Research Task - Refine HQTA intersections

[edasmalchi]

Research Task - Refine HQTA intersections

• Question or Goal: Further improve HQTA accuracy for Major Bus Stop intersections

Currently, segments are tagged as either North-South or East-West, and if the underlying segments qualified as major stop precursors the intersection of an N-S and E-W segment creates a major stop. Before #1330, this could happen where a single route made a turn, which is contrary to the intent of the statute (the consensus being that bus intersection major transit stops are where a rider has the freedom to travel along multiple corridors in at least three directions).

While prep_pairwise_intersections now excludes comparisons within the same route, a similar issue can occur where we aggregate multiple routes in a corridor.

In this example, local route 22 and rapid route 522 both run on El Camino Real, which runs broadly Northwest-Southeast. Thus small turns can change the segment direction from N-S to E-W or vice versa, and with two routes being present still create an erroneous major stop in our dataset.

• Possible Solutions (could implement multiple)

  • exclude comparisons between routes that can be aggregated together. Expedient, but would throw out legitimate intersections where multiple frequent routes share a corridor and then diverge (T intersections).
  • Apply some sort of sequential smoothing on segment directions. Routes that briefly jog arguable don't offer the freedom to travel in a different direction, and it would fix some of these cases.
  • Replace N-S/E-W comparison with a measure of "segment heading", compute intersections based on the heading difference between the two segments in degrees. Would solve this and be more resilient to weird street grids, too.

• Expected Outputs / Findings: what do you hope to have created when this is complete?

HQTA dataset with improved accuracy for bus intersection Major Transit Stops.

[edasmalchi]

https://stackoverflow.com/questions/54873868/python-calculate-bearing-between-two-lat-long
https://pyproj4.github.io/pyproj/stable/api/geod.html#pyproj.Geod.inv