Abstract: Abstract: Watershed storage capacity (Wm) is an important basis for the classification of dry or wet soil in the calculation of mountain flash flood warning index of small watershed in hilly regions. Based on the three major hydrological regions in Shandong hydrological atlas, the storage capacities of the three major hydrological regions (hereafter refer to as regional storage capacity) have 3 different values. Because of the larger area of major hydrological region and the small area (generally 20 km2) of small watershed in hilly region for mountain flash flood warning, there may exists a large difference between regional storage capacity and the storage capacity of a small watershed in hilly region. In addition, the regional storage capacity was calculated based on hydrological and meteorological data in different regions before 1975. In the past 40 years, the underlying surface conditions of regions have changed greatly, and the regional storage capacity may have also changed accordingly. Therefore, in order to determine storage capacity of small watershed more precisely in the semi-arid and semi-humid northern region, especially in the hilly region that either lack data or no hydrological data, the SCS-CN model was used to determine storage capacity of the watershed with about 20 km2. Taking Zhudonghe watershed as an example, 6 rainfall events (with large rainfall, long duration and high flow) and runoffs from 1990 to 2008 were selected, and simulated by SCS-CN model, through which 6 CN numbers of each subbasin of Zhudonghe watershed were determined which were 116.49, 90.53, 84.57 121.48, 111.34, 94.95 and 76.52 mm, respectively. The average storage capacity of Zhudonghe watershed was obtained by the method of area weighting. The storage capacity of Zhudonghe watershed was 95.82 mm. To test the reliability of this SCS-CN method, the methods of measured rainfall runoff analysis and SCS formula are used and obtained the average storage capacity of Zhudonghe watershed that was 88.97 mm and 98.98 mm respectively. The relative difference between SCS-CN model and above two methods to determine the Wm was 7.7% and (3.2%, respectively. Therefore, comprehensive analysis of three methods, SCS-CN model to determine the watershed storage capacity is more reasonable and feasible. The SCS-CN model here provides a new way to determine the storage capacity of small watershed with data shortage in hilly regions, and provides a technical support for accurate mountain flash flood warning in small watershed.