** -0.0002** -0.009*** -0.007*** (0.00005) (0.0001) (0.0001) (0.0001) (0.0001) precipitation_mean -0.006*** 0.153*** 0.066*** -0.013*** 0.046*** (0.0004) (0.002) (0.003) (0.0005) (0.004) precipitation_mean2 -0.00004* -0.022*** -0.004*** 0.00005** -0.005*** (0.00003) (0.0004) (0.0005) (0.00002) (0.0005) cloud_mean -0.172** -0.818*** -0.369* -0.577*** -0.759*** (0.068) (0.120) (0.199) (0.031) (0.056) Cell fixed effects Yes Yes Yes Yes Yes Cell linear time trends Yes Yes Yes Yes Yes Adjusted R squared 0.87 0.87 0.87 0.93 0.94 Observations 637511 267605 266093 343794 169091 SPEI = Standardized Precipitation–Evapotranspiration Index. Notes: This table presents the estimation results of Equation (1) for the crops at the head of each column. All regressions include constant, cell-fixed effects, and cell-specific linear time trends. Standard errors clustered by cell appear in parentheses. * p<0.10, ** p<0.05, *** p<0.01. The models are estimated using the Stata reghdfe package developed by Correia (2016). Source: Authors. Robustness of the findings. The preferred specification is superior to tested alternatives, including (i) a regression that omits temperature, precipitation, and cloud cover variables; (ii) a regression that uses time and unit dummies, but not time by unit interactions (as Equation 2 below); (iii) a regression that also controls for CO2 (at 5-degree resolution), irrigation, and air pollution;8 and (iv) a regression that also controls for irrigation. The results are presented in Appendix 2. Regression (i) achieves nearly the same R2 as the preferred specification, but has larger coefficients, suggesting that small differences in temperature and rainfall are also being captured by the SPEI-6 variables. This makes it difficult to consider as additional to gridded crop growth model results. Regression (ii) has a lower R2 than the preferred specification, but similar coefficients. Regression (iii) has implausible coefficients on CO2 (reverse of the expected sign from CO2 fertilization) and an R2 that is not improved. CO2 and pollution variables are, however, 8 See Appendix 2 for a detailed presentation of these variables. 13 observed with low granularity and temporal resolution.9 Regression (iv) also does not improve R2 by including irrigation. Notably, the results, particularly the effect of SPEI on crop yields, remain consistent whether year dummy variables are used instead of cell-specific linear time trends or regressions control for additional factors. ln TWSOc,t = X’c,t βi + αc0 + αt + εc,t (2) Results by subregion. Effects of climatic extremes are estimated on crop yield for different subregions. Table 4 presents the estimated coefficients from Equation (1), derived from regressions for maize, winter wheat, spring wheat, wet rice 1, and wet rice 2 across different subregions of the world. For simplicity, only the estimated coefficients associated with SPEI are reported: column (a) for the variable spei_6_mean and column (b) for the variable spei_6_mean2. However, all regressions include the variables temp_max_mean, temp_max_mean2, precipitation_mean, precipitation_mean2, and cloud_mean, as well as the fixed effects described above. The coefficients suggest substantial variation in drought vulnerability for crops cross subregions. For example, maize is more vulnerable to drought in Northern America than the Caribbean, and wheat is more vulnerable in Southern Asia than in Western Asia. B. Projection of Future Losses Under a High-End Emissions Scenario To translate the coefficients presented in Table 4 into effects on crop production under climate change, they are used with SPEI-6 projections and historical SPEI-6 observations to calculate effects on yield. These projections take into account both the subregion-specific relationship between yield and SPEI-610 and the grid cell-specific changes in SPEI-6 expected under climate change. First, for each grid cell, the historical average SPEI-6 value over 2003–2015 is calculated as a point of