A snow cover mapping algorithm based on a multitemporal dataset for GK-2A imagery

Abstract

Snow plays a crucial role in the climate. In particular, snow controls the surface temperature and stabilizes the energy budget due to its spectral characteristics. However, snow can be affected by climate change. If the snow cover extent (SCE) diminishes due to global warming, the rate of introducing radiance to the surface increases because non-snow areas can absorb more solar radiance. Therefore, it is highly important to accurately detect snow-covered areas. In this paper, we develop an algorithm for deriving a daily snow cover map of East Asia using Geostationary-Korean Multi-Purpose SATellite-2A (GEO-KOMPSAT-2A, GK-2A) imagery. After processes for cloud masking and misclassification removal are conducted, a threshold with consideration for the characteristics of a geostationary satellite is applied to the normalized difference snow index (NDSI), and a daily snow cover map is generated utilizing a stacking process. In the case of the cloud mask, the angle-time variation and displacement by clouds proposed in this study were used to enhance the accuracy of cloud detection. For the quantitative validation, the F1 score was 0.89 for Landsat-8 and 0.79 for the interactive multisensor snow and ice mapping system (IMS). In addition, when the snow cover extent calculated from the IMS was compared, the correlation reached 0.91 from December 2020 to January 2021.