Branching and Transfer Breaking

This is the winning option in a poll I conducted among my Patreon backers. Thanks to everyone who participated. Another option, about commuter rail infill stops, came a close second, and I will likely tackle it later this month.

The New York City Subway is unusually branched. I’ve written about the general concept here, and specifically criticized reverse-branching on the subway here. In this post, I want to talk about a more specific feature of complex branching arrangements: they have station locations that make it hard to disentangle the branches without breaking transfers.


The left image is a common way junctions are set up. In this image, it’s possible to travel from any leg to any leg; an example of this is BART, with its three-way junction in Oakland between the East Oakland Line carrying trains to Fremont and Pleasanton, the line to the north carrying trains to Berkeley, and the line to the west carrying trains to San Francisco. In many other cases, the branching is simpler, with a clear trunk and two branches, and it’s often not possible for trains to travel between the two branches without backing up; this is like the depicted image with one of the connections missing.

New York has one current example like the left image: the A/C/F/G junction in Downtown Brooklyn has a northern leg (A/C/F), an eastern leg (A/C/G), and a southern leg (F/G). All legs have four tracks and not every track pair connects to every other track pair, but each leg connects to both other legs. It has one former example: the junction between Sixth Avenue Line and 53rd Street Line, with the B/D going south-to-west (then north), the E going east-to-west (then south), and the F going south-to-east. The E/F shared tracks to the east, but neither service shared tracks with the B/D to the south or west.

The problem with this arrangement is that it makes the schedules more fragile. A delay on one branch can cascade. Toronto at one point ran its subway line this, with an eastern and western leg under Bloor Street (continuing to Danforth to the east), and a southern leg under University (looping back north under Yonge); it subsequently ended branching by extending the University leg to the north along the Spadina Expressway right-of-way and operating two independent lines.

The rub is that such an extension usually breaks transfers. Look at the right image: running the lines without branching means no transfers, since there is no station located at the crossing. Toronto dodged this problem because of how the original branching was laid out – in fact, there are two adjacent transfer stations. But usually, it is not hard to convert a branching like the left image into two lines with a simple transfer in the middle.

The 53rd/Sixth situation in New York is a good example of the problem. New York realized it needed more capacity going east, toward Queens, since there were only three track pairs – 53rd Street, plus two more disconnected from the system depicted. For this, it built a tunnel under 63rd Street, and connected it to Sixth Avenue, routing the F through it and creating a new service for Sixth-East 53rd trains, then called the V and now called the M. The junction now looks like an incomplete version of the right image, missing the two upper arcs. The F continues north under Sixth, and only diverts east under 63rd, and has no transfer with the E, which runs east-west under 53rd. The next transfer between the two services to the south is at West 4th Street; the next transfer to the east is at Roosevelt Avenue/74th Street, well into Queens, since the alignment of 63rd Street Tunnel into Queens prevents it from intersecting the E closer in, at Queens Plaza in Long Island City.

The highly-branched nature of the subway in New York makes sure that it is possible to travel between legs even when there’s no transfer, provided one is okay transferring between lines with not-great frequency. The first station south of the junction on Sixth, 47th-50th Streets-Rockefeller Center, lets passengers transfer wrong-way, between southbound and northbound trains. I have used this before to transfer from the B/D to the F on my way between Columbia and Queens, which are not well-connected to each other. Going from east to south is already easy on the M; going from east to north is possible via the M and F, but is unusual, since ultimately both legs lead into the same line in Queens.

However, it is hard to disentangle this to reduce branching. If one believes that reducing branching is useful for reliability and capacity, then one must believe it is necessary for New York to figure out how to split branching in the least painful ways. Partial data from the London Underground is suggestive (see international benchmarking, PDF-p. 15) – the non-branching Victoria and Piccadilly lines are more reliable than the complexly-branching Northern line. Moreover, the intensive service in Moscow, topping at 39 trains per hour without any automation, only works since none of the lines branches. This compels New York and other cities with highly branched systems to disentangle lines.

In the Bay Area, the situation is relatively easy, in the sense of requiring relatively little capital construction. There is no real need for a one-seat ride between East Oakland and Berkeley. The reason there are any trains running on that leg is that Downtown Oakland is on the leg to Berkeley and not on the leg to San Francisco. This was bad planning, and was noted as bad planning even in the 1960s.

What is required is a short bypass tunnel. There are two options. First, a tunnel from the east, replacing the Lake Merritt station with a station a few blocks to the north, effectively moving the junction one station north, so that 12th Street-Oakland City Center can be on the western leg toward San Francisco. Second, a tunnel from the west, between West Oakland and Downtown Oakland. This would not move any station, and put 12th Street on the eastern leg toward East Oakland; Downtown Oakland has a second station, at 19th Street, which would stay on the northern leg, for Berkeley-Downtown Oakland service. Either option would break East Oakland-Berkeley transfers, but make the remaining system more robust.

