30 March 2024

Level Boarding is Legal in California

Arrow level boarding platforms
at San Bernardino, CA

Comments to old posts on this blog are stored in a moderation queue that your author doesn't visit often enough. Over a year ago, commenter jpk122s discovered quite a gem: an official resolution by the CPUC (California Public Utility Commission) that level boarding station platforms are not bound by General Order No. 26-D section 3.4. This means it's nerd time.

Some California Background

The CPUC regulates all railroads in California, including their clearance dimensions under General Order 26-D. This regulation, originally published in 1948, requires all mainline train platforms to be no higher than 8 inches above top of rail per section 3.3. If you want to build a station platform higher than 8 inches, it needs to be set back at least 7'6" from the track center line per section 3.4. This requirement is deeply inscribed into the built environment of train stations around California, including Caltrain's.

  • The taller platforms used for boarding passengers with reduced mobility, known as "mini-highs" and cluttering the north end of most Caltrain station platforms with ramps and railings (see diagram below), must be set back at least 7'6" per section 3.4. This is quite far from the track, requiring the use of bridge plates to cross the wide (~3 foot) gap between the mini-high and the train.
  • The 48" level boarding platforms used by SMART (in Sonoma and Marin counties) are closer than section 3.4 requires, but as mitigation, a set of gauntlet tracks allows freight trains to stay clear.
  • The 23.5" (ish) level boarding platforms used by Sprinter (Oceanside to Escondido) are closer than section 3.4 requires, but as mitigation, they have folding edges that tilt up and out of the way of freight trains that pass during the night.

Current Caltrain platform standards
These examples are all Rube Goldberg solutions that are expensive, clunky and inconvenient – especially when considering that nothing physically precludes freight trains operating past high platforms, as is common practice on the east coast.

Then, along came the San Bernardino County Transportation Authority, with a request for an exemption from section 3.4.

Level Boading for Arrow

Arrow is the brand name for a new passenger rail service linking San Bernardino to Redlands. This service uses Stadler FLIRT diesel multiple units, of a standard vehicle design sold in more than 2500 copies around the world. The platforms are built for level boarding at 23.5" to comply with the accessibility requirements of the Americans with Disabilities Act (ADA). Rather than contrive a new technical solution to comply with GO 26-D section 3.4, the parent agency did something unusual: they asked for an exemption.

The May 5th, 2022 resolution adopted by the CPUC, an agency known for its conservatism and dogged focus on safety, was surprising: "The RSD [Rail Safety Division] has determined that an exemption from General Order 26-D, Section 3.4 is not necessary since it is preempted by the federal Americans with Disabilities Act (ADA)." Section 3.4 (a state regulation enacted in 1948) is preempted by the ADA (a federal law enacted in 1990). The resolution continues:

General Order 26-D, Section 3.4, sets forth a minimum clearance requirement for station platforms. However, this provision of General Order 26-D is preempted by the ADA, which requires a different platform height and distance from track center line to accommodate the introduction of the Multi Unit (MU) equipment– and thus, results in a smaller clearance area – than what is set forth in General Order 26-D, Section 3.4.

Interestingly, the freight railroads that usually complain about the slightest infringements to their operating environment did not comment on the resolution before it was adopted by the CPUC.

Implications for Caltrain Level Boarding

Perhaps Caltrain already knew this all along, but this CPUC order implicitly relieves one of the key regulatory constraints to platform heights and level boarding, discussed numerous times in the past 15 years of this blog. It turns out that no waiver of GO 26-D section 3.4 is ever needed.

It may take a year or two before Caltrain finds out the hard way why they need level boarding, but this is a positive development. For that, we have the San Bernardino County Transportation Authority to thank.

03 February 2024

The Cost of EMU Maintenance

Caltrain recently published a strategic financial plan update, where we learn that maintaining each EMU in the new electric fleet in good working order is expected to cost $1.2 - 1.5 million per year, a significant increase from last year's estimate. This post seeks to answer the question: is that crazy?

This analysis revisits and updates an older post here.

