Wheat Yield Forecast Based On Assimilation Of Remote Sensing And Crop Model

The monitoring of crop growth dynamics and yield prediction on regional scale are of great significance to ensure food security and the formulation of agricultural policies.Crop model can simulate the growth of crops under specific environment,and has been widely used in crop yield prediction,climate change impact estimation and other fields.However,because of the difficulty of obtaining the parameters of regional scale,crop model is limited in regional application.The yield prediction based on remote sensing fully reflects the advantages of remote sensing in spatially and temporally,and greatly improves the timeliness and accuracy of crop yield estimation.But remote sensing can only obtain instantaneous crop information,and the method of yield estimation does not involve the growth and development mechanism of crops.The coupling of remote sensing data and crop growth model can monitor the crop growth and development continuously in spatially and temporally,improve the prediction accuracy and application range of the model,and solve the key problems of regional crop yield prediction.Based on four wheat production types areas in Henan Province,this study selected 17 experimental sites of Neihuang,Huaxian,Yuanyang,Jiaozuo and Xiuwu in the Northern Henan irrigation area;Xuchang,Kaifeng,Luohe,Zhoukou and Shangqiu in the supplementary irrigation area of central Henan;Luoning and Ruyang in the dry farming in the west of Henan Province;Dengzhou,Fangcheng,Luoshan,Xixian and Pingyu in rainfed area of South Henan Province as the research object,and carried out the coupling study of MODIS remote sensing and WOFOST(WOrld FOod STudies)model and constructed a assimilation scheme of remote sensing data and crop model under different regions and different production conditions.The main results and conclusions of the study are as follows:1.The sensitivity of 43 crop parameters to yield,biomass and leaf area index(LAI)in WOFOST model under potential and water limited conditions was analyzed by EFAST.It is found that the sensitive parameters among regions are consistent when the yield and biomass are the target outputs under potential conditions,and the parameters related to photosynthesis and dry matter conversion efficiency(such as EFFTB,AMAXTB,CVO,CVS etc.)are the main sensitive parameters.Under the water limited conditions,the sensitivity of the parameters related to light interception(such as SLATB0,KDIFTB0)increased with the increase of water stress.The parameter sensitivity of LAI was analyzed,and the sensitivity of the parameters was different in different growth stages.Under the condition of water limitation,the parameter sensitivity of early extinction coefficient(KDIFTB0),maximum root depth(RDMCR)and leaf mortality(PERDL)increased with the increase of water stress.Therefore,water stress is the key factor affecting the sensitivity of parameters.2.Based on the sensitivity analysis,the model calibration strategy under different production conditions is proposed.The WOFOST model is calibrated by using the SUBPLEX optimization algorithm.WOFOST model has effetely on the simulation of phenology,and the simulation error of flowering and maturity is within 2 days.The R~2 of the LAI calibration result is between 0.87-0.98,the RMSE is between 0.34-0.79,the R~2 and RMSE of the verification result are 0.77 and 1.06 respectively;the R~2 of the biomass calibration result is between 0.92-0.97,and the RMSE is between 1.73 t/ hm~2-2.36 t/hm~2,the R~2 and RMSE of the verification result are 0.94 and 2.22 t/hm~2,respectively;the error of the yield calibration result is between 0.1 t/hm~2-0.34 t/hm~2,and the R~2 and RMSE of the verification result are respectively 0.66 and 1.43 t/hm~2.The results show that the calibrated model has large errors when verified in other regions.Therefore,when the model is applied in a region scale,it is necessary to correct the model with remote sensing data.3.By analyzing the characteristics of parameter sensitivity to water response,assimilation strategies under different water supplies are proposed.With LAI as the assimilation variable,the En KF algorithm is used to assimilate the corrected WOFOST model and the MODIS remote sensing data corrected by the S-G filter algorithm.Analyzed the results of different assimilation strategies under potential conditions and water limited conditions.Under the potential conditions,the R~2 of the assimilation results considering the water supply and not considering the water supply are 0.50 and 0.48,respectively,and the RMSE is 1.47 t/hm~2 and 1.56 t/hm~2,respectively.Under the condition of water limited conditions,the R~2 of the assimilation results with and without considering the water supply are 0.86 and 0.79,respectively,and the RMSE are 0.51 t/hm~2 and 0.64 t/hm~2,respectively.The results show that: according to the water supply in different regions,the assimilation results with the corresponding assimilation parameters perform best.Therefore,when remote sensing and crop model assimilation are applied on a regional scale,corresponding assimilation strategies are adopted in combination with the water supply and production characteristics of different regions,which can effectively improve the assimilation accuracy.