A global database of the projected effects of climate change on four major crops

Data collection

As shown in Figure. 1We used two methods to create this dataset: The first method is based in part on Aggarwal’s meta-analysis. et al.⁵, which includes studies published before 2016. (Aggarwal DS, below). This meta-analysis builds on data from the 5^ThAssessment report of the IPCC^4,14You can also search through three other types of databases: Scientific database (Scopus. Web of Science. CAB Direct. JSTOR. Agricola etc), journals. open access repositories and institutional Websites (FAO Database, AgMIP Database. World Bank etc.). Google Scholar. Aggarwal et al.⁵ for details. Aggarwal’s search terms are briefly summarized below et al.⁵ include “agriculture” or “crop “or “farm” or “crop yield” or “crop yields” or “farm yields” or “crop productivity” or “agricultural productivity” or “maize” or “rice” or “wheat” and “climate change assessment” or “climate impacts” or “impact assessments” or “climate change impact” or “climate impact” or “effect of climate” or “impact of climate change”. There are 166 papers that cover the four major crops. We also screened them based on the availability of local temperature rise information and geographic information. This resulted in 99 studies published between 2014 and 2016.

Diagram showing paper collection and selection using both search strategies. N is the number or studies.

The second method uses a Scopus-based literature review from March 2020 that covers four major crops (maize rice, soybean and wheat). It includes peer-review papers that were published in 2014 in line with the sixth IPCC assessment cycle. This method used a variety of terms to retrieve relevant studies that report simulations of climate change’s impact on crop yields.

For maize, the following search equation was used: PUBYEAR > 2013 AND TITLE-ABS-KEY((maize OR corn) AND ((“greenhouse gas” OR “global warming” OR “climate change” OR “climate variability” OR “climate warming”)) AND NOT (emissions OR mitigation OR REDD OR MRV)).

Similar search formulas were used to find the other crops. The search returned 4703 references from 2014 to 2020. This includes 1899 for maize, 1790 wheat, 757 rice, and 257 soybean. Some duplications were possible because some papers examined multiple crops. The number of studies reduced to 3816 after removing duplicates.

To collect climate-scenario-based simulations, we then selected a subset of studies including the following terms related to climate scenarios in titles, abstracts, or authors’ keywords; “RCP”, “RCP2.6”, “RCP6.0”, “RCP4.5”, “RCP8.5”, “CMIP5”, and “CMIP6”. RCP stands for Representative Concentration Pathways¹⁵Each RCP corresponds a greenhouse-gas concentration trajectory that describes future greenhouse gas emission levels. The number following RCP is the level radiative forcing (Wm).⁻²) reached at the end of the 21^stCentury, which increases with the amount of greenhouse gas emitted into the atmosphere¹⁶. CMIP5¹⁷CMIP6¹⁸Phase 5 and Phase 6 of the Coupled Model Intercomparison Project Phase 5 respectively, where different earth system models (ESMs), provide global-scale climate projections that are based on different RCPs. Additionally, “process-based model” was used to search in the authors’ keywords to select for studies that use crop simulation models under CMIP5 or CMIP6 climate scenarios. As of March 2020 there were no results for CMIP6 in any of the search results.

The screening process yielded 207 references for four major crops. These studies were further screened for data availability and variables. Studies that did not report yield data were excluded. The projected yields of adaptations and the baseline period’s yields were extracted along with geographic coordinates, species and greenhouse gas emission scenarios and adaptation options. We also tried to find local and global changes in temperature and CO.₂Concentrations as high as possible. We also obtained data from the literature. We contacted several grid simulation authors to obtain aggregated results for specific countries or regions. The three grid simulation study authors responded with baseline and projected yields, annual temperatures and precipitation data aggregated for countries or regions.^19,20,21. The results of different ESMs were then averaged.

We removed duplicates from the two datasets and finally obtained 202 unique studies. Both datasets contain studies at different spatial scales: site-based and regional, as well as global. Among these, the results from the global gridded crop models were aggregated to country levels, and we focused on top-producing countries, which account for 95% of the world’s production of each commodity as of 2010 (FAOSTAT, http://www.fao.org/faostat/en/Retrieved September 4, 2020. The dataset contains 8,703 sets yield projections for the 21.^stCentury from studies published between 1984 & 2020 (Online-only Table 1).

Impacts on relative yield

To calculate the impact of climate change (YI) on yield (Simulated grain mass per unit area):