Operational Data Fusion Framework for Building Frequent Landsat-Like Imagery

An operational data fusion framework was built to generate dense time-series Landsat-like images by fusing MODIS data products and Landsat imagery. The spatial and temporal adaptive reflectance fusion model (STARFM) was integrated in the framework. Compared with earlier implementations of the STARFM, several improvements have been incorporated in the operational data fusion framework. These include viewing angular correction on the MODIS daily bidirectional reflectance, precise and automated coregistration on MODIS and Landsat paired images, and automatic selection of Landsat and MODIS paired dates. Three tests that use MODIS and Landsat data pairs from the same season of the same year, the same season of two different years, and different seasons from adjacent years were performed over a Landsat scene in northern India using the integrated STARFM operational framework. The results show that the accuracy of the predicted results depends on the data consistency between the MODIS nadir bidirectional-reflectance-distribution-function-adjusted reflectance and Landsat surface reflectance on both the paired dates and the prediction dates. When MODIS and Landsat reflectances were consistent, the maximum difference of the predicted results for all Landsat spectral bands, except the blue band, was about 0.007 (or 5.1% relatively). However, differences were larger (0.026 in absolute and 13.8% in relative, except the blue band) when two data sources were inconsistent. In an extreme case, the difference for blue-band reflectance was as large as 0.029 (or 39.1% relatively). Case studies focused on monitoring vegetation condition in central India and the Hindu Kush Himalayan region. In general, spatial and temporal landscape variation could be identified with a high level of detail from the fused data. Vegetation index trajectories derived from the fused products could be associated with specific land cover types that occur in the study regions. The operational data fusion framework provides a feasible and cost-effective way to build dense time-series images at Landsat spatial resolution for cloudy regions.