| Affected by the main factors such as meteorology,soil and field management,the spatial correlation of crop yield has the characters of non-uniformity and heterogeneity.However,the background error of most existing crop growth model and remote sensing data assimilation systems is often set to be uniform and homogeneity,which cannot meet the application requirements of high-precision crop growth monitoring and yield estimation.Therefore,in order to improve the output accuracy and computational efficiency of the data assimilation system,a non-uniform and heterogeneous background error,i.e.,the flow-dependent background error was constructed in crop growth model and remote sensing data assimilation system in this research.This research focused on the construction and implementation of the winter wheat yield estimation system based on the flow-dependent background error assimilation scheme,and deeply discussed from the following aspects,such as LAI inversion of winter wheat,the sensitivity analysis of crop model parameters,the construction of flow-dependent background error assimilation scheme and the estimation of regional winter wheat yield.The research results could further improve the estimation accuracy and stability of the assimilation system,and provide a technical support for timely,accurate and wide-ranging crop growth monitoring and crop yield forecasting.（1）A look-up table consisting of model parameters and band reflectances was generated based on PROSAIL radiation transfer model and multispectral remote sensing data,and the regional winter wheat LAI were retrieved using this look-up table.The retrieval LAI results showed that the determination coefficient and the root mean square error between the inversion and the measured values were 0.967and 0.369 in the four winter wheat growth stages,respectively.The winter wheat LAI inversion results accorded with the winter wheat actual growth situation in Hengshui City,and the inversion results met the accuracy requirements of data assimilation input data.（2）According to the characteristics of winter wheat growth and morphological in Huanghuaihai Plain,the winter wheat LAI inversion model based on radar remote sensing data was studied.The LAI inversion results based on Radarsat-2 remote sensing image showed that the determination coefficient and the root mean square error between the inversion and the measured values were 0.918 and 0.675 in the three winter wheat growth stages,respectively,which proved that the winter wheat LAI inversion model had certain feasibility and applicability,and the inversion results met the accuracy requirements of data assimilation input data.（3）The Sobol’sensitivity analysis method was used to analyze the sensitivity of WOFOST crop growth model parameters.Combined with the ground observation data and related literatures,the crop parameters were calibrated according to the sensitivity analysis results,and then the winter wheat yield were simulated.After parameters calibration,the determination coefficient and the root mean square error between the yield simulated and the measured values were 0.430 and 229.8 kg.ha-1,respectively,which proved the feasibility of the WOFOST model in winter wheat yield simulation.（4）The flow-dependent background error assimilation scheme was designed and applied in the winter wheat yield estimation system.The winter wheat LAI and yield in the typical experimental area were used as the validation data to discuss and analyze the assimilation scheme feasibility and the parameter settings.The results showed that the determination coefficient and the root mean square error between the yield simulated and the measured values based on DFD-4DVar were 0.706 and 167.8kg.ha-1,respectively;the determination coefficient and the root mean square error between the yield simulated and the measured values based on DFD-EnSRF were 0.711 and 173.6 kg.ha-1,respectively.It could be seen that the DFD-4DVar and DFD-EnSRF assimilation algorithms had significant effects on improving the yield accuracy of winter wheat yield.（5）The winter wheat LAI was used in the combination of remote sensing information and crop growth model,the regional winter wheat yield simulation based on flow-dependent background error assimilation scheme was implemented and the yield simulation accuracy was evaluated.The simulated mean winter wheat yield in Hengshui City based on DFD-EnSRF was 6910kg.ha-1,and the root mean square error between the simulated yield and official statistics was 6.46%,which further proved the feasibility and effectiveness of crop growth model and remote sensing data assimilation system based on flow-dependent background error at regional scale. |
