When are Express Trains Warranted? Part 2: Subways

Last week, I wrote about which regional rail lines should run local and express trains and which should only run locals. I gave various guidelines, specific to that mode of transportation. The same question makes a lot of sense for subways (or elevated lines), and is especially important for large cities building new metro networks, typically in developing countries. The key difference is that subway trains almost always run so frequently that the only way to have local and express services is to build four-track mainlines, like in New York. So really, this is a question about when it’s useful to build four-track subways instead of two-track subways.


The cost of a four-track subway is higher than that of a two-track subway. However, the difference depends on the method of construction. Cut-and-cover four-track subways do not appear to be much more expensive than two-track subways; in the 1900s, New York had little to no cost premium over London or Paris. The First Subway’s underground segments, slightly less than half of which are four-track, cost $39 million per km in today’s money, compared with about $29 million per km in Paris in the same era (see sources in this post). The Metropolitan line in London cost about the same: £1.3 million for 6 km in the early 1860s, or about $30 million per km.

In contrast, boring four tracks appears to cost twice as much as boring two. It’s hard to find examples, since four-track bores are extremely rare; the examples I do know, such as the East River Tunnels and the tunnel carrying the Lexington Avenue Line under the Harlem River, are short. There are bound to be efficiencies in engineering and sitework reducing the cost of boring, just as there are with cut-and-cover, but the majority of the cost of tunneling is the boring and the systems. The majority of the rest is stations, and the local stations can be built for not much of a premium over stations on a two-track line, but the express stations require considerably more excavation.

What this means is that cities that build cut-and-cover should probably aim to build four tracks rather than two. In retrospect, Paris should have built Metro Line 1 with four tracks: the narrowest street segment, Rue de Rivoli, runs for 3 km and is about 20-23 meters wide, which can take four tracks if there are only local platforms, and everything else is at least 30 meters and could take four tracks and express platforms. The only express station under Rivoli would probably be Chatelet, where there is a wide square where the station footprint could expand.

The question of whether to use cut-and-cover today is a separate issue. It’s easier for the first few lines than for subsequent lines, which have to cross under the old lines. Nonetheless, it’s still in use, for example in China; the lack of express tracks on Chinese subways has led to criticism on some railfan forums, particularly by Japanese railfans, who are used to the fast express trains of the JRs through urban areas. But in India, the longest underground segment in Delhi, on the Yellow Line, appears to be bored, running deep under Old Delhi. The one potential pitfall is that bored tunnels, while generally more expensive to construct than cut-and-cover tunnels if wide streets are available, are nimbler, making it easier to build lines to high standards, with wide curves.


The two obvious benefits of express subways are speed and capacity. New York averages a decent speed, just under 30 km/h, buoyed by express trains that average about 36 km/h. When frequency on both the local and express tracks is adequate, which it regrettably isn’t most of the time in New York, it’s possible to use cross-platform local/express transfers to improve trip times even for people on the local stations.

The capacity benefits are sometimes compromised by transfers. In New York, the express tracks are consistently more crowded than the local tracks on the main lines, and the Upper West Side in particular sees the city’s most overcrowded trains on the express tracks run alongside the second least crowded on the local tracks. However, on the East Side, the local and express trains under Lexington are both quite full. The number of passengers they carry would overwhelm any two-track line: to carry the same number of passengers on just two tracks at the current peak frequency, each train would need about 2,100 passengers, more per unit of train floor area than the most crowded Tokyo Metro lines.

There are also indirect effect of four-tracking. The most distant stations are presumably express-only, or, if not, passengers will transfer to an express train at the first opportunity. This means that adding local stops does not increase trip times for passengers far out, which in turn argues in favor of tighter stop spacing on local trains, to provide more coverage. Evidently, New York has one of the smallest interstations (on local trains) of any major metro network in the world, surpassed only by Paris, which built the Metro without regard for the suburbs. Second Avenue Subway, built with just two tracks, has wider stop spacing than other Manhattan north-south mainlines, missing 79th Street, with the planned phase 2 stopping at 106th instead of 103rd and 110th.

Network effects

I have hammered repeatedly that metro systems need to avoid three pitfalls of network planning:

  1. Tangential lines, starting as radial farther out but becoming circumferential closer in rather than staying radial and serving city center.
  2. Reverse-branching: a common trunk in an outlying area splitting into branches in city center, the opposite of the more normal branching arrangement.
  3. Missed connections, in which two lines intersect without a transfer.

