Greenland Daily NDSI Data Reconstruction Based on Spatio-Temporal Extreme Gradient Boosting Model

The distribution of snow in Greenland has important effects on the energy balance and climate change in the Arctic. The spatio-temporally continuous data of normalized difference snow index (NDSI) are key to understanding the mechanisms of snow occurrence and development as well as the patterns of snow distribution changes. However, frequent clouds in the Arctic region lead to a large number of missing pixels in the MODIS NDSI daily data. In order to generate the complete NDSI dataset, this paper proposes a spatio-temporal Extreme Gradient Boosting model. Various independent variables, including topographic, geometry-related, and surface attribute variables, were selected, and spatio-temporal variation information was employed to the reconstruction model. The results of simulation experiments show that the reconstructed data have R
²
= 0.953 and RMSE = 0.031. Moreover, the proposed model has better predictive power than the classical regression model.