Net primary productivity (NPP) of grassland ecosystem in Inner Mongolia from 1982 to 2006 was estimated using improved light use efficiency model using NOAA/AVHRR GIMMS NDVI, land use classification data and meteorological data. And the relationship between NPP and climate in different temporal unites (annual, season and month) and scales were analyzed based on correlation and partial correlation coefficients of the two-group elements. Results indicated that NPP tends to increase during 1982-2006, with an average of 0.861Mt C/a. The relation between NPP and climate in different unites and scales can be interpreted as follows: annual NPP of Inner Mongolia correlated more with precipitation than temperature, while interannual spring and summer precipitation influence NPP more seriously than that of autumn, but taking account of temperature, the correlation of summer NPP was the weakest. Correlation between month NPP and climate rises significantly than annual level. Multi-year average month NPP within a year strongly correlated with precipitation and temperature. Eliminating influence of precipitation, correlation coefficients of NPP and temperature dropped obviously and spatial distribution changed dramatically, which revealed that precipitation contributed more to vegetation growth than in month unite and within a year scale. Taking April, July and October for example, though NPP was affected more by precipitation, correlation coefficients of interannual April and October NPP with precipitation were larger than that of July and that showed contrary results in temperature.