In a two-part series on The Transport Politic, I previously argued that to improve Greater New York’s commuter rail service, the agencies controlling it should orient their capital plan to emphasize good service on existing lines instead of spending on outbound extensions, with a special focus on through-routing. Such a system would remodel New York’s commuter rail along the lines of the Paris RER or a German S-Bahn.
In the six months since my articles were published, I have continued to refine some of the points in the proposal. Some of those refinements come from tweaks proposed in the comment threads; others come from reading more about good commuter rail operations in France and Germany, as well as about the state of tracks in New York, for which Rich Green’s maps are an invaluable resource.
The basic premise of the plan remains the same, and almost the entire map of the proposal and most of the details I gave in the previous posts could stay the same. I believe a few of the route choices should be tweaked, but beyond this, most of the changes would be in station layout and in operations and scheduling.
All proposed improvements here have a unified theme, which is that New York regional rail should look more like the RER or an S-Bahn. The previous two posts emphasized through-routing and service to city neighborhoods; this coda will stress seamless operations, highlighting transferring and schedule convenience.
The best transfer is one that is timed and cross-platform. Timing reduces waiting time, and cross-platform configurations simplify walking from one train to another. The transit planning literature recognizes this fact: ridership projections for future New York City subway lines assign a time penalty to transfers, recognizing the fact that walking from one platform to another is inconvenient for commuters beyond the extra time cost; those projections, however, do not assign any transfer penalty to cross-platform transfers beyond the waiting time for the connecting train, which transfer timing reduces to zero.
The proposed Fulton Street station, where Yellow, Orange, and Blue lines will meet, should be converted to cross-platform operation. In the initial proposal, the tracks are laid in a cross shape. The north-south tracks (Blue Line) could stay the same, but the east-west tracks (Yellow and Orange Lines) could be tweaked: the tunnel from Flatbush to Manhattan would be moved further south to give the tracks time to curve north, and then the tracks would curve west to the Village as in the first plan.
In addition, if possible, the underground Hoboken station for trains to Fulton should be at the same level as PATH, with cross-platform transfers. This is little different from the practice in Paris, which configured the central transfer station, Châtelet-Les Halles, to allow cross-platform transfers from the north-south RER B to the east-west RER A.
The other transfers in the proposal—Secaucus, Tonelle, Jamaica, and Sunnyside—either are already cross-platform or cannot be converted. Those that are cross-platform should always be configured with two platforms, four station tracks, and possibly two bypass tracks; as much as possible, each route should stop reliably at the same platform, and schedules should be coordinated for timed transfers. This would allow cross-platform transfers between the LIRR-Morristown and Northeast Corridor trains at Sunnyside and Secaucus, relieving Penn Station.
At Secaucus and Tonnelle, the cruciform two-level transfers between the trains to Penn Station and those to Hoboken cannot be converted to cross-platform, but can simplified by tearing down or not building faregates. But they could still be timed if trains wait for one another for a minute at each station, a process that can be performed off-peak without straining capacity; this is done on the Berlin U-Bahn for wrong-way transfers between the U6 and U7 at Mehringdamm.
Finally, three additional infill stops should be considered, two in New Jersey and one in Brooklyn. The West Shore Line (part of the Orange Line) should have a new stop at 51st Street, near the Tonnelle Avenue stop of the Hudson-Bergen Light Rail. The Morristown Line (Purple Line) should have an infill stop at Orange Street in Newark, intersecting the original Newark subway, which has no direct connection to Newark Broad Street Station. And the new Flatbush-Fulton tunnel (Yellow and Orange Lines) would pass under the Jay Street and Court Street-Borough Hall subway stops, permitting a new Borough Hall station to be constructed; this stop would offer transfers to both Court Street and Jay Street stations.
The above-described change in the Fulton Street station layout suggests a second route for the Hoboken-Fulton segment (Yellow and Orange Lines) through Manhattan. Instead of going north under Hudson or Greenwich Street and stopping at Houston Street, it could go north on the same route as the Staten Island-Harlem connection (Blue Line), on separate tracks, and curve west north of Houston, stopping below the existing West 4th Street subway stop.
