Washington, D.C., has got too many neighborhoods. The city’s Office of Planning recognizes 131 distinct neighborhoods and even acknowledges that the list is, at best, incomplete. Residents of the District love nothing more than arguing over where Eckington ends and Bloomingdale begins (or LeDroit Park or Truxton Circle or Shaw, for that matter).
Wouldn’t you rather live in Boston? By the city’s own count, Boston boasts around two dozen neighborhoods. Sounds cozy. Some of them don’t bear especially accurate names: West Roxbury isn’t adjacent to Roxbury, and Mid Dorchester is north of regular Dorchester. But no matter. (For the record, I would not rather live in Boston, thank you.)
Pull up D.C. on Google Maps or Apple Maps, and labels for most of these dozens and dozens of so-called neighborhood names won’t render, no matter how wide or narrow the zoom. “Stronghold,” possibly the city’s tiniest enclave, appears on either map. Yet the label for mighty “Petworth” doesn’t show up on Google Maps at any resolution—try as I might to find it on my phone and desktop. (Petworth residents: Switch to Apple.)
But even in Boston, a city that claims a low density of neighborhood names on paper, digital maps indicate otherwise. On Bing Maps, little greyscale titles pop up all over Dorchester: names like Bowdoin North, Four Corners, Ashmont, Fields Corner (and Fields Corner East and Fields Corner West), and so on. (Imagine arguing with Bostonians over neighborhood boundaries, or anything.)
So how do apps draw up neighborhood names if they’re not getting the data from the city? And how do maps render neighborhoods when they get too much data from a city? I got hung up on this mystery, so I asked someone in the map game—AJ Ashton, the lead cartographer at Mapbox. A slightly edited and condensed email conversation follows:
How does a map app deliver the name of a neighborhood?
This varies a lot depending on the map you are looking at. For Apple Maps, the primary source is TomTom, as well as information from their own internal data collection team. Google’s data is combined from sources such as government data, information collected by their StreetView vehicles, information gleaned from the search and directory parts of their business, and feedback submitted by users through the Google MapMaker program. At Mapbox, we use OpenStreetMap where neighborhood names are gathered by volunteers based on their own local knowledge of the city.
Who or what dataset informs the decision to delineate between Petworth and Park View [in D.C.] in a map presentation, to name one example?
With many maps, there often is no clear delineation. In both OpenStreetMap and TomTom, the neighborhood names are usually associated with only approximate center points, not fully drawn boundaries.
For places where actual boundaries are available, there are many possible sources. Some cities may have neighborhood boundaries published on open government data portals. Some companies such as Flickr and Foursquare have aggregated crowd-sourced geolocation data—for example, if 1,000 users upload photos from their phone tagged as being in Petworth, the GPS coordinates of those photos could be used to draw a fuzzy boundary of Petworth. Where neighborhood boundaries exist in OpenStreetMap, it’s often simply the consensus of local contributors in that area about what the boundaries of the neighborhood are.
Neighborhoods in D.C. are often small and obscure: Swampoodle, for example, or the nexus of LeDroit Park/Bloomingdale/Shaw/Truxton Circle/Eckington. Who makes the call?
D.C.’s neighborhood boundaries are available from the government on their open data portal. As far as I know, the boundaries are only an approximate reference and have no official definition or purpose. And in reality, the boundaries of many neighborhoods are quite fuzzy, and there are many areas of the city that could reasonably fall into two or more neighborhoods.
If you search for different neighborhoods in San Francisco on Google Maps, you’ll see that many overlap. Google’s boundary for South Beach lies entirely within their boundary for SoMa. Their boundary for Civic Center overlaps a large portion of Tenderloin.
Does digital mapping introduce new or different problems in defining neighborhoods?
I think the most important and most interesting thing about digital mapping and neighborhoods is the ability for the residents of a city to define their own neighborhood boundaries and have them shift with the development of the city and even bring new ones into existence on the map simply by consensus.
As I mentioned above, companies like Foursquare and Flickr and Google are collecting similar data on a much larger scale—but in more passive ways. Neighborhoods are not official entities in most cities, so their extents are nothing other than what most people agree on. With a GPS-capable phone in most people’s pockets and data collection through social media, it’s actually possible to get a sense of what many (if not most) residents of a city think their neighborhoods look like.
There are also more purposeful examples of projects where city residents are specifically asked to draw their idea of neighborhood boundaries—in Portland, New York, Burlington, and Boston. And many other cities.
I imagine there are both technological and philosophical questions to address, right?
Yes. One example of both is how much weight or trust to give certain sources. For example, in a real-estate context, people try to stretch the boundaries of neighborhoods with better reputations to their advantage. Another example is tourists in a city with misconceptions about well-known neighborhoods and poor knowledge of smaller ones. Errors from these situations could potentially come through in aggregate data from social-media sources, but they would be difficult to filter out.