Abstract: By using retrieved LAI from remotely-sensed imagery, this paper studied the regional winter wheat yield estimation in Huanghuaihai Plain of North China. In order to improve the quality of remotely sensed data for winter wheat yield estimation, a Savitzky-Golay filter was used to smooth the MODIS-NDVI time series data to reduce the cloud contamination and remove the abnormal data. Then, a Gaussian model was used to simulate the daily crop LAI which was corrected by interpolating the measured LAI to get the average LAI values for each phenological stage. Using these LAI data, the relationships between LAI and crop yield at the main phenological stages of winter wheat was established. After optimizing the yield estimation model, the optimal time period and the best model parameters for winter wheat yield estimation in the study area were selected out. Finally, the established model was applied to estimate winter wheat yield based on the retrieved LAI from MODIS-NDVI, and the model accuracy was tested. Through the comparison of the predicted yield with the measured yield in the field, the mean relative error was 1.21%, and the RMSE was 257.33 kg·hm^-2. The model and the method proposed in this study were promising, and could help to get the accurate estimated yield of winter wheat in about 20-30 days ahead of the harvest.

Key words: remote sensing, winter wheat, yield estimation, leaf area index, MODIS, NDVI