This option reduces the amount of necessary construction in Lower Manhattan, as well as the total route-length of tunnel to be built, which correspondingly lowers costs. It also serves the Village in a more central location. Unfortunately, West 4th is a three-level station, so crossing under it would require diving deep underground, substantially increasing costs. In Tokyo, one of the reasons for substantial subway cost escalation in recent years is that to cross existing lines, new lines have to burrow deep underground, as this new tunnel would have to. I believe this option would be worth it if the cost were the same or lower than that of the route proposed in the original plan.
At the same time, I am no longer convinced by some of the outbound extensions I had previously proposed. It may not be cost-effective to run improved regional trains on their respective commuter lines’ full length. The original plan already cut out some low-ridership branches and line segments; however, there may be room for more cuts, for examples west of Raritan on the Raritan Valley Line, east of Ronkonkoma and Babylon on the LIRR, and west of Dover on the Morristown Line.
On the other hand, there should be more double-tracking of single-track bottlenecks, such as the single-track bridge over the Hackensack over the Erie Main Line, which is otherwise fully double-tracked.
At least according to the comments on my posts, the most controversial idea I suggested was the tunnel from Staten Island to Manhattan. This tunnel would be expensive, at $7.4 billion, using the estimated costs for a Brooklyn-Jersey City freight tunnel as a baseline. The main benefit of the Staten Island tunnel is not cost per rider, but commute shortening. Residents of Staten Island are in a near-tie with those of Queens for the longest average commutes in the United States. However, Staten Island’s situation is worse: unlike in Queens, where neighborhood retail is often within walking distance, on Staten Island most people need a car to run errands, so shopping trips take much longer.
In either case, it might be useful if expensive to extend the proposed Staten Island lines west to meet New Jersey Transit. The existing Staten Island Railway would have to be extensively modified, complete with a new railway bridge, an elevated line in Perth Amboy, and a raised Tottenville station on the bridge’s approach; this would connect the line with the Perth Amboy commuter rail station, where there could be a cross-platform transfer. At a much lower cost, the North Shore Line could be extended west on an existing freight rail bridge, follow the Morristown and Erie and Conrail lines to cross the Northeast Corridor at an infill station north of Linden and then join the Raritan Valley Line at Cranford.
Penn Station Pedestrian Flow
While through-routing is enough to eliminate the capacity problems resulting from Penn Station’s limited track space, there remains the serious issue of pedestrian capacity. One of the arguments I have heard proponents of the under construction Access to the Region’s Core project use is that the platforms at Penn are narrow and have narrow stairways to the concourses, so a new station is necessary (and will be built according to current plans for the ARC tunnel).
There are multiple solutions to the circulation of pedestrians at Penn Station besides the new connections and stations proposed in my plan. First, Penn Station does not use its existing tracks as efficiently as it could. The LIRR recently remodeled its platforms and the lower concourse so that each of its platforms has four or five staircases leading up to waiting areas. NJT has done no such thing, and each of its platforms only has two such staircases. Remodeling the NJT tracks would be expensive, as it was for the LIRR, but building a new station would be much more pricey.
In addition, today’s station has 11 island platforms, each flanked by two tracks, with only one track adjacent to two platforms. Paving over half the tracks so that each track is adjacent to two platforms would not only widen the platforms and allow the installation of wider staircases and elevators, but also double the number of usable doors on the train. This would leave Penn with 11 or 12 tracks, of which only nine would connect to both the North River Tunnels under the Hudson and East River tunnels.
For reference, with four tracks to the east and six to the west (four to New Jersey, two through an upgraded Empire Connection), Penn would not need more than six to eight through-tracks; it would run out of access tunnel capacity before it would run out of station track capacity. This solution would be more radical than remodeling existing platforms but might be cheaper for a given capacity.
Finally, the concourses should be stripped of back offices immediately, and space-consuming concessions should be eliminated as traffic increases. George Haikalis of the Institute for Rational Urban Mobility notes that only 54% of the lower concourse is used for passenger circulation purposes; the rest is consumed by Amtrak back offices and concessions. This goes against standard practice worldwide. As train stations get too busy, sometimes even existing retail gets kicked out, as was necessary at Shanghai Metro’s busiest station, People’s Square.
Compare the following two off-peak train schedules for Monday, January 4th. Both schedules only list departure times.
I selected these two schedules at random, based on trips I had taken recently. The TER schedule is clockface: trains leave at regular intervals, at the same time every hour. It is easy to remember. The Metro-North schedule has some clockface patterns as well, but they are less regular and break down on the shoulders of rush hour.
