284 — The Morphology and Circuity of Walkable and Drivable Street Networks

Geoff Boeing (10.2139/ssrn.3119939)

Read on 31 May 2018
#urban-design  #graphs  #graph-theory  #networks  #roads  #python  #circuity  #open-street-map  #GIS  #walking  #walkability 

I’ve previously read other work by Geoff Boeing (#39): This paper builds upon prior work in the field of urban design by automatically comparing the circuity of different urban environments. Circuity (which I keep typing incorrectly as “circuitry”) is “the ratio of network distances to straight line distances,” and essentially captures a measure of how roundabout a path between two points is.

Using OSMnx, a Python library that converts OpenStreetMap downloads into embedded networkx graphs, Boeing compares 50,000 random routes between two random points in each of 40 US cities (for a total of 2 million routes).

My city, Baltimore, scores a $\mu_d$ of 1.232 and a $\mu_w$ of 1.221, for a $\varphi$, how much driving circuity exceeds walking circuity, of 4.8%. (For reference, San Francisco’s is 46.2%, Manhattan is 47.6%, San Diego is -13.3%, and Cleveland is 2.6%.)

One of the core contributions of this work is the fact that walking and driving networks are considered separately: One way streets and park paths are not always traversable by car (hopefully), but that shouldn’t affect walking-circuity.

It’s interesting to look at the tables in this paper and compare cities based upon intuition (I assumed Manhattan would be very walkable and Baltimore would be even more-so). I think that a large contributor to this metric is the scale and breadth of the city: Baltimore’s “core” downtown is very walkable, but I imagine that the majority of the city is pretty annoying to walk (based upon personal experience).