Abstract: Data scarcity poses a critical challenge to water resources development, utilization, and protection in mountainous basins, where high spatio-temporal resolution monitoring systems were not widely established until around 2010. Multi-source precipitation products, which offer long-term coverage and high spatio-temporal resolution, can be integrated with hydrological models as a near-term solution to this issue. This study targeted a typical small-to-medium karst watershed and employed the SWAT distributed hydrological model driven by ground observation data (OBS) and three precipitation products (MSWEP, ERA5, CHM_PRE). A systematic analysis was conducted from three perspectives: multi-scale assessment of runoff simulation accuracy, consistency between precipitation accuracy and runoff accuracy, and the sensitivity and rationality of model parameters. The results indicate that daily runoff simulations driven by the three precipitation products generally achieved good accuracy, with significantly higher performance during the rainy season (May-October) than in the dry season. CHM_PRE exhibited better stability across different months and demonstrated the best overall performance, whereas MSWEP performed poorly during the dry season. Although precipitation accuracy largely determined model performance and showed good consistency with runoff simulation results, runoff accuracy exceeded precipitation accuracy when evaluated by rainfall intensity classes. This discrepancy is attributed to the integrated influence of antecedent hydrological processes on runoff, in contrast to the relatively independent nature of precipitation events, suggesting limitations in such error analysis. The set of significantly sensitive parameters remained stable across different precipitation products and effectively reflected key hydrological processes; however, parameters calibrated under OBS and CHM_PRE exhibited greater physical plausibility. These findings provide a basis for optimizing and merging precipitation products, thereby helping to mitigate hydrological data scarcity in mountainous regions.