The Study On Assimilation Method Between Modis Lai And Crop Growth Model

Remote sensing data and crop growth models can complement each other foragricultural monitoring and yield estimation at a large scale. As we known, remotesensing (RS) technology can depict land surface for a large area, however, onlyinstantaneous information of crop canopy can be observed using RS, and it is hard toderive information of the entire growing process based on RS. Crop growth modelshave the ability to document crop growing process based on the conversionmechanism between matter and energy at day step. But the single pointcharacterization leads it impossible to be propagated to a large area. As a consequence,integrating RS and crop growth models can improve the research level of remotesensing agriculture.Data assimilation is a good way to combine remote sensing data and crop growthmodels. This article attempted to assimilate RS data and crop models using MODISLAI data, field data and ALMANAC crop growth model, and then evaluated thefeasibility of these methods. The main works are concluded as the following:(1) The localization and regionalization sensitivity analysis of crop modelsThe Hailun Agricultural Experiment Station was the study area of this paper.First, the sensitivity analysis of ALMANAC model was performed by SOBOLalgorithm to determine the sensitivity genetic parameters and finished the localizationwork. And then, based on the localization model, a whole analysis of sensitivity forparameters was conduct, so that to recognize the parameters for the extension of localmodel.(2) Assimilation algorithmâ‘ Direct minimization method: The simulated annealing algorithm as employed toassimilate MODIS LAI and ALMANAC model to determine the optimal value ofsensitive model parameters for yield simulation. By comparing with the actual production, here we discussed the impact of the number and types of externalassimilation data on the accuracy of assimilation. At the same time, the simulationresults by using field measurements for external assimilation data were alsocompared.â‘¡Variational assimilation algorithm: This paper induced the variational method andadjoint model into the solution of cost equation. When the cost equation reaches theleast value, then that value is the optimized one. The difference between modelsimulated LAI values and MODIS LAI was evaluated, and the ability betweenoriginal model and calibrated model in assimilation of crop yield estimation was alsocompared.(3) Ensemble Kalman filter method: The article introduced the ensemble kalman filterprinciplely to obtain the optimal LAI value of model simulated LAI and MODIS LAI,and to obtain the optimal solution of sensitivity parameters. The simulated field cropyield was finally compared with the original model.