Springtime soil moisture variability and its changing environmental drivers: a CMIP6 multi-model ensemble analysis for the subtropical East Asian region

Soil moisture strongly influences land–atmosphere interactions, yet regional-scale analyses of future changes and shifting environmental drivers for the subtropical East Asian region (STEA) remain under-represented compared to global studies. This study uniquely integrates multi-layer soil moisture analysis with machine learning-based driver attribution to reveal temporal shifts in climate controls across climate-vulnerable STEA in response to future warming, where springtime transitions are crucial for water security. Using outputs from 14 CMIP6 models, evaluated against ERA5-L, we find that 86.7% (73.3%) of models capture historical surface (total) soil moisture patterns, providing confidence in projections. Projected changes are assessed using the multi-model ensemble under SSP5-8.5 scenario. Our Random Forest Importance Score analysis reveals a critical hydrological regime transition: rainfall and runoff dominate historical (1995–2014) and mid-future (2041–2060) periods, while near-surface temperature becomes the dominant environmental control by far-future (2081–2100), demonstrating non-linear responses where temperature effects overwhelm rainfall changes. Regional projections indicate progressive drought vulnerability across STEA, with surface (total) soil moisture decreasing by 3.1% (2.0%) by mid-future and 9.1% (5.7%) by far-future, driven by this fundamental reorganization of the environmental drivers. This quantitative assessment provides essential insights for temperature-informed water management strategies, revealing that traditional rainfall-centric approaches become inadequate as warming intensifies across climate-sensitive STEA.