The code in this repository processes Open Street Map data to calculate, for each location in a specified region:
d_road, the distance from that location to the nearest road, andd_park, the distance from that location to the nearest park.
The results are used for both visualization and analysis.
We visualize the data using techniques that are traditionally employed for rendering topography.
First we define for any point p a height z by:
z(p) = d_road(p) - d_park(p)
where d_road(p), d_park(p) are the distances from p to the nearest road and park, respectively.
We then plug this z(p) heightmap and process it with digital elevation map tool (see technical details below) to generate a map like this:
 |
|---|
| The hillshaded relief layer. |
Next we take a typical roadmap:
 |
|---|
| A typical roadmap. |
And composite the two into our final map:
 |
|---|
| The final map. |
By analyzing d_road we identify the parks in a city which are farthest from roads.
These are typically large parks but can also be smaller parks bordered by industrial sites or water.
Similarly, we use d_park to find the urban points which are farthest from a park.
The data is loaded from OpenStreetmap (OSM) into a database (PostgreSQL with the PostGIS extension) using the the osm2pgsql program.
The OSM extracts from Mapzen or Geofabrik are useful for this.
Distances are calculated with respect to a rasterized grid of a specified resolution using nearest neighbor search with spatial indexing.
The gdaldem program is used for processing the raw distance data into separate hillshade, slopeshade, and color relief layers.
The final maps are rendered using Mapbox.