Abstract Accurate field‐scale soil moisture (SM) monitoring in heterogeneous alpine regions is essential for hydrological modeling and satellite validation yet remains challenging due to scale disparities between point sensors and remote sensing. The cosmic‐ray neutron sensor (CRNS) offers a promising mesoscale solution, but its performance in complex and highly heterogeneous alpine terrains remains poorly understood. This study presents a 2‐year evaluation (2021–2022) of a stationary CRNS on the Tibetan Plateau, corrected using a nested in situ sensor network. We aimed to (i) develop a scalable kernel normalized difference vegetation index (kNDVI)‐based correction for aboveground biomass hydrogen and (ii) compare three vertical‐extension methods—multiple linear regression, exponential filter, and random forest. Results showed the application of kNDVI‐based biomass correction significantly enhanced the precision of CRNS SM estimation, with root mean square error (RMSE) values dramatically reducing from 0.021 and 0.043 cm 3 cm − 3 (in 2021 and 2022) to a consistent 0.010 cm 3 cm − 3 . Among the vertical‐extension approaches, the random forest model outperformed others in estimating root‐zone SM, achieving a testing R 2 of 0.94–0.96 and an RMSE of 0.018 cm 3 cm − 3 after biomass correction. Our multiscale analysis revealed that CRNS‐derived mesoscale soil water storage exhibited lower temporal variability (standard deviation SD = 7 mm) compared to point‐scale (SD = 10–50 mm), demonstrating a significant scale‐dampening effect on heterogeneity. We conclude that integrating CRNS with satellite‐derived vegetation indices and machine learning provides a robust framework for mesoscale SM estimation in alpine catchments, bridging the gap between point‐scale observations and landscape‐scale hydrological requirements.
Kang et al. (2026) studied this question.