Machine Learning based Estimation of High-Resolution Snow Depth in Alaska using Passive Microwave Remote Sensing Data

Snow depth (SD) knowledge is significant in many applications related to hydrology, climate, and disaster management. Many snow models are developed using multifrequency spaceborne passive microwave (PMW) brightness temperature (Tb) observations because of their sensitivity SD. The Tb SD affected by metamorphism, which constrains the utility several empirical conceptual for estimating For first time, Extremely Randomized Trees (ERT), a machine learning algorithm less suceptible data noise used this study at high resolution (1 km x 1 km) Alaska. Different ERT (i.e., Alaska wide model, zonal model) Advanced Microwave Scanning Radiometer-2 auxiliary datasets various regions during 2012-21. These evaluated three different cross-validations sample, spatial, temporal). Further, models' predictive power assessment performed independent temporal datasets. results indicate that (1) inclusion parameters improves accuracy estimates; (2) there no substantial difference between model (3) when $\gt $ 30cm, have outperformed AMSR-2 product, GlobSnow Chang with error (4) mean absolute estimates increases decrease latitude, increase elevation, from early winter late across Overall, shows has good potential improving moderate deep estimates.