Calibration and validation of the AquaCrop model for simulating cotton growth under a semi-arid climate in Uzbekistan

Cotton is a crucial fiber crop, but its conventional production methods are heavily water intensive. In regions where water availability already limits cotton yields, there is a growing need to explore alternative field management practices that stabilize yields while reducing irrigation demands. Crop models, such as the AquaCrop model, are instrumental in these efforts, enabling simulations of the complex interactions between field management, water dynamics, crop growth, and yield. However, the variability in calibrated parameter values reported across AquaCrop studies for cotton raises concerns about the transferability and reliability of previously calibrated models. In this study, we calibrated and validated the AquaCrop crop model, developed by the Food and Agriculture Organization (FAO), to predict canopy growth, biomass accumulation and yield of cotton. The calibration protocol developed here is rather conservative, adhering strictly to the guidelines provided in the AquaCrop documentation. The calibration involved approximately 1500,000 simulations per treatment, employing a Monte Carlo (MC) protocol to systematically assess the effects of varying input parameters across multiple evaluation criteria, including their impact on water stress. The calibrated AquaCrop model delivered good to acceptable performance levels in simulating canopy growth, biomass accumulation, and yield under various irrigation treatments, comparable to previous AquaCrop cotton applications. Additionally, the MC protocol uncovered a previously undiscover bug in the model, which shifted the crop's planting date by approximately two weeks without user awareness, when the minimum rooting depth is set below 0.18 m. Furthermore, the rigorous calibration protocol clearly depicted compensatory interactions between parameters, where changes to one parameter can be offset by adjustments to another, highlighting the subjectivity and limitations encompassed in trial-and-error calibration approaches.