Functional diversity among crop models reveals contrasting resilience of soybean and maize to heatwaves and CO₂ enrichment

Abstract Sustaining food production under increasing heat extremes requires models that accurately capture crop responses to multiple interacting stresses. We compared three process-based crop models—WOFOST, BioCro, and DSSAT (CROPGRO-Soybean, CERES-Maize)—to assess how structural differences influence simulated resilience of irrigated soybean ( Glycine max ) and maize ( Zea mays ) in the U.S. Midwest. Scenarios included uniform warming, elevated atmospheric CO₂, and reproductive-stage heatwaves. All models reproduced historical yields, yet projected sharply different sensitivities. WOFOST showed strong CO₂ fertilization responses but moderate warming effects; DSSAT produced steep heat-induced declines linked to explicit grain-fill temperature functions; BioCro displayed intermediate behavior driven by biochemical canopy photosynthesis. Across models, maize yields declined by 25–46% at + 4°C, while soybean losses (8–25%) were partially offset by CO₂ gains (+ 18 to + 44%). Short heatwaves reduced yields by 1–5%, especially during flowering and seed filling. These consistent structural “fingerprints” demonstrate that model divergence reflects underlying physiological formulations rather than random uncertainty. Advancing functionally transparent and physiologically consistent crop models will improve the credibility of climate impact assessments and support sustainable intensification and adaptation planning in a warming world.