By making timetables easier to remember, clockface scheduling makes travel easier for passengers, increasing ridership. While the clockface example above is of half-hourly service, there is no lower limit to frequency: in New York, some buses already run clockface, even if they operate every five minutes.
Best industry practice is in Germany, where the S-Bahn not only maintains clockface scheduling, but also rationalizes the additional rush hour service. The regularity is such that in Stuttgart, there is no need for a comprehensive timetable; instead, a system map indicates at how many minutes after the hour each line arrives at each station. Each line has two departure times, spaced exactly half an hour apart, with additional peak hour trains at the quarter-hour marks. Berlin, whose services are more complex, does have a timetable, but each of its lines maintains clockface scheduling with intervals of five, ten, or twenty minutes; further, the schedule shows that on the Stadtbahn, the S3 and S5 arrive at the shared stops simultaneously, allowing cross-platform transfers.
Even today, New York has the track capacity to maintain clockface schedules with regular intervals on each line. The local/express train alternation is not a problem for two-track railroads with passing sidings, let alone four-track railroads such as the Northeast Corridor and the inner portions of the LIRR and Harlem Line mainlines. Once a new pair of tracks under the Hudson River is in place, clockface scheduling will become even easier.
On New York’s commuter rail systems, as on the RER, not all trains stop at all stations. This does not worsen service as long as express trains are run on a limited-stop basis like express subway trains and if schedules are regular. As on the subway, regional rail express trains should enable people to make diagonal travel, going from suburb to suburb without passing through Manhattan, switching instead at an outlying transfer point such as Jamaica. While transit’s greatest advantage over cars is over straight trips that end in or pass through Manhattan, it can also serve useful purpose for a substantial number of diagonal trips. The current train service pattern squanders this opportunity: for example, the New Haven Line trains skip all stations in the Bronx, making it difficult to travel to stations on the Harlem Line.
A better way of treating diagonal trips would be to require all or most trains to stop at stations located such before splits, as far as track arrangement permits. The LIRR does this at Hicksville; other important junction stations include Woodlawn, Floral Park, Rahway, Valley Stream, Summit, and Newark Broad. At those stations, as far as possible the schedule should time outbound and inbound trains to facilitate diagonal transfers: where platform arrangements permit cross-platform transfers, for example at Valley Stream, the trains should arrive at the same time, and where they do not, for example at Woodlawn, the outbound train should arrive one minute after the inbound train.
No system mainline should have less than two trains per hour at any hour of operation; ideally, the minimum frequency should be three trains per hour. Branches and low-ridership outlying segments should have no less than one train per hour. When there is too much branching to run hourly trains to all branches without running them empty on the common trunk lines, the branches could be served with shuttles with timed transfers off-peak.
In the urban areas, frequency should be higher, starting at six trains per hour. This could cause problems on the Northeast Corridor, the LIRR lines feeding into East Side Access, and the lines feeding into the Hudson Line, which begin to branch out in inner-urban neighborhoods. On the lines feeding into East Side Access, timed transfers at Sunnyside could be enough. But on the Hudson Line’s two branches and the Northeast Corridor, off-peak service should include short-turning trains serving just those branches: for example, the Northeast Corridor could be served by local trains running from New Rochelle to Newark or Penn Station.
None of this applies to peak hour, when there is enough demand to permit one-seat rides to Manhattan from every branch. The system should still avoid mixing lines, for example running Montauk Branch trains to Penn Station instead of Fulton, but on the Northeast Corridor, Hudson Line, and LIRR Main Line, direct trains should serve both inner-urban branches from all outlying corridors.
My original proposal called for faregates, on the models of Paris and Japan. However, it may be better to use a German- and Swiss-style proof of payment system, in which stations would be barrier-free and passengers would have to present tickets at fare inspections to be conducted at random. Such a system could even extend to bus service, and would go a long way to reducing operating costs. The MTA’s recent Making Every Dollar Count report says that out of every dollar the agency obtains in revenue, it needs to spend fifteen cents on fare collection.
The tradeoff between faregates and proof of payment is an issue of ridership. At the passenger density of the RER or Tokyo’s commuter rail system, or for that matter the New York City Subway, fare inspections are infeasible. But at lower passenger density, fare inspectors cost less than station agents. The busiest lines in New York straddle the boundary between RER and S-Bahn ridership. But either faregates or proof of payment would cost much less than having multiple conductors per train collecting tickets.