Optimization of the irrigation scheduling for underground drip-irrigated walnuts in Southern Xinjiang of China based on the AquaCrop model

To investigate the optimal subsurface drip irrigation regime for walnut in southern Xinjiang, field experiments were conducted during 2023 and 2024 in the Aksu region, Xinjiang of China. The study focused on 16-year-old "Wen 185" walnut trees with four irrigation quota treatments: 75%ET_c (crop evapotranspiration), 100%ET_c, 125%ET_c, and 150%ET_c. An optimization model for the walnut irrigation regime was developed based on the AquaCrop model and non-dominated sorting genetic algorithm II (NSGA-II) algorithm, aiming to maximize yield and minimize irrigation amount. Using the technique for order preference by similarity to an ideal solution (TOPSIS) method, the optimal irrigation regime for subsurface drip-irrigated walnuts in the southern Xinjiang was evaluated. Results showed that moderately increasing the irrigation quota could enhance walnut yield. However, the water use efficiency and irrigation water use efficiency decreased with higher irrigation quotas. The AquaCrop model's coefficient of determination (R²) for canopy coverage over two years was not less than 0.94, with root mean square error (RMSE) and normalized RMSE (NRMSE) ranging from 3.01% to 8.52% and 5.02% to 17.71%, respectively. Nash-Sutcliffe efficiency (EF) and goodness-of-fit index ranged from 0.57 to 0.89 and 0.90 to 0.99, respectively. For soil moisture content, R² was not less than 0.82, with RMSE and NRMSE varying from 9.02% to 21.51% and 4.31% to 9.81%, respectively, and EF and goodness-of-fit index ranging from 0.41 to 0.79 and 0.89 to 0.95, respectively. Yield simulation values versus actual measurements had R² of 0.89-0.95, RMSE of 114.57-178.73 kg/hm², NRMSE of 0.04%-0.05%, EF of 0.14-0.48, and d, 0.84 to 0.88. AquaCrop simulation results for different irrigation scenarios showed that scenario T9, with an irrigation quota of 6 500 m³/hm² and an irrigation interval of 7 days, yielded the highest production at 4 080.34 m³/hm². The dual-objective optimization model based on AquaCrop-NSGA-II revealed that scenario S7, with an irrigation interval of 11 days and a total irrigation amount of 3 520.22 m³/hm², produced a yield of 4 021.75 kg/hm². Compared to scenario T9, which had the highest yield in the AquaCrop model, scenario S7 resulted in a yield decrease of 1.44%, but saved 2 979.78 m³/hm² of water. It is recommended that the subsurface drip irrigation regime for walnut in the southern Xinjiang adopts an irrigation interval of 11 d, with 10 irrigation events and an irrigation quota of 3 520.22 m³/hm². The approach integrating AquaCrop model simulation and multi-objective optimization algorithm developed in this study can be used to optimize the subsurface drip irrigation regime for walnut in the southern Xinjiang of China.