Rethinking the Urban Bike Map

November 12th, 2015

Gosh. I’m going to do one of those academic blog posts where I self-promote by telling you that I’ve just published a paper and that you should go read it. I hate those. But, I had actually been meaning to make my thoughts into a blog post or two, and without the intervention of my academic advisor at the time, I would have; now the thing is a paper instead of a post, a full year after it would have been a post, and have I mentioned that you should read it? It’s about bike maps, and what I hate about bike maps and how I attempted to make them better as a genre, by example, and then by overly formal peer-reviewed explication.

The paper is in Cartographic Perspectives which is the journal of the very cool North American Cartographic Information Society who’s conference I attended last year. Cool people. Very friendly. If anyone out there is considering being a cartographer in North America, I recommend you do it, and not just for the conferences and open access journals.

P.S. Here is a link to the article PDF just in case the above link breaks when the issue gets published.

4 responses to “Rethinking the Urban Bike Map”

  1. Thanks for sharing this paper (and authoring it!). It’s fascinating and relevant on many points to the kinds of techniques my colleagues and I want to enable at Mapbox through our tools. As I share this paper internally, my colleagues will be pleased to see a discussion of many issues they personally care about: contour lines, colorblindness, real versus official speed limits, and more.

    On the issue of deemphasizing poorly connected streets, one explanation may be that the map’s users are already familiar with this treatment in general-purpose maps such as Google Maps. What probably went unnoticed was that the same treatment was novel in bike maps and indeed paper maps, where the cartographer may take pride in high-density road labeling.

    Your work on the bike map has made OSM’s coverage of Cincinnati so much better, not only as a data source for bike maps but for all kinds of purposes. Have you considered giving a talk at State of the Map (either the international conference or the U.S. one)? A lot of folks would be interested to hear your perspective.

    • Nate Wessel says:

      What?? Google is doing that already? I just looked in a couple places, and I’m not totally sure I believe that. Though, maybe they should. Though again, I’d rather see someone else do it first.

      I’d actually love to make it to the next State of the Map! I can probably find some travel funding this year, as long as it’s not too crazy far away.

      Anyway, I’m glad to hear you’re finding the paper useful. Maybe I’ll try to withhold my general cynicism about academic publishing for a few more months ;-)

      • Google deemphasizes dead-ends and other poorly connected subgraphs based on zoom level. It’s more pronounced in suburbs where cul-de-sacs dominate. Take this neighborhood for example:

        https://www.google.com/maps/@39.265236,-84.299147,16.25z

        All the streets here have identical build quality, but Google considers Woodcrest, Ravinewood, Fallenoak, and Shingleoak to be higher priority streets than the dead end streets that hang off them. You can see those dead ends drop out as you cross from z14 to z13:

        https://www.google.com/maps/@39.2677035,-84.2953995,14z (zoom slightly out)

        Then, if you zoom slightly out from z13 to z12, the entire subdivision drops out:

        https://www.google.com/maps/@39.2678702,-84.2940918,13z

        Until a couple years ago, you would see dead ends drawn in a more muted style than the through streets. I may be mistaken, but the results I saw pointed to an algorithmic approach rather than a more manual road classification approach. The difference now is that roads are either included or excluded, without an intermediate style. That could be for aesthetic reasons, but regardless it’s definitely a performance optimization.

      • Nate Wessel says:

        Hmm. I suppose it would be, now that they’re serving vector tiles.

        It appears that I need to look a lot more deeply into this… I’d been toying with the idea of writing a paper (it’s all about the papers!) looking more deeply at the possibility of deemphasizing dead-ending subgraphs and other dangly bits, but if google is already doing it, then my research looks a bit less like a contribution to the literature ;-) I had never heard of anyone doing that before!!

        So… yeah. Thanks for the heads up on that! I may still submit an abstract somewhere just to give me the excuse to do the analysis. Possibly I’ll take a more empirical approach though, just to mix things up. Some crazy part of me want to run Tajan’s and a couple other algorithms on the OSM planet file and just see what patterns turn up.

        P.S. Commenting is fixed, for now at least. Thanks for the tip on that! I think it was a WP plugin update that broke something.