Study area: The study area is the inland river watershed of northwest China, the upper Manas River basin. Study focus: Baseflow serves as a critical component for water resource management during dry seasons in arid regions. The formation mechanisms of baseflow are governed by multiple factors including watershed characteristics (area and topography) and runoff generation processes (rainfall versus snowmelt dominance). To investigate the drivers of baseflow dynamics in northwest China's Manas River watershed, we employed multi-source remote sensing data to identify the optimal baseflow separation method. Furthermore, the impacts of hydroclimatic variables and vegetation conditions on baseflow were quantitatively evaluated through explainable machine learning approaches, specifically Categorical Boosting (CatBoost) coupled with SHapley Additive exPlanations (SHAP) analysis. New hydrological insights for the region: The results demonstrate that the Eckhardt filtering method provides optimal performance for comparative evaluation of baseflow separation in the watershed. Our analysis reveals significant influences of multiple hydroclimatic variables on baseflow dynamics: Normalized Difference Vegetation Index (NDVI), snow depth (SD), precipitation (Prec), surface air temperature (Temp), solar radiation (SR), profile soil moisture (PSM), and snow cover area (SCA). Among these, NDVI exhibits particularly pronounced effects.These factors collectively induce an approximate two-week delay in baseflow formation processes. Quantitative analysis shows SD has the shortest lag period (12.45 +/- 8.50 days), while SR (14.16 +/- 6.67 days) and Temp (14.89 +/- 7.58 days) demonstrate comparable delay durations. Notably, all observed lag times fall within a 22-day window without statistically significant differences. Furthermore, spectral analysis identified 3-month to semi-annual oscillations between hydroclimatic variations and baseflow responses, potentially explaining observed seasonal patterns in baseflow dynamics. This study provides valuable insights into baseflow generation mechanisms in inland river basins, with particular relevance for water resource management in arid regions.