Prediction of vegetation biomass and biochemical composition using PROBE-1 hyperspectral imagery in Yellowstone National Park, Wyoming

Mirik, M.

North Dakota State University, Fargo, North Dakota, USA, PhD thesis


Plant above-ground biomass and biochemical contents were predicted using ratio-based vegetation indices of PROBE-1 1 m2 resolution hyperspectral imagery acquired in August 1999 over Yellowstone National Park, WY. The PROBE-1 remote sensing detector gathers information of the earth's surface from the 423 to 2507 nm portion of the electromagnetic spectrum with an average, minimum, and maximum of 15, 10.7, and 19.8 nm bandwidth, respectively. The detector consists of 4 channels, each of which has 32 bands, for a total of 128 bands. Aboveground plant biomass was clipped and stripped in fifty-six 1 m 2 plots, with vegetation stratified manually into components consisting of sedge, grass, forb, willow, sage, and litter. After drying and weighing, samples from each plot were combined for further chemical analysis. Nitrogen, phosphorus, ash, and nutrient detergent fiber analyses of biomass samples were determined using standard nutrient analysis procedures. Ground study plots were matched to the corresponding image pixels for the regression analysis. Very weak to very strong correlations were found by regressing the biomass and biochemical components on custom-built, ratio-based vegetation indices. The coefficients of determination (R2) of the simple regression models were 0.88 and 0.85 for the total and live biomass, respectively. Values of R2 were 0.88 and 0.71 for the nutrient detergent fiber and nitrogen, respectively. The conclusion from these analyses is that the PROBE-1 1 m2 resolution hyperspectral imaging system can be used with an acceptable degree of accuracy to estimate, at the landscape level, both aboveground biomass and its biochemical constituents.

Patrick Cross2001