24/7 Rapid Transit
It’s a commonplace in New York that the New York City Subway is almost the only one that runs 24/7, and that the rest – PATH, PATCO, and two lines of the Chicago L – are small operations. The reason for this operating plan is that the main Manhattan trunklines have four tracks, making it feasible to shut down tracks for weekend and late-night maintenance and skip a few stations in one direction. Occasionally, even midday midweek service is disrupted. This leads to complaints from passengers who actually ride transit in the off-peak, as well as various politicians, and exhortations from political defenders of the MTA that it’s a necessary byproduct of 24/7 operation.
In fact, there’s one additional system not mentioned above: the Copenhagen Metro, which began 24/7 operation in 2009. Although around-the-clock operation on weekends is common in some European cities, such as Berlin, Copenhagen took the extra step to run 24/7 reliably. It has only two tracks, like some lines in New York, but made sure it would be possible to single-track at night for maintenance. Late-night headways in Copenhagen are 20 minutes, like in New York, and this gives enough time to reduce long segments to a single track and run wrong-way service. Copenhagen’s trains are automated and this helps with wrong-way signaling, but it’s not a prerequisite and wrong-way operation is already done late at night on the subway in New York.
What this means is that there’s a technical solution to the problem of late-night and weekend service disruptions: make sure that there are crossovers placed at regular intervals to allow 20-minute service on single track. Installing switches requires extra capital construction money, but is orders of magnitude cheaper than building extra tunnels, and would make late-night maintenance much easier. Headways are such that a switch would be required every 7 or 8 minutes, which means every 2.5-5 km. At some places, crossovers already exist at that density, for example at all four tunnels from Queens to Manhattan, and all that’s required is schedule modification.
The result would still not be as satisfactory as in Copenhagen, ironically because of the multi-track trunklines. Under the slow-fast-fast-slow system used in New York, as well as most other four-track lines, it’s impossible for a local train to cross over to the opposite track without fouling the express tracks. This would create serious problems even on the three-track lines in Queens and the Bronx, since extra switching moves would be required, shortening the acceptable crossover spacing. It would still be possible, say with crossovers 6-7 minutes apart, but the maintenance requirements would be higher.
On the four-track mainlines, I don’t see any solution that unequivocally improves on the status quo. It’s possible to have the same crossovers, but at even tighter spacing, and without any express traffic. Weekend express traffic could possibly still be retained, but not late-night express trains, and late-night frequency would be reduced to 20 minutes even on combined lines, for example the local 1/2 in Manhattan.
What this means for future trunklines is that, if four-tracking is required for capacity or for express service, it should not run as was built in New York a hundred years ago. Instead, the slow tracks should be in the middle, and the fast tracks on the outside; this allows more operational flexibility as well as short-turning local trains, at the cost of making it harder to build infill stations. While the subway short-turns some local trains, for example the C at 168th and at Euclid, this requires flying junctions, which contributed to the IND’s excessive cost.
Maximum flexibility could be obtained by building every station with two island platforms, as if it were an express station, and having express trains skip low-traffic stations. This way, two tracks could be shut down for maintenance along the entire line with no ill effect on reliability, except that retaining express service would required timed overtakes. The problem is of course the much higher cost of such a line, especially if it is underground.
For underground lines, there’s very rarely a reason to four-track. Washington may complain about lack of flexibility and express service, but modern subway lines with good rolling stock and wide curves can achieve acceptable average speed even with medium stop spacing. The Copenhagen Metro averages 40 km/h, a speed previously reserved for systems with very long (~1.6 km) interstations such as the Moscow Metro, even though its stop spacing is just 1 km. Capacity is the only serious drawback of two-track lines, but if it is so pressing then the city should built two separate two-track lines, which with tunnel boring machines cost about the same as one four-track line.
Pedestrian Observations 
I think I’ve encountered A service in Washington heights that was scheduled every half hour during late nights due to construction or whatever, north of 168th street. It seems that this would allow complete single track operation on the 2-track segment between 168th and 207th – so I think this is what they did there. It’s possible that they are already doing what you are proposing, but maybe not to the extend you’d like to see.
