Methods And Application Of Assimilating Remote Sensing Data And Crop Growth Model

A key problem of quantitative remote sensing is estimating land surface parameters. In recent years, there have been considerable interests in developing algorithms for retrieving leaf area index (LAI) from remotely sensed observations. Many efforts have been made to retrieve LAI through its statistical relationship with spectral vegetation indices (VI), physical model inversion,Look-up Table or other nonparametric methods. Using VIs to retrieve LAI is simple, but its limits are also obvious. Nonparametric methods have the limits of physical theory. Model inversion is a reverse process which starts from the inversion cost function. Efforts have been made to retrieve leaf area index (LAI) from remotely sensed observation using prior information. But an obvious drawback is the difficulty in obtaining prior information. Assimilation is a possible method for retrieving LAI by assimilating more useful information sources about LAI. One of potential information of LAI is its temporal profile. The development of crop growth model provides the feasibility of using this kind of information. Crop growth model can simulate crop growth continuously, and give explanation of reasons and essence of environmental factor impacted on crop. Assimilating remote sensing data and crop growth model has become a noticeable way for retrieving LAI from remote sensing data. Currently, two main assimilation algorithms are used extensively, they are Ensemble Kalman Filter (ENKF) and Variation algorithm, but Particle Filter (PF) is another potential method for retrieving LAI from remote sensing data, the reason is it can estimate posterior probability of inverted parameter effectively.Based on present research, this paper play more attention on the assimilation algorithm using remote sensing data and crop growth model; Another focus is placed on the regional use of different assimilation algorithms.In this paper, main research and conclusion included1. Assimilating remote sensing data and LOGISTIC model for LAI retrieval using ENKFIn this study, simple crop growth model LOGISTIC and canopy reflectance model are assimilated into inversion cost function. The LAI is concerned parameter, and the feasibility of algorithm is validated.2. Validating the Variation algorithm assimilating remote sensing data and LOGISTC modelIn this study, LOGISTIC model is used for the validation using Variational algorithm as assimilation method. The maximum LAI in winter's life was chose as adjusted parameter, and LAI as the aimed parameter. The results have shown that this algorithm is a promising method for LAI retrieval from remote sensing data.3. Assimilating remote sensing data into crop growth model using Variational algorithm to retrieve regional LAIIn this study, crop growth model CERES-Wheat was used because it has physical advantage. Inverted LAI results were obtained in Shunyi, Beijing. Aimed on improving LAI results'drawback, these statistical results of LAI from spectrum library were used as prior information according to expert's advice, and assimilated into inversion cost function.4. Assimilating remote sensing data and CERES-Maize into inversion cost function using PFConsidering many assimilation algorithms'Gaussian hypothesis of posterior probability of inverted parameter, PF method was used in this study. The goal is using PF method to estimate the posterior probability of inverted LAI. The PF algorithm is used to retrieve maize's LAI in Yushu area, Province Jilin, and the yield is also estimated for tentative research.Finally, some potential spaces are proposed for future research, and some drawbacks are also discussed.