Taking Shanxi Plateau of China as an example, this study aims to quantitatively analyze the vertical error distribution of the ASTER GDEM V2 data using ICESat/GLA14 data as the benchmark data. The ASTER GDEM V2 and ICESat/GLA14 data were downloaded from USGS and NSIDC respectively, and the ICESat/GLA14 data was used to derive distribution of vertical errors in ASTER GDEM V2 data in the study area firstly; then, the study area was divided into geomorphic units at third class levels, the whole Shanxi Plateau, three regions and eleven subregions. In addition, vertical error distributions were analyzed in these divided geomorphic units; finally, the vertical error distributions among the topographic, landuse and NDVI classes were analyzed. The results show that the vertical error distribution presents a normal distribution in the study area, whose mean and RMSE values are -0.1m and 11.8m respectively. The study area is divided into three geomorphic regions: Central Basin, western Lvliang Mountain and eastern Taihang Mountain; the mean values of the vertical error are 1.1m, 0.0m and -0.6m respectively in the three regions. As to the RMSE values, 9.1m, 12.2m and 12.6m respectively; so the vertical error is low in the middle basin region, but high in the mountain regions of the two sides.With the values of the topographic factors (elevation, slope and relief) increasing, the value of the vertical error becomes higher accordingly. As to the aspect factor, the vertical error is symmetrical along the NE-SW axis; besides, the NE and SW directions has highest positive and negative values respectively. To the NDVI factor, the vertical error decreases with the NDVI values increasing generally; to the land use factor, the highest vertical error distributes in the forestland and grassland.