Spectral Mixture Analysis of EO-1 Hyperion Imagery in the Channeled Scablands of Eastern Washington

This research applied multiple endmember spectral mixture analysis (MESMA) onto a Hyperion Earth Observer 1 (EO-1) hyperspectral image of the channeled scablands of eastern Washington. The goal of this research was to quantify sub-pixel abundances of three endmembers: sagebrush, soil, and grass. Quantification of land cover abundance would greatly benefit wildlife habitat conservation, rangeland health monitoring, and fire danger studies within the research area. Accuracy was assessed through comparison of the MESMA results with manual image classifications using a stereoscope and NAIP imagery acquired within approximately one month of the EO-1 scene. A coefficient of correlation statistic (R) was used to evaluate the effectiveness of this method. The R value for sagebrush was 0.0713, 0.1466 for grass, and 0.1027 for soil. Based on these results, MESMA, as implemented in this study, has little potential for estimating arid vegetation cover in the Channeled Scablands of eastern Washington. However, utilization of specific spectral characteristics, which are amplified in lateSeptember to mid-October, to isolate each endmember may yield contrasting results. Therefore, future studies should isolate specific spectral characteristics of sagebrush, grass, and soil in order to maximize class separability.