It is easier to miss connections when the stop spacing is wider, because city centers are so small that line 1 can easily miss the road that line 6 will later run under. The Paris Metro has just one missed connection (M5/M14), and the city made a failed effort in the 1900s to bring Line 5 to what would be the transfer station, Gare de Lyon. New York has only two missed connections among the lines built from 1900 to the 1920s (Junius/Livonia on the 3/L, Bowling Green/Whitehall on the 4-5/RW), both of which are easy to fix; the city’s tens of misses come from a deliberate decision to avoid transfers in the lines built in the 1930s. Systems with wider stop spacing have more opportunities for missed connections, such as Tokyo, Shanghai, and Moscow.

Four-track mainlines short-circuit this problem doubly. First, fewer lines are required to provide the same coverage and capacity. On the eve of the Great Depression, New York had six subway mainlines, three with four tracks and three with two, and 2 billion annual riders. Even if a city insists that all transfer stations be express (which is the case in New York among lines built before the 1930s, though at one station express platforms were only retrofitted decades later), there are fewer potential intersections; at city center, even the express lines are unlikely to have very wide stop spacing.

Moreover, it’s fine to miss transfers on express lines. The reason missed connections are so bad is that the alternative option typically requires a three-seat ride. For example, the 1 passes between Columbus Circle-59th Street and 50th Street without a connection to the E train, leaving connecting passengers with one of two options: ride to Times Square and transfer there, which is both a detour and a very long walk between the platforms, or change to the B or D at 59th and then again to the E at 53rd; the latter option is the faster one, saving about 5 minutes of in-vehicle time and 3 minutes of transfer time.

But if one of the transfers is cross-platform, then a three-seat ride is less onerous. This is especially true if this transfer is local/express on a long shared trunk line, where the transfer stations usually don’t get as crowded as between two separate trunks. This means that it’s possible, if there is no better option, to have local-only transfers. Two of the top stations in New York, 53rd/Lex and Columbus Circle, are local-only transfers.


Small cities don’t need to think about express trains very much. Transit cities with about 2 million people in their metro areas, such as Vienna, Stockholm, Prague, Budapest, and Hamburg, make do with a handful of metro lines, sometimes supplemented by regional rail. But as they grow, the number of urban rail lines they need grows to the point that a coherent two-track network is too difficult, not to mention too slow. Paris manages to make do with the RER acting as the express layer, but outside India and Pakistan, cities that are just beginning to build their metro networks don’t have large legacy commuter rail networks to leverage for express service. In Africa between the Sahara and South Africa, nearly everything has to be built from scratch, and the same is true of Bangladesh and much of Southeast Asia and Latin America.

Moreover, in developing countries, a large number of cities either are megacities or can expect to grow to that size class soon. This is less true in Latin America, nearly all of which has at- or below-replacement birth rates, already high urbanization rates, and negative net migration (thus, the main case for four-track subways in Bogota hinges on the fact that the city has absurdly wide roads for cut-and-cover, not on future scale). However, in poor countries with high birth rates and low urbanization rates, cities can expect to increase their population many times over in the next few decades. Lagos and Dhaka are already huge, and midsize cities such as Nairobi, Kampala, Dar es Salaam, Addis Ababa, Khartoum, Accra, Abidjan, Douala, and Sanaa can expect to grow to that size class within a generation, as can midsize cities in countries with lower birth rates, such as Chittagong, Hanoi, and Ho Chi Minh City. In such cities, transportation planning should presume much greater scale than may seem warranted today.

A smaller-scale version of this principle can be found in cities that are growing rapidly due to immigration, especially Singapore but also Sydney, Melbourne, Toronto, Vancouver, and Stockholm. Such cities might want to consider infrastructure investments based not on their present-day population levels, but on those that they can expect after 30 more years of high migration rates. Vancouver and Stockholm are too small for express subways no matter what, and Toronto has already built its central spines with two tracks (and is upgrading its regional rail network to act as an express layer), but Singapore, Sydney, and Melbourne should all think about how to plan for the possibility that they will have 10 million people within 50 years.


  1. Untangled

    Paris manages to make do with the RER acting as the express layer, but outside India and Pakistan, cities that are just beginning to build their metro networks don’t have large legacy commuter rail networks to leverage for express service.

    Yet if you look at Delhi, they have an extensive legacy network to use for regional rail but they’re not going to use it. Instead of using legacy lines as an express layer to the metro, they’re going to build an entirely new high speed regional rail network, called the RRTS. Is it because they want a more metro-style service that only dedicated infrastructure can provide? I don’t know but they are completely skipping over the legacy network.


