The Cheap Way to Improve Reliability of the New York Subway

First, apologies for the crayon, but I hope to make a point.

Several recent power outages have each led to a nearly complete loss of service to the B-division of the New York City Subway. Even on a day-to-day basis, the subway is scourged by inconsistent service and crowded trains. The MTA is using this recent publicity to push for upgrading the subway’s signal infrastructure, making that very expensive and long-term solution seem like the way forward. While signal upgrades have merit, there is a much simpler way to avoid propagated delays in the subway: de-interlining, or as Alon Levy puts it, eliminating reverse branching. If the B, D, and E never crossed paths with other subway lines, the power outage at 7th Ave/53rd would have never affected the entire subway network. In addition to preventing system meltdowns, Alon’s posts point out that de-interlining allows for simpler scheduling, easier wayfinding, more frequent trains, reduced wear on switches, and reduced bunching of trains.

Such a de-interlined service could look like:

The primary downside to such a service pattern is the need for customers to connect between services where a one-seat ride is currently provided. However, almost all of these connections can be made easily cross-platform, and service reliability and frequency would generally improve, improving journey times for almost everyone. Inwood and 53rd St line customers whose current one-seat express journeys would only directly served by a local would have slightly longer journeys. So would customers from the Broadway and 6th Avenue lines in Manhattan having to connect at Atlantic Av-Barclay Center or to a local R train to reach their desired service to Brooklyn.

Dekalb Junction

Here, the B, D, N, and Q currently weave together from the 6th Ave Express and BMT Broadway Express lines in Manhattan to the Brighton and 4th Ave Express lines in Brooklyn. There are two options to eliminate this weave, sending Brighton trains down either Broadway or 6th Ave Express. The latter is preferable, as it allows cross-platform connections at Dekalb to either 6th Ave or Broadway via the local.

Central Park West

Here, the A,B,C, and D weave twice. As noted in the MTA A and C line review from 2013, scheduling of all these trains is constrained by the Central Park West segment. Along this segment, there is significantly more demand for express trains than local trains, with up to 19 trains per hour express and 13 local during the peak. At the north end, both station entries and existing AM peak frequency along the Concourse Line are approximately 50% greater than along the Inwood line, thus Concourse trains should go express and Inwood trains local. At the south end, one of the lines will continue down the 6th Ave. express, and onto Brighton. The 2015 Hub Bound Travel Survey shows that ridership for the Brighton and Central Park West express lines are nearly identical, thus these should through run along the 6th Ave express tracks.

Broadway Lines

With Dekalb junction sorted, the Broadway express links the 2nd Ave Subway and the 4th Ave Brooklyn express lines while the local goes to Astoria, Bay Ridge, and potentially Queens Boulevard. The peak loads at either end of the express line are well matched, with approximately 20,000 customers during the peak hour, while the local line can short turn trains at Whitehall in order to balance having significantly more demand from Astoria than from Bay Ridge.

Queens Boulevard

The Queens Boulevard express service is the most crowded service in the B division, and needs to run at its full capacity of 30 trains per hour. Crowding and pass-ups occur by 74 St-Roosevelt Av, while customers at stations closer to Manhattan do not have a reliable wait when attempting to board an express train during the AM peak. Queens Boulevard local trains thus provide crowding relief, and have to serve the busiest stations in Long Island City. Queens Plaza and Court Square both have important connections to other lines and serve more than twice as many customers as Roosevelt Island and Queensbridge, thus local service should run to the 53rd St tunnel, leaving express service to run down the 63rd St tunnel.

Fundamentally, sending all trains down the 53rd and 63rd St tunnels cuts off the R from it’s main yard at Jamaica, and breaks the Lexington express connection to Queens boulevard trains. The former results in an increase in operational cost due to deadheading. Until the Second Avenue Subway is extended to 125th, the latter forces either a 3 block walk for and out-of-system transfer, or a connection to the 6 which is overcrowded at peak hours. A peak hour extension of Broadway local service onto Queens Boulevard can mitigate these effects. As all the capacity of the Queens Boulevard local is required to serve busy East midtown stations, Penn Station, and the Cranberry St tunnel, the 60th St tunnel service would continue onto the Queens Boulevard express tracks, crossing the local trains on the level at Queens Plaza. Cross-town G trains could also make this level crossing, allowing the line to be extended from Court Square to Queens Plaza using the center pocket track after Queens Plaza. The track layout, with cross-overs immediately after the station in both directions, is optimized for capacity, and should be able to handle 36 trains per hour in the peak to cross on the level.

Off-peak Queens Boulevard Express ridership does not warrant more than 15 trains per hour.  In order not to split this service frequency to thinly, Culver and either Broadway local service or Williamsburg bridge service should run through off-peak.  To minimize branching, Williamsburg bridge service is preferred.

West 4th St

Here, the 8th Ave and 6th Ave locals have a grade separated junction. Trains from the Central Park West local and from Queens Boulevard to the north may serve the Culver line, the World Trade Center terminus, the 2nd Ave terminal tracks, or the Williamsburg Bridge to the south. There are two reasonable choices for the service pattern.

1) As today, 8th Ave trains continue to WTC, and 6th Ave trains continue to the Culver line, Williamsburg Bridge, or 2nd Ave terminus. 8-car trains from the Williamsburg Bridge would either have to be used on the Queens Boulevard express, or be terminated on the Broad St subway.
2) 8th Ave trains, continue to the Williamsburg Bridge and 2nd Av terminal tracks with 6th Ave trains alternating between WTC and the Culver line. This requires tracks to be shared between two services though Broadway-Lafayette station.

Option 1 is chosen to minimize track sharing.

‘A’ Division

Interlining occurs at two locations on the ‘A’ division: at Rogers junction in Brooklyn, and between 149th St/Grand Concourse and 180th St in the Bronx. Neither location is suitable for de-interlining without new capital projects being performed.

At Rogers Junction, it is not possible for express trains to serve Nostrand Ave and Kingston Ave stations or access the Livonia Ave elevated without crossing other services on the level. At 149th St/Grand Concourse, the 5 train service could be eliminated, with 3 trains continuing to Dyre Ave. However, there is no elevator to connect the Jerome Ave and Lenox Ave line platforms, thus the 5 train is the only means of access for people with limited mobility.

Service levels

A workable service pattern is presented below. By changing which lines to Manhattan are paired together, some trans from over-served lines, such as on the Central Park West local, Williamsburg Bridge, and 4th Avenue Subway, may be reallocated to more crowded lines including the Second Avenue Subway, Queen’s Boulevard local, and Brighton lines. In order to prevent delays where lines merge, frequency should be split in a simple 1 to 2, 1 to 3, or 1 to 4 ratio at each merge point. While on paper this produces some inefficiencies on paper, such as providing excess capacity to the G, gains in service reliability will compensate and provide a more efficient system overall.

Conclusion

It is possible to have a more reliable New York City subway without spending much money.  However, some long-held scheduling principles of providing as many one-seat rides as possible will have to be abandoned.