Uncertainties in simulating wheat growth responses to waterlogging

Waterlogging is an increasingly important stress in wheat growing regions, yet its representation in process based crop models (CMs) remains limited. We used a standardized AgMIP experiment to diagnose how 15 wheat CMs differ in their response to waterlogging timing, duration, and rainfall driven soil saturation. Before fitting waterlogging specific crop response coefficients to the eight waterlogged Lleida treatments in the current exercise, the ensemble median reproduced these treatments with an rRMSE of 17.0%. We interpret this as a pre fit extrapolative benchmark rather than strict independent validation because CMs retained their existing formulations and parameterizations and one CM reported prior calibration of soil water variables. Fitting waterlogging specific parameters to the same dataset reduced rRMSE to 6.3%, which is calibration goodness of fit rather than an independent test. Large differences among CMs remained when the calibrated parameters were transferred across phenological stages. Stem elongation was the single most frequently identified stage of maximum yield sensitivity, occurring in 4 of 15 CMs, although maxima occurred across all seven stages. ... mehrMean inter model variability in relative grain yield, expressed as the coefficient of variation across CMs, increased from 26.7% under 4 day waterlogging to 41.4% under 24 day waterlogging. A family balanced ensemble that first averaged related MONICA and SIMPLACE versions produced nearly identical mean responses (r = 0.998; maximum difference 2.8 %age points), indicating that related model versions did not drive the ensemble pattern. Most CMs showed weak yield responses to synthetic rainfall perturbations, including very large events, but these scenarios lack matched observations and therefore diagnose sensitivity rather than model bias. An exploratory structure response analysis found no single directly affected crop process that separated the severe stress response typologies. The strongest result is the large structural disagreement among current wheat CMs under standardized waterlogging perturbations, pointing to the combined importance of hydrological stress generation, crop process representation, parameterization, and development dependent damage and recovery for future model improvement.