In New York, disentangling reverse-branches is considerably more difficult. On the numbered lines, it isn’t too difficult to shuffle the 2, 3, 4, and 5 so that the only track sharing is between the 2 and 3, and between the 4 and 5. On the lettered lines, first of all one key connection has to be severed: 11th Street Connection, letting the R go between 60th Street Tunnel toward Manhattan and the Queens Boulevard Line. All trains via 60th Street would go to Astoria; in comments, Alexander Rapp suggests flipping the connection at Queensboro Plaza, letting trains from 60th Street (such as the R) go to Flushing and the 7 go to Astoria, matching the busier line in Queens (Flushing) with the more popular route into Manhattan (60th Street). Queensboro Plaza and Queens Plaza have no transfer, and one would need to be constructed, but even with moving walkways, transferring would involve several minutes of walking between platforms.

Then, the Queens Boulevard Line would be left with local and express services, feeding 53rd and 63rd Street Tunnels. Trains on 63rd have to go to Sixth Avenue. This requires all 53rd Street trains to serve 8th Avenue – the east-west line shown in the images. No more M, just a more frequent E train, with implications for how the A/C run (probably both express between 145th Street and Chambers, where the E terminates). This breaks the transfer, and there is no possible way to create a new one. Transfers between the E and trains on 63rd would only be at Roosevelt and West 4th, and trips from East 53rd to Sixth would require a wrong-way transfer on the western leg using the B/D.

It’s possible to keep the limited reverse-branching and have Queens Boulevard trains of either type, local and express feed either 53rd or 63rd Street Tunnel. Local-local transfers would then be available immediately east of Queens Plaza. The problem is that this still introduces schedule dependence, on what is most likely the most crowded line in the city now that Second Avenue Subway has taken pressure off of 4/5/6 on Lexington. Conversely, without reverse-branching, both choices of how to match lines have drawbacks: sending locals to 63rd and expresses to 53rd means there is no connection between local stops in Queens and Long Island City, whereas doing the opposite makes the connections better but matches the busier Queens trunk (the express tracks) with the less desirable Manhattan connection (63rd).

That said, despite the drawbacks, something like this disentanglement is requires. New York needs more capacity, and shuffling trains like this effectively creates another half a tunnel’s worth of capacity between Queens and Manhattan and allows higher frequency on Second Avenue Subway, useful given the high population density in the part of the Upper East Side that it serves.

For other cities, let this be your lesson: do not build infrastructure that looks like the left image, unless you know how you can convert it to two intersecting lines with a transfer, the way Toronto did. Branching may look like a nifty way to provide one-seat rides between more pairs of origins and destinations, but it will reduce your capacity, and in the distant future force you into difficult choices in which anything you do, including the no action alternative, will screw someone over. What looked like good planning when the IND built subways under Sixth and 53rd in the 1930s turns out to be bad planning today with what we know of how subways operate around the world.


  1. Lawrence Velázquez

    You don’t consider 53rd Street / 8th Avenue (between Columbus Circle, 7th Avenue, and 50th Street) to be a relevant example of a three-way junction?

  2. FDW

    Your post was spot on about NYC, but what you’re suggesting is physically impossible for Downtown Oakland. 12th and 19th Street stations are extremely close to each other, and you’re asking for an extremely sharp turn in no distance, or moving the Wye to around West Oakland.

    My crayons for Downtown Oakland have instead been focusing on nuking the wye dead.

    • Alon Levy

      Do you have a track map for Oakland? How far apart are the two stations? Even 250 meters between platforms should be okay for the average subway train, and 400-500 should be doable even for a mainline train.

      • FDW

        The Northern edge of the 12th St station (around 14th St), and the southern edge of the 19th St Station (17th St) are about 250-300 meters apart. But the problem is that it really wouldn’t make the system more robust at all. If you want a cheap fix for the system, build the 4th Track through Downtown Oakland. This allows you to rework the system into 2 Lines: Millbrae-SF-OAK-Pittsburgh, and Richmond-OAK-San Leandro/Castro Valley-Livermore/Fremont-San Jose. Of course you’d have to have multiple service patterns, but I’d think this would be far more robust than what you’re proposing. (And your plan wouldn’t get rid of Lake Merritt, but might move it closer to said Lake.)

        My crayons involve converting the Concord line to Mainline rail, connecting it to a 2nd Transbay Tube that connects to Caltrain/HSR at the new Transbay Terminal. But I increasingly despair at the possibility of any real movement on transportation investment in this region, especially as it seems like Johnny Sanphillippo (and others) are starting to look right about where California is going.

        • Alon Levy

          That would break the one-seat rides from both Berkeley and East Oakland to SF. It’s especially bad given that Berkeley’s connections to Downtown Oakland aren’t strong – travel volumes to FiDi are way higher.

          Getting rid of just the circumferential line would still involve a lot of branching, but the scheduling would get simpler. And yes, a second tube could just connect to either East Oakland or Berkeley; Berkeley is easier for a second tube running south of the existing tube, but if the second tube goes to Geary in SF, then East Oakland results in straighter lines (East Oakland-the Richmond, Berkeley-Millbrae).

