In order to study the sensitivity of WRF-Hydro model parameters in the Hexi Inland River Basin, the area above the Jiutiaoling Hydrological Station in the Xiying River Basin was taken as the research area.The WRF-Hydro model was driven by global forecast system (GFS) data from 2015 to 2018 in standalone mode. The latin-hypercube one-factor-at-a-time (LH-OAT) method was used to analyze the six main parameters of the WRF-Hydro model and was carried out with confluence grids of different evaluation objectives and different spatial resolutions. The sensitivity index of each parameter was quantitatively calculated and classified. Combined with a typical flood, the specific influence of each parameter on flood hydrograph was analyzed. The results show that the Manning roughness multiplier (MANNFAC) is an extremely sensitive parameter. The infiltration coefficient (REFKDT), the soil saturated hydraulic conductivity (REFDK), and the deep hydraulic conductivity (slope) are sensitive parameters. The surface roughness multiplier (OVROUGHRTFAC) is a general sensitive parameter, and the surface water holding depth (RETDEPRTFAC) is an insensitive parameter. The sensitivity of each parameter is not static, and the parameter sensitivity can be affected by the evaluation index, and a certain degree of change exists. With the variation of the spatial resolution of the confluence calculation grid, the global sensitivity of each parameter can also change slightly, but the global sensitivity level of the parameter does not change. MANNFAC has the best adjustment effect on peak time, REFKDT and REFDK have significant adjustment effect on flood volume and peak, and slope can adjust the shape of flood hydrograph well.