    I can’t think of any other network that will be like the RRTS apart from Seoul’s GTX network or the Pearl River Delta’s high speed system. (Beijing is getting express lines soon (R1, R2, R3) but they’re not going to be as long or average as fast as RRTS or GTX.)

    • Eric

      I’m guessing it’s because of agency turf battles, not genuine need.

      Though it’s conceivable that they expect future intercity traffic (currently extremely small, relative to city size) to take up all the space on the existing tracks.

    • Alon Levy

      I don’t know why Delhi is doing that. Mumbai and Kolkata both have extensive regional rail networks; it’s not like in the US, where there’s no domestic example. My best guess is that the Mumbai network is so overcrowded and unsafe that Delhi associates its high ridership with failure and is looking for extending its own lesser-used mode.

      South Asia generally is making a series of mistakes, often induced by foreign consultants who recommend standard-gauge metros instead of broad-gauge ones for mainline compatibility. Kolkata has a beautiful corridor for four regional rail tracks connecting commuter lines at the two ends, but instead it’s building a two-track standard-gauge metro on it.

      • Untangled

        IIRC the Delhi Metro is standard gauge because the original head of it, E. Sreedharan, wanted it standard gauge. He didn’t get it orginally but got it with more recent lines. He was not a consultant, in fact he was a well known engineer. And later other Indian cities copied Delhi’s standard gauge so I don’t think the gauge thing is entirely the work of foreign consultants.

        • Alon Levy

          I read that the choice of standard gauge for the more recent lines came from private pressure, but I don’t have any access to sources that you don’t on this (it was probably from Wikipedia).

          • Matt Hutton

            The other advantage of standard gauge is that it makes the subway trains cheaper to buy.

      • Nicolas Centa

        I think they prefer to leave the main line network to obsolescence and build something new from scratch instead of improving it.

        It allows to keep poor people from using the new service by setting a high price, and I think in many countries this is the reason.

  2. Joseph

    What should the station spacing be on new 4-track metros? New York has 400 to 800 meter local station spacing and 1600 meter express spacing., right? But many of the newer metros are planned with 1500meter average spacing. Would the local stops be 800m apart (except perhaps in the CBD) and the expresses every 3km?

    • anonymouse

      When I crayoned out a local/express service for the Blue Line in LA, that was the stop spacing I used. It seems like a reasonable compromise, especially if it works out with local conditions (important transfer points and such).

  3. sg

    It seems to me that the load distribution between the 1-2-3 is worse than it should be because the 1 is so slow. If ~half the local stations on the 1 train were shut down, probably many fewer people would transfer to the 2-3 at 96th. (I guess this is the same argument as people usually make for wider stop spacing on buses.)

  4. Eric

    “Second Avenue Subway, built with just two tracks, has wider stop spacing than other Manhattan north-south mainlines, missing 79th Street, with the planned phase 2 stopping at 106th instead of 103rd and 110th.”

    I think that’s more because station construction is so much more expensive when they are in caverns far underground, as opposed to being a few steps down from the street.

  5. Eric

    What do you think of taking the current 4-tracks in Mumbai and converting them to 6- or 8-tracks?

  6. Bgriff

    Another benefit to the 4-track system, related to its higher capacity, is what it enables for density. In order for Midtown Manhattan to work, you need to be able to get a ton of people to specific places like Rockefeller Center and Grand Central. With 2-track systems, you would need to have two (or more) lines converge on a single spot, which may be tricky to achieve depending how the streets are laid out and where else the lines need to go—given the small number of available north-south routes for the subway in New York, it seems likely that if the system had been built all with 2 tracks, there might have been a call to stack lines on top of each other under the same avenue or something. Or just live with much lower density than we now have. Modern 2-track systems like the Hong Kong Island Line, with its huge trains and extremely high frequency, can get a lot of the way to the kind of capacity New York’s 4-track lines are currently able to offer, but of course a 4-track Island Line could offer even more capacity.

    It also, in a way, makes the system more comprehensible. The NY Subway is famously hard to understand for newcomers, but at the same time, for people who kinda get it — say those who have lived here a few months, or tourists on their third visit — the NY system is much easier to understand than if the 5 4-track Manhattan trunks were replaced by 10 2-track lines. Tokyo is kind of an example of this: they continually need more capacity to key areas and so keep building new 2-track lines attacking those areas from slightly different angles. It might be more efficient and easier to understand for a newcomer if some of the existing lines just had more capacity with 4 tracks.

    That said I’ve never really understood why express subway arrangements are so uncommon. Which raises the question of why other cities mostly didn’t build them, though given how rare they are, maybe the better question is why New York took such a big chance on them.

