We conducted a study to examine the spatial variability and vertical characteristics of soil erodibility (K-factor) and its relation to vegetation types based on a case of study on Dengxia catchment, China. Using traditional statistical and geostatistical methods, soil erodibility (K-factor) was calculated by EPIC model. The results showed: (1) The spatial variability of K value at the studied area was significant difference, with the range from 0.1498 to 0.4981, mean of 03316, and variance coefficient of 22.11%; (2) The whole distribution tendency of K value presented belt-shaped obviously and was higher in the northwest than in the southeast, with several high K value centers embedding in the middle and south part of the area. It showed that soil anti-erodibility of forest covering region mainly in the north part is stronger than that of farmland and human resident living region in the middle and south part of the area; (3) The vertical variability characteristics of K value by different vegetation types all followed the order of K0-20cm＜K20-40cm＜K40-60cm at the area while farmland followed the order of K0-20cm＜K40-60cm＜K20-40cm. It showed that soil anti-erodibility in the topsoil (0-20cm) was the strongest except for farmland. For soil erodibility (K-factor) of different vegetation types in the topsoil(0-20cm), the order was as follows: fallow land >tea garden land >farmland >grassland >broad-leaved forest land >microphylla shrub land >coniferous forest land > bamboo forest land.