The 19th Avenue Subway

The Market Street Subway is a bottleneck for the Muni Metro system, since each of its lines eventually merges into it. Early documents show plans for a four-track subway under Market; alas, we didn’t commit to that project and now we are stuck with a two-track subway. Combine this with the fact that the surface segments of each line force Muni to run one- and two-car trains where the stations could handle a lot more and it’s easy to see that we’re leaving tons of capacity on the table.

The Central Subway pulls half a line out of the Market Street Subway, which probably won’t improve the situation since we still have five lines in the subway at all times. In addition, the new Parkmerced developments are going to add thousands of daily riders to the M Ocean View line, which runs in mixed traffic on 19th Ave and is prone to some of the worst gaps and delays in the whole system. Muni decided to look at improving this line next. They did some studies, offered some suggestions that seemed lacking in the long term, and collected a bunch of community feedback.

Last month, something interesting happened: Muni greatly increased the scope of the project. The details can be found here, and I encourage you to go read those before I poison your thoughts with my editorializing.

If completed, Muni will have built its first fully grade-separated metro line, and it would improve the situation throughout the system. The M would be free of just about all gaps and delays, and it would use 4-car trains, greatly expanding the capacity of the Market Street Subway. West Portal will be able to handle many more trains and people. St. Francis Circle will be free of the crazy signaling, the intersection will be less complicated, and pedestrians will be able to make more use of the metro there. 19th Avenue will generally become more pedestrian-friendly and the line will better serve SFSU and Parkmerced. It’s a fantastic proposal, in my opinion. There are a few interesting details I’d like to point out as topics for further discussion or clarification.

J Ocean View

If this project completes as planned, the J will take over service in Ocean View. (We might need to rename some lines, but I digress.) The choice of the J to take over this part of the journey is interesting because it has a number of features limiting its capacity at the moment. One of the more frustrating issues with the J is the inability to run even 2-car trains along the line, because some of the stops would render the second car inaccessible. However, the plan explicitly talks about running 2-car J trains. Muni plans to remove one of the problem stops (Liberty St on the private right-of-way), but I’m not sure that this is the only one (21st St might be too short, and possibly Glen Park). This might boil down to adding a few square feet of concrete here and there, but it’s something that Muni hasn’t mentioned to date in its other Muni Forward plans.

This also means that the J will spend a much higher percentage of its time above ground and in mixed traffic, since it doesn’t look like there are plans to give it a red lane south of 16th St. This might not be an issue for Ocean View commuters trying to get downtown because I assume they will take the J outbound to SFSU Station and transfer over to the M, and it will probably be faster than what they have now. However, we have to be careful that we aren’t getting rid of delays on the M by simply passing them on to the J. The J will need extra reliability improvements so it can remain a useful part of the rapid network.

The Curious Case of St. Francis Circle

If you look at the rail map in the project details, you might notice the very strangely laid out subway station at St. Francis Circle.

st-francis-circle.png

At first, I took this to mean that the M trains coming through would always stop on the west side of the platform no matter which direction they are headed towards, and similarly for the K on the east side.

one-level-platform

This seemed like a strange interpretation because Muni is making this immense capital expenditure to get rid of delays and yet this seems like it might cause a lot of them. This limits the maximum frequency of trains through St. Francis Circle because they must maintain stopping distance while a train from the other direction is passing through. It also means that the M and K have the same maximum frequency in the station, which is not the most efficient way to do things—as extra capacity is needed, it would be more useful to swap out K trains for M trains and have some K trains terminate at St. Francis Circle. Riders could then transfer to the longer M trains for the rest of the journey. The K seems to have the same problem as the J in that it can’t support trains longer than 1 car, so running more frequent M trains through seems like a no-brainer.

I did some digging around in NYC Subway track maps to see if I could find an analog for this sort of station, and the closest thing I found was Queensboro Plaza:

queensboro-plaza

From the map alone it seems like Queensboro Plaza is structured the same way, with one track going through the station for the 7 and <7> and the other track for the N and Q. This doesn’t make sense when you’re running trains every two minutes in both directions. However, Queensboro Plaza is actually a two-level station, where the Manhattan-bound trains from each line stop on one level and the outbound trains stop on the other level. Each train has two tracks going through the station. It’s possible SFMTA is planning a similar structure for St. Francis Circle station!

two-level-platform

The structure is a bit harder to see here, but the western side of the station would have M trains, and the eastern side would have K trains. This would allow them to run trains in both directions without a bottleneck, and would provide a way for the M and K to cross over each other without sharing a single switch. There are some drawbacks—building this station and track layout is probably way more expensive than the one-level layout, and the inbound and outbound tracks for each individual line don’t actually meet anywhere, meaning we couldn’t terminate K trains at this station like we wanted to. In the one-level world, we can pull a train into the station and just back it out when it’s ready to turn around. With the two-level design, we’d probably have to run them to West Portal and turn them around there, which defeats the point. (This might still help since we can pull the K trains back before the L trains merge into the subway towards Forest Hill, but M trains might be stuck waiting for K trains to turn around at West Portal anyway.)