    • Alon Levy

      The first city to build a subway was London, in the 1850s and 60s. It was a short line and it didn’t need that much capacity for the most part; it was getting 12 million annual riders. Even then, one part of the system did have four tracks, between King’s Cross and Farringdon: two tracks were used by the Met and are still used by the Underground, two were used by other railroads and are now used by Thameslink. By 1900, when there was a wave of other big cities building subways, cities had grown and there was way more demand for CBD-bound travel, but Berlin already had four-track spines in the Stadtbahn and the Ringbahn, and Paris was digging up streets by hand and didn’t really have room for four tracks except on Lines 1 and 2+6. Paris was also aggressive uninterested in serving the suburbs, which were poor, which is why the Metro is maximally incompatible with mainline rail (it runs on the right, French mainline rail on the left). On one choice segment Paris did build four tracks, carrying M8 and M9, but they’re all local and don’t have cross-platform transfers.

      • Bgriff

        I can also see how it is hard to make room for 4 tracks in a lot of European cities (although many do have big boulevards that would leave plenty of room, but I’m not sure if all of those boulevards were there at the time when the subways were being built).

        I’m more curious why Asian cities didn’t go for 4 tracks. Some places simply didn’t have the money to be that forward-thinking, but the subways in Beijing and Shanghai could definitely have used it on some lines, and those systems don’t exactly have modest ambitions.

        • adirondacker12800

          Stack them like the NYC subway is under Lexington Ave, and Central Park West.

          • Alon Levy

            Sure, but that’s uncommon in New York – for the most part, four-track subway lines are single-level.

          • adirondacker12800

            The legend is that Lexington isn’t wide enough for four tracks. I dunno how true that is but that is the legend. Central Park West, expect for a few days a year, all of the demand is on the west side of the street. Other things being equal I suspect a deep hole is cheaper to dig than a wide hole – for instance 8 units of exterior instead of 10. ….unclear how things are arranged under Flatbush Avenue where there is 6 tracks of subway. Sixth Avenue between Herald Square and West 4th is odd but that likely comes from three phases of construction with decades in between each one. How much stuff can be wedged in depends on how hard you want to think about it and how much money you have.

          • Alon Levy

            Lex is borderline. It can fit 4 tracks but only with local stops, not express stops; the same is true of Rivoli, which is a hair narrower. I think this is why Lex goes back to a single-deck 4-track configuration north of 96th and south of 125th: all the stops are local anyway, so there’s room.

      • Michael James

        Paris was digging up streets by hand and didn’t really have room for four tracks except on Lines 1 and 2+6. … On one choice segment Paris did build four tracks, carrying M8 and M9, but they’re all local and don’t have cross-platform transfers.

        Isn’t the main issue to do with the stations, ie. on line 1 you might fit in 4 tracks (though there are a few pinch points) but surely not 4 (or 2 & wide central-shared) platforms and 4 tracks. However inn 1900 no planner would really be thinking about having express and stopping trains over a approx. 10km route through the heart of the city (it has since expanded to 16km) with only a few stations that do not intersect other lines; such a scheme would mess up the concept of the dense network. And look at it: 120 years later this line still handles its 214m pax p.a., ie. it continues to serve this heavily trafficked (and heaviest tourist) route very well. Of course it has had a major upgrade to equip it for the future–and that is what is wrong with most of NYC’s system, whether 2 or 4 tracks it has been starved of necessary investment and upgrades.

        Re Paris M8 and M9, we’ve discussed that recently. They only overlap for a relatively short segment of 6 stations; the reason for the overlap is probably related to the limited possible routes which had to use the new wider boulevards created by Haussmann. (In a very real sense the Paris Metro is Haussmannian; London did not have that option and had to do deep bore tunnels*.) The reason M8+9 are double-decked is because of engineering issues, ie. the ground is soft and unstable requiring deep underpinning and it was much cheaper to double-deck than to have to underpin a double-width structure. Similar issues may account for the double-decking adirondacker talks of in NYC; or feasibly it has to do with the hi-rise buildings foundations (or sub-floors) combined with geology in particular locations (often water courses or ancient marais etc).
        *trivia for peridromophiles: talking about deep-bore, I happened to catch the Michael Portillo railway tv series this week. He was in Kiev which has the deepest, longest descent to a metro in the world, at 100+m. But just this second I remembered the Gotthard Base Tunnel which has an emergency escape shaft in its middle that runs vertically for 800m. Apparently there were serious thoughts about making this a station for the HSR that will use the tunnel but it has been abandoned. Too bad.

