Estimation of winter wheat yield by assimilating MODIS LAI and VIC optimized soil moisture into the WOFOST model

Abstract

Accurate and timely crop yield prediction is essential for effective agricultural management and food security. Soil moisture (SM) is a major factor that directly influences crop growth and yield, especially in arid regions. Hydrological models are often used to determine SM, which can be incorporated into crop growth models to estimate crop yield in large-scale areas. However, in existing studies on the coupling of hydrological models and crop models, there is little integration of remote sensing observation indicators into the coupled models, and few studies focus on selecting the most effective depth of SM and the number of SM layers. In this study, we developed a framework for integrating the Variable Infiltration Capacity (VIC) model and the WOrld FOod STudies (WOFOST) model to estimate winter wheat yield in the Yellow River Basin (YRB). The framework first selected the optimal SM layer from three layers and then jointly assimilated this SM as well as the leaf area index (LAI) from the Moderate Resolution Imaging Spectroradiometer (MODIS) model into the WOFOST model using a genetic algorithm (GA). Results showed that the VIC model had a high performance in the validation period across the four subregions, with the Nash Sutcliffe Efficiency (NSE) of the simulated daily runoff and the observed runoff ranging from 0.31 to 0.73 and the corresponding Root Mean Square Error (RMSE) ranging from 256.55 to 467.21 m³ /s. The first SM layer (SM1), with a depth of 0–10 cm in the Longmen-Toudaoguai subregion and 0–26 cm in the Huayuankou-Longmen subregion, was found to be optimal, and jointly assimilating SM1 and LAI resulted in the best performance at the point scale (coefficient of determination (R²) = 0.85 and 0.87 in 2015 and 2018, respectively). The R2 improved by 0.11 and 0.06 in 2015 and 2018, respectively, compared to assimilating LAI alone, and the R2 improved by 0.04 and 0.02, respectively, compared to assimilating SM1 alone. Moreover, joint assimilation significantly improved the estimation of winter wheat yield compared to a model without assimilation (open-loop model) at the regional scale, with the R2 increasing by 0.57 and 0.59, respectively, and the RMSE decreasing by 1808.12 and 859.20 kg/ha in 2015 and 2018, respectively. The yield estimated by the joint assimilation of SM1 and LAI showed more spatial heterogeneity than that estimated by the open-loop model. This study shows that assimilating the optimal SM layer from the VIC model into the WOFOST model enhances the reliability of crop yield estimation, providing policymakers with information to improve crop management.