Assimilating Satellite-Based Biophysical Variables Data into AquaCrop Model for Silage Maize Yield Estimation Using Water Cycle Algorithm

Accurate crop yield estimation is critical to successful agricultural operations. Current crop growth models often overlook the spatial and geographic components of the lands, leading to suboptimal yield estimates. To address this issue, assimilation of satellite vegetation products into these models can account for spatial variations in the land and improve estimation accuracy. In this paper, the AquaCrop model, a water-driven crop growth model, was selected for recalibration and assimilation of satellite-derived biophysical products due to its simplicity and lack of computational complexity. To this end, field samples of soil (sampled before cultivation) and crop features were collected during the growing season of silage maize. Digital hemisphere photography (DHP) and destructive sampling methods were used for measuring fraction vegetation cover (fCover) and biomass in Qaleh-Now County, southern Tehran, in 2019. Based on our proposed workflow in previous studies, a Gaussian process regression–particle swarm optimization (GPR-PSO) algorithm and global sensitivity analysis were applied to retrieve the fCover and biomass from Sentinel-2 satellite data and to identify the most sensitive parameters in the AquaCrop model, respectively. Here, we propose the use of an optimization water cycle algorithm (WCA) instead of a PSO algorithm as an assimilation method for the parameter calibration of AquaCrop. This study also focused on using both fCover and biomass state variables simultaneously in the model, as opposed to only the fCover, and found that using both variables led to significantly higher calibration accuracy. The WCA method outperformed the PSO method in AquaCrop’s calibration, leading to more accurate results on maize yield estimates. It has enhanced results, decreasing RMSE values by 3.8 and 4.7 ton/ha, RRMSE by 6.4% and 10%, and increasing R2 by 0.17 and 0.35 for model calibration and validation, respectively. These results suggest that assimilating satellite-derived data and optimizing the calibration process through WCA can significantly improve the accuracy of crop yield estimations in water-driven crop growth models, highlighting the potential of this approach for precision agriculture.