        • Alon Levy

          Champs-Elysees and Cours de Vincennes are both much wider than Broadway in New York, so there’s clearly enough space. In New York they routinely fit 4 tracks and 2 central platforms under 30-meter avenues, and the entire course of M1 is that wide except the narrows under Rivoli between Bastille and the Louvre. Even then the narrows are about 22-23 meters wide, comparable to Lexington, which does host 4 tracks and local platforms at 103rd, 110th, and 116th. The only express stop under Rivoli on M1 would be Chatelet, in an area where there would be room to expand to 4 tracks and 2 platforms.

          M1 has a lot of transfer stations in rapid succession today, but that wasn’t the case when it was built. It took until the 1920s to build M9, the first line to cross M1 at a place that wouldn’t have been an express stop in 1900 (Nation, Gare de Lyon, Bastille if it was still impossible to drag M5 to Gare de Lyon as was planned, Chatelet, Concorde, Etoile).

          And yes, the line gets 215 million annual riders. It’s a 2-track metro with very high peak frequency and high turnover. This is what metro lines are capable of even with short-ish trains (cf. the Central line at 289 million). But then you have the Lexington Avenue Line in New York, with not great frequency and asymmetric demand, getting maybe a million riders per weekday. If you count tails, then Paris has 10 radial lines (omitting 2, 5, 6, 10) of which 8 are two-tailed (omitting 11 and 14), so a total of 18 Metro entries to the CBD, defined to stretch from Chatelet to Saint-Lazare; a 19th tail is under construction in the M14 extension to the north. New York has 7 to Uptown and the Bronx, 4 to Queens, and 9 to Brooklyn, so in theory 20 and in practice also 18 because the branching structure ends up wasting space in some of the tunnels. Without the express lines New York loses 3 Uptown/Bronx and 1 Queens route, which are more or less the 4 busiest.

          • Michael James

            Alon Levy 2018/04/29 – 12:34

            Champs-Elysees and Cours de Vincennes are both much wider than Broadway in New York, so there’s clearly enough space.

            Doh, yes, but I was thinking about both pinchpoints (on rue de Rivoli) and the narrowing as you go east, on rue St Antoine and IIRC, on the other side of Bastille the r. du Faubourg St Antoine; without consulting the history, I reckon both these sections (in total >1.5km) are pre-Haussmann.

            M1 has a lot of transfer stations in rapid succession today, but that wasn’t the case when it was built. It took until the 1920s to build M9,

            Again, doh, because you know M1 was the first built. But, again without checking history, a large amount of the Paris network was designed before any of it was built. Surely FDR station was built at the time M1 was built, regardless of M9?

            And yes, the line gets 215 million annual riders. It’s a 2-track metro with very high peak frequency and high turnover. This is what metro lines are capable of even with short-ish trains (cf. the Central line at 289 million).

            Right, but my main point was, why would planners have considered making 4 tracks? If you were designing it today, would you do that? (Hmm, perhaps as part of a different line that branches off north or south before those pinchpoints? IIRC you have written on that specific strategy, radial lines that turn into cross-lines?) It’s interesting that you mention the Central line because it is one of the few London undergrounds that was built under the streets, and thus follows the streets, and broadly it is the equivalent of M1 running east-west thru the main centres of the city but almost 5x longer than M1 and reaching out to the suburbs it is almost the RER-A of its day. But that didn’t stop the building of CrossRail (of course modelled on RER-A) which I predict will parasitise a lot of Central’s pax (in the same way, I presume RER-A took pax from M1 for those heading to La Defense and perhaps Gare de Lyon?). The other way to phrase this is: if M1 had been 4-tracked (and despite my concerns about pinchpoints etc, all the way to Gare de Lyon) would they have built RER-A?

          • Alon Levy

            If I were designing M1 today? Absolutely! Evidently in the 1960s they did build express tracks, they just called them the RER A. But even in the 1890s there was a decent case for it, as evidenced by two things: first, Paris designed many Metro lines at once, including two separate east-west ones, suggesting there was a lot of pent-up demand; and second, New York, not a much larger city than Paris at the time, did build four tracks on its one line.

          • Michael James

            Alon Levy 2018/04/29 – 22:55

            If I were designing M1 today? Absolutely! Evidently in the 1960s they did build express tracks, they just called them the RER A. But even in the 1890s there was a decent case for it, as evidenced by two things: first, Paris designed many Metro lines at once, including two separate east-west ones, suggesting there was a lot of pent-up demand; and second, New York, not a much larger city than Paris at the time, did build four tracks on its one line.

