Correcting LAI Saptial Scale Differences To Impove Remote Sensing Assimilation Based Estimation Of Winter Wheat Yield

Remotely sensed data and crop growth model have complementary advantages in the crop yield estimation.However,scale mismatch between model's state variables andremotely sensed observations has great influence on the accuracy of assimilation model,which would increase the uncertainty of the model inversion and data assimilation.This study focuses on the estimation of winter wheat yield by assimilation of remotely sensed data into crop growth model.Accounting on scale issue,statistical models,physical methods and data fusion technology were employed to analyze the mainsprings of scale effect of the assimilation parameter(leaf area index,LAI),and also quantitatively describe and correct the scaling bias caused by spatial heterogeneity and model nonlinearity.After discussing the differences of multi-source remotely sensed data and the mechanism of which,the scale transformation method of Landsat and MODIS data was built.Based on them,the basic correction framework of spatial difference was established and verified on the study area.Data fusion technology was applied to temporal scale expansion of the assimilation parameter from remotely sensed data,and then coupling WOFOST model with multiple spatio-temporal remotely sensed data to estimate the winter wheat yield in Hengshui city,Hebei Province.The research work and main conclusions of this study are as follows:(1)Based on the analysis of scale effect and the differences of multi-source remotely sensed data,the overall difference of LAI inversed from multiscale remotely sensed data was quantitatively analyzed.It's shown that the difference caused by multi-source remotely sensed data was higher than that from scale effect.(2)Considering the spatial heterogeneity of LAI and refining the production of scale effect,the influences on scale effect from different inversion processes of winter wheat LAI were quantitatively analyzed based on the combination of wavelet transform method and fractal theory.The scaling bias due to scale effect was effectively corrected.(3)Based on the system mechanism,the differences of Landsat and MODIS data were summarized and analyzed,and the relevant information that could be simulated by mathematical methods were extracted.The differences of remotely sensed observation data were quantitatively corrected by the point spread function and particle swarm optimization,and then the scale transformation model was established to reduce the uncertainty from scale difference.(4)According to the quantitative analysis of LAI inversion differences from multiscale remotely sensed data,the correction framework of spatial scale differences was built by mathematical models.As a result,the overall uncertainty was decreased by more than 50%.(5)The winter wheat yield was estimated by assimilation of spatial scale differences corrected and temporal scale expanded remotely sensed data into the WOFOST model,using for-dimensional variational assimilation and SCE-UA optimization algorithm.On the premise of sufficient accuracy,the efficiency of data assimilation method was greatly improved.