27 September 2013

Beyond Level Boarding

It thankfully happened: after all these years, Caltrain has finally stated something (anything!) about level boarding.  This is an improvement that has long been neglected in favor of the more flashy electrification project.

Around here, level boarding has been discussed several times already, so here are a few oldies but goodies from the archives:
  1. Platform height: background and compatibility issues
  2. How platform incompatibility results in oversize stations
  3. Platform sharing issues between Caltrain and HSR
  4. How platform incompatibility is making a mess at SF Transbay
  5. Platform height: the full collection of articles from this blog.  Read it and weep.
There is also a nice APTA overview presentation that goes into some detail about the conflicts between level boarding and freight trains.

Caltrain's presentation sums up the main advantage of level boarding in one bullet point, "Operating efficiencies," without clearly stating why it's so important.
  • Level boarding reduces trip times.  Without steps, passengers can board and alight far more quickly and the train can save about 15 seconds of dwell time at every station stop.  For an all-stops local, this starts adding up, and saves about half again as much as much time as saved by switching from diesel to electric!
  • Level boarding reduces delays.  Wheelchairs, bikes, strollers, hordes of drunk Giants fans, airport travelers with suitcases, everybody can get on and off far more quickly and easily.  As any rider will tell you, station dwell times suffer from a statistical "long tail" of agonizingly long station dwells due to a variety of these factors, a sort of station dwell Russian Roulette.  The only way to run this railroad is to add plenty of timetable padding to absorb these anomalies without making a mess of cascading delays.  Fast forward to 2030: under a blended scenario where hourly traffic is doubled on largely the same tracks, this is no longer acceptable.  Station dwell times must be predictable and reliable to avoid making a mess of the morning rush, and level boarding isn't just a nice-to-have; it is a must-have.
Caltrain's tentative plan is to study the possibility of converting all station platforms to 25 inches above rail (up from today's 8 inches) by some unspecified time in the future, and to let high-speed rail have its own set of separate platforms, most likely at 48 inches above rail.

Beyond Level Boarding: Platform Sharing

Level boarding as envisioned by Caltrain is not a trivial undertaking.  It will be expensive, because dozens of platforms will have to be entirely rebuilt.  It will be politically difficult, not in the least because state regulations will need to be overhauled over the strenuous objections of freight railroads.  It will be time consuming, because so many platforms can't be rebuilt overnight.  It will be logistically complicated, because the system will have to continue operating during the transition with a mix of platform heights.

Why not go a tiny bit further to achieve far greater benefits, at far lower cost?

Platform sharing is the next step beyond level boarding, where Caltrain and HSR can share the same platforms at San Francisco, Millbrae and San Jose.  Platform sharing has distinct advantages that have been discussed at length on this blog.  Platform sharing between conventional trains and high-speed trains is already practiced everywhere in the world, except in those countries where track gauge precludes it.
  • San Francisco Transbay is where all the action will be.  There are more jobs within a half mile of Transbay than there are within a half mile of each and every Caltrain stop, combined!  Service into and out of this tiny six-track terminal station will be inherently inefficient but crucially important for both Caltrain and HSR.  Platform sharing would increase Transbay capacity for decades into the future.
  • Millbrae is where the HSR consultants once expressed interest in building a $1.9 billion dollar tunnel to shoehorn new platforms between a handful of residential properties and the existing BART station, which incidentally cost only about $0.1 billion to build.  Never mind the sheer fiscal insanity of it: platform sharing would allow the station to serve both Caltrain and HSR within its existing footprint.
  • The mid-peninsula HSR stop would be nothing more than another platform where a high-speed train happens to stop, should that be deemed advantageous.
  • San Jose is where the HSR plans get truly loopy: they want to build a massive station up in the sky over the existing station, hovering on a forest of concrete straddle bents, with a six-mile approach viaduct to the north and an "iconic bridge" (Bay Bridge east span, anyone?) to the south.  Platform sharing would enable both HSR's and Caltrain's needs to be met on a single level, at grade, within the existing footprint of the station.  This would save billions of taxpayer funds that the HSR authority seems to be rather short of.
Why are the respective agencies here not talking about platform sharing?  There is bureaucratic inertia and a general unwillingness to coordinate, sometimes verging on outright hostility.

