Nonlinear water stress response functions can improve the performance of the DSSAT-CERES-Wheat model under water deficit conditions

Crop growth simulation models are valuable for understanding and managing agro-ecological systems, especially in arid regions. The DSSAT-CERES-Wheat model is widely used to simulate wheat growth and development, but its accuracy diminishes under water stress conditions. This study evaluated the effects of different types of water stress response functions on the CERES-Wheat model, focusing on unit grain weight, biomass, and yield to improve the model's accuracy under water stress conditions. Our findings have disclosed that the calibration process demonstrated average Relative Mean Absolute Error (RMAE) and Relative Root Mean Square Error (RRMSE) values hovering at 5 %, whereas the evaluation process values surpassed 15 %. It was found that water stress occurring before the jointing stage significantly influenced model simulations of biomass and grain yield. This study assesses the performance of modified water stress response functions stepwise. Optimal parameters for these functions were identified, with the smallest RMAE and RRMSE observed at specific parameter values for different curve types. Comparative analyses using the modified CERES-Wheat model with six water stress response curves (WSRF) demonstrated improved WSRF 1, 2, 3, and 6 performance over the default function. The Penman-Monteith method for estimating potential evapotranspiration (E0) outperformed the Priestley-Taylor method, particularly in simulating unit grain weight, aboveground biomass, and grain yield. Convex functions (WSRF 1–3) were superior to concave functions (WSRF 4–6), enhancing overall simulation accuracy by 13.7 %. Furthermore, time-series output variables, including soil water and biomass dynamics, were evaluated. The original model underestimated early soil moisture and overestimated soil water stress, leading to significant biomass simulation errors. The modified model showed substantial improvements, particularly under low irrigation. However, simulation errors persisted under severe early-stage water stress. Validation results using DSSAT v4.7 databases confirmed the enhanced performance of the modified model under water stress conditions. The finding of this study proved that the improved model for the vast arid regions of China and similar environments globally can offer valuable insights and improve the accuracy of simulations, aiding in more effective agricultural management and planning under water-limited conditions.