Fuzzy vs. Crisp Land Cover Classification of Satellite Imagery for the Identification of Savanna Plant Communities of the Oak Openings Region of NW Ohio and SE Michigan
The Midwest Oak Ecosystems Recovery Plan, establishes the need to identify current savanna land cover throughout its range including within the Oak Openings region of Northwest Ohio and Southeast Michigan. Oak savanna is a transitional plant community falling within the continuum between forest and prairie. Transitional plant communities are more difficult to accurately identify using traditional crisp remote sensing classification techniques. The purpose of this study was to determine if a fuzzy classification approach would better identify savanna plant communities within this region than a traditional crisp classification. The area was classified using both techniques on a Landsat 7 ETM+ image from August 21, 2001. The resulting classifications had accuracies below the acceptable overall accuracy of 85%. The overall fuzzy classification accuracy ranged from 70-73%. The overall crisp classification accuracy was 62%. With such low accuracy results, an adequate comparison between the two methods could not be made, however they do show the fuzzy classification to be significantly more accurate than the crisp classification. More work will need to be done to improve both classifications before any conclusions can be drawn as to which method, fuzzy or crisp, will better identify savanna land cover within this region.