Abstract: Based on the DEM terrain data of Zhanghe reservoir in Jingmen City on Hubei Province with a spatial resolution 90 m×90 m, twenty flood processes selected from 2012-2015 (sixteen of which were used for simulation and four were used for verification), the operational forecast of WRF model over triple-nesting domains with spatial resolutions of 3 km×3 km, 9 km×9 km and 27 km×27 km in Central China was combined with results of the lumped Xin'anjiang model and the semi-distribution hydrological model, i.e. the Topmodel, in the present study for experimental prediction of flood. Results of comparative experiments are analyzed and major conclusions are given below. In the case when spatial and temporal distribution of rainfall was uniform in the basin, the lumped Xin'anjiang model could accurately forecast the peak flow and peak time; when the spatial-temporal distribution of rainfall became uneven, the forecasting error also increased. The Topmodel based on the DEM data could reflect the difference in flood forecast for rainfall with different spatial and temporal distributions. The results show that the outputs of WRF at the 3 km×3 km and 9 km×9 km domains were close to each other for flood forecasting and both were better than the output at the 27 km×27 km domain based on the deterministic coefficient and relative error of peak forecast. However, the result of 27 km×27 km domain was good at forecasting the peak time of flood. Further analysis also found that the WRF at the three spatial resolutions couldn't realistically forecast the spatial and temporal distributions of rainfall, but the WRF forecast could still be accurate when the forecast error of temporal distribution offset the error of spatial distribution. Therefore, the model forecast of rainfall with high spatial and temporal resolution may not be able to yield more accurate flood forecasting results. The optimal spatial and temporal resolution for the coupled hydrological model and numerical weather forecast model needs to be determined by numerous experimental studies.