I guess the moral of the story here is that I’m not really sure what they want to do with this station, but the track layout they’ve picked makes me a little worried that we will again run into capacity issues sooner than we’d like.

The Daly City Extension

The plan leaves room for the extension of the subway to Daly City BART. BART and Muni already have plenty of connections at Balboa Park, but this one would be pretty useful because data from the Transit Effectiveness Project/Muni Forward on how people use the 28 and 28R (formerly 28L) show that Daly City BART is a good anchor for the line. Linking SFSU with Daly City by a direct subway connection is probably a really good idea since those are the busiest stops along the corridor.

This would also offer a kind of redundancy with BART in that there would be two discrete grade-separated lines from Daly City to Embarcadero, which would improve the resilience of the whole system in San Francisco. If BART is having problems in the city, the M might be a good substitute for people commuting up from the south. Of course, this might cast the lack of integrated fares into sharp relief, but that’s a much more complicated issue for another post.

Similarities to New Muni Metro

Back at the end of 2014, Nextransit published the first part of their New Muni Metro plan. The principles behind the Ocean View plan and the New Muni Metro plan are similar, though New Muni Metro didn’t expect to launch into multi-billion dollar capital projects for a long time. But they realized that we already have a fully-grade separated line if we end the Market Street Subway lines at West Portal. All of the other lines are streetcars only and people transfer to the Market Street Subway at West Portal (K, L, M), Church (N, J), or Embarcadero (T). The New Muni Metro plan also involves some line fusion, with the J taking over the surface portion of the K, and the L taking over the surface portion of the M. This means we could run all of the surface lines more often without worrying about whether we’re overloading the subway, and the subway will be able to run longer trains all of the time because there won’t be any short stations along its path. Higher surface line frequencies plus long, extremely frequent subway trains will mean way more capacity and generally shorter trips, even if you have to transfer.

I think the most interesting part about the similarity is that it means Muni could probably experiment with this same kind of service, should capacity become an issue. I alluded to this in terms of substituting K trains for M trains in the St. Francis Circle station design, but we could theoretically do this with any type of train. Forcing people to transfer probably wouldn’t make everyone happy, but the trains would be less crowded this way, so it might work out!

However the details fall out, this is incredibly exciting news. Transit agencies often avoid making grand plans like this because the funding for grand plans has dried up, but if we know that we’ll be getting something this useful out of it, the political situation might be workable after all. We shouldn’t be afraid to dream big—even if these projects seem expensive, they’re necessary, and nobody will look back on it in the future and think it was money poorly spent. Let’s see how Muni builds on this in the coming years!

Advertisements

Missing the point with a second tube

I figured I’d talk about BART some more because they’ve made it into the news recently. They’re currently fighting a lot of issues, and people are understandably mad, but I don’t think the absurd set of constraints the agency has to operate under gets enough airtime. It’s made even worse by the fact that people are demanding more and more of them as time goes on. Please be patient with them!

As I wrote about last time, talk of running BART around the clock seems to be reaching a fever pitch. One of the big topics here has been the second Transbay Tube, and it seems like people believe that this is necessary and sufficient for 24-hour BART. I think we badly need the second tube, but for completely different reasons. It also seems unlikely to me that this is all we’d have to do to get the trains running all night and we’ll be sorely disappointed if we use this as a justification.

The second tube is an absolutely necessary capacity upgrade, and we can see why from BART’s planning documents. Right now, the Transbay Tube carries 24 trains per hour during peak commute time. This means we can put about 23,000 people across the Bay in one direction in an hour. Unfortunately, it seems like we’re coming close to hitting this number already! This is a huge problem, because we won’t be laying another tube for decades. In the meantime, BART has some ideas to squeeze a little extra capacity out of the current tube. One is to make all Transbay trains 10 cars long. They currently don’t have enough train cars for this one, but the Fleet of the Future promises a massive expansion. They’ll also have to work out those crazy voltage spikes and power issues. When all is said and done, it’ll allow a modest capacity increase to 25,000 or 26,000 people per hour. Another relies on signal upgrades which should be included in the next BART bond. The benefits include an increase in frequency from 24 trains per hour to 30, pushing the capacity further up to around 32,000 people per hour. This is a big help, and is definitely the best short-term strategy, but the Bay Area is expanding so quickly that this will only help for so long.

Enter the second tube: BART projects we’ll be approaching the 32,000 passenger per hour limit in about 35 years, so we should probably begin planning this tube now. It will roughly double Transbay capacity and keep BART humming along even if it surpasses 1 million daily riders. It would also provide some badly-needed redundancy in case something goes wrong with the current tube or a station approaching it. Because of the improvements to capacity and robustness of the whole system, it’s definitely a worthy investment.

