University of Redlands
Legends are a common map element, but differences between legends are not well articulated. This project analyzed over 400 legends and identified legend types in printed 2D maps to support legend creation in GIS software. Legends were characterized using 31 attributes then clustered based on those attributes. While characterizing the legends, a dictionary of common legend terms was created. Clustering was done using a Gower number matrix and the DBSCAN clustering method through a Python script. The automated clusters were manually refined to identify eleven legend types: simple, centered patch, reference multipatch, complex reference, descriptive polygon, thematic ramps with text, thematic classes with numbers, bivariate, proportional symbol, complex thematic, and natural.