How To Pull It Off
  1. Get HSR to back off from the airport-like notion of separate, dedicated platforms.  Building a TSA security perimeter around the HSR system while leaving Caltrain open won't prevent mass carnage, any more than locking the doors on only one side of your car will prevent theft of your belongings.  HSR can be built far more cheaply with platform sharing.  Of course, none of these measures would be favored by the for-profit Transportation Industrial Complex.
  2. Get Caltrain to back off from the idea of 25 inch platforms.  Not all bilevel EMUs must have low doors; some of the latest models work with high platforms, and there are smart ways to facilitate a height transition by adding additional doors to the EMU fleet.  Or try another way: get HSR to adopt a 25 inch standard; good luck with that.  And please, don't myopically plan everything around the Bombardier fleet staying until 2030.
  3. Get the Transbay JPA and San Francisco to understand the economic advantages of platform sharing: a busier, more efficient and more commercially successful terminal, and the opportunity to reclaim much of the under-utilized train parking at 4th and King for redevelopment.
None of these challenges are technical.  They are institutional and political.


  1. Thanks for this post Clem! Your documented and well written articles provide an excellent reference for those that continue to try to advocate for changes. If level boarding ever happens (and shared platforms!), it will be due in large part to your articles and dedication to the cause! These changes come slowly (as we now all know), but I'm hopeful they'll come. Bravo!

  2. Unfortunately, it looks like they're "considering" inclusion of the Coast Daylight service, though that doesn't rule out the possibility that it's eliminated in the end.

  3. The Coast Daylight has proposed using the tracks, so Caltrain needs to list it. Whether a non-time-sensitive vacation train should be allowed to impact the schedule of highly time-sensitive commuters seems highly dubious. A reasonable conclusion is that it shouldn't.

  4. At this rate, I wouldn't be surprised if the mid-peninsula stop gets dropped from HSR plans.

    1. If Caltrain and HSR have the same platform height etc. all the Caltrain stations could be HSR stations.

  5. "A reasonable conclusion is ..."

    Stop right there.

  6. Fremont BART handles 8 train/h with one island platform (2 trucks). Milbrae BART handles 4 train/h just with one track. How many trains per hour will Caltrain runs?
    Other option is extend turn-around trucks beyond the station. Pararell parking can be consider given long platform as HSR but shorter Caltrain's consist.

    1. Several key things:

      First, Transbay in SF has been designed ("designed"), at quite literally every single point, to minimize train and passenger throughput. It really is that bad! Fremont BART is luxurious compared to it.

      Getting trains into and out of the busiest mainline station in Northern California is going to be un unmitigated clusterfuck due to route conflicts, choke points, and the way in which PTG/TJPA/PCJPB have quite deliberately maximized the temporal and physical extents of the conflicts.

      Getting passengers on and off the platforms likewise. Massive, completely unnecessary, impediments to circulation. It will make NY Penn look good. And no, I am not exaggerating.

      Second, Caltrain's "Chief of Rail Transformation" freely made an explicit choice that he wanted Caltrain to operate, foreever, under US freight railroad regulation. (Because "it would be a fun project to work with freight." And no, I'm not making this up.) The sort of train turnback times and train make/break times you see on BART -- and BART is a slouch compared to other metros, and even some other mainline oeprations -- is neve rgoing to happen on an Olde Tyme Commuter Railroad operated under FRA regulation by the BLE and UTU.
      20 minutes (3tph per platform) turnback is going to be pushing it at any level of reliability. In other words, Caltrain is going to be exactly as bad as it is today, forever. Because that's the way Caltrain Modernizaton Program ("CalMod"!!) people like things. Works for them just fine!

