Assimilation of the leaf area index and vegetation temperature condition index for winter wheat yield estimation using Landsat imagery and the CERES-Wheat model

To improve the accuracy of regional winter wheat yield estimation in the Guanzhong Plain, China, the field-measured leaf area index (LAI) and soil moisture at the depth of 0–20cm (θ) and both Landsat-retrieved LAI and θ were assimilated into the CERES-Wheat model with an ensemble Kalman filter (EnKF) algorithm. The correlation between the assimilated LAI and the measured yield at each main wheat growth stage, including the green-up, jointing, heading-filling and milk stages, was compared with that between assimilated θ and yield. Then, five types of assimilation schemes were investigated to test the effects of assimilating different state variables at each wheat growth stage on wheat yield estimation. The results showed that the correlations between LAI and wheat yield at the jointing and heading-filling stages were higher than those between θ and wheat yield; moreover, the correlations between θ and wheat yield were higher at the green-up and milk stages. Among the five assimilation schemes, the accuracy of the yield estimation obtained by assimilating LAI at the jointing and heading-filling stages as well as by assimilating θ at the green-up and milk stages was the highest (R2=0.76, root mean square error (RMSE)=548.97kgha−1), followed by the accuracy of the yield estimation obtained by assimilating LAI and θ simultaneously at each growth stage (R2=0.67, RMSE=610.67kgha−1). Conversely, the accuracy of the yield estimation obtained by assimilating LAI at the green-up and milk stages as well as by assimilating θ at the jointing and heading-filling stages was the lowest (R2=0.41, RMSE=928.95kgha−1). Thus, the assimilation of more yield-related state variables at each wheat growth stage in an agricultural data assimilation framework provides a reliable and promising method for improving wheat yield estimation.