Restoring continuity: exploration of techniques for reconstructing the spatial distribution underlying polygonized data

Polygonized data, e.g., census tract data, are hard to relate to other factors. We developed a procedure that: (a) dissects a continuous 'original surface' into discrete polygons, and (b) reconstructs the original surface from the polygons. Five reconstruction algorithms were tested. We conclude that (a) degrade-and-restore techniques are an effective and intuitive way to test restoration skill; (b) resolution is more important than choice of algorithm; (c) results depend on the interplay of the original surface, the polygon mesh, and the restoration algorithm, and (d) sophisticated algorithms such as Kriging are best left to sophisticated users.