Intensification and Spatial Reorganization of Drought Risks for Global Rainfed Crops Under Climate Change

ABSTRACT Rainfed agriculture covers roughly 80% of global cropland and supplies more than half of global food production, yet uncertainty in how meteorological drying propagates into root‐zone water deficits and crop losses impedes robust risk assessment. Here, we establish an integrated, production‐weighted assessment framework coupling SPEI, SSI, and 12 GGCMI3 crop models to trace how meteorological drying propagates into root‐zone deficits and crop losses, explicitly separating affected area, conditional yield loss, and accumulated production loss across rainfed maize, soybean, and wheat. By 2070–2099, opposing‐direction SPEI–SSI change classes occupied 6.8%–28.4% of rainfed area across crops under SSP1‐2.6/default CO 2 and 26.6%–48.2% under SSP5‐8.5/default CO 2 . Comparison with fixed‐2015 CO 2 experiments (7.5%–27.6% and 21.7%–34.9%, respectively) indicates that elevated CO 2 substantially expanded the spatial footprint of this hydrometeorological decoupling. Historically, SSI‐defined drought affected 12–16 million ha year −1 and was associated with annual production losses of 5–12 Mt year −1 (1.6–2.0% year −1 ). Under default CO 2 , affected area increased to 14–21 million ha under SSP1‐2.6 and 19–27 million ha under SSP5‐8.5, with production losses of 6–12 and 9–12 Mt year −1 , respectively. Under SSP5‐8.5, median loss rates were higher under fixed‐2015 than default CO 2 for maize (3.33 vs. 1.96% year −1 ), soybean (4.22 vs. 2.84% year −1 ) and wheat (5.41 vs. 2.90% year −1 ), as elevated CO 2 reduces stomatal water loss and thus mitigates drought impacts, whereas SSP1‐2.6 estimates were similar. Severe drought became the largest loss contributor under both pathways and CO 2 settings. Under default CO 2 , loss increases clustered in Central America, northeastern Brazil, western Europe and southern Africa; fixed‐2015 CO 2 broadened the area of robust grid‐cell loss increases. These results show that drought‐related production losses are spatially uneven and sensitive to the modeled CO 2 response, highlighting the value of crop‐ and region‐specific risk assessment.