Global Crop Failure and Associated Changes in Agroclimate Conditions

Shifts of local-scale climate in crop production regions may threaten global food security. While consensus results have been reached regarding the impacts of mean climate on crop yields over different crop production regions, little is known about how agroclimate conditions, based on daily weather, are changing and how these changes impact crop failure risks. In this study, we analyze variability and trends of crop failures at high resolution in eleven different crop production regions over the globe from 1982 to 2016 for four staple crops, maize, rice, soya, and wheat, as well as their linkages with agriculture-significant climate conditions using a machine learning model. We derive twelve agroclimate indices as indicators of agriculture-significant climate conditions for each crop growing season over global croplands. Crop failure series were derived from satellite-based crop yield data at 50-km over the globe. The results show on average the eleven regions contribute over 50% of global crop failure events. Drought events globally can be described by low growing season total precipitation (TP), a high number of dry days (DD), and above average number of heat stress days (HS). To different extents, agroclimate indices can well predict and explain crop failure events over the past decades depending on regions and crop types. Overall, at the global scale, TP is the most important agroclimate index influencing crop failure risks nonlinearly in maize, rice, soybeans, and rice the probability of crop failure increases in extreme dry or wet conditions. Similarly growing degree days (GDD), DD, and field conditions mid-season influences crop failure nonlinearly when the ideal threshold is not met or surpassed, while accumulating HS increases crop failure probabilities. Significant positive and negative TP trends exist in 4% to 52% of crop-specific cropland. Important variables GDD and HS have the largest extent of significant increasing trends ranging from 12% to 77% and 2% to 73%, respectively, likely increasing crop failure risks in the future. Regions having last spring frost or first fall frost as the most important variables may decrease crop failure due to earlier and later frost dates. The findings of this study will be useful for informing the global climate adaptation strategies for agriculture and food security.