Optimizing watershed ecological management for water conservation by nonlinear effects of meteorological and vegetation drivers

Abstract

Water conservation is one of the most important ecosystem services in watershed ecosystems. However, optimizing watershed ecological management to enhance water conservation is a critical challenge, complicated by the nonlinear and spatially heterogeneous responses to meteorological and vegetation changes. This study investigates the nonlinear effects of meteorological and vegetation factors on water conservation by integrating machine learning and hydrological modeling. The results show that: (1) the multi-year average water conservation in the watershed was 179.69 mm, with an overall increasing trend from 2010 to 2023. Spatially, water conservation decreased from the upstream to the downstream regions, with the western and southern regions functioning as key water conservation areas. (2) Meteorological and vegetation factors explained water conservation well across the entire watershed (R² = 0.939, p < 0.05). Vegetation evapotranspiration reduced water conservation when vegetation exceeded a critical threshold. Precipitation promoted water conservation across all regions, but the degree of promotion varied due to differences in environmental factors among the subregions. (3) We propose differentiated vegetation management strategies: maintaining the NDVI threshold at 0.65 for mountainous areas (elevation > 950 m) and at 0.45 for plain areas (elevation < 400 m). These findings enhance the protection of watershed ecosystems by optimizing the vegetation configuration of the watershed to optimize water-holding services.