Research Request - Segment stability in speed time-series / quarterly and yearly averages

[tiffanychu90]

Complete the below when receiving a research request, and continue to add to this issue as you receive additional details and produce deliverables. Be sure to also add the appropriate project-level label to this issue (eg gtfs-rt, DLA).

Research Question

Single sentence description: We have aggregated speeds for segments. Explore how well we can match a segment over time (start with roughly 16+ month period) so we can see how speeds change over a route geographic entity.

Detailed description:

• Start with stop_segments, which handles intra-day shape variation and combine that with how we've standardized route naming across months.
  • time harmonization Research Request - refactor segment speed exports  #992
  • route_ids typically change in a 6 month period....shape_ids change within 2-3 months` Epic - GTFS Digest (exploratory portfolio) #989 and done in Digest ymls #1074
• Try aggregating to quarterly and yearly segment speeds
  • segment geometry is a bottleneck...the gdf is huge, but if we go over longer time periods, we want to dedupe first
• If this is promising, we can borrow this and use for roads to aggregate even further across operators and routes to actually identify the same geographic entity Epic - Transit speeds on road segments #916

How will this research be used?

What will be achieved or informed? What is the intended impact?

Data sources

• Cal-ITP data sources:
• aggregated segment speeds

Deliverables

• Notebook showing how segment geometries line up over the 16 month period (segment line-up as time-series #1209)
• Scripts to aggregate speeds by quarter or year
• Speed averages (year and year-quarter grains) #1267: refactored the single day and multi day averaging function to be more flexible with grouping columns
• Quarter averages #1297: sketched out how we want this done, but even doing it for at a year grain is taking ~25 min, and segment geometry is getting merged in for every day we want included. Is this the best way to do it?
  • TODO should be to investigate how we set up the aggregation / time_series_utils functions