          • FDW

            Yes, they would. But it could be a cross-platform transfer between frequent trains. And I would say that the weak connections within the East Bay now are a really a result of BART and AC Transit have different fare structures and generally poor connections and scheduling, so most people don’t bother.

            My idea, which I’ve floated before, would have the tube land in Alameda, and take over half of Existing BART tunnel in Downtown Oakland. North of there, it would split, with one branch taking over the Concord line to Stockton, and the other branch swinging to San Pablo before joining the Capitol Corridor to Sacramento. (Which I’d reroute to Vallejo.) I’m operating on the assumption that there just isn’t enough money, given needs elsewhere, to do more 2-tracks under the Bay.

            Within Downtown Oakland, my other key idea was to create a “Soviet Triangle”.

  3. Si Hollett

    The biggest problems in NYC for segregating lines are the Queens area that you’ve highlighted, and the lack of full interchange at DeKalb Av in Brooklyn. There’s also the awkwardness of the Myrtle-to-Midtown service being popular, adding complexity.

    However most connections currently done on the lettered lines by messy services can be done by cross-platform interchange already. I really don’t know what the IND were thinking with building 7 Av in a convoluted way to make useful cross-platform changes and then rendering it relatively useless by running direct services that make the connections provided by crossing the platform.

    • Joey

      I would actually switch the two transbay lines – route the Richmond and Pittsburgh/Bay Point lines via a new tunnel coming from Alameda Point, and connect the new east/west tunnel under Downtown Oakland to the existing tube.

      • Eric

        That involves either the Dublin/Pleasanton+Warm Springs line skipping downtown Oakland, or else a lot of extra construction and abandoning existing tunnels.

  4. Fbfree

    I’m not certain that it’s right to call the connection at 53rd/7th Ave a wrong-way transfer. In my mind, it works more as a line crossing transfer, with the 53rd line continuing south-west and the 6th Ave express continuing south east, with same platform transfers in the southward directions.

    Also, a shameless plug. I recently posted an NYC subway service map. implementing many of the ideas Alon ideas.

  5. Joey

    Does BART actually have any schedule reliability problems resulting from the Richmond – Fremont line and the reverse branching that entails? It seems like (1) old, failure prone equipment and (2) the downtown SF bottleneck are much more serious issues.

    Then again (1) should hopefully be solved by the new trains and (2) might be solved by an additional transbay tunnel. And with that additional capacity frequency on the East Bay branches could increase to the point where reverse branching might actually be an issue.

  6. Joey

    How would you handle phases 2 and 3 of the 2nd Ave Subway? I see a few options, neither of which seem ideal:

    1) Build as designed, with one branch serving the lower part of 2nd Ave and one branch serving the Broadway express tracks (current Q). Both would have limited frequency due to the reverse branch and there is no transfer between the 63rd St tunnel and lower 2nd Ave

    2) Build the originally envisioned connection between the 63rd Ave tunnel and lower 2nd ave. This involves even more reverse branching and the Broadway express’s frequency is still limited

    3) Keep all 2nd Ave trains on 2nd Ave and all 63rd St trains on 63rd St, taking the current connection used by the Q out of service. You still have the transfer problem though, and you’ve eliminated the track connections that might help with that. It might be possible to build infill stations on the 60th and/or 63rd St tracks serving 2nd Ave (although probably not simple in either case) and then have a new station on 2nd Ave Subway serving one or both of them.


    Regarding the Queens BLVD local/express, I wonder if it would be possible to convert 36th St into an express stop so that transfers would be possible.

    • Fbfree

      Would it make more sense to scrap phase 3 of the 2nd avenue subway, and tunnel a 3rd avenue subway instead? This would initially run from 63rd/60th south, with future segments running express to 125 then and up 3rd avenue in the Bronx. The cost of these several miles of boring may be cheaper than any station work required to make connections at 2nd avenue.

        • Eric

          If NYC subway ridership continues to grow at the recent pace, then absolutely. Remember that the Lexington line is still super crowded.

    • Eric

      I would build a new express line, with stops in the Bronx, 125th St, 86th St (connecting to existing SAS), Midtown, and further south as desired. Since it would be built with a TBM, and only have one stop shared with the current SAS, it would be almost as cheap+good as the original 4-track SAS plan.

      • Fbfree

        Hi Eric,
        I was figuring that one would complete SAS phase 2, and possibly a cross-town 125st extension, then build the 3rd avenue subway instead of Bronx and southern extensions.

  7. Bobbo

    Would extending the G on either the Queens Blvd. express or local be helpful in untangling the Western Queens lines? Assuming the 11 St. Connection was taken out of service, the E could run local, and the F and G could run express.

      • Jack

        What about extended the G tunnel from Court Square to new platforms to connect Queens Plaza and Queensboro Plaza and then hook west to transfer to the F at 21st Queens bridge?

      • Bobbo

        I see. Is there any merit to the idea of using the G tracks to just turn locals around after Queens Plaza then, in the way the Fulton St. local was proposed to turn at Court St. after Hoyt-Schermerhorn?

