Estimating Winter Wheat Yield By Assimilation Of Remote Sensing Data Into Crop Growth Model

Winter wheat is the most widely distributed and productive crop in the global wide.Jiangsu province has become one of the main winter wheat planting areas in southern China because of its favorable topography and climate conditions.It is very important to obtain the accurate information of winter wheat and estimate its yield effectively.In this study,we simulated the development stage of winter wheat accurately by using PCSE/WOFOST model.Then we assimilated remote sensing data into crop growth model with Ensemble Kalman Filter(EnKF)in both point and regional scale based on the medium resolution remote sensing satellite data.The analysis and evaluation of assimilation results were provided to realize agricultural yield estimation.Main contents and conclusions are as follows:(1)Main parameters in crop model were calibrated in order to simulate crop growth in study area.To calibrate the parameters,we collected the meteorological,crop and soil data of agricultural stations in Jiangsu,combining the FSEOPT optimization method and previous researches.The RMSE of anthesis and maturity were 2.80 days and 3.48 days,which showed that the calibrated model can accurately simulate the main development stage of winter wheat.(2)The study on assimilating remote sensing data into crop growth model in point scale was carried out on the basis of calibration.Take Xinghua station as an example,we assimilated MODIS Leaf Area Index(LAI)into PCSE/WOFOST with EnKF to estimate winter wheat yield.After preprocessed and filtered MODIS images,we fitted filtered LAI with double Logistic curve and got the corrected LAI.Leaf area index was selected as assimilation state variable,and the effect of different step length and ensemble size were discussed to determine the possibility of its application.(3)The assimilation research on regional scale was carried out on the basis of point scale.Inverse Distance Weighted(IDW)method was used to interpolate the meteorological data.Every pixel in study area was simulated one by one,and the final results were verified by statistical yield.The RMSE decreased from 1881.68 kg/ha to 458.28 kg/ha after assimilation.Results showed that assimilating remote sensing data into crop growth model can effectively improve the prediction of yield estimation,which is a feasible method of winter wheat yield estimation.