            I’m unconvinced. You do have a pronounced tendency to apply today’s century-plus of knowledge retrospectively to earlier planning decisions, even ignoring all the politics and finances involved in addition to other techno etc fortune-telling tricks. Did anyone, including demographers, really envision such mega-cities or that a single transit line would carry almost 300m pax? And my musing was really about what flow-on consequences there might have been: no pressure for the RER system? Indeed you could take your points about NYC and London’s Central Line as factors in why those cities waited for so long (>40 years in London’s case; how much longer for NYC?) before building the equivalent. (Not trying to imply great wisdom on the French transit planers here; a lot is serendipitous and contingent on so many unforseeables…).
            I don’t think your argument about NYC 4-tracking is as clear-cut as you say (though NYC was a bigger city back then, with rampant immigration, than either Paris or London, or indeed any city in the world). A big factor I see is that of the long linear arrangement of Manhattan–it is 22km long, about double of east-west Paris. A bit like HK and the multi-tracking under Nathan road–they were driven by geography as much as anything else to fit in as much as they could. Hmm, did Shanghai adopt 4-tracking on those central lines easily predicted to be extremely busy?

          • adirondacker12800

            The West Side had the Ninth Ave. El, the East Side had the Second Ave. El and the Third Ave. El.

            Apparently the third track on all of them was part of the Dual Contracts.


            Alternately, the technology didn’t exist. The first electric streetcars run in 1888. Takes them a few years to figure out how to do multiple units.

      • Nicolas Centa

        How about the Grand Paris Express?

        It is labeled as a “métro”, which in France means, as you said, maximal incompatibility with mainline train, not only running on the right but also on tires on many lines including the most recent one, but also no express service, and also now full automation.

        • Untangled

          The Grand Paris Express is really quite different to the original metro. While it does stop all stations along the route like the original metro, the stations are much further apart. So it’s really a suburban metro. At 68 stations for 200km, the wide 2.9km stop spacing also means that it’s effectively an express service compared to the original metro.

          • Michael James

            Well, the GPX is actually a mix of Metro and RER lines. Of course the name is a bit confusing in that really “Metro” in Paris/Ile de France consists of “the Metro” + RER, in which Metro is narrow-gauge light-rail and RER is heavy-rail that is operationally equivalent to Metro.
            The new Metro lines of GPX are modelled directly on M14, the first Paris driverless trains + platform-aligned station doors, opened in the mid-90s. It too had stations more spaced apart than existing Paris Metro. In fact one of the “new” GPX lines is a very long extension of M14. Obviously to cover the bigger suburban distances the stations need to be further apart or it will take an age to get anywhere (as it does on London Underground). In as much as old Metro lines extend deep into the suburbs, there is the same thing: M7 was extended about 5km into the Petit Couronne in the mid-80s and this is one of the few bits of Metro where the trains get up quite a speed.

    • adirondacker12800

      They went with four tracks because the Elevateds were already overcrowded.

    • Nicolas Centa

      There is a lot of four tracking in Tokyo for railways such as the Joban line between Toride and Kita-Senju, the Chuo line between Ochanomizu and Mitaka, the Sobu line, and it is being expanded on the Odakyu Odawara line over as well as underground.

      But it is true that the subway operators are not doing it : on many lines there is through service but a four tracks line connects to a two tracks subway, so only locals have through service and therefore always get overcrowded.

      For example on the Joban line in the morning expresses without through service to the subway are quite empty, but locals that continue to the Chiyoda line on the subway are so crowded that delays are frequent.

      Also, maybe building lines in slightly different places is probably considered better by the city than increasing capacity on the existing ones, because it increases coverage (the number of properties that are really close to stations), and maybe also because it is difficult to properly organize giant stations with huge walking traffic (you can see it with Shinjuku and Shibuya that have hundreds of exits and impossibly complicated and/or distant connections).

      • anonymouse observer

        The reason why quad-track line are connected to double-track subway for through service in Tokyo is because that is how the through service scheme was originally intended. The main objective of the through service between commuter rail and subway was to relieve overcrowding issue the the CBD-end terminal itself, which are typically stub-ended at Yamanote Line, but the congestion relief on the main tracks near the CBD-end terminal are more like an afterthought. Most of these stub-end-at-Yamanote terminal stations are located where it is very difficult to expand due to right of way availability or purely high price of real estate in Tokyo CBD: look at how Odakyu (ended up with double-decking the whole station) and Keio (needed to actually reduce number of tracks from 5 to 4 then to 3 because it is land-locked) dealt with this issue at their Shinjuku terminal, even with through service to subway lines that allows off-loading some traffic volume.

