Data mining approach for estimating cloud-covered areas in MODIS satellite images

One of the main remote sensors used for monitoring snow covered areas is the Moderate Resolution Imaging Spectroradiometer (MODIS) employed by NASA's Terra and Aqua satellites. Using MODIS-derived snow cover images is limited under cloud-covered regions due to the sensor's capabilities. This paper presents an automated process based on K-Nearest Neighbor (KNN) algorithm using spatiotemporal features, to estimate the pixel cover for cloud-covered regions. The algorithm was tested using MODIS's daily snow-cover datasets obtained from Terra (MOD10A1) and Aqua (MYD10A1) satellite for the Lebanese territories as a study region. Several experiments were implemented to test the accuracy of the proposed algorithm, and the results were highly acceptable (>90%).