The WRF-Hydro modeling system (WRF-Hydro model) was used for flood forecast in semi-humid small and medium catchment in China, proposed by the National Center for Atmospheric Research Center (NCAR) in 2013. The Global Land Data Assimilated System (GLDAS) and observed rainfall data are adopted to drive WRF-Hydro model in Chenhe Basin located in the south-central Shanxi Province. This study examine the practical features and applicability of the model, and also compare with Xin'anjiang (XAJ) model in structure, parameters, inputs & outputs and simulation results. Given the excessive time consume in model running, the stepwise approach for calibration is used for five main parameters in this study. The multiplier for subsurface layer thickness (ZSOILFAC) is introduced to modify this thickness to establish contact between subsurface layer and aeration zone to meet the actual situation of the basin, and the good agreement between subsurface layer modified by ZSOILFAC and the aeration thickness evaluated by XAJ model is achieved. The results show that the time step is supposed to be reduced from 6 s recommended by User's Guidance to 1 s when spatial resolution of routing grid is 100 m in Chenhe Basin. WRF-Hydro model is good at simulating the details of flood, while XAJ model performs well and stably. The qualified rates of runoff depth and flood peak for the former are equal or slightly inferior to those for the latter. The mean root mean square Error (RMSE) of the former is 21.5% less than that of the latter for the flood events with qualified runoff depth and flood peak; while mean RMSE of the former is 56.2% greater than that of the latter for others. WRF-Hydro model has good skills in simulating the start time of observed hydrograph and has promising potential for hydrological simulation, flood forecasting and water resources evaluation for small- and medium-sized catchments.