  8. newtonmarunner

    Thanks, Alon, for this.

    Does the 63rd and Lex stop have 2 or 4 tracks for the 6th Ave./7th Ave. Lines (currently F and Q)? Also, can the 7th Ave./BWay Lines go into the 7 and vice versa as Mr. Rapp wants?

    I’m beginning to see your reasons for 2nd Ave. Subway Phase III hesitancy. Phase III either exacerbates the reverse branching problem or provides zero connections. As an alternative to 2nd Ave. Subway Phases III and IV, I like the idea of hooking the mythical Utica Ave. subway into another mythical 2nd Ave./50th St./Amsterdam Ave. subway (to 72nd St.) to relieve the L/A/C/1/2/3 Lines with a more linear connection (if local). This of course would be circa 2317.

    Thanks again, Alon.

    • adirondacker12800

      Extend the L up 10th Ave, across 86th Street and out Northern Blvd.

      In usual service the 6th Ave trains go through 63rd and Lex to Queens Blvd and the Broadway trains serve Second Ave.

      They,, have moderately detailed information on each station too.

      • newtonmarunner

        That’s way too circuitous a path for Northern. Most would still take the 7. For Northern, I’d take the G, or if there’s room, a Queens Blvd. local line, or a 63rd St. tunnel line.

        Also, not enough on UWS/Hell’s Kitchen work around 10th Ave. or 14th St. (which, they’d take the 1/2/3 or A/C/E anyway) compared to the 40s and 50s (50th St. subway).

        • adirondacker12800

          they would take the Flushing line if there were teletranporter booths on Northern to get them to the Flushing line. It would make more sense to have teletransporter booths in Manhattan too. They could avoid the Flushing line.

          The 1, 2, 3, A, C or E don’t serve Northern Blvd. Someone on Northern wants to get to Union Square they can use the skills they have honed by using the bus to an Astoria line station or the bus to a Flushing line station and changing at Grand Central or a bus to a Queens Blvd. station to figure out they can change trains at 86th and 3rd. To the Lex on the west end of the station or the 2nd on the east end of the platform. if they want to go to Rockefeller Center they can change at 86th and Central Park West for the 6th Ave local or 72nd and Broadway for a 7th Ave local.

          • newtonmarunner

            The 1/2/3 @ 72nd and 4/5/6 @ 86th St. stops are way too crowded towards Midtown to have another line transfer there, and the L (itself needing relied going into Manhattan from Brooklyn) isn’t exactly much relief. You can hook a Northern Line into an 8th Ave. Local line via the 53rd St. Tunnel if there’s room. If you have Northern go through the 63rd St. tunnel and 6th Ave. Lines, they can (reverse) transfer to the B/D @ 47-50th Rockefeller Center to get to UWS/Harlem (or SAS Phase 2.5 @ 63rd for 125th St. Harlem) and 14th/W 4th to get to 1/2/3 and A/C/E lines downtown. I think the extended L Line fix is too expensive for what it does (which still isn’t bupkiss but not enough to justify costs). I just like my idea a little more.

          • newtonmarunner

            P.S.: Also, for Northern, 63rd St./6th Ave. or 53rd St./8th Ave. have a magnitude more jobs and connections than 86 St./72nd St./10th Ave./14th St.

          • adirondacker12800

            The three dozen people from Brooklyn who would be stupid enough to wander through that much of Manhattan to get to Queens wouldn’t be a significant problem. Nor would they be very many Brooklynites on a northbound train north of 59th and almost none once it gets past Central Park West.
            No there isn’t enough room in any of the other tunnels which is why the city would building express tracks under the existing L line and extending them back out to Queens.

          • newtonmarunner

            The A in Brooklyn can transfer to the G at Hoyt-Schermerhorn to get to LIC without going into Manhattan (albeit on a local line). The B41/B44 buses on Flatbush/Nostrand also go to the G. If the G is extended to Northern, that will make for a two-seat, Manhattan-free ride from Brooklyn to Northern. Triboro also should help. If the 6th or 8th Ave Lines are used for Northern (which would give Northern better access to Midtown as a magnitude more people work in Midtown than on the A/C/E or 1/2/3 lines in Brooklyn), a three-seat, Manhattan-free ride is still in the cards. I do think the L to Northern is counterproductive in that it increases demand on the overburdened 1/2/3 and 4/5/6 lines as they become the most linear path for many Northern commuters (though SAS and the Astoria and Flushing Lines help).

  9. Untangled

    Apart from reliability and capacity, another benefit of the simplification of the network, it is much easier to automate trains on self-contained lines. Once that’s done, there’s a strong case to even fully automate lines, introduce line-specific (or 2) rolling stock, install platform screen doors (much-discussed issue nowadays) and provide even more capacity (automated Paris Metro does 42 tph). Of course, getting rid of train operators is part of it as well.

    Sounds like strong a case for the Second Avenue Subway to be upgraded to a stand-alone line once it’s full length in Manhattan is completed, considering the new infrastructure, this shouldn’t be too hard. L and 7 are also easy targets for this but only CBTC and ATO is being introduced here.

