Extending post-classification change detection using semantic similarity metrics to overcome class heterogeneity: A study of 1992 and 2001 U.S. National Land Cover Database changes

The use of post-classification change methods for the analysis of land cover change provides intuitive and potentially reliable results. A recurring problem is the difference in land cover nomenclature that can occur over time or across space when multiple data sources are required. Building on work that uses category semantics as a foundation for reasoning with land cover classes, this paper uses a fuzzy sets based approach to develop attribute based prototype definitions of land cover classes. These formalized category descriptions are used to look at land cover changes as a semantic change evaluated through semantic similarity metrics. The methodology is illustrated on the U.S. National Land Cover Data from 1992 to 2001 over Chester County, PA. The results demonstrate that the proposed method is more versatile than the standard post-classification method in that it can both provide an overall, spatially explicit evaluation of land cover change, as well as nuanced assessments of graded changes for heterogeneous land cover types.