reference. Yield loss from reference SPEI-6 is then calculated using linear and squared term coefficients from Table 4. Yield losses under SPEI-6 projected values are subsequently calculated for 2030, 2050, and 2090, from which historical losses are netted, so as to give the effects of climate change on yield. Climate change effects on yield are then multiplied by average historical 2003–2015 production in each grid cell (area of each crop multiplied by the average 2003–2015 yield), summed by subregion, and divided by total historical 2003–2015 9 Pollution and irrigation are observed annually, while CO2 levels are measured at a resolution of 5 degrees. In the case of CO2 levels, which are observed with much less detail (i.e., have much larger cells) compared to the crop data, the crop data are matched to the CO2 data using the centroid of the crop data. Specifically, this matching occurs when the centroid of a crop grid cell falls within the area of a CO2 data cell. 10 Subregions lacking enough observations for coefficient estimation use global SPEI-6 coefficients from Table 3. 14 production in each subregion to give relative effects on subregional production. Similar results have been generated by the authors on a country basis and interpolated annually, but are not included in the paper for ease of presentation. The results show high yield vulnerability to the effects of pronounced climate change on precipitation-evapotranspiration extremes (Table 5). In contrast to gridded crop growth model results showing yield gains or modest reductions of only a few percent under a high-end emissions scenario, these results show reductions that exceed 50% for selected subregions and staples within the 21st century. Maize is the crop most vulnerable to changes under climate change, followed by wheat and rice. Losses are highest in tropical and subtropical regions, especially those that have the lowest food security, such as Southern Asia and Southern Africa. The rate of loss appears to accelerate with time, as the 40 or so years from the reference period to 2050 lead to losses that are generally far below additional losses from 2050 to 2090. 15 Table 4: Estimates of Effects of SPEI-6 on Crop Yields by Subregion Region Maize Winter Wheat Spring Wheat Wet Rice 1 Wet Rice 2 (a) (b) (a) (b) (a) (b) (a) (b) (a) (b) spei6_mean spei6_mean2 spei6_mean spei6_mean2 spei6_mean spei6_mean2 spei6_mean spei6_mean2 spei6_mean spei6_mean2 Australia and New Zealand 0.159 -0.035 (0.003)*** (0.001)*** Caribbean 0.000 0.003 (0.003) (0.003) Central America 0.045 -0.011 (0.003)*** (0.002)*** Central Asia 0.077 -0.020 0.116 0.026 (0.003)*** (0.002)*** (0.006)*** (0.007)*** Eastern Africa 0.011 -0.014 (0.002)*** (0.002)*** Eastern Asia -0.002 -0.020 0.025 -0.014 0.000 -0.005 -0.016 -0.012 0.000 -0.010 (0.001)*** (0.001)*** (0.001)*** (0.001)*** (0.004) (0.004) (0.000)*** (0.000)*** 0 (0.000)*** Eastern Europe 0.018 -0.002 0.013 -0.005 0.036 -0.017 (0.001)*** (0.001)* (0.001)*** (0.001)*** (0.003)*** (0.002)*** Middle Africa Northern Africa 0.115 0.001 0.011 -0.005 (0.004)*** (0.002) (0.002)*** (0.003)** Northern America 0.040 -0.019 0.126 -0.041 0.049 -0.037 0.024 -0.009 (0.001)*** (0.000)*** (0.004)*** (0.002)*** (0.002)*** (0.001)*** (0.002)*** (0.001)*** Northern Europe 0.071 -0.022 (0.003)*** (0.002)*** South America 0.003 -0.013 0.071 -0.065 0.020 0.000 (0.001)*** (0.001)*** (0.003)*** (0.001)*** (0.002)*** (0.001) South-eastern Asia 0.019 -0.010 -0.018 0.001 0.067 -0.007 (0.001)*** (0.001)*** (0.001)*** (0.000)*** (0.001)*** (0.001)*** Southern Africa -0.006 -0.023 (0.003)** (0.003)*** Southern Asia 0.024 -0.015 0.008 -0.031 0.042 -0.033 0.002 0.000 0.046 -0.003 (0.001)*** (0.001)*** (0.004)** (0.003)*** (0.001)*** (0.001)*** (0.001)** (0.000) (0.002)*** (0.002)* Southern Europe 0.030 0.001 0.045 -0.052 0.023 -0.027 (0.002)*** (0.001) (0.002)*** (0.002)*** (0.005)*** (0.003)*** Western Africa 0.011 -0.006 0.001 -0.003 (0.003)*** (0.001)*** (0.002) (0.002)* Western Asia 0.001 