Xiaohua Hao’s research team developed a MODIS-based fractional snow cover product for High Mountain Asia for 2000–2022 using machine learning. Fractional snow cover quantifies snow coverage at the subpixel scale. Compared with binary snow cover, it better represents snow distribution in complex mountain terrain and is an important input to mountain snowmelt runoff modeling and climate change prediction. The study developed a MODIS FSC retrieval algorithm based on a Multivariate Adaptive Regression Splines (MARS) model with land-cover-specific feature selection. The resulting FSC product was processed with a cloud-removal algorithm to produce fractional snow cover data for High Mountain Asia. Overall RMSE and MAE were 0.148 and 0.093, respectively, outperforming the NASA-NSIDC MODIS FSC product (RMSE 0.178; MAE 0.096) and the ESA MODIS FSC product (RMSE 0.247; MAE 0.131). The daily MODIS FSC product provides important data support for climate change, hydrology, and water resources research in High Mountain Asia.

Figure 1 shows the study area and mountain regions; Figure 2 shows land-cover types and the distribution of samples; Figure 3 shows the model-retrieved fractional snow cover in High Mountain Asia on November 11, 2018; and Figure 4 shows the locations of terrain-validation areas.

The study was supported by National Natural Science Foundation of China Joint Fund project U22A20564, General Program project 41971325, and National Key Research and Development Program project 2022YFF0711702-05.

Figure 1. Study area and mountain regions