Study regionThe source regions of the Yellow River and Yangtze River in the central-eastern part of the Tibetan Plateau. Study focusHydrological model is an important tool in the simulation of watershed hydrology. However, as more and more basic geographic data becomes publicly available and shared globally, researchers are developing a 'symptom' of difficulty in choosing data, so a systematic comparative analysis for model input data selection is necessary. We tested the effects of different types, sources, and resolutions of input data on the model output results based on the SWAT model, and focused on the mechanism of the role of different input data in the model and how to select an appropriate input data for similar studies. New hydrological insights for the regionThe results show that: the meteorological data is crucial in the model's runoff simulation, and ground meteorological observation station data outperforms reanalysis data such as CFSR. Optimizing CFSR significantly improves the model's performance. The DEM resolution minimally impacts runoff simulation, as the difference in results is primarily due to the topographic characteristics of the watershed. DEM selection should consider TWI complexity and its compatibility with the watershed network. The selection of LULC data has little effect on the simulation, and the best input data combination is OBS + 90 m DEM + CNLULC. These findings assist input data selection for similar watersheds using the SWAT model.