Per-edge Elevation Profiles

[mattwigway]

[As of late 2021 this is a development priority.]

[Original description from 2015]: This is a low priority, but we eventually want to import elevation data for use in routing and Modeify.

[rafapereirabr]

Hi @mattwigway and @ansoncfit , are there any updates on this issue?

[rafapereirabr]

Hi guys. Just for the record, @mvpsaraiva did a quick analysis a couple months ago looking at how elevation affects walking and cycling travel times for different routing engines, with OTP, R5 and dodgr. The text is in Portuguese, but only by looking at the maps you will see some expected results:

• Ignoring elevation will generally underestimate travel times, and consequently overestimate accessibility
• When ignoring elevation, dodgr and R5 have similar results
• When accounting for elevation, dodgr and OTP have similar results

[xtimbeau]

Thanks for the link. The results are quite striking and elevation is indeed an important factor for bikes (however, not much can be done interm of infrastructure, except with rent and ride systems and transport of parked bikes from low to high elevation.
Any idea on if elevation wise calculation has been/will be implemented in R5 ?
XT

[ansoncfit]

Hi @rafapereirabr and @xtimbeau,

Thanks for the notes/questions. We have preliminary workflows in place to account for slope in walking/cycling results, and we are hoping to refine them.

In general, our approach is to read an elevation raster in a pre-processing step, add relevant attributes to OSM edges, then use the custom OSM edge attributes in R5. Our proof-of-concept uses the elevation and generalized cost scripts in https://github.com/RSGInc/ladot_analysis_dataprep, as described in https://docs.conveyal.com/learn-more/generalized-cost. The slope bins are somewhat crude, but they were needed to meet requirements of a specific project. We eventually plan to store elevation profiles for each edge, and use a continuous function to adjust walking/cycling speed rather than these bins, but we are still working through funding and collaboration options for this effort.

[abyrd]

An update for @mvpsaraiva @rafapereirabr and @xtimbeau - I am currently working on elevation raster loading, producing per-edge elevation profiles and applying speed coefficients. This is similar to your DEM raster processing at https://github.com/ipeaGIT/r5r/blob/master/r-package/R/elevation_utils.R, but samples the elevation at regular intervals along the edge and calculates slope and coefficient independently for each segment. Hiking functions such as Tobler's are not linear or symmetric around 0 slope (gently sloping downhill walking is easiest) so the combined effect of these small segments can differ significantly from that of the net elevation change across the whole edge. This also allows re-splitting the edges and recalculating their coefficients when we apply scenarios in Conveyal (insert transit stops or roads).

[ansoncfit]

⛰ implemented in #782