EVALUATING LANDSAT THEMATIC MAPPER DERIVED VEGETATION INDEXES FOR ESTIMATING ABOVEGROUND BIOMASS ON SEMIARID RANGELANDS

Ground data from the Central Plains Experimental Range in northeast Colorado and Landsat satellite images of that area acquired in August 1989, June 1990, and September 1990 were used to evaluate the level of association that can be expected from a univariate model relating spectrally derived vegetation indices (difference, ratio, and normalized difference vegetation indices) and dried green vegetation biomass. The vegetation indices were related to the ground sample estimates using a sample point, spectral class, and greenness strata approach. No strong relationships were found between the vegetation indices and sample estimates of dried green biomass using the sample point approach. The spectral class approach produced significant results only for the June 1990 sample period (r2=0.96). Significant relationships were found for the August 1989, June 1990, and September 1990 samples periods (r2=0.95, 0.71, and 0.95, respectively) when the data were aggregated by greenness strata. The high degree of association between green biomass and the NDVI, obtained when the data were combined into greenness strata, indicated that it is possible to predict green biomass levels on semiarid rangelands using univariate regression models.