        When such quad-tracking are desperately needed in subway, they build another new subway line as bypass in Tokyo. These bypass lines are built with wider station spacing and different standards from the “legacy lines” (longer trains in wider and longer car body) along with through-running capability.

        A few more “Doochey mcNitpicking” (sorry Comradefrana for using it without consent, but boy it is a good way to phrase it!):

        2. Majority of commuter rail operators with through service to subway lines in Tokyo actually send a lot of express trains into subway in these days, even during peak hours. For example:
        – Every third inbound express train on Keio Line during AM peak hour are through train to Toei Shinjuku Line;
        – Keisei sends only 1/3 of all limited express trains (all commuter limited express trains) to Keisei Ueno whereas the rest (or all rapid limited express trains) goes to Oshiage Line as through service to Toei Asakusa Line during AM peak hour;
        – All through service trains from/to Keikyu Line (to Toei Asakusa Line) are in one of express patterns while all local trains turns back at Shinagawa.
        – All through trains from/to Tokyo Metro Hanzomon Line from Tobu side comes from Tobu Isesaki/”Skytree” Express Line

        The only exceptions are Tokyo Metro Hibiya Line (through service to Tobu Isesaki/”Skytree” Local Line), Tozai Line (through to JR East Chuo-Sobu Local Line), and Chiyoda Line (through to JR East Joban Local Line).

        3. On Joban Line, crowding on the express service (intra-JR East through service to Tokyo and Shinagawa) and the local service (through service to Tokyo Metro Chiyoda Line) appeared to be quite similar according to the MLIT statistics Alon mentioned (http://www.mlit.go.jp/common/001099727.pdf). Also, trains on express line are 15-car long whereas local line trains are only 10-car long.

        4. Again on Joban Line, primary cause of delay on Joban Local-Metro Chiyoda-Odakyu through service are unreliable Odakyu Line service according to a recent Toyo Keizai article (https://toyokeizai.net/articles/-/207222?page=2). I think the situation has changed in positive way since the completion of quad-track project on Odakyu Line, which allowed Odakyu to double the number of through service trains from/to Chiyoda Line. I don’t have statistics, but opening of Ueno-Tokyo Line might have an negative impact on reliability of Joban Express Line service due to inter-lining with Tokaido, Utsunomiya, and Takasaki Lines.

        • jack (@jlichyen)

          As a native New Yorker, I love the opening line of that Toyo Keizai article: 「首都圏の鉄道では当たり前となっている相互直通運転。」 or, “It goes without saying that train lines in the city center are doing reciprocal through-running.” Oh yeah, it goes without saying? Really?

          • Nicolas Centa

            It means it now goes without saying in the capital city (= Tokyo) area.

            Which does not mean it was always that way in the past, nor in other cities even in Japan (almost none happens in Osaka or elsewhere).

          • anonymouse observer

            Yes, it is. Through-running is fairly common among operators outside of Tokyo. Some subway lines built relatively later in smaller cities are designed with intention to do through-running between subway and commuter rail lines (e.g. Fukuoka Subway Airport Line, Nagoya Subway Tsurumai Line, and both lines in Kyoto Subway).

            One of differences is that outside of Tokyo, most of them are between intercity trains (e.g. any trains running through JR company territory limits including Tokaido, Sanyo, and Kyushu Shinkansen HSR lines between JR Central, West, and Kyushu) or between commuter rail operators via underground section owned and maintain by non-subway operators like one between Kintetsu Nara Line and Hanshin Electric Nishi-Osaka Line (via Hanshin Namba Line) or the other between Hanshin Electric Main Line, Hankyu Electric Kobe Line, and Sanyo Electric Main Line via Kobe Rapid Transit lines.

  7. Comradefrana

    “However, in poor countries with high birth rates and low urbanization rates, cities can expect to increase their population many times over in the next few decades. Lagos and Dhaka are already huge… ”

    Just to be Doochey mcNitpick: While definitely poor and lowly urbanized, this makes it sound that Bangladesh also has high birthrates. That is not really the case. Bangladesh is at replacement fertility and only has medium birth rate due to population momentum. In that regard it is similar to Latin America and not even comparable with Sub-Saharan Africa.


    • Alon Levy

      Yeah, you’re right, and evidently I did put Chittagong in the low-TFR category alongside Saigon and Hanoi. Bangladesh has a low urbanization rate (as does Vietnam), though, so cities can expect to grow by a large factor in the next 50 or so years even without African (or Afghan, or Yemeni) natural growth rates.

