The temporal variations of normalized difference vegetation index (NDVI) were analyzed using MODIS NDVI time series data in Inner-Mongolia typical steppe and desert steppe from 2000 to 2016. Effects of drought on NDVI for typical steppe and desert steppe were studied. The NDVI had a relatively small interannual variability, with C V of 0.2 less for both typical steppe and desert steppe. The NDVI models of typical steppe and desert steppe were set up using stepwise regression combined with meteorological data,including annual rainfall,rainfall from May to August,mean annual temperature,mean temperature from May to August,annual aridity index and the aridity index from May to August were used in this study. The results showed that NDVI fluctuated in typical steppe and desert steppe from 2000 to 2016,and the typical steppe and desert steppe had low interannual coefficients of variation of NDVI. Compared to the normal year,drought significantly reduced NDVI by approximately 23% for Leymus chinensis community and Stipa grandis community of typical steppe,respectively. The main factors that affected NDVI for L. chinensis community and S. grandis community of typical steppe were mean precipitation from May to August and aridity index from May to August,respectively. NDVI for the L. chinensis+S. breviflora community and S. plareosa community of desert steppe were mainly affected by mean annual temperature and mean temperature from May to August,respectively. The key factors that affected NDVI were mean precipitation from May to August and mean annual temperature at the regional scale. We assessed the precision of NDVI models,and the high accuracy of NDVI models for typical steppe and desert steppe was observed. Precipitation during the growing season is the key factor that affects NDVI in typical steppe,and temperature significantly impacts on NDVI in desert steppe in Inner Mongolia. The results suggest that NDVI for typical steppe would be more susceptible to drought compared to desert steppe under future climate change scenarios.