Abstract: In order to seek effective method for estimation of grassland biomass in macroscopic view and improve the estimation precision, Landsat5 TM imagery acquired in 2010 was applied for grassland biomass based on field survey data in Zhenglan Banner of Inner Mongolia, China. The author analyzed the relationship of biomass with image bands, principal component analysis (PCA), tasseled cap transformation (TC), vegetation indices, topographical data and meteorological data. And then the regression model between the biomass and all the variables in each category of data was established by mult-stepwise regression. It was shown as follows: the relationships of biomass with EVI, PCA, TC had higher significant correlation. It should choose index of elimination soil background information to estimate grassland biomass. Grassland biomass was 1196412.9 t and carbon storage was about 0.54 TgC in Zhenglan Banner. Carbon density of temperate steppe was 517.49 kg C/hm², and carbon density of the lowland meadow was 683.30 kg C/hm².