Rainfall estimation from surface soil moisture using SM2RAIN in cold mountainous areas

Rainfall across mountainous areas is vital for the water supply and ecosystem services of arid watersheds. Rain gauges are the most common method to measure rainfall, but these are sparse in mountainous areas. Satellite and reanalysis products can provide rainfall information over a large area but often have large uncertainty in high elevation environments. A recently developed "bottom-up" approach (SM2RAIN, Soil Moisture to Rain) estimates rainfall from soil moisture dynamics and provides a novel method to estimate rainfall. However, the reliability and accuracy of this method in high-altitude mountainous areas are currently not well understood. This study evaluates the SM2RAIN method under different environmental conditions based on data from 9 in-situ soil moisture and rainfall observation stations in the Qilian Mountains in Northwest China. Subsequently, we compare the Rsim (rainfall estimated using the SM2RAIN in-situ), the global SM2RAIN rainfall product (SM2RAIN-ASCAT) and the reanalysis rainfall product (China Meteorological Forcing Dataset, CMFD) with the in-situ rainfall observations. Results show that the performance of SM2RAIN decreases with increasing elevation. SM2RAIN performs well in alpine meadows, but underestimates rainfall in forestland due to strong interception, and overestimates rainfall in farmland due to irrigation. Meanwhile, SM2RAIN has the potential to evaluate the interception capacity of forestland and the irrigation of farmland. The SM2RAIN-ASCAT and CMFD have similar performances in estimating daily rainfall in the study area. Calibration of SM2RAIN using high spatio-temporal resolution soil moisture products and an advanced bias-correction method can significantly improve rainfall estimation performance in data-scarce mountainous areas.