      Third, without level boarding -- and remember that one single phone call at any time in the last two decades would have fixed this, a call that Caltrain staff explicitly chose never to make -- and with no designed-in infrastructure which can "buffer" pile-ups of arriving trains (Mission Bay station should offer that, but PTG's "design" for the downtown extension is as bad as possible, and no I'm not making this up) there is no such thing as service reliability along the line, and hence huge amounts of schedule padding (meaning reduction in trains/platform/hour) are mandatory at the terminals.

      Fourth, given Caltrain's insane and bloody-minded determination to operate unpredictable, ad-hoc, complex, non-clockface, irregular-gap schedules forever, the opportunities to dynamically reassigned equipment and crews to fill in for late runs and service disruptions.

      Fifth, the BART end stations are suburban and comparatively low-use, and passenger flow is very largely tidal (all-on or all-off) at peaks. The highest ridership BART stations are through-running. Transbay, in contrast, will have heavy bi-directional passenger traffic at peaks, but the platforms and vertical circulation at the station are catastrophically bad. Every peak turnback will have to add several completely unnecessary minutes just to clear the platforms. It really will be that bad!

      Sixth, there is no plan for double-berthing at Transbay. Exactly the opposite, in fact. Instead of, as they easily could have done, building six 400m long fully-interchangiable platform tracks, and adding mid-platform scissors crossovers to allow four of those platforms tracks to function as eight quasi-independent 150m (six-car) platforms, the are instead truncating the platforms to work around the insane "architecture" of the park-above-a-bus-station, and configuring the trackwork to minimize operating flexibility. And of course, separate platforms for HSR and Caltrain -- separate technically, separate logistically, separate Vaterlandssicherheitsdienstliche -- with Caltrain getting the crumbs.

      It's a terrible terrible terrible situation. They've deliberately painted themselves into a box in which it's going to be very hard indeed to reliably get 4tph of Caltrain in and out of a "Caltrain Downtown Extension" terminal built at a cost of over four billion dollars.

    2. 20 min turn around time, FRA, BLE, UTU, trash cleaning.... are fine, but those have to be done at turn-around tracks outside the station. Station platoform should be immediately clear after disembark the passenger and train have to move trun-around track. I saw this type of practice at San Jose Diridon before. Assume only 2 tracks assigned for Caltrain, one for arrival, the other for departure. So, they can still handle more than 8 train/h.

    3. 20 min turn around time, FRA, BLE, UTU, trash cleaning.... are fine, but those have to be done at turn-around tracks outside the station. Station platoform should be immediately clear after disembark the passenger and train have to move trun-around track. I saw this type of practice at San Jose Diridon before.

      Good Lord...are you being sarcastic? Why not add a turntable and coaling station while you are at it.

    4. "Turnback tracks" -- out-of-revenue-service extra track extending beyond passenger platforms -- are about the single best way to run up capital and operating costs.

      Let's see: massive extra capital cost for the tracks (usually in tunnel, for extra benefit) themselves. Huge cost for extra vehicles to account for the added non-revenue hours they waste going out of service, running out and back on non-revenue tracks, and finally reappearing back in service. Large ongoing extra operating costs for the vehicles and crews that waste time in non-revenue moves.

      Think about it: if you had to think of the single worst possible way to spend money, could you do better than digging tunnels in a CBD to effectively park out of service trains? (Same argument to a lesser extent even for suburban endpoints.)

      Pathetic sloppy irresponsible unprofessional operators (ie all US ones) love this stuff, because it inflates capital budgets (always the prime motivation) while allowing the most dismal undisciplined operations. Win-win! The transit-industrial mafiosi love it even more, because they're the ones who build the unnecessary structures and tunnels and trains.

      And as icing on the cake, these "tail tracks" generally function as advance starts on unjustifiable further line extensions. "We're already in Burlingame, might as well keep going to Santa Clara."

      A single scissors crossover, placed as close as possible to the outbound end of wide two-track island terminal platform can in fact work extraordinarily well. There are downsides, and single points of failure, and throughput is in fact slightly lower, but the huge "gift that keeps on giving" disadvantages of tail track turnback operations means that only those with no care in the world about costs or service make that choice.