0.004 -0.013 -0.007 0.124 -0.088 (0.005) (0.004) (0.001)*** (0.001)*** (0.005)*** (0.004)*** Western Europe 0.025 0.001 (0.002)*** 0 SPEI = Standardized Precipitation–Evapotranspiration Index. Note: This table presents the estimation results of Equation (1) across different regions of the world for the crops at the head of each column. Only the results associated to the variables spei_6_mean (column (a)) and spei_6_mean2 (column (b)) are displayed. All regressions include constant, temp_max_mean, temp_max_mean2, precipitation_mean, precipitation_mean2, cloud_mean, as well as cell-fixed effects and cell-specific linear time trends. Standard errors clustered by cell appear in parentheses. * p<0.10, ** p<0.05, *** p<0.01. The models are estimated using the Stata reghdfe package developed by Correia (2016). The geographic regions follow the grouping system utilized by the United Nations Statistical Division. Source: Authors. 16 Table 5: Effects on Crop Production of Precipitation-Evapotranspiration Extremes Under a High-End Emissions Scenario Region Rice Production Wheat Production Maize Production 2030 2050 2090 2030 2050 2090 2030 2050 2090 Australia and New Zealand -9.2% -21.4% -49.8% Caribbean -4.0% -8.1% -19.2% 1.1% 3.2% 10.6% Central America -1.3% -4.8% -16.6% -3.7% -8.9% -26.3% -11.0% -21.8% -46.9% Central Asia -0.6% -1.3% -3.8% -3.5% -8.9% -23.8% Eastern Africa -3.5% -10.7% -33.4% -0.8% -3.4% -14.0% -4.6% -13.4% -40.5% Eastern Asia -6.0% -13.2% -32.5% -7.5% -16.4% -37.0% -3.3% -10.3% -33.7% Eastern Europe -2.6% -4.4% -9.7% Middle Africa -5.3% -12.2% -28.6% -8.6% -22.0% -54.4% Northern Africa -5.8% -12.2% -30.0% -11.9% -19.6% -36.6% -2.4% -7.1% -23.5% Northern America -4.8% -11.1% -30.3% -4.0% -9.1% -23.3% -7.9% -17.6% -43.6% Northern Europe -4.8% -15.9% -46.4% South America -2.9% -5.2% -10.0% -1.4% -5.6% -18.0% -1.9% -7.7% -29.1% South-eastern Asia 4.3% 8.0% 17.5% -6.4% -14.5% -34.3% Southern Africa -10.4% -27.7% -65.0% -1.6% -10.0% -41.0% Southern Asia -0.2% -0.2% 0.3% -6.1% -23.7% -65.2% -4.2% -14.7% -43.6% Southern Europe 0.5% -0.4% -5.1% -2.5% -7.3% -27.1% -4.0% -6.2% -10.8% Western Africa -1.5% -4.8% -14.7% -4.2% -10.4% -27.2% Western Asia -13.9% -35.5% -73.4% 0.3% 1.3% 5.8% Western Europe -2.1% -7.2% -27.5% -2.7% -4.3% -7.7% Note: This table presents the difference in yields predicted by coefficients presented in Table 4 applied to ensemble mean values of SPEI-6 under RCP8.5 generated by 25 general circulation models in each future period from yields predicted for the mean 2003–2015 reference period SPEI-6, divided by yields predicted for the mean 2003–2015 reference period. The geographic regions follow the grouping system utilized by the United Nations Statistical Division. Source: Authors. The patterns of results are generally consistent with previous findings on the effects of drought and extreme climatic events. Matiu, Ankerst, and Menzel (2017); Hendrawan et al. (2022); and Santini et al. (2022) also empirically find limited impact of drought on rice, compared with wheat and maize. Here, similarly, rice is positively impacted in South-eastern Asia and hardly impacted in Southern Asia, which are two core production breadbaskets. The unique puddled production system for rice, which creates a soil hardpan that reduces percolative water losses; the use of transplanting, which enables adaptation to delayed onset of rains; and the production in intensive wet seasons, may explain the relatively lower vulnerability of rice to drought. In contrast, wheat is negatively impacted in the top producing regions of Eastern Asia and Southern Asia, with impacts especially large in the latter. Large impacts in Eastern Asia are consistent with findings of Yao et al. (2022), while large impacts in Southern Asia are consistent with findings of Kumari et al. (2023). 