The redundancy aspect has people setting their sights on 24-hour BART. After all, if you can run the trains in one tube, you can close the other one for maintenance. If you switch off between the tubes on consecutive days, you should always be able to keep one of the tubes open, and we can run trains all night, every night!

There are a number of problems with this line of thought. For example, just because we have four tracks crossing the Bay doesn’t mean we have four tracks anywhere else in the system (save MacArthur, but that’s not really useful for this aspect). The new tube could play out in one of two ways. One is that it crosses the Bay, connecting up to the existing Market Street Subway in San Francisco. Another is that it creates a new BART line within SF city limits with some connection to the current subway. The most popular formulation of this concept has the connection at Montgomery and the new line running down Post or Geary.

Both of these plans have choke points that aren’t the tube itself—namely, they dump out into two-track subways, and if you close those down, you still won’t be able to run trains. While studying the New York City Subway, I realized that the decision to quadruple-track the trunk lines was one of the most brilliant decisions they made, as it allowed them to run trains along those lines no matter what happened. Unless we have quadruple-track subways, we’ll never reach that same level of availability.

Of course, New York City runs the trains all night on every line, not just the ones with four tracks. Plenty of lines run on two tracks when they leave Manhattan. Some lines, such as the L, have only two tracks the whole way through! I wasn’t sure what to make of this, so I asked my dad (an ex-MTA employee) how it was possible. He gave me two answers: Bus substitution, and the dreaded Planned Service Changes.

Anyone who has been in a subway station in New York has likely seen the enormous boards detailing all of the ways in which service will deviate from the official patterns in the near future. Trains might skip particular stops or there might be breaks in service along the line, requiring free bus bridges along the gap, or transferring to other trains. Trains from one line might use tracks from another line for some duration of their trip to allow maintenance on their own tracks. These service changes allow the trains to run all night, every night, but they have some cost. First, they rely on the network’s massive amounts of redundancy to continue functioning. Second, they make everything really complicated. The NYC Subway is already this huge, complex system with many lines and hundreds of miles of track, and remembering the standard service pattern is hard enough. It must be impossible to keep up if they have to change it all the time!

In either of the scenarios for the second tube, this same strategy of service changes to avoid maintenance areas isn’t going to cut it. The system simply doesn’t have enough redundancy to support the kind of service changes the MTA can pull off, and building that infrastructure now would require many billions of dollars. Let’s look at what we’d have to do even if we only built extra tracks between MacArthur and 24th St Mission, which is basically the smallest area for which it makes sense. We’d be paying for an extra 3 miles of two-track subway on the SF side and 2 miles of one-track subway on the Oakland side, and we’d have to rework the wacky subway-to-elevated transition as the C tracks move into the 980 median. This would add at least $4 billion to the total, and that’s in the absolute best case scenario. Reworking the transition and construction in the stations would probably cause service outages. It would also only grant us 24-hour service on a very limited section of BART. (This isn’t necessarily a bad thing, since we may only want to run trains all night in the areas dense enough with the kind of activity to support them. I assume demand for all-night BART to Pittsburg-Bay Point is pretty low. It is meant to highlight the immense expense of adding this infrastructure.)

What can we do? I know I was something of a killjoy last time when talking about the problems with running BART all night on weekends, but I think the conclusions there are valid here as well. If we can’t rely on service changes and we don’t have the infrastructure to do maintenance concurrently with service, then we’re going to have to turn to our old friend, bus substitution. As I stated before, even the NYC Subway resorts to this, so it shouldn’t feel like a cop-out. Bus substitution isn’t ideal because it drops capacity and doesn’t provide service as rapid as with trains. If we’re going to provide 24/7 service, we might need to do bus substitution at any hour, even during the commute rush. This could be really painful for a lot of people, especially while BART is already having capacity issues! We can do a little damage control by adding more switches in between stops so we’ll only have to substitute over one station, but this requires some extra infrastructure and is probably a ways off if so. We can see examples of this in the wild, though: Chicago’s L has switches between most stations on the two-track segments of its Red and Blue lines, and this is no doubt an integral component to their ability to run trains all night on those lines.

I think the new tube and 24-hour service are often tied together needlessly in discussions about the future of BART in a way that makes it hard to talk about what the former is actually necessary for, and what is actually necessary for the latter. My position on it is:

  1. If we want to be running trains all night, every night, then we do need to wait for the second tube, but that’s not even close to everything we need. Miles of extra subway and lots of new infrastructure will need to be built. Whether or not that’s a worthy investment is beyond the scope of this post, but we shouldn’t make the mistake of thinking we can build the extra tube and be done with it.
  2. If we’re comfortable with bus substitution, then we can mostly already get 24-hour service, if we are willing to tolerate substitution across arbitrary segments of lines and possibly at arbitrary times. This is, again, a value judgement, but given the outcry over the recent issues on the PBP line, I think this would be unpopular.

