Categories
Urban

Urban Street Network Orientation

My new article, Urban Spatial Order: Street Network Orientation, Configuration, and Entropy, has just been published in one of my favorite journals: Applied Network Science (download free PDF). This study explores the spatial signatures of urban evolution and central planning. It examines street network orientation, connectivity, granularity, and entropy in 100 cities around the world using OpenStreetMap data and OSMnx for modeling and visualization:

City street network grid orientations, order, disorder, entropy, rose plot, polar histogram made with Python, OSMnx, OpenStreetMap, matplotlib.

So, who’s got a grid and who doesn’t? Each of the cities above is represented by a polar histogram (aka rose diagram) depicting how its streets orient. Each bar’s direction represents the compass bearings of the streets (in that histogram bin) and its length represents the relative frequency of streets with those bearings. The cities above are in alphabetical order. Here they are again, re-sorted from most-ordered/gridded city (Chicago) to most-disordered (Charlotte):

City street network grid orientations, order, disorder, entropy, rose plot, polar histogram made with Python, OSMnx, OpenStreetMap, matplotlib.

Note that these are cities proper (municipalities), not wider metro areas or urban agglomerations. Some cities, like Seattle, Denver, and Minneapolis, have an offset downtown with a relatively small number of streets, but the rest of the cityā€™s much larger volume of streets swamps the histogramā€™s relative frequencies. Above, we can see that Chicago is the most grid-like city here and Charlotte is the least. To illustrate this more clearly, in Manhattan for example, we can easily see the angled, primarily orthogonal street grid in its polar histogram:

Manhattan, New York City, New York and Boston, Massachusetts street network, bearing, orientation from OpenStreetMap mapped with OSMnx and Python

Unlike most American cities that have one or two primary street grids organizing city circulation, Boston’s streets are more evenly distributed in every direction. Although it features a grid in some neighborhoods like the Back Bay and South Boston, these grids tend to not be aligned with one another, resulting in a mish-mash of competing orientations. If you’re going north and then take a right turn, you might know that you are immediately heading east, but itā€™s hard to know where youā€™re eventually really heading in the long run. What Boston lacks in legible circulation patterns, it makes up for in other Lynchian elements (paths, edges, districts, nodes, landmarks) that help make it an “imageable” city for locals and visitors.

This study measures the entropy (or disordered-ness) of street bearings in each street network, along with each cityā€™s typical street segment length, average circuity, average node degree, and the networkā€™s proportions of four-way intersections and dead-ends. It also develops a new indicator of orientation-order that quantifies how a city’s street network follows the geometric ordering logic of a single grid. These indicators, taken in concert, reveal the extent and nuance of the grid.

Across these study sites, US/Canadian cities have an average orientation-order nearly thirteen-times greater than that of European cities, alongside nearly double the average proportion of four-way intersections. Meanwhile, these European cities’ streets on average are 42% more circuitous than those of the US/Canadian cities. North American cities are far more grid-like than cities in the rest of the world and exhibit far less orientation entropy and street circuity. We can see this with a cluster analysis to explore similarities and differences among these study sites in multiple dimensions (full methodological details in paper):

Cluster analysis of urban street networks via hierarchical agglomerative clustering: OpenStreetMap, OSMnx, Python, scikit-learn, matplotlib

The clustering dendrogram above shows how different cities’ street networks group together in similarity. We can also visualize this in two dimensions using t-SNE, a manifold learning approach for nonlinear dimensionality reduction. Here is a scatterplot of cities in two dimensions via t-SNE, with cluster colors corresponding to those above (triangles represent US/Canadian cities and circles represent other cities):

t-SNE visualization of urban street network clusters via OpenStreetMap, OSMnx, Python, scikit-learn, matplotlib

Most of the North American cities lie near each other in three adjacent clusters (red, orange, and blue), which contain grid-likeā€”and almost exclusively North Americanā€”cities. The orange cluster represents relatively dense, gridded cities like Chicago, Portland, Vancouver, and Manhattan. The blue cluster contains less-perfectly gridded US cities, typified by San Francisco and Washington (plus, interestingly, Buenos Aires). The red cluster represents sprawling but relatively low-entropy cities like Los Angeles, Phoenix, and Las Vegas.