17 V. CONCLUSION It long has been conventional wisdom that agriculture is directly vulnerable to the effects of climate change, especially large levels of change under a high emissions scenario. This research empirically affirms this perspective by taking into account both the subregion-specific relationship between climate extremes and yields and the location-specific predictions of climate extremes under a high-end emissions scenario. In so doing, the regressions find that a quadratic specification of the climate extreme indicator, along with grid cell trends and intercepts, explains a vast majority of yields observed by remote sensing. The research is the first study to estimate yield effects of climate extremes in a manner that is applied to climate projections to show future yield loss globally. The study’s findings of large yield losses under climate change across all three cereals stand in stark contrast to those of recent gridded crop growth modelling studies that yield changes are positive for wheat and rice and modestly negative for maize because of the overwhelming effects of CO2 fertilization. An attempt to include the effects of CO2 concentration anomalies finds negative coefficients, rather than positive coefficients consistent with CO2 fertilization expectations. This may be an artefact of reverse causality, as faster growing crops may deplete CO2 faster during photosynthesis, creating CO2 reduction anomalies. It may be noted that the ISIMIP3b yield gains under a high-end emissions scenario are mostly below 10% for rice and below 20% for wheat by 2070 in all regions, which is generally a fraction of the magnitude of losses from this study. As the magnitude of precipitation-evapotranspiration extreme induced losses is far larger than the magnitude of gains found by ISIMIP3b from changes to temperature, rainfall, and CO2, they imply that crop production is still likely to experience strongly negative effects of climate change. It should be noted that the analysis here is of effects of the changes to the mean of an index of extremes. Effects in future extreme years will be far more pronounced, as what is considered now extreme will be typical in many locations. 18 APPENDIXES Appendix 1 The results presented in Table 3 are illustrated graphically in Figures A1.1, A1.2 and A1.3. These figures illustrate the marginal effects of the Standardized Precipitation–Evapotranspiration Index (SPEI), temperature and precipitation on crop yield, using the minimum and maximum observed values of these variables (as shown in Table 2) as the relevant lower and upper bounds for analyzing these effects. Figure A1.1: Marginal Effects of SPEI on Crop Yields for Various Crops SPEI = Standardized Precipitation–Evapotranspiration Index. Source: Authors. 19 Figure A1.2: Marginal Effects of Temperature (° Kelvin) on Crop Yields for Various Crops Source: Authors. 20 Figure A1.3: Marginal Effects of Precipitation (Millimeters) on Crop Yields for Various Crops Source: Authors. 21 Appendix 2 Equation (2) is estimated considering year dummy variables instead of cell-specific linear time trends, along with cell-fixed effects. Furthermore, this incorporates additional explanatory factors: pollution, irrigation, and carbon dioxide (CO2) levels. • Pollution data come from the NASA’s Socioeconomic Data and Applications Center (Hammer et al. 2022). This platform offers global annual PM2.5 microgram per cubic meter concentration grids at a spatial resolution of 0.01 degrees (Hammer et al. 2020; Hammer et al. 2022). • The irrigation dataset provided by Nagaraj et al. (2021) is included. The data are in the form of annual grids at 5 arc-minute spatial resolution and take values between 0 and 2, where 0 represents no irrigation, 1 is low to medium irrigation, 2 is high irrigation. • CO2 data comes from the Copernicus Climate Change Project. The column-average dryair mole fraction of atmospheric carbon dioxide (XCO2) grids are applied, which are monthly parts per million concentration data at a spatial resolution of 5 degrees. More specifically, the average irrigation (irrigation_mean), the logarithm of the average XCO2 (lnxco2_mean), and the average PM2.5 pollution (pollution_mean) during the growing phase are used. The results are presented in Tables A2.1 through A2.5 for maize, winter wheat, spring wheat, wet rice 1, and wet rice 2, respectively. Columns [1] and [2] show the results using year dummy variables instead of cell-specific linear time trends. Columns [3] and [4] present the results with additional explanatory variables included. Column [5] displays the results from the primary specification used in the main body of the paper for comparison. 