Spectral Mapping of Shortgrass Prairie Biomass

Multispectral scanner data have been processed to yield biomass maps of imagery from shortgrass prairie vegetation. The results of the image processing of these data were compared to actual biomass values measured at the time the aircraft data were acquired. The comparison demonstrated that image processing predicted 1.15 times the actual biomass present with a correlation coefficient of 0.98 for 26 biomass ground-truth areas sampled from a j7.ight line containing a large range of biomass values. A simple, hand-held device has been constructed which utilizes a spectral ratio between two specific wavelengths, 0.68 and 0.80 J.Lm, to accurately estimate grass biomass. Several field experiments have demonstrated correlation coefficients between 0.95 to 0.98 for the hand-held device in estimating undisturbed grass canopy biomass. The hand-held device has been shown to be an accurate and expedient method for estimating grass canopy biomass. This type of device could be used to gather greater amounts of ground-truth informationfrom overj7.ight areas and thus would add greater statistical significance to image processing results. remeasuring of the same plot again at a later date to obtain an estimate of the change in * ow with the Public Service Company of Colorado, Denver. CO 80207. t Now a National Academy of Sciences Fellow, Goddard Space Flight Center, Greenbelt, MD 20771. Analysis of in situ collected spectral reflectances from a wide range of blue grama grass plots has yielded the method for a simpler approach to measuring grassland biomass non destructively. This approach uses remote sensing of the grassland area in question coupled with a double sampling PHOTOGRAMMETRIC ENGINEERING AND REMOTE SENSING, Vol. 42, No.3, March 1976, pp. 317-323. 317 318 PHOTOGRAMMETRIC E GI EERI G & REMOTE SENSI G, 1976 procedure in order to establish the relationship between canopy radiance of reflectance and the amount of grass biomass present in the canopy. Airborne multispectral imagery and ground-based measurements were used to evaluate these remote sensing methods. The utility of airborne multispectral scanner data to assess grassland biomass distributions were first evaluated. In the course of this analysis it became apparent that one of the limiting factors was the ground-truth collection of biomass data. These data were collected by hand clipping plots of known area which were subsequently weighed. The tedious and inefficient method of collecting multispectral scanner ground-truth data by clipping led to the development of a hand-held radiometer designed to spectrally estimate the biomass present in the grass canopy. The hand-held radiometer system, called the biometer for biomass meter, was designed for rapid, nondestructive estimation of rangeland biomass after a double sampling approach had calibrated the instrument. After completion of the calibration procedure, which entailed the radiometric measurement and hand clipping of 20 to 30 plots, the biometer could be used to accurately measure up to one plot per minute in a field situation. The versatility and speedy operation of the biometer would then enable greater ground-truth to be collected with a high degree of accuracy and small investment of person-hours. This greater groundtruth data base would allow for more accurate classification of the multispectral scanner imagery. The development and preliminary testing ofthe hand-held radiometer will be described first. METHODS, MATERIALS, AND EQUIPMENT Spectral reflectance and total dry biomass data of40 V4-m2 sample plots of blue grama grass (Routeloua gracilis) were subsampled into two sets representing 20 plots each. The data from the first subset were input into a regression routine to establish the relationship between the ratio of the reflectance at 0.800 JLm and 0.680 JLm and the dry biomass (Figure I). A ratio approach was taken because the ratio of the reflectances in two narrow wavelength intervals in close proximity is directly proportional to a first approximation of the ratio of the radiances in those same intervals. The ratio was selected for specific reasons. The radiance or reflectance at 0.680 JLm and the amount of biomass present in the grass canopy exhibits a strong, statistically l( ' ..rtssi.. ':.91