Validation of Cloud Reduction Algorithms Over MODIS Snow Products on Andes Mountain

The Andes Mountains span a length of 7000 km and are important for sustaining regional water supplies in South America. Rivers flow from the west side of the Andes to the Pacific Ocean and are the main source of water supply for energy generation, irrigation, and drinking water. Snow variability across this region has not been studied in detail due to sparse and unevenly distributed instrumental climate data. The optical remote sensing approach has been developed as a great tool to avoid this limitation. However, cloud cover reduces the ability to use it in the northernmost and southernmost portions of the Andes and the winter and spring in the central part of the Andes. We tested the performance of temporal, spatial algorithms in consecutive and simultaneous steps over daily MODIS snow cover products (Aqua and Terra). We evaluated the cloud reduction (effectiveness) and accuracy using simulated experiments from MODIS data by selecting low cloud cover images (“truth”) and cover with artificial clouds, then we ran all the algorithms and tested them based on the “truth” dataset. On clear sky days, we include higher spatial remote sensing data (Landsat and Sentinel) and in-field data from UAV. The combination of Aqua and Terra reduced the cloud cover by 10-15% on a yearly scale. The temporal combination with previous and following days yielded a substantial improvement in cloud removal but is usually less effective for large-area cloud cover. Developing a new dataset with cloud reduction can help to increase the performance of the snowmelt runoff model and extend a large latitude range across the Andes Mountains to use optical remote sensing data for seasonal snow studies. The use of machine learning, fusion with other snow products (e.g. radar), and more intense use of UAVs point to the next research.