Regional Yield Estimation Research Based On Remote Sensing Data And Crop Model Assimilation

As an important grain crop in China,it is important to forecast the yield of winter wheat accurately and in real-time.Assimilation of crop growth models with remote sensing data is an important tool for yield estimation,but the low-resolution remote sensing data widely used for area-wide assimilation has mixed image element problems.Therefore,in this study,we use high-resolution remote sensing data to optimize the yield estimation effect of assimilation based on the assimilation of model multi-state variables.In this study,the WOFOST crop growth model was used to simulate the growth of winter wheat in Hebi city as the study area.The study firstly conducted rigorous calibration of the model for Hebi city,then coupled the model with MODIS data for assimilation,and finally used Sentinel-2 data to optimize the assimilation yield estimation results.The main contents and conclusions of this study are.(1)The study collected observations and statistical data from meteorological observation sites in Henan Province,and used the K-Means method to screen out areas with similar meteorological and winter wheat cultivars in Xinxiang,Zhengzhou,Kaifeng,etc.and Hebi City,and optimized the WOFOST model parameters using the site data and ground observation experimental data from these areas,and calibrated the sensitive parameters using the DREAM method.The simulations of LAI and yield of the rigorously calibrated crop growth model were significantly improved,in which the simulated LAI was 61.11% less in error than when using the default parameters of the model,and the simulated yield was 68.54% less in error than when using the default parameters of the model.(2)The study assimilated the WOFOST model and MODIS remote sensing data using SCE-UA global optimization algorithm with LAI and ET as assimilation variables.The study carried out assimilation studies according to four assimilation strategies: different assimilation of any data,assimilation of ET,assimilation of LAI,and simultaneous assimilation of ET and LAI.The results showed that the slopes of total yield and statistical total yield correlation in the study area under the four assimilation strategies were 1.4417,1.3151,1.3172,and 1.1969,respectively,and the correlation coefficients R2 were 0.8938,0.8953,0.8945 and 0.8989,and the root mean square errors were 75014,65619,65903,and 56173 tons,respectively.Assimilating both ET and LAI can improve the correlation and reduce the root mean square error while assimilating the slope of both remote sensing information closest to 1,the highest correlation and the lowest root mean square error.(3)Based on using MODIS low-resolution remote sensing data to assimilate yield estimation,the study used Sentinel-2 high-resolution remote sensing data to extract crop planting information of the study area,superimposed on MODIS assimilated image elements to calculate the proportion of winter wheat planted in this image element,and combined the two to optimize the assimilated yield estimation effect.The results showed that the total yield of the study area after optimization using Sentinel-2data was 28.19% less than the total assimilated yield error without optimization.