Influence pathways of hydrological processes: Perspectives from the pathway probability in different types of...

Water availability is an important determinant of regional sustainability in arid and semiarid areas. The exploration of influence pathways of hydrological processes and differences across different watersheds is vital for understanding the key influencing factors and improving the conditions of vulnerable ecosystems. This study reveals the influence pathways and associated probability of hydrological elements, including daily average evapotranspiration (EVP) and average soil moisture content (MC), in two different types of watersheds, namely grassland dominated Xar Moron River watershed and the forestland dominated Laoha River watershed, in the arid and semiarid regions of northern China. Hydrological processes during the growing season were simulated by the Distributed Hydrology Soil Vegetation Model (DHSVM) and validated against observed monthly runoff. Model performance was evaluated using the Nash-Sutcliffe efficiency (NSE) and coefficient of determination (R²). Correlation analysis was conducted to analyze the relevance of the spatial heterogeneity between each of the hydrological elements and the possible influencing factors. Boosted regression trees identified the dominant influencing factors, and structural equation modeling combined with Bayesian belief networks quantified the influence pathways and their occurrence probabilities. In the grassland-dominated watershed of Xar Moron River, the NSE values were 0.7004 (calibration) and 0.11 (validation), with corresponding R² values of 0.8542 and 0.8326. Topsoil sand plays a key role in shaping the EVP and the MC in upper layer soil. The dominant pathway was topsoil sand fraction → topsoil texture → EVP/MC. Under sandy soil and grassland conditions, the MC in upper layer had the highest probability of being very low (63.16%). In the forestland-dominated watershed of Laoha River, the NSE values were 0.5644 (calibration) and 0.7031 (validation), and R² values were 0.7526 and 0.9112, respectively. Vegetation type was the primary driver, and the dominant pathway was vegetation type → EVP/MC. When vegetation type was forest and precipitation was high, the EVP had the highest probability of being very low (70.59%). These findings reveal distinct soil-vegetation-hydrology coupling mechanisms across watershed types and provide quantitative support for differentiated water resource management strategies.