Identification of Surface-Disturbed Features through ISURSL Non-Parametric Analysis of Landsat MSS Data

In response to recent state and federal legislative mandates, the Indiana State University Remote Sensing Laboratory (ISURSL) has initiated a research program applied to evaluation of coal strip mine features in Indiana, Illinois, and Ohio using machine-assisted processing of Landsat MSS data. Specifically, two large strip mines in western Indiana were analyzed implementing both supervised and unsupervised non-parametric classification algorithms which were partially or totally developed at ISURSL. Nine classes of strip mine features were identified which included bare mine spoil, revegetated mine spoil, and water features in various physical states. An estimation of accuracy was made through comparison of the Landsat classification results with 1/30,000 scale aerial photographs taken the same day as the Landsat pass. Class accuracies ranged from 73% to 96% with an overall accuracy of 85%. The non-parametric approaches to classification used at ISURSL provide coal strip mine feature information of comparable quality to that generated by commonly used parametric classification systems, but they require as little as one-fourth the computer time for analysis