Sprawling, high-entropy Charlotte is in a separate cluster (alongside Honolulu) dominated by cities that developed at least in part under the auspices of 20th century communism, including Moscow, Kiev, Warsaw, Prague, Berlin, Kabul, Pyongyang, and Ulaanbaatar. Beijing and Shanghai are alone in their own cluster, more dissimilar from the other study sites. The dark gray cluster comprises the three cities with the most circuitous networks: Caracas, Hong Kong, and Sarajevo. While the US cities tend to group together in the red, orange, and blue clusters, the other world regionsā€™ cities tend to distribute more evenly across the green, purple, and light gray clusters.

For more information on my methodology and findings, check out the open-access article, or check out OSMnx for the Python tool used for these analyses and visualizations. Some of my preliminary work on this (and links to source code) appears in two blog posts from last summer.

26 replies on “Urban Street Network Orientation”

I would be an interesting follow-up to factor in actual usage (traffic loads)

I wonder if there is a correlation between bike accessibility and lower order entropy?

Having grown up in Atlanta I am surprised it’s considered so grid-like. The streets I know meander quite a bit and I can think of at least one street that crosses another twice. I gather you may have used a small subset of the city, for example, an urban core. How did you define this and was the size of such an urban core the same for all cities? Having visited Tokyo I would have expected it to be the most disorganized. Surely the megalopolis of Tokyo would have a larger core area to be measured than the relatively small Charlotte.

I love your work. Insightful and beautiful. I think roses are a great visual. I wonder if other visuals would work, like a histogram? Assigning some sort of spectral reading to each city would allow a different sort of colour grading. I’m reading this on my phone, so haven’t dug deeply to see if some South African cities are included, but would love to see your work applied to my home country. Thanks again for sharing your work.

It’s interesting to compare Manhattan with Montreal, which have very similar grid orientations (both on islands too), and to note that while Manhattanites use the words “East” and “West” to refer to streets that lie along the axis that is closer to the true East-West, Montrealers think of that same axis as “North-South”, and treat the other axis as their East-West axis for the purposes of street naming. That is, in Montreal, the Saint-Laurent river is thought of as “south” of downtown, while Mount Royal is thought of as “north”, and the northerly tip of the island is thought of as the “east island” while the southerly tip is called “west island”.

Thanks for the beautiful visualizations!

I’d be interested in seeing a plot of Canberra, Australia. Canberra is renowned for having a non grid like street layout and I’d be interested in seeing how it fared in your analysis.

[ā€¦] Š½ŠµŠ“Š°Š²Š½ŠµŠ¼ ŠøссŠ»ŠµŠ“Š¾Š²Š°Š½ŠøŠø с ŠøсŠæŠ¾Š»ŃŒŠ·Š¾Š²Š°Š½ŠøŠµŠ¼ Š“Š°Š½Š½Ń‹Ń… Š¾ Š³Š¾Ń€Š¾Š“сŠŗŠøх уŠ»ŠøчŠ½Ń‹Ń… сŠµŃ‚ŃŃ… [ā€¦]

[…] Š½ŠµŠ“Š°Š²Š½ŠµŠ¼ ŠøссŠ»ŠµŠ“Š¾Š²Š°Ń‚ŠµŠ»ŃŒŃŠŗŠ°Ń рŠ°Š±Š¾Ń‚Š° с ŠøсŠæŠ¾Š»ŃŒŠ·Š¾Š²Š°Š½ŠøŠµŠ¼ Š“Š°Š½Š½Ń‹Ń… Š¾ Š³Š¾Ń€Š¾Š“сŠŗŠøх уŠ»ŠøчŠ½Ń‹Ń… сŠµŃ‚ŃŃ… […]

Your sample is limited to gridded parts of the city laid out before WW2. Today, grids make up less than 25% of the built urban environment. Cities are instead organized by cul-de-sac spines rather than continuous grids. Is your model adaptable to the majority of urban streets?

Leave a Comment