    • newtonmarunner

      The problem with doing Second Ave. Subway (SAS) as a stand-alone is the lack of connections. Using a branch of the 7th/Broadway Line (or the 6th Ave. Line) for means (SAS) means connections to the A/C/E/1/2/3/N/R/W/B/D/F/M/4/5/6. An exclusive SAS would be a tough transfer at 42nd and Lex, and wouldn’t provide transfers to many of the other lines I mentioned. An exclusive SAS also wouldn’t provide a one-seat ride from UES to Midtown West or a less congested two-seat ride for those on the 6 in the Bronx to Midtown West (likely to get a seat at 125th St.). Also, the 4/5/6 northbound to Midtown East doesn’t need relief like the Lexington Ave. Lines (though the M14 and M15 buses have a combined 80K ridership). Finally, mythical regional rail can be extended from Grand Central to Fulton St. via Union Sq. to relieve the Lex Lines. I’d rather have the focus be on 125th St., Utica Ave., Triboro, Nostrand Ave., and possibly Northern (and relieving the L via Utica Ave. Subway) subways as well as regional rail than SAS Phases III and IV.

      • Untangled

        I should expand on my stand-alone thought bubble a bit more. I agree that Bronx and Brooklyn connections are needed once the full length is completed, but I’m imagining these connections being upgraded/converted to become part of the new SAS stand-alone system rather being part of the existing system. The links between the old system and the new system will be severed from the old section, except for transfer tracks to move equipment at certain places.

        Essentially, I’m imagining an entirely new division anchored by the SAS, a C division, taking over some some existing lines in the Bronx and Brooklyn. This C division system will be fully automated and have platform screen doors at all stations including on the upgraded/converted stations. I think that steps like this are needed to take the subway out of the early 1900s and correct the branching problem described in the article. Any other action is really just patching up the failures of the existing system but to bring it world-class standards, you need to reinvent it and it should begin with the C division thought bubble (Which, tbh, I only came up with when reading this article) as a dry run. Once that’s done, work should be made to bring the rest of the system up to C division standards.

        Now, as to a one seat-ride to Midtown West, I’m not sure what the obsession with a single seat ride is but I don’t think such a thing is to be expected in a city like New York, even some LIRR commuters don’t have a single ride to Manhattan and has to change at Jamaica, it shouldn’t be a difficult ask to change for subway commuters, the best place for a transfer in this case looks to be at Lexington Av/53 St. Also, except for the Triboro, some of your lines could be built as part of the new C division. That said, I agree that regional rail to Lower Manhattan is a better option that Stage 3 and 4 even if it means that my dreams of a C division is flushed down the toilet. There needed to be serious densification in the suburbs within the station’s vicinity to make it worthwhile though.

        • Alon Levy

          The transfer at Jamaica on the LIRR is cross-platform. In contrast, an SAS/53rd Street Line transfer would involve walking from 2nd and 55th to 3rd and 53rd, a distance of 350 meters at the lowest, potentially going as high as half a kilometer depending on which part of the train you stand on. This makes the one-seat ride provided by SAS is valuable in ways one-seat rides on the LIRR aren’t. It’s also useful not just for UES-West Midtown commutes, but also UES-UWS/West Harlem commutes, which are now down from a three-seat ride via the 42nd Street Shuttle, with annoying transfers at Grand Central and Times Square, to a two-seat ride, with just one annoying transfer at Times Square. My commute in 2009-10, 72nd/York to Columbia, was 50 minutes then and would be 37 minutes now via SAS; without the Q going to Times Square, it wouldn’t have been any better.

          • adirondacker12800

            Q across 125th and L across 86th would relieve crowding at Times Square and Grand Central a bit..

          • Untangled

            If anything, your comment about UES to UWS/West Harlem commutes on the subway really highlights how poorly the bus network is performing. If the buses were up to scratch, you really shouldn’t need the subway for that sort ot trip.

            Also, I don’t think that 350 metres is too bad, when I was in Singapore, I remember one time I had to walk ages in a tunnel to change lines and both lines had the same station name. Besides, is there not a way you can get the SAS to pivot to 3rd Ave and then back on to 2nd Ave after that stop? I can’t see why you have to stick to one road when you’re using TBMs.

            That said, I think this also raises the importance of getting the buses up to scratch to support big changes and simplify the subway network.

          • Alon Levy

            In Singapore, those long transfers were a deal-breaker for me. In undergrad, my commute by MRT would be Orchard-Buona Vista. I could get a cross-platform transfer at Raffles Place, or a pair of long transfers at Dhoby Ghaut and Outram Park. I almost always went for the two-seat ride rather than the three-seat ride, even though it was somewhat more circuitous. (I actually rode the bus more than the MRT, since it got me closer to home in the expat ghetto than the MRT did.)

