The NRCS-CN method is widely applied in the estimation of precipitation runoff. It primarily relies on soil mapping and land use data as input parameters, although land use data is not frequently updated. This study replaced such data with land cover classes based on satellite images, and found this approach favorable to investigate how land use affect runoff. The study further determined the relationship between typhoon events (between 2005 and 2009) and natural hazards using Chenyoulan watershed located in central Taiwan. to The Nash–Sutcliffe model efficiency coefficient was used to assess the accuracy of the NRCS-CN method, which revealed an efficiency coefficient as high as 90.8%. Numerous studies have applied the mean curve number of the whole watershed as the unit of analysis, resulting in poor fit between estimated and observed values. Therefore, this study employed the estimated values of the grids where the monitoring stations were located to form a relationship curve with observed values. This corrected the runoff values and improved the estimation of runoff in the watershed. Furthermore, the model can be applied to simulate runoff depths of grids, and estimate flooding risk, providing a reference for watershed disaster management.