It’s great to see the Bay Area taking aim at New York here, but we do have to realize that their system is much better suited to 24-hour service than ours is, and we have a lot of work to do if that’s something we want as well. Either way, we are going to need the new tube. We simply shouldn’t delude ourselves into believing it will solve our 24-hour issue as well.

The all-night compromise

As the Bay Area grows, I’ve been hearing more and more talk about how we need all-night BART. Running trains all night, every night is a complicated problem that requires a lot of extra infrastructure to solve, so people are asking what it would take to enact a less ambitious proposal: running the trains 24 hours on weekends only. This seems like a reasonable middle ground, since there will still be time for overnight maintenance during the week, and people will be able to use BART to get in and out of the city on weekends.

BART makes a number of claims about why it doesn’t run all night:

  • Every hour of scheduled maintenance is necessary to keep the system in a good state of repair;
  • This work has to be done every night;
  • The longer maintenance windows on the weekends allow them to perform more complicated maintenance work that can’t be done during the week.

Even if we could postpone the night work on weekends, we would need to move those hours somewhere. Trains are still running along Pittsburg-Bay Point tracks until about 1:30 AM every day. During the week, they start again at about 4 AM. On Saturdays, they start at 6 AM, and on Sunday the line opens at 8 AM. This means our weekend maintenance window totals 11 hours each week. We can distribute these hours almost evenly throughout the week, though I’d suggest extending the Sunday night maintenance window with the leftover hour in order to try and make up for the backlog on the weekend. This means we’ve got 4.5 hours of maintenance from Monday night to Thursday night, and 5.5 hours on Sunday night.

It turns out we can’t really postpone BART’s opening, since the 4-5 AM hour is an unexpectedly big commute period. This means we’ll have to pull those maintenance hours from night service, and close BART at 11:30 PM five nights a week. It’s likely this service change will have a much bigger and more negative impact than the addition of late-night trains on weekends. It will almost certainly fail the Title VI analysis that requires BART to study the impact of service changes on minority and low-income riders. Past attempts by BART to extend weekend service were nixed for this very reason.

It’s also worth asking if this service change is going to help a lot of people. Boston’s MBTA recently voted to wind down its late-night subway hours, citing maintenance backlogs that are piling up, high expense, and low ridership, as expected. However, I think this article brings up one of the most overlooked points in this discussion:

“There aren’t that many jobs that are only two nights a week. We think this late-night service is not a broad solution to economic access because it’s not a seven-night-a-week service.”
Charles Planck, MBTA Assistant General Manager

If transit is supposed to be a great economic equalizer, then this plan for weekend-only 24-hour service isn’t what we want. People will be able to stay out later on weekends, but we pay for it by getting rid of late weekday service that people are probably using to get home from work, and we’ll likely need extra maintenance windows during normal service hours when more people need BART.

This isn’t to say late-night transit isn’t important—it is! The question I’d rather consider is whether it should be on rails, and I don’t think we can justify that. AC Transit’s 800 and 822 bus lines provide owl service along BART lines, but they come once an hour. I know late-night ridership is low, but if they’re that unreliable, people might act like they’re not even there!

Muni runs owl bus service on a number of lines. Some of them are not very frequent, but the most frequent show up about every 12 minutes. The N-Owl substitutes for an important rail line and shows up every 15 minutes. Philadelphia’s SEPTA runs owl buses on weekends along its subway lines, and these buses also come every 10-15 minutes. Why do the BART owl routes have such poor frequency in comparison?

The reason may be the length of the line. The 14-Mission has a route length of about eight miles. The N-Owl clocks in at around nine and a half. SEPTA’s Market-Frankford Line is just under 13 miles long. By comparison, the 800 route is almost 27 miles long! This means that service at a certain frequency on the 800 will require roughly twice as many buses (and drivers, and so on) as on the MFL, or three times that of the N-Owl. (This is a fuzzy approximation since about eight miles of the 800’s route are spent on freeways, but that still takes time.) As such, huge frequency increases on these routes might be expensive, but they are definitely much cheaper than the work required to bring all-night rail to BART. I would also argue that it’s wasteful to build that infrastructure if we can’t get decent ridership on buses with 20-minute headways or better. Not only that, but this liberates us from the weekend-only stipulation, and that probably serves everyone better.

The all-night BART discussion is often framed as a question of how to run trains all night. This tunnel vision has prevented us from looking at other modes of transportation that are better-suited to solving the problem in the near term. When you think about this problem, I urge you to instead look at it as a question of how to provide service to the areas covered by BART, regardless of the kind of vehicle that does it.