Bicycles/Pedestrians

Safety Toolbox: Problem identification: Analysis: Geocoding SWITRS Data

 

Most GIS packages come with an automated geocoding service. When using SWITRS data, expect to spend some time cleaning the data and manually geocoding the small percentage of collisions that the automated service will fail to geocode properly.

Geocoding to Intersections

SWITRS data has variables with primary and secondary street location, as well as distance and direction of the collision along the primary street. Primary and secondary streets provide the closest intersection to the collision. This information is sufficient to map most intersection collisions and events that occur within 20 to 30 feet of an intersection. The police usually report distance from the curb-line of the street secondary to the collision, but they sometimes measure from the edge of the painted crosswalk.

Primary and Secondary Streets are obtained directly from SWITRS for each collision. These need to be cleaned for misspellings and for proper street/avenue designations. Street cleaning (pun intended) is best automated using MS Excel or more sophisticated software.

Cleaned street names are then concatenated with a ‘ & ‘ in between them to create an intersection variable for each collision. The data file is stored as a *.dbf with a unique identifier for each collision or injured party, then geocoded directly from the intersection variable by the GIS software's geocoding service. The GIS software can apply X and Y coordinates to the data during geocoding, so that the collision events can later be mapped directly from the *.dbf file. X, Y coordinates allow for easy manipulation of data using different GIS software.

The geocoded file then needs to be checked for errors. Unmatched collision intersections can be hand geocoded using the GIS software's geocoding service.

San Francisco Department of Public Health's experience with ArcGIS is that pedestrian injury-collisions can be geocoded 98% of the time, but motor vehicle-motor vehicle collisions geocode to a lesser extent (80%). Property damage only collisions have the least complete intersection information and geocode correctly 50% of the time.

Geocoding Distance Offsets from Intersections

Further precision is obtained when street addresses are provided in SWITRS. Street addresses can be geocoded directly onto a street location outside of an intersection, but the police rarely supply this information.

Collision location can be offset from the intersection by using distance and compass direction from the intersection. This offset can be geocoded and the collision location placed on the street by automated means.