28 Kumari, M., A. Chakraborty, V. Chakravarathi, and P. S. Roy. 2023. Spatio-Temporal Trend of Crop Phenology, SPEI, and Their Interactions over Different Agro-Ecological Regions of India. Theoretical and Applied Climatology 154. Leng, G. 2021. Maize Yield Loss Risk Under Droughts in Observations and Crop Models in the United States. Environmental Research Letters 16. Matiu, M., D. P. Ankerst, and A. Menzel. 2017. Interactions Between Temperature and Drought in Global and Regional Crop Yield Variability During 1961-2014. PLoS ONE 12. Mohammed, S. et al. 2022. Assessing the Impacts of Agricultural Drought (SPI/SPEI) on Maize and Wheat Yields Across Hungary. Scientific Reports 12. Monfreda, C., N. Ramankutty, and J.A. Foley. 2008. Farming the planet: 2. Geographic distribution of crop areas, yields, physiological types, and net primary production in the year 2000, Global Biogeochem. Cycles, 22, GB1022, doi:10.1029/2007GB002947. Nagaraj, D. et al. 2021. A New Dataset of Global Irrigation Areas from 2001 to 2015. Orlov, A. et al. 2024. Human Heat Stress Could Offset Potential Economic Benefits of CO2 Fertilization in Crop Production Under a High end emissions Scenario. One Earth 7. Santini, M., S. Noce, M. Antonelli, and L. Caporaso. 2022. Complex Drought Patterns Robustly Explain Global Yield Loss for Major Crops. Scientific Reports 12. Tang, F. H. M. et al. 2023. CROPGRIDS. figshare. Dataset (accessed 15 May 2024). Vogel, E. et al. 2019. The Effects of Climate Extremes on Global Agricultural Yields. Environmental Research Letters 14. Wit, A. et al. 2022. Crop Productivity and Evapotranspiration Indicators from 2000 to Present Derived from Satellite Observations. Copernicus Climate Change Service (C3S) Climate Data Store (CDS) (accessed 15 May 2024). Yao, N. et al. 2022. Response of Wheat and Maize Growth-Yields to Meteorological and Agricultural Droughts Based on Standardized Precipitation Evapotranspiration Indexes and Soil Moisture Deficit Indexes. Agricultural Water Management 266. ASIAN DEVELOPMENT BANK ASIAN DEVELOPMENT BANK 6 ADB Avenue, Mandaluyong City 1550 Metro Manila, Philippines www.adb.org EMPIRICALLY ESTIMATED IMPACTS OF CLIMATE CHANGE ON GLOBAL CROP PRODUCTION VIA INCREASING RECIPITATION– EVAPOTRANSPIRATION EXTREMES David A. Raitzer and Joeffrey Drouard ADB ECONOMICS WORKING PAPER SERIES NO. 759 December 2024 Empirically Estimated Impacts of Climate Change on Global Crop Production via Increasing Precipitation–Evapotranspiration Extremes To assess climate change effects on crop yields, remote sensing-derived yield and agrometeorological reanalysis data are used to construct a panel at 0.1-degree resolution for 2003–2015. Regressions controlling for grid cell-specific intercepts and time trends, temperature, rainfall, and cloudiness estimate the subregional relationships between yields and precipitation-evapotranspiration extremes for rice, wheat, and maize. Results imply that climate change will cause global yield reductions for all crops, with losses highest for wheat and maize, especially in South Asia and Southern Africa. About the Asian Development Bank ADB is committed to achieving a prosperous, inclusive, resilient, and sustainable Asia and the Pacific, while sustaining its efforts to eradicate extreme poverty. Established in 1966, it is owned by 69 members —49 from the region. Its main instruments for helping its developing member countries are policy dialogue, loans, equity investments, guarantees, grants, and technical assistance.