Geomorphometric analysis of the Summit and Ridge classes of the Geographic Names Information System

This research aims to conduct a geosemantic comparison of landforms classified in the Summit and Ridge feature classes in the Geographic Names Information System (GNIS). The comparison is based on a 2D shape analysis of manually delineated polygons produced by USGS staff to correspond to 33,304 Summit and 8,006 Ridge features. Five shape measures were chosen for this specific geomorphometry-based analysis. Univariate and bivariate statistics are first calculated to compare the two feature classes. This is followed by unsupervised learning with k-means cluster analysis to identify two major geomorphometric clusters corresponding to Summit and Ridge features. Although this supports sufficient internal homogeneity to have stable Summit and Ridge feature classes, more than 7,500 (18%) special features were also identified, which were assigned by k-means to the cluster not corresponding to their given GNIS class. These features remain to be analyzed further to decide if GNIS features should be reclassified based on geomorphometric analysis of available polygonal representations.