A study of four raster-based data generalization procedures

Abstract

Strategies for generalizing raster-based land cover data were investigated. Generalization strategies were studied as part of the U.S. Geological Survey's (USGS) Federal Land Information System (FLIS). Area filtering, category aggregation, resampling, and modal search and replace comprised the strategies tested on USGS Alaska Interim Land Cover data. Generalization of the land cover data was deemed necessary for two reasons: 1) reduction in the volume of homogeneous land cover regions required for computer memory storage, and 2) simplification of the highly complex land cover map. The generalization strategies were evaluated based upon how well they maintained the integrity of land cover information while minimizing the number of homogeneous land cover regions. Maintenance of land cover information after application of a generalization strategy was measured by omission and commission errors, percent unchanged, Cohen's statistic, and the number of land cover regions.