Assimilation of Leaf Area Index and Surface Soil Moisture With the CERES-Wheat Model for Winter Wheat Yield Estimation Using a Particle Filter Algorithm

Abstract

To improve winter wheat yield estimates in the Guanzhong Plain, China, the daily leaf area index (LAI) and soil moisture at depths of 0–20 cm ( θ ) simulated by CERES-Wheat model were assimilated from field-measured LAI and θ and from Landsat-derived LAI and θ using a particle filter algorithm. Linear regression analyses were performed to determine the relationships between assimilated LAI or θ and field-measured yields to identify highly yield-related variables for each growth stage of winter wheat, which were used to establish an optimal-assimilation yield estimation model. At the green-up and milk stages, assimilated θ was highly correlated with the measured yields, and at the jointing and heading-filling stages, both assimilated LAI and θ were highly correlated with the yields. The optimal-assimilation yield estimation model was then established by combining the regression equations relating assimilated θ to the yields during the green-up and milk stages with the equations relating assimilated LAI and θ to the yields at the jointing and heading-filling stages, which resulted in better estimation accuracy than the yield estimation model established based on dualistic regression equations relating the assimilated LAI and θ to measured yields for each growth stage. Moreover, establishing different yield estimation models for irrigated and rain-fed farmlands improved the yield estimates compared with the established estimation model that did not take into account whether the farmlands were irrigated or rain-fed. Therefore, the assimilation of highly yield-related state variables at each wheat growth stage provides a reliable and promising method for improving crop yield estimates.