passive microwave remote sensing, snow depth inversion, machine learning, fine resolution, northeast china, Mathematical geography. Cartography, GA1-1776

Seasonal snow cover is a key component of the global climate and hydrological system, it has drawn considerable attention under global warming conditions. Although several passive microwave (PMW) snow depth (SD) products have been developed since the 1970s, they inherit noticeable errors and uncertainties when representing spatial distributions and temporal changes of SD, especially in complex mountainous regions. In this paper, we developed a fine-resolution SD retrieval model (FSDM) using machine learning to improve SD estimation quality for Northeast China and produced a long-term, fine-resolution, daily SD dataset. The accuracies of the FSDM dataset were evaluated against in-situ SD data along with existing SD products. The results showed the FSDM dataset provided satisfactory inversion accuracy in spatiotemporal evaluation, with the root-mean-square error (RMSE), bias, and correlation coefficient (R) of 7.10 cm, −0.13 cm, and 0.60. Additionally, we analyzed the spatiotemporal variations of SD in Northeast China and found that snow cover was mainly distributed in the Greater Khingan Range, Lesser Khingan Mountains, and Changbai Mountain regions. The SD exhibited high-low distribution patterns with the increased latitude. The annual mean SD slightly increased at the rate of 0.029 cm/year during 1987-2018.