    • Eric

      “Bangladesh is at replacement fertility”

      Bangladesh is below replacement fertility. Replacement rate is 2.1 for Western countries. However in countries with higher mortality, the replacement rate is higher because more people die before reaching/completing reproductive age. (The same reason replacement is 2.1 rather than 2.0 for Western countries). Worldwide the rate is 2.33, and I suspect Bangladesh is a bit higher.


      • Comradefrana

        Fair point. I usually consider 2.1 as replacement rate because a) country specific rates are hard to find a b) achieving that rate should be a goal anyway. Looking further, the best data I found (from the 2017 World Population Prospects) shows an estimated Net reproduction rate of 1.022 and a fertility rate of 2.22 in the 2010-2015 period and a projected Net reproduction rate of 0.964 and a fertility rate of 2.07 in the 2015-2020 period.

        • Comradefrana

          Probably should add: Net reproduction rate (NRR) = number of daughters an average woman would have if she was conformed to the age-specific fertility and mortality rates of a given year through her life.

          tl;dr NRR = 1 means replacement fertility.

  8. anonymouse observer

    Alon, I think there is another way of running express trains on subway lines: with standing overtakes (double-track subway line with 3 or 4-track stations instead of building continuously 4-track subway line). Even though these 3 or 4-track stations need to be built at right locations, I think it can reduce the construction cost significantly but still works if it is done right.

    I have seen such service only in Tokyo (Fukutoshin Line, Tozai Line, Toei Shinjuku Line, and underground section of Tokyu Den-en Toshi Line, if you count it as “subway”), but it works as intended (reduced trip time).

      • F-Line to Dudley

        The Orange Line third track from Community College to Wellington would’ve only been used if service were extended to Reading outright displacing commuter rail (outer-zone CR service from North Wilmington to Haverhill would’ve been retained via the Downeaster’s Lowell Line routing instead). In that displacement scenario the express track would’ve been continued through Oak Grove station eating the lone commuter rail track and taking up residence on Malden & Oak Grove’s CR platforms. Stops from Wyoming Hill to Reading would’ve then reverted back to 2-track. When the extension was truncated at Oak Grove the express track lost all its planned utility, and never ended up being used for anything. The reason for the ROW having 4 track berths from Sullivan Square to Wellington but only 3 from Wellington to Oak Grove is that one RR track would’ve had to be retained crossing the Mystic River even in a commuter rail displacement scenario to access a thicket of (now long perished) freight customers who were still holding strong by the mid-70’s next to Wellington station.

        MA State Transportation Library doesn’t seem to have scanned and put up online the old 45-year-old scoping docs for the Reading extension like it has for several other studies from that era, so it’s still a little bit of a mystery precisely how the express service patterns would’ve broken out during the service day or managed to balance the traffic flow on the Orange Line. We have a ballpark idea of how it would’ve worked, but so far not the actual online archive of that study’s blow-by-blow traffic sim data plots to reference for posterity’s sake.

  9. johndmuller

    Quite a few of NYC’s subway lines have 3 tracks in their outer sections, becoming 4 tracked after merging with another line. The third track can of course be used to run a peak hour express service and as long as two separate lines are paired like this (and have reasonable similar frequency), there will be no slowdown penalty on the the off-peak direction sinigle tracks at either end of a through run (where they are carrying the express trains in the off-peak portion of their runs); there will likely be some unnecessarily high frequencies of service though.

    I believe that a similar 3-track – 4-track – 3-track system, without the partnership with the other half of the paired lines above, could be made to work with higher frequencies than than on the individual lies of the paired lines above by running the off-peak single track sections at a slower speed with closer spacing, assuming that the signal system was sophisticated enough to handle alternative block sizes or some other mechanism for achieving higher density. This is essentially the future for systems that are approaching overcrowdedness – change the signaling system to allow for more people to be carried, but at a reduced speed.

    Much as signal system improvements are touted as being the answer to all our problems, in practice, once you are running the maximum number of trains at their fastest speeds, all the advanced signal systems may in fact be able to deliver is the ability to vary the packing of trains in the tunnel to optimize the number of people being moved at the cost of reduced speed. The final solution here would be with the tunnels completely full of sardine-can full subway cars bumper to bumper – stop and go, moving forward one platform length at a time.

  10. Pingback: Little Things That Matter: Circulation at Transfer Stations | Pedestrian Observations

Leave a Reply

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

WordPress.com Logo

You are commenting using your WordPress.com 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.