CitiBike Estimated Routes
Alex Chohlas-Wood, NYU CUSP
Curious where CitiBike rides go? This map shows potential routes for most rides from July 2013 through February 2014.
These bike routes were calculated using Google Directions, so they are only approximations of travel paths, rather than actual travel paths. Also important to note: there is no evidence to believe CitiBike riders represent the typical bike user in New York City.
On one hand, these routes demonstrate how bike lanes (shown in dashed black lines) are already structured to handle much of the potential bike traffic. Lanes on, , , , and could each be absorbing upwards of 600 CitiBike rides/day. Similar ridership numbers exist for long sections of the . The network is so complete that much of the estimated high ridership seems to never leave a path.
But I wanted to see if this data could also illustrate where demand may exist for new bike lanes. Even though Google Directions steers riders towards safe bike facilities, I thought that particular streets might lack bike lanes but feature high demand.
A few streets in Manhattan jump out. Perhaps the biggest surprise was, a busy midtown arterial with no bike lanes. At one point just south of Grand Central, the street has demand for almost 250 CitiBike rides/day, or just under 60,000 rides from July 2013 to February 2014.
Many Midtown Avenues, like, also feature high demand (over 30,000 CitiBike rides total, or about 120 CitiBike rides/day) but little in the way of bike facilities. As opposed to most crosstown streets – which are narrower and quieter – users on these avenues may benefit from additional facilities.
South of Midtown is a similar story.is a busy corridor that (among other things) handles traffic coming off the Brooklyn Bridge. It is among the few arterials connecting Chinatown/Lower East Side with the Financial District.
There also seem to be some links that are underserved. From personal experience, I ride fairly frequently onsoutheast of Washington Square Park. This road – covered in uneven pavement and potholes, and lacking a bike lane – may have seen around 40,000 rides over this time period.
I downloaded the entire CitiBike dataset, available here. I accessed the data on April 1st, 2014. The period of analysis runs from July 1st, 2013 to February 28th, 2014.
Next, I compiled a list of all start and end stations for all rides. Many of the most popular departure/arrival pairs were the same exact stations, presumably for people taking more of a joy ride. Since there was no predictable path for these rides, I excluded them from analysis. Also, since Google’s Directions API only allows for 2,500 queries/day, I was unable to get directions for every possible station pair -- but still captured the vast majority of rides.
Google Directions provides a “polyline”, which you can decode into straight segments of lat/long pairs. Luckily, segments from two routes that overlap are identical. After disaggregating Google’s route polylines into straight-line segments, I re-aggregated the segments if they were spatially identical. This included segments that were spatially identical but inverted. Finally, the segments were exported to GeoJSON and mapped with CartoDB.
If you’re curious or want to expand on this yourself, here’s my repository for this project. The processed data, in various stages of aggregation, is also available to download.