Improvement Of Crop Yield Estimation Under Drought Condition Based On WOFOST Model And Remote Sensing Data Assimilation

Agriculture is crucial to China.Soil drought is the most serious natural disaster that affects agricultural production,and even leads to no harvest,so it can threaten to national food security and sustainable agricultural development.However,the impact of soil drought on crops is a complex process,and crop growth models cannot respond well to it in many cases,especially under extremely drought.As a result,the production forecast is inaccurate under this circumstance.Therefore,it is necessary to improve the crop growth model so that it can forecast crop yield better under drought condition.This study takes corns as the research object and aims to improve the accuracy of estimated production based on the crop growth model and data assimilation algorithm from three different ideas:mechanism supplement,learning from other models and algorithm improvement.Accordingly,we selected WOFOST model as the main model to simulate water-limited growth,GF1-WFV and HJ-1 A/B to drive model operation at two space scale,and use the ensemble of Kalman filter assimilation algorithm to combine crop model and remote sensing images.Before simulation,parameter sensitivity analysis was done by using statistical parameters and remote sensing inversion.Then we designed three methods(field crop growth stage stress method,the method of Double Stress Factors and based on double crop coefficient method to calculating surface soil evaporation)to simulate the growth and production of corn.Finally,the simulation results were evaluated from two spatial scales-field scale and city scale.The main conclusions are listed as follows:For the methods based on field-oriented modelling of crop growth period stress,the accuracy of estimating yield under drought conditions is better than that of the original model,and the spatial heterogeneity between fields is improved.Compared with the original model,R~2 increased to 0.71,RMSE decreased to 1335.91 kg/ha,and the coefficient of variation increased from 0.213 to 0.254 at the same time.As for the method of Double Stress Factors,the precision of the simulation is improved on the scale of the field,but there is still a certain gap between the statistical results and the county scale.On the scale of fields,the range of yield distribution and the average yield of fields are slightly lower than those of the original model,which is more similar to the measured data.However,this method still does not have a good simulation of the status of crop failure.R~2 has increased to 0.588,RMSE was reduced to 1906.49 kg/ha.At the city-county scale,simulation results using the coupled model shows that Horqin Left Wing Rear Banner was severely affected,and the total maize yield in Tongliao city was overestimated by 1.66 million tons,and the average yield was overestimated by 723.03 kg/ha.For the method of yield simulation using the WOFOST model based on the double crop coefficient method,a more satisfactory result was achieved.Although this method still does not describe the phenomenon of crop failure due to extreme drought at the field scale,it has greatly improved the accuracy of the simulation results and the original model while maintaining RMSE.Compared with the original model,R~2increased 18.46%.At the city-county scale,the simulation results indicate that the yield of maize severely affected by soil drought at the southwestern part of Tongliao City,and the total production was slightly overestimated,but the average yield was reduced.