Abstract: It is an important approach for understanding the spatial variation of biomass at large scale to predict the spatial distribution of biomass based on remote sensing data and typical field investigation data. The Pearson correlations between RVI,NDVI,SAVI,DVI and IPVI and plot biomass were made separately using the Landsat 7 Enhanced Thematic Mapper data and the biomass investigation in the typical sample plots,the linear regression models between biomass and vegetation indexes were established for the Xiaodonggou forest region in the Altay Mountains. The spatial distribution map of aboveground biomass of arbor species in the Xiaodonggou was produced by the regression model and by using the spatial analysis function in ArcGIS 9.1 software. The residual type analysis showed that the strongly predicting area,moderately predicting area and lowly predicting area occupied 66.60%,30.31% and 3.09% respectively,which revealed that the predicted results of the spatial distribution of biomass in the study area were ideal. The results of overlay analysis of spatial distribution of aboveground biomass of arbor species with the slope,aspect and elevation revealed that the aboveground biomass of arbor species was in an order of steep slope (15°-35°,200-250 t·hm^-2) > gentle slope (0°-15°,150-200 t·hm^-2) > very steep slope (>35°,100-150 t·hm^-2). In slope aspect,the biomass on the southeast slope and south slope was lower than that on other slope aspects. In elevation,the biomass was in an order of low elevation (1 042-1 400 m,100-150 t·hm^-2) > high elevation (>1 900 m,150-200 t·hm^-2) > moderate elevation (1 400-1 900 m,200-250 t·hm^-2). This study showed that the aboveground biomass of arbor species can be well predicted with the vegetation indexes derived from remote sensing images,and the spatial distribution of biomass is significantly related to the topographical factors.

Key words: vegetation index, arbor species, biomass, regression model, Xiaodonggou, the Altay Mountains