Today’s blog is the final post in our geocoding overview series and addresses the scenario where a collision cannot be offset from the intersection due to an invalid offset direction. Although this situation can occur at any intersection, it is more likely to occur along longer roads such as state highways where small segments vary from the general orientation of the route.
A great example is the Pacific Coast Highway (California state route 1) in Los Angeles. The Pacific Coast Highway is typically oriented north/south, but in some locations it becomes east/west or slightly opposite. This can create a mismatch with the offset direction in the collision record. For example, this map shows the intersection of the highway with Bay Club Dr:
The Pacific Coast Highway clearly crosses this intersection in an east/west oriented direction. However, many collision records will be shown as occurring north or south of the intersection. This makes sense because the official Caltrans direction designation for state highway 1 is north or south and officers would keep that in mind for the general offset direction. Where does that leave us then?
Well, our offset code that checks the direction (explained towards the bottom of this blog post) is not going to interpret a north/south scenario correctly given the actual geometry of the road. Therefore, there are two ways to resolve this situation:
- You could develop a linear referencing system for geocoding state highway collisions. Linear referencing refers to measuring locations based on their positions relative to known locations. Without going into too much further detail, this is actually what is typically done for large scale processes and the steps I used at UC Berkeley are outlined in this report. Most of those steps are no longer necessary in California, however, as Caltrans now provides a complete free download of their state highway network and postmile system. Oh how I wish that data was available years before… Despite all the powers of a linear referencing system, it is actually overkill for our scenario where we mostly focus on local roads and it also lacks the ability to assign collision locations to the nearest intersection. That leaves us with option #2.
- You can add a directional exception to the specific intersection. Finding these directional exception locations are a manageable task if you are not trying to do it for an entire state. If we locate a problem intersection, we simply tell our geocoding process for this specific intersection if you see a N direction replace it with a W direction or if you see a S direction replace it with an E direction. The process then runs as expected and all the cases will be handled appropriately.
How do you pull these cases out of the haystack of data? The initial review of the geocoding results will show collisions that could be geocoded to an intersection, but failed to offset properly. A quick look at their directions and you can usually identify a pattern that needs investigating. Intersections with fewer collisions will be harder to locate initially, but you can address the biggest problems first before deciding whether it makes sense (time/resources vs. overall geocoding match percentage improvements) to address the rest of the intersections.