Research On OLI Data Assimilation Based On The Earth Surface Reflectivity Database

Series of Landsat satellites are famous for their high spatial resolution, rich information and high positioning accuracy. They play an incomparable role in the global scale change of the ecological environment monitoring, but their biggest drawback is low time resolution. Because Series of most Landsat satellites’ temporal resolution is 16 days, the number of every year’s Landsat satellites’ images is about 23 scenes, what’s more, the images are affected by the cover of cloud and mist in different degree, the actual available images are less each year. This paper takes surface reflectivity of earth as joint point. Aimed at the characters of OLI and MODIS images, we make full use of MODIS images’ advantage of high time resolution. So we built the database of MODIS images’ earth surface reflectivity. We used the ensemble kalman filter algorithm to assimilate OLI images and MODIS images data, predicting OLI images whose date are same to the MODIS images, which make up for the loss and the insufficiency of original data.Aimed at the differences in the space coordinate projection, band structure and band width, the spatial resolution and temporal resolution, and many other aspects of the differences between OLI images and MODIS images data, we studied the pretreatment process before assimilation, including geometric correction of MODIS images, OLI images’ radiation calibration, interesting areas designated, spatial resolution rebuilt, normalization of reflectivity. Then we executed the data assimilation operation based on the condition of the finish of a series of pretreatment process. We completed the single point ensemble Kalman filter assimilation study in the form of quarterly. The results show that the estimate of OLI images’ surface reflectivity data is close to the actual surface reflectivity, and experiments prove that the predict OLI images have good consistency with statistical OLI mean surface reflectivity of several years. In addition, the ensemble Kalman algorithm including various forms and various models of assimilation, we used regional ensemble Kalman assimilation QG model, then select optimal assimilation results by changing model corresponding parameters. After the data assimilation completion of OLI images with MODIS images, the predict OLI images have good texture, close to the real images spectrum characteristics. Compared with the real images through sampling, we made quantitative estimates of OLI images precision evaluation.The results show that the predict OLI images have high similarity with the truth, which achieved the desired results.The experimental results show that it’s feasible to assimilate OLI images and MODIS images’ reflectivity data by using ensemble Kalman filter for predicting OLI images’ reflectivity which can improve the temporal resolution of OLI images. So getting OLI images without cloud whose date is same to the MODIS will come true, effectively improving the validity and availability of OLI images data which has very important realistic basis.