ArcGIS Weighted Voronoi Diagram

Generate a weighted Voronoi diagram (Thiessen polygons) from a set of points and attribute values, using an intermediate raster for distance calculations. Instead of assigning raster cells to source points based solely on distance, the distance is divided by some numeric attribute provided with the point data.

Parameters/Usage

All parameters are required.

• Point source is a point feature class containing the points around which to build polygons. It must contain at least one numeric field, the weight field. The feature class must be in a geographic coordinate system.
• Weight field is a numeric field in the point source that the distance between any two raster cells is divided by to determine assignments.
• Diagram output feature class contains the resulting polygon features. OIDs from the source are carried through to the output.
• Buffer point source is the number of decimal degrees to buffer the extent of source by.
• Raster cell size is the number of decimal degrees in both X and Y to use when creating the assignment raster.
• Geodesic distance methods must be one of Haversine or Vincenty. Distances between cells are calculated between the latitudes and longitudes of cell centers. The Haversine method calculates distance on the spheroid (assuming 6371 km as the average earth radius) while the Vincenty method calculates distance on the ellipsoid (using WGS84 datum values for semi-major and semi-minor axis lengths). Note that the Haversine method is significantly faster than the Vincenty method, roughly 6x-8x.
• Smooth output polygons passes through to the Raster to Polygon tool at the end of the process.

Environment variables

The arcpy.env.extent environment variable is temporarily overridden to allow for buffering of the point source extent.

Python packages included

The Haversine 0.4.5 and Vincenty 0.1.4 Python packages are included with the toolbox.

Compatibility

Tested using ArcGIS for Desktop 10.4.1 using 64-bit background geoprocessing.

Known Issues and Limitations

• It requires the source data to be in a GCS (e.g., WGS84 or NAD83).
• It's slow for higher-resolution runs. Better use of NumPy may help.
• Both the Haversine and Vincenty geodesic distance methods have fixed parameters for earth radius and ellipsoid axis lengths.