Developing a Cloud-Reduced MODIS Surface Reflectance Product for Snow Cover Mapping in Mountainous Regions

摘要: Cloud obscuration is a major problem for using Moderate Resolution Imaging Spectroradiometer (MODIS) images in different applications. This issue poses serious difficulties monitoring the snow cover mountainous regions due to high cloudiness such areas. To overcome this, cloud removal methods have been developed past where most of them use MODIS products and spatiotemporal dependencies estimate undercloud coverage. In this study, new approach adopted that uses surface reflectance data cloud-free pixels estimates cloudy pixel as if there were no cloud. estimation obtained by subsequently applying k-nearest neighbor dynamic time compositing methods. The modified are then utilized inputs Normalized Difference Snow Index (NDSI)-based algorithm map study area. results indicate suggested able appropriately bands 2, 4 6, can with 97% accuracy, which substantial improvement over conventional method an accuracy 86%. Finally, although clear underestimation (about 15%) observed proposed approach, still, it much better than 30% method.