            There are a lot of arguments for and against transitioning to Third to build phase 3 of SAS. The main arguments for are (as you note) easier transfers, and better service to the Midtown job cluster. The main argument against is that even with a TBM there may be construction difficulties at the transitions, say around 59th-63rd and 3rd-7th to avoid having to go under foundations of tall buildings.

        • newtonmarunner

          A few points:

          If you do an independent SAS, and have both branches hooked into the B/D in the Bronx (the only lettered lines in the Bronx), you lose the 125th St. Subway, which provides a linear two-seat ride from Washington Heights to UES. You also lose Highbridge, etc. one-seat ride to Midtown West, UWS, and West Village.

          With equal fare zoning regardless of mode, plenty in the Bronx who live off Webster, 3rd Ave., etc., will take RR (or super-express subway lines as I like to call them in NYC) to Grand Central and FiDi.

          And the subway lines aren’t mine — they’re Alon’s. He’s the one with all the good ideas.

          • Untangled

            I don’t think taking over a letter or number for the C division should matter too much especially since disruptive and major work will be required for any route. I would keep the 125th St subway, I’m imagining there will be two branches in the new system, 125th is one of them and the other one will a taken over Bronx line. I think the 5 might be a good candidate although this one will require some work. From Eastchester, it’ll travel down to Morris Park, then in a new tunnel, pivot over to underneath Third Ave, then stop alongside the existing Lexington Ave Line 125th station and then pivot back to Second Avenue for its trip down Manhattan. But 4 or 6 should work as well, in fact, 4 might be the least disruptive conversion option since it already parallels the B/D. I don’t envision a conversion that is not majorly disruptive for any route so you might as well make the most of it and convert a numbered line.

            Yes, equal fare zoning is important too but I don’t see that happening soon. However, for now, I would like to see the more expensive express bus Metrocard fares becoming valid on the LIRR and MNR, especially since the pricing difference isn’t as drastic. I would love to see an NY RER eventually.

          • newtonmarunner

            @Untangled — I don’t like the idea of one SAS branch to 125th St. and one under the river to the Bronx. Transit rules are that given the capital/operating costs of going under a river, you need to have the full trunk line go under the river.

            My personal preference would do the reverse of the SAS Q(/R) line with a mythical Utica Ave. Line. Alon has proposed running Utica parallel to the L, entering Manhattan at Houston, and hooking Utica into SAS. I say instead of hooking Utica into SAS, hook Utica into a 50th St./9th/Amsterdam/145th St. to Hunts Point Subway that among other things,
            (1) relieves the L and A in Brooklyn from a more linear (if local) path to Midtown
            (2) provides a two-seat ride to LIC for Flatlands, etc. (Utica to G) without going into Manhattan and one-seat ride from Williamsburg to Midtown
            (3) relieves the M14 and M15 buses with SAS sorta-Phase III
            (4) relieves the 2/3 @ 72nd St. and A/B/C/D @ Columbus Circle by giving UWS a one-seat ride to Midtown East, and
            (5) relieves the 6 by giving those on the 6 in the Bronx a two-seat ride to UWS.
            Definitely not cheap, but the combination of regional rail, Triboro, SAS, and this Utica Ave.+ Subway Line will solve a lot of big problems.

          • newtonmarunner

            Sorry, should have written instead of “reverse SAS” a UWS/Midtown East/LES crosstown vs. a E. Harlem/UES/Midtown West crosstown line.

          • Untangled

            “I don’t like the idea of one SAS branch to 125th St. and one under the river to the Bronx. Transit rules are that given the capital/operating costs of going under a river, you need to have the full trunk line go under the river.”

            I know that you have to build an expensive underwater bit but the high cost of construction in NYC strikes again. I’m only proposing this to help introduce driverless trains to New York as a starter measure and the 125th Street branch would have really only been the equivalent of a short turn of the SAS not exactly a full line so most SAS trains would go over the Bronx River. Some of the high cost of construction can be recouped by long term savings from getting rid of on board staff. The trunk rule seems a bit silly, considering that the full SAS capacity of the underwater crossing probably won’t be required unless you’re doing some Hong Kong style new town redevelopment in the Bronx. If you really needed that extra capacity across the river, then maybe do 125th Street as a standalone shuttle.

            I’m not sure what a subway underneath Amsterdam/9th Ave achieves, you’re proposing the west side of Central Park to have 10 tracks and east to have 6 (with 2 of which that do not run the whole length), when the east is denser than the west. And if you’re going to connect it to a Bronx Line, that’ll make it much more circuitous. Instead of more criss-crossing lines aimed at providing one-seat rides in Manhattan (I’m off on another tangent here), I would rather make the subway a grid and get everyone to change for Midtown or Midtown East. Step one would be to cut the 53rd Street Line off any North-South connections, build a new terminus at 53rd/9th or around there and upgrade transfer links. I’m entirely too sure how to re-route lines but I’m just putting it out there.

          • newtonmarunner

            @Untangled — Converting the 125th St. Subway into a shuttle means a three-seat ride from Wash. Heights to UES (or a more circuitous two-seat ride if SAS goes to Times Sq. and 7th Ave./Broadway Lines).

