Supervised classification of the volcanic rocks in the vicinity of El'gygytgyn impact crater, Chukotka, Russia with Landsat ETM+ and MODIS data

Abstract

The northern Chukotka sector of the Okhotsk-Chukotka Volcanic Belt, located in northeastern Siberia, is a mid-Cretaceous acidic volcanic complex. The remote and rugged terrain in this region makes geologic mapping difficult through conventional mapping approaches. The purpose of this project was to assess the capability of satellite imagery to discriminate volcanic unit boundaries under real-world conditions. Training areas were selected from Belyi and Belaya's (1998) geologic map and supervised maximum likelihood classifications were preformed on Landsat ETM+ and MODIS data. Various masks were applied to reduce effects of vegetation, clouds, and other environmental factors that interfere with rock unit classification. Results show overall accuracies of 1-km MODIS and degraded 1-km Landsat classifications are nearly twice as high as the 30-m resolution Landsat classifications. This suggests that the lower spatial resolution MODIS data tends to average the high spectral variability of surface features, as seen in Landsat data, and produces better classification results. A major finding of this study was that masking vegetation and other environmental obstructions prior to classification of Landsat data did not significantly improve the classification results.