I’ve seen this on a few other two-track lines. As far as I can tell, the problem with that segment is that current practice is to slow down trains operating adjacent to a work crew; on NYCT, it’s 10 mph if I remember correctly. The schedule is 9 minutes between 168th and 207th. If trains can cruise at their normal speed, about 50 km/h, then single-tracking north of the existing crossover just south of 181st should allow 20-minute service with a 6- or 7-minute turnaround, which is reasonable.
Another potential source of problems is schedule adherence. The subway has very little; it needs very little during the day, when constant headways are more important. But late at night, timed transfers, overtakes, meets, and what not suddenly become valuable. Even during the day, it would be nice if they ran the Z express and the J local with timed overtakes at places with room for multi-tracking, instead of the current skip-stop pattern. But the Jamaica Line is marginal, whereas at night all lines need schedule discipline.
On that note, it should be much more easy to keep an impossible schedule at night, impossible in terms of overtakes and passing at sidings etc., because the night is effectively only 4 hours long (~1:30am-4:30am). Even if you build a schedule where any delay of any train ripples throughout the whole system, it doesn’t really matter. If at the end of the 4h, every single train is 10minutes late, who cares.
Alon, I agree with most of this except:
1. What if a line is at capacity even with stop-skip spacing? Philadelphia’s MFL is a good example of this.
2. What if later line additions create a situation where the new service and the old service exceed previous capacity? The extensive Orange Line/Silver Line interlining that will go into effect on the Washington Metro suggests a need for a new express tunnel between the two lines’ junction and Rosslyn…or else a new alignment for the Silver Line into Rosslyn. And where these two lines interline with the Blue Line? Fuhgeddaboudit! WMATA’s next major Metro plan will absolutely have to disentangle either the Blue or Silver Lines from the triple interlining since double-interlining, not triple-interlining, is the network’s design capacity.
In other words, if a line was not historically built with express capacity, but has reached, through time or the weight of system extensions, a capacity ceiling, other techniques (such as improved signaling) have already been used, and the ceiling still exists, there does need to be serious thought about putting in an express tunnel.
In Washington, the stop spacing is so wide, and there’s so much untapped potential for rapid transit-ified commuter rail, that the solution to the Orange/Silver/Blue Line question should be to either build inner-city trams to relieve the worst segment, lengthen trains, or build a separate alignment for the Blue Line. This is what I proposed for relieving capacity by building two two-track lines rather than one four-track lines.
Look at the history of the Piccadilly Line in London for an example of this kind of thinking. In fact, many of the London rail routes went over capacity and “relief routes” had to be built for them. (In some cases, former branches were cut and given their own trunk lines.) They’ve ended up with no proper four-track lines but a lot of semi-overlapping routes.
This seems to be what usually happens when a system evolves. Why does NYC have any four-track lines at all? *Because in NYC the subways were being built to replace elevateds and streetcars*.
The elevateds were almost all two-track, but there were a lot of semi-redundant routes. It clearly seemed more efficient to provide local / express service on a single trunk rather than digging up every avenue at once. Nobody ever builds four-track “fresh”, only when upgrading or replacing an existing service. 😛
Cities like DC and Boston need to enter the 19th century and provide 24 hour service. Thats right, when the Boston subway first opened, there was 24 hour service, and today, it closes at 1am.
No need to even use the damn trains…run buses, like every other major city in the world.
If transit is to be a credible alternative to the auto, “Cinderella Liberty” must end. As New Yorkers well know, many parts of the subway are only 2 tracks, yet maintenance gets done and with some exceptions service continues. Worst case a bus bridge for a short segment can be deployed, but mostly “single tracking” in the overnight hours is sufficient. The lame excuses given by BART and WMATA mask a disinterest in serving late night riders.
I hesitate to say that late-night service is a must, but it’s clearly possible. The explanation I hear from commenters in Japan is that transit can’t be competitive late at night, but I just don’t buy that in cities with low car ownership they can’t fill trains all night long. (And JR East reliably runs 15-minute service on single track with a timed meet every 5-6 minutes at the outer end of the Yokosuka Line, so it’s technologically feasible. Miles, am I getting something wrong here?)
Just a note to to this:
If you are speaking about bored tunnels, it makes sense to build them as two parallel double-track tunnels – I guess that such arrangement would minimize costs of both tunnelling (as smaller bores are significantly less technically demanding) and infill stations that would mean boring just a third tunnel between the existing ones and cross tunnels.