Improving wheat yield simulation by integrating UAV-based multispectral data in the AquaCrop model for crop seasons with varying temperature and rainfall patterns

Crop yield simulation models are crucial for decision-making and resource optimization and for guiding small-scale farm adaptations toward sustainable food production. Overlooked in the existing literature, this study assesses the robustness of the AquaCrop model in predicting wheat yield for three crop seasons where temperature and rainfall patterns varied significantly. For this purpose, controlled experiments were conducted on winter wheat crops using both drip and flood irrigation treatments. Furthermore, it compares biomass and grain yield (GY) simulations using canopy cover (CC) curves derived by a calibration and validation approach with those based on UAV (Unmanned Aerial Vehicle)-derived CC curves for the third season, rarely explored in previous studies. Based on the calibration and validation approach, the coefficient of determination (R2), Nash–Sutcliffe coefficient (NSE), and root mean square (RMSE) between observed and simulated biomass ranged from 0.91 to 0.94, 0.85 to 0.94, and 0.29 to 0.33 t/ha, respectively, across the seasons. Similarly, for GY, these values ranged from 0.89 to 0.95, 0.87 to 0.94, and 0.26 to 0.31 t/ha, respectively. Model performance evaluation indicated that using the traditional calibration and validation approach results in significant errors, and recalibrating the CC curve in seasons having variability in temperature and rainfall patterns is indispensable for improving biomass and yield predictions. In the third season, UAV-based multispectral data were used to precisely derive the CC curve that was used as input to the AquaCrop model. As compared to the recalibration of the CC curve approach, the UAV-derived CC curve improved the AquaCrop simulation of biomass (R² = 0.97, NSE=0.98, RMSE=0.21 t/ha) and GY (R² = 0.98, NSE=0.97, RMSE=0.17 t/ha). The use of the UAV-derived CC curve is recommended over the cumbersome trial-anderror recalibration approach to improve the AquaCrop yield simulations in crop seasons in which there is notable variability in temperature and rainfall patterns.