            Subway on 50th/9th/Amsterdam gives UWS/Morningside Heights a one-seat ride to Midtown East. Connecting this subway to Hunts Point Av. gives Soundview and others on the 6 a two-seat subway ride to UWS. With transfers @ Columbus Circle and 72nd/Broadway 1/2/3 and A/B/C/D are relieved.

            UES also has commuter rail service that UWS doesn’t have. With fare integration and more frequent service from regional rail, those regional rail lines on Park count as super express subway lines for UES.

          • Untangled

            Again, I’m not really sure why a one seat ride is a must when you can change trains like in the rest of the world. The west side already has more subway tracks than east, it’s less dense, is not in need of crowding relief and does not connect as much to the Bronx as the eastern side lines. Even if you hooked a Bronx line onto this new 9th/Amsterdam line, your idea would be more circuitous than going via the east. Using the Park Ave as rapid transit doesn’t change those factors, even if you used that line for some Bronx relief.

          • adirondacker12800

            Using Metro North tracks, for something other than Metro North, would get the people already crowding them a little annoyed.

          • newtonmarunner

            @airondonacker — Yes, there would be NIMBYs opposing transforming Metro North into Regional Rail because of lost skip-stops and “those people” crowding “their” overcrowded trains. But headways would be more frequent and commuters on MNR would have direct connections (or at worst cross-platform transfers to) under RR to Penn Station, Union Sq., Fulton St., etc. That’s the benefits of a more logistically efficient fully integrated system with integrated fare zoning. There are definitely winners and losers, though.

          • adirondacker12800

            if the train is full of people who got on in Westchester, people in the Bronx can’t get on. All it would do is annoy everyone.

          • Alon Levy

            They can stand, same way people who get on the 4/5 at 86th Street stand after everyone in the Bronx got on and took all the seats.

          • adirondacker12800

            IN 1987 when there was standing room.It’s not 1987 anymore.

        • Eric

          The problem with multiple-seat rides is that when you often have to wait 15 or 20 minutes for a particular line, transferring has a massive time penalty. Of course ending interlining will allow a higher frequency on each line, but this future benefit is not immediately obvious to people now. And the system’s general reliability has to improve.

    • newtonmarunner

      Forgot, also the other big problem with making SAS exclusive is you revert back to the lettered lines (A/C/E, B/D/F/M, and N/Q/R/W) having 12 tracks going from Brooklyn to Midtown but only 10 tracks going from Bronx/Queens to Midtown (see branching, reverse). SAS Phases I and II helped mitigate this problem.

      I’m with you on keeping things simple, though. -:)

      • Untangled

        Yeah, the complexity of the subway network really staggers me when I look at other places. I only suggested that SAS become seperated with full automation and PSD because it was new so it was an easy target, along with the already relatively independent L and 7 but the ship has sailed on those two with CBTC.

        • Eric

          The only remotely comparable system I can think of is London’s. The District/Circle/Hammersmith&City/Metropolitan Line is a giant mess, and the Northern Line is interlined too.

          After that, all I can think of is Washington DC (Blue/Yellow lines). And of course a bunch of European cities have interlined tram or suburban rail networks, but I assume we aren’t counting those. Beyond these, I’m aware of virtually no interlining anywhere (excluding radial branches of a core line).

          • adirondacker12800

            it’s difficult to achieve interlining if your system is two or three lines.

          • threestationsquare

            Other examples of reverse-branching on full metros:
            – Caracas lines 2 and 4
            – Munich rush-hour line U7 (reverse branch of U5)
            – Munich rush-hour line U8 (reverse branch of U3)
            – Delhi Green Line Inder Lok and Kirti Nagar branches
            – Singapore Circle Line Dhoby Ghaut and Marina Bay branches
            – Shanghai lines 3 and 4
            – Tokyo Yurakucho and Fukutoshin lines
            – Tokyo Namboku and Mita lines

  10. BikeGuyEmoji

    “…sending locals to 63rd and expresses to 53rd means there is no connection between local stops in Queens and Long Island City…”

    If the structure for the old el was reused, would sending a tram from Northern Blvd across the Queensboro bridge upper deck help? The tram would elevate before leaving Northern and take over the Astoria line tracks. The Astoria line would shift north, roughly in line with the 60th st tunnel. The Queensboro Plaza station simply expand to two platforms. The tram would take the two center lanes of the bridge and run across 57th or 59th and/or branch elsewhere.

    A dealbreaker here is there might not be enough clearance on the lower level of the station for a level platform. Also the crossover between A & B division would have to be redesigned or eliminated.

  11. Pingback: Commuter Rail Infill Stops | Pedestrian Observations
  12. Greg Jordan-Detamore (@GJordanDetamore)

    Excellent blog post! This is something I’ve been thinking about for years, and actually drew a map similar to the concept here (see below).

