Towards large-scale daily snow density mapping with spatiotemporally aware model and multi-source data

Abstract. Snow density plays a critical role in estimating water resources and predicting natural disasters such as floods, avalanches, snowstorms. However, gridded products for snow are lacking understanding its spatiotemporal patterns. In this study, considering the strong heterogeneity of density, well weak nonlinear relationship between meteorological, topographic, vegetation, variables, geographically temporally weighted neural network (GTWNN) model is constructed daily China from 2013 to 2020, with support satellite, ground, reanalysis data. The leaf area index high total precipitation, depth, topographic variables found be closely related among 20 potentially influencing variables. 10-fold cross-validation results show that GTWNN achieves an R2 0.531 RMSE 0.043 g cm−3, outperforming regression (R2=0.271), (R2=0.124), product (R2=0.095), which demonstrates superiority capturing performance state amount snow, more stable plentiful would result higher estimation accuracy. With benefit map, we able obtain knowledge pattern China. proposed holds potential large-scale mapping, will beneficial parameter resource management.