The D.C. Metro system and its hundreds of thousands of daily riders are routinely tested with singletracking, delays, and train breakdowns.
In its fifth Metro Hack Night last week, members of the Transportation Techies Meetup group presented a number of ways to track and communicate these disruptions, from fires to inaccurate arrival predictions. Congestion modeling, in particular, offered new insights on how riders navigate the system and where crowding could be alleviated.
Exactly how crowded is that station platform?
At the previous Metro Hack Night at the WMATA headquarters, speakers from WMATA said they were interested in congestion modeling for the Metro system. Presenter Dan Larsen did just that, focusing on morning congestion on the red, blue and silver lines. Why these lines? Largely, to test some Metro conventional wisdom that the Red and Blue lines are notoriously crowded and that the Silver line runs more trains than necessary.
Using publicly available data from PlanItMetro, Larsen created a simulation that measured when a train was completely full during 12 to 15 minute increments – the maximum level of specificity given current data sets. In order to estimate the number of people on each train, he combined the data for number of people swiping into a station with the train schedules.
Clik here to view.
Orange and red indicate train cars at their highest capacity.
For westbound Red line trains at 9:12 am, trains were at capacity from Silver Spring all the way to Metro Center. Capacity typically means 720 passengers for six-car trains and 960 passengers for eight-car trains. Gallery Place, the most popular station for onloading, boarded 211 passengers in the 7:45 am time slot. For offloads, the highest number was 252 at Metro Center at 8:45 am.
Image may be NSFW.Clik here to view.
Moving to the Blue line, eastbound congestion was consistently high between 7:12 am and 9:24 am, with fully-loaded trains stretching from Pentagon City to Foggy Bottom. Pentagon station was by far the busiest with 1,340 passengers offloading between 9:30 am and 10:00 am. Larsen also looked at “balks” at the Pentagon station, the number of people who were waiting for a train but couldn’t find space. He found that from 9:00 am to 9:12 am, 1,286 passengers were unable to board at Pentagon, showing the crushing demand on just one Blue line stop.
The problem of congestion is so serious – especially with lower train frequencies during SafeTrack – that WMATA’s planning team developed a Metrorail Capacity Analysis. Justin Antos, a member of that team, said that prior to this analysis, Metro employees were placed at the most congested stations and asked to manually count the number of people on each car. Each person had approximately 40 seconds to count up to 120 people per car – and WMATA had no way to verify the accuracy. Now, using a modeling method similar to Larsen’s, WMATA is counting passengers digitally but with a much more specific data set.
Clik here to view.
How WMATA’s modeling tool estimates individual routes and train crowding.
WMATA can combine exact data from each SmarTrip card with train arrivals to assign each rider to a specific train. Obviously, transferring riders do have some route choices between their starting and ending point, so WMATA looked at “bursts” of rider exits at certain stations to determine the lines that riders took.
Clik here to view.
Justin Antos of WMATA
In testing the tool, the simulated passenger counts have been very similar to the manual counts. Overall, the model shows that Orange line trains are more crowded than Silver line trains by the time they arrive in Rosslyn, due to a demand mismatch after the last combined stop of East Falls Church.
So far, congestion data is only available after trips are completed, but the desire for real-time congestion data is there. One major issue uncovered by the model so far: the clocks at the Rhode Island Metro faregates are consistently wrong, causing issues with the accuracy of the data from the station. Antos noted that this and other potential roadblocks are being smoothed out so that the analysis tool can be employed on a more reliable basis in the near future.
How long is 10 minutes in WMATA time?
Joseph Haaga, a freelance software developer, presented his attempt at a command line utility to track train arrivals. While his utility was ultimately inconclusive due to an inability to reliably run every 10 minutes, it did document an additional timing issue for Metro: in at least one instance, a train was slated to arrive at Brookland station in 18 minutes. But 10 minutes later, it was recorded as being 10 minutes away. A new arc in the space-time continuum, perhaps? This all-too-common issue opens another avenue for future projects.
James Pizzurro and Jennifer Hill, creators of the MetroHero train-tracking app, showed off some new features, including a mapped visualization of all trains currently running on the system. They’ve also started compiling tweets of Metro problems that MetroHero can then funnel into an alert system for app users.
Pizzurro also noted that, despite forthcoming real-time train positions from WMATA, they still see a demand for their own, proprietary API on train arrivals. The WMATA data will provide the base, but MetroHero then will add in its own predictions based on delays reported via Twitter and other sources.
There’s also the tongue-in-cheek ismetroonfire.com, created by Nick Stocchero, which scrapes Twitter for data on whether or not Metro is “on fire.” Though he notes it started as a joke, yes, Stocchero found there is demand for that information. Unfortunately, the keywords system used for scraping data and then spitting out a tweet isn’t always accurate. Take, for example, one recent day in which the @ismetroonfire Twitter account reported that all lines were on fire after a tweet mentioned the range of stations affected by singletracking.
Tell all your friends, Twitter is reporting fire/smoke on the red, orange, yellow, green, blue, silver lines. https://t.co/Ke5MT5fnwt
— Is Metro On Fire? (@IsMetroOnFire) June 27, 2016
While not entirely successful at being a reliable news source, Stocchero said it opened up additional possibilities for tracking where smoke is seen on the Metro and other details from Twitter.
Although MetroHero does have an information page regarding SafeTrack, it was striking that none of the presenters mentioned the program’s ongoing impacts to Metro’s well-documented congestion and timing issues.
For more on past and future events from Transportation Techies, sponsored by Mobility Lab, visit the group’s Meetup page.
Photos, from top: A crowd waits for an Orange Line train in Ballston (Sam Kittner for Mobility Lab, www.kittner.com). Justin Antos presents at WeWork Crystal City (M.V. Jantzen, Flickr).
The post Crowd control: Simulating congestion in the D.C. Metro appeared first on Mobility Lab.