    There are a few things you didn’t discuss and I’m wondering what you think about them:
    — The IRT. You said it’s fairly easy to split the 2/3 and 4/5 (I agree). I take it that’s running on the assumption that the Rogers Ave Junction (near Franklin Ave) in Brooklyn gets rebuilt and the 2/3 go down Nostrand while the 4/5 go down Eastern Parkway. But what about Harlem and the Bronx? My idea was to kill the 148 St terminus for the 3 and instead run the 2/3 to E 180 St (then splitting), while sending the 4/5 up Jerome Ave. Did you have the same idea in mind?
    — Brooklyn BMT. My idea was to have the N/R be local down through Whitehall St and under 4th Ave (splitting at 59 St) while having the D be 4th Ave express then the West End Line. The B would be eliminated and the D and Q would run at a “double” service level, with the Q doing Broadway express, Manhattan Bridge, and the Brighton Line (alternating between Brighton local and express, so basically the the Q and Q-diamond setup). What do you think?
    — Would the M return to Nassau Street?

    In my concept, the line namings would actually be simplified so that there’s only one letter or number going on each Manhattan track. Here are maps to illustrate:
    A division:
    B division:

    Here are textual descriptions:
    (1) As-is: 7 Av Local.
    (2) Wakefield/Eastchester to Flatbush via 7 Av Express and Eastern Pkwy Local. Bronx trains would alternate branches (off-peak all local; there could be some peak-direction express aspect).
    (4) Jerome Av Local, Lex Av Express, Eastern Pkwy Express (then local past Franklin Av), Liviona Av Local/Utica Av Local (alternating).
    (6) As-is: Lex Av Local.
    (7) As-is: Flushing Line.
    (A) Concourse Local (with half of peak-direction trains being express?), 8 Av Express, Fulton St rotation (rotating between Fulton Local, Express to Ozone Park, and Express to Far Rockaway).
    (D) Inwood Local, Central Park West Local, 6 Av Express, Manhattan Bridge, 4 Av Express, West End Local.
    (E) Queens Blvd Express (alternating branches), 53 St, 8 Av Local. (Or could be Queens Blvd Local.)
    (F) Queens Blvd Local, 63 St, 6 Av Local, Culver Line (during daytimes alternating between locals to Church Av and expresses to Coney Island).
    (G) Crosstown, same as-is except would terminate on the Queens-bound local track at Queens Plaza (if they can pull off the timing).
    (J) Nassau St Local, Williamsburg Bridge, Brooklyn Broadway Local (alternating between Middle Village and Jamaica). Unsure about the eastern part of the line/what to do about peak service patterns (skip-stop, using the center track, etc.).
    (L) As-is: Canarsie Line.
    (N) Astoria Local, 60 St, Broadway Local via Whitehall, 4 Av Local (alternating between Bay Ridge and Sea Beach Line).
    (Q) 2 Av Local, Broadway Express, Manhattan Bridge, Brighton Line (alternating between local and express).
    (S) Shuttles as-is. Possible 148 St shuttle to replace former (3) service, though it’s probably too expensive/not worth it.
    No Second Avenue Subway phase 3 or 4.

    • newtonmarunner

      I like a lot of it. I really like getting rid 145th and 148th Harlem stops for more service in the Bronx, no reverse branching, and less track sharing. I, however, think having Washington Heights go virtually all local to get to parts of FiDi (“D” to “N”) close to the ferry with a two-seat ride will be a tough sell. Also, the G — does it share tracks with the F?

    • Fbfree

      Hi Greg,

      There’s one fatal flaw, and a few issues with the B-division plan.
      The fatal flaw is that WTC won’t have a turnback capacity of 30 tph that you need to run all QB express services to it.
      Other issues include poor load balancing on the N and D, and the standard issues with Myrtle line and QB local services connecting to the rest of the network.

      My own take on a simpler network is here,

      • newtonmarunner

        I keep coming back to this post. Always love new transit configurations.

        I do think that Washington Heights — not the Bronx — should get the CPW Express Line. Grand Concourse is right by the express 4 train, and the Grand Concourse line has during rush hour an express until 155th St.

        I’d probably send QB express to the 63rd St. tunnel and 6th Ave. local and the QB local to the 51st St. tunnel 8th Ave. local terminating at WTC. I don’t like the parabolic QB Local/8th Ave./Fulton St. local/Express, either, and prefer a more “L-shaped” line if a line cannot follow a linear path. But neither option is great.

        Definitely agree with Greg on the numbered lines and the 3 in Harlem’s insufficient usefulness when compared to serving the Dyre Ave. Line.

        At any rate, good job, folks.

  13. Pingback: Infrastructure for Mature Cities | Pedestrian Observations
  14. J BRANT

    Would it be possible to build a tunnel between West Oakland and 12th St, but have the new tunnel approach 12th St. from the north? This way all trains can use the existing stacked platforms but both levels will have westbound trains. It would look like something from the Taipei metro, which uses stacked platforms to allow fast cross platform transfers. It would also continue to allow Richmond – San Jose trains since you keep the portion of the wye between 12th St. and Lake Merritt.

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

This site uses Akismet to reduce spam. Learn how your comment data is processed.