      From a June 2005 article in International Railway Journal:

      The track layout at each end of a line is a bottleneck in most metros. It controls the ultimate through-put of the line. By optimising the layout of the terminus it is possible to reduce the number of trains required to operate the peak-hour service.

      There are two main types of terminus design. The so-called in-front-of-terminus design, whereby a cross-over is installed on the approach to the final station, avoids the need for turn-back tracks, and will save one train. However, this requires a centre platform which may not be possible in all cases. Table 2 shows the potential saving of implementing an in-front-of-terminus design at both ends of the line.


      Area of saving: Estimated saving ($US millions)
      12 cars at $US 2 million/car: 24
      600m overrun tunnel: 60
      Interest only charges at 7% a year over 10 years: 60
      Maintenance costs at $US 50,000/car/year: 6
      Total saving over 10 years: 150

      An after-the-terminus design has the benefit of providing a place to store a defective train or a train taken out after the peak hour. It may also serve as a connection to the depot. However, it requires an additional train, and 300 to 400m of track to turn the train around.

    5. It lets the train platform at a higher speed. Not terribly important on a low frequency line but there are cases where they can squeeze a few extra trains onto the line because the end station isn't the place where they turn the train around. Very few places.

    6. I'm in São Paulo at the moment. Rode the #1 Metro to its northern terminus the other day. As I was merely joyriding, I alighted on the platform looking for how/where to catch the next train back going back the other way. In my terrible Portuguese, I asked the driver who was alighting with me (I was in the first car behind the driver's booth) and he indicated I'd better get right back on ... which I did just as the doors closed and the train zipped off going the other way. Total dwell was right around 1 minute or so.

      The subway system here carries around 4 million per day on only 5 or so lines. Jam packed is an understatement. 1-3 minute headways on most lines during the peaks which are extraordinarily long. Of course, in city of ~20 million, they need a hell of a lot more subway lines ... which they are (interestingly) trying to build like mad now that Brazil (and São Paulo) will be hosting World Cup soccer (football) soon. Weirdly, some of the newest lines I noticed under construction were ... mono rail! Can't wait to see how that works in a city where transit is run very hard, frequently and jam-packed nearly all the time.

    7. Richard, I did not explain correctly. My point is reduce platform occupancy and vacate for next train immediately. If we want to run 8 train/h into Transbay with 2 tracks, Caltrain need to trun-around 10~13 minutes. 12 train/h needs 5~7 min of turn-around time. (You know Caltrain's "US commuter rail" culture and how difficult to archive this?)
      Caltrain mangement want to reduce the number of train into Transbay because of "Capacity" with their standard.
      If we build trun-around tracks beyond the station, Caltrain cannot make any excuse.
      With 12 train/h, we can archive very good schedule of local and two types of express each runs every 15 min frequency.

  7. @Richard, so, how and why is it that obviously incompetent station and track layout designers were allowed to f*ck up the train station portion of the TTT? I would think that only people (consultancies) with a track record of best practices state-of-the-art design skills would be chosen and allowed to determine the track and train station layout design.

    I believe that you (and others?) tried like hell -- and in vain -- to point out the numerous and substantial deficiencies in the TTT track and train station design and that you were essentially blown off or placated. But it seems that a major mistake was made well before that point -- the people/groups/stakeholders who "owned" that design task were obviously not up to the task -- let alone aware of and able to apply "world class" best practices.

    I realize that probably nothing that gets said or posted now will have much chance of changing or improving the outcome ... but nevertheless, it is (or seems to be) somewhat valuable to conduct and understand an "autopsy" of how and why we are left with such a terrible-seeming design when a far better one could have been specified in its place for little or no (or less?) cost.

    What do you see as the root cause of the lousy design, and how would/should it be avoided in any such future project(s)?

  8. @Richard: continuing from my prior comment: of course, I assume that nobody that had any influence or control over the track and station design/layout had any incentive to choose and advance a sub-optimal design ... so that just leaves me to assume that they were unwilling or unable to recognize the weaknesses in their own design(s) and do better -- even with outside input telling them so from outside sources/stakholders (such as yourself). So yes, just to be clear ... my question assumes they didn't know and/or care to design a better track and station layout. That they were not *deliberately* and *knowingly* submitting and promulgating/defending what they knew to be a crappy/sub-optimal track & station design.

  9. Re "For an all-stops local, this starts adding up, and saves about half again as much time as saved by switching from diesel to electric," which trains are you comparing here? Because in YouTube videos, a FLIRT has a 14-second stop penalty accelerating to 100 km/h, vs. about 33 for a Talent, 43 for the Colorado Railcar with half the cars powered, and 70 for the diesel loco-hauled trains the MBTA runs. That's a difference of way more than 10 seconds per stop. If the top speed is 160 km/h instead of 100 the time saving only grows.

    1. On the peninsula corridor, you get about 30 seconds per stop by substituting a Stadler double-deck EMUs (running at a pedestrian 4MW, the max continuous rating) compared to the regular F40 + 5 cars consist. I estimated level boarding would save around 15 seconds per stop, hence my point that level boarding saves about half again as much time as electrification.

      You can try different combinations of power and top speed using simulated run times for the peninsula corridor at our handy service pattern generator.

      You'll see that throwing more speed or more power at an all-stops local is not nearly as effective as level boarding.

    2. Ah. "Half again" means "1.5 times as much." I misunderstood.

    3. My bad, english wasn't my first (or even second) language. I meant to write that level boarding is about 50% as effective as electrification, and because it will be considerably cheaper than electrification, it may provide similar bang-for-buck.

    4. "it may provide similar bang-for-buck"

      If electrification infinitely effective? Because that is what it would take.

      In the last decade or so Caltrain has rebuilt platforms at all of the following stations without regard for level boiarding. If there had been any brains anywhere, all of them would today be ready for low-floor level boarding. (High floor? Expensive, not necessary. Let HSR work with Caltrain and Metrolink, not the other way around. 40 stations vs 5!)

      * Tamien
      * San Jose
      Not College Park, but should be closed, so who cares?
      * Santa Clara
      * Lawrence
      * Sunnyvale
      * Mountain View
      * San Antonio
      * California Avenue
      * Palo Also (not completely rebuilt, but lots of platform work done)
      * Menlo Park
      Not Atherton, but should be closed, so who cares?
      * Redwood City
      * San Carlos
      * Belmont
      * Hillsdale
      * Hayward Park
      * San Mateo
      * Burlingame
      Not Broadway
      * Millbrae
      * San Bruno (rebuilt three times so far, and counting!)
      Not South San Francisco (killed by third stupid San Bruno rebuild.)
      * Bayshore
      * 22nd Street
      Not Fourth and King.

      Conclusion: today we could have level boarding at zero cost.
      Level boarding is guaranteed to make Caltrain run cheaper and faster.

      Compare to electrification at $1.5 billion (and counting)
      Electrification? Some have said that Caltrain's number show it will make Caltrain more expensive to run.
      Caltrain's "blended plan" shows that it will make trains run slower.
      But Atherton and Broadway will get to have trains stop. Good investment for $1.5 billion!

      1500000000 / 0 = infinite! (OK, budget $50 million to rebuild unrebuilt Fourth and King. 15000000000 / 50000000 = 30 times the cost of level boarding, and maybe no benefit at all!)

    5. That's a pretty sad state of affairs... I hadn't stopped to consider that they had rebuilt that many platforms.

      So will it be a money pit? Yes. Par for the course.

      I can see both sides of the high vs. low platform argument, but rebuilding an 8-inch platform for level boarding is going to cost pretty much the same regardless of height. The transition is also a challenge because once you get past a height change of more than one step (which we will, regardless of height) you can't operate any of the current fleet without modification, which will break any reverse-compatibility with 8-inch platforms. All the platforms can't spring forth overnight, so it becomes a problem to retain any of the old fleet during the platform transition period as Caltrain seems to be planning.

    6. What old fleet? As I recall, planning for these platform projects happened at the same time Caltrain was doing fleet "modernization" of the gallery cars.

    7. Old fleet = 25x Bombardier bi-level with 25 inch floors, as opposed to new EMUs. These cars are the reason for Caltrain picking 25 inches as their level-boarding height. They are probably worth what, $1 to $1.5 million each today? So what we have is this: they are planning service for decades into the future, involving several billion dollars, all based around a 25 to 40 million (with an M) dollar investment.

    8. And won't they be decades old reefs by the time anything gets done?

    9. Any bi-level is going to have a lower floor within spitting distance of 25 inches above top of rail. (Usually lower, but that's because most other places don't have the vertical space to play with that you do in the Western USA.)

      Every big European train builder can deliver single-level commuter trains with low floors within spitting distance of 25 inches. Some even come close with single level high speed trains.

      Every big global train builder would be willing to listen, listen very very attentively, to somebody who comes asking about scores or hundreds of double deck high speed cars with doors somewhere around 25 inches above top of rail.

      The North American Bombardier cars aren't freaks. There are many hundreds of things closely resembling them (but less tall) all over Central Europe as well.

      Platforms that match up with the lower level of Bombardier cars aren't freakish.

      What's freakish is supposing that the historical way of building rail cars -- boxes mounted on top of wheels, with some tech-y stuff slung in the space between the wheels -- is the only way to do things today or tomorrow. It's not necessarily a bad way of doing things, but transportation is about engineering systems in contexts, not about saying "platforms are 48 inches above the top of the rail now deal with it."

      Signed, Not Richard, but I know he's on the same page.

    10. That would be ideal. Now how do we make it real? How does one get the CHSRA and its consultants to consider this possibility for even a second? (this is a serious question)

    11. Clem: Persuading a set of entrenched bureaucrats, the CHSR Authority, to rescind their decision to use high level boarding platforms (42 to 49 inches) will never happen if lower floor rolling stock threatens to impede the necessary effort to achieve HSR train operating stability. Train modifications needed to achieve high speed operating stability, in addition to reducing wheel surface conicity from 1/20 to 1/40, careful research has determined that lengthening the wheelbase of the bogies and reducing the non-suspended weight on the bogies is effective for increasing the speed range where bogy transverse instability or hunting will not occur.
      Quoting Bernard de Fontgalland Secretary General of the International Union of Railways points out on page 35 of his book The World Railway System, Cambridge University Press: “This last point has particular application in a new conception of the proper positioning of traction motors, which are placed under the wagon bodies, and no longer incorporated into the bogies, and consequently in the development of transmissions between motor and axle which are compatible with this kind of construction. The final result was the development around 1970 of rail motor sets capable of traveling at 300 km/h on traditional tracks -with as much safety and comfort as conventional trains traveling at 150 km/h.”
      So how much room will be left for passengers on a drop-frame double-deck car with platform doors 25 inches above the rail that also accommodate extra-long-wheel-base bogies plus traction motors mounted within the car’s lower body?
      There are additional substantial economic reasons a HSR system designer would be encouraged to choose single-deck-high floor level rolling stock. HSR balancing-speed data implies that on straight-level-track at 220 mph wind resistance is over 20 times all other friction sources combined. Air viscosity times a train’s surface area times the rate air velocity changes as one moves away from a moving train’s side dominates a long train’s total air resistance. Rolling through tunnels increases air resistance for passenger trains, but a drop-frame double-deck train’s closeness to the track-way exacerbates a train’s air resistance losses which are particularly acute during high-speed operations.

  10. Did anything non-awful get said by anybody other than Clem at the don't worry your pretty little heads we have it all under control now sign over a billion and a half fest?

    Correction: it's possible they said not-entirely-awful things ("level boarding is important", "we value your contributions", "please feel free to contact us", "we have the best people working on this", "electrification is awesome", "wonderful things are happening, just wait and see", "give Caltrain more money") but was there any evidence they meant any of them (other than "give Caltrain more money")?