Historical vehicle maintenance costs

Note these figures are in constant 2023 dollars
The National Transit Database is a fantastic resource provided by the federal government, charting facts and figures for every transit operator in the United States. For the period 2000 - 2022, we look up vehicle maintenance costs and vehicle revenue miles for Caltrain as well as for two regional rail operators in the New York City area (Metro-North and LIRR), who operate the largest "heavy rail" EMU fleets in the United States in a region with similarly high costs as the Bay Area. Note that New Jersey Transit is not included because teasing out their large bus fleet from the overall agency figures is complicated. Dividing vehicle maintenance expense (adjusted for inflation to 2023) by vehicle revenue miles, we get the approximate per-mile cost of vehicle maintenance. One can guess that Caltrain's increase after 2017 is related to operating the legacy fleet way past its retirement age. In constant 2023 dollars, eyeballing this chart, the cost of maintaining EMUs might be around five bucks per vehicle revenue mile -- let's charitably say four because Caltrain's fleet is brand new and won't break down as much as older fleet mixes used by the other operators.

Caltrain EMU vehicle revenue miles

The new service plan published by Caltrain makes it easy to calculate the number of annual EMU vehicle revenue miles. There are 66 trains per weekend day (33 in each direction, for 16 hours of half-hourly service, with every other train serving Tamien) and 104 trains per weekday (52 in each direction, with 18 hours of half-hourly service = 36 plus 8 hours of additional peak express service = 16). It's 46.7 miles from SF to SJ, and 48.4 miles from SF to Tamien. That adds up to 2*(16*48.4+17*46.7) =  3137 revenue train miles per weekend day, and 2*(18*48.4+18*46.7+16*46.7) = 4918 revenue train miles per weekday. Each train has seven vehicles, as defined by the FTA. Assuming each year has 6 holiday weekdays with weekend-like service, that all works up to 11.2 million vehicle revenue miles per year, which is... a lot. For context, the most service Caltrain ran pre-covid was 7.9 million vehicle revenue miles. The increase of 41% arises from running half-hourly service all day, every day, with long seven car trains.

Caltrain projected vehicle maintenance expenses

Note these figures are in year of expenditure
The 2024 strategic financial plan helpfully breaks out projected annual vehicle maintenance expense by fleet, with the EMUs charted separately from the diesels. The EMU costs are shown at right (MoE = Maintenance of Equipment). These figures are in year of expenditure, not inflation adjusted, so we need to make some assumptions before we can compare apples to apples. Taking 5% inflation and deflating these figures back to 2023, the previous forecast (in red) was $12M, while the new forecast (in blue) is closer to $21M with 19 EMUs climbing to $25M when all 23 currently on order are delivered. We can guess that in the out years, Caltrain is assuming that ridership has bounced back enough that the FTA will require them to operate six trains per peak hour per direction (104 + 32 = 136 trains per weekday) as originally planned, further increasing to 13.9 million vehicle revenue miles per year.

Putting it all together

We've made some assumptions that are not completely valid -- namely that vehicle maintenance cost scales directly with the number of revenue miles operated. To first order, this is true, but vehicle maintenance cost has time-based components (such as mandated inspections, or replacement of ultraviolet-crazed window glazing) and distance-based components (such as wheel and brake wear). Not everything scales proportionally to revenue miles. With this caveat in mind, let's see what happens.

If you multiply 11.2 million vehicle revenue miles by $4 of vehicle maintenance cost per vehicle revenue mile from the National Transit Database, you end up at $45M per year (again, with everything in 2023 dollars.)

Caltrain's latest figure is half that, so what looks like a large increase in their latest strategic financial plan may still be an underestimate. Their estimate of $21M divided by 11.2 million vehicle revenue miles gives just $1.90 of vehicle maintenance per vehicle revenue mile for the EMU fleet (in 2023 dollars), a value lower than Caltrain has ever achieved with its legacy fleet.

Seen another way, 11.2 million vehicle revenue miles operated with 19 trains works out to 590,000 vehicle revenue miles per year (or, since each train has 7 vehicles, 84,000 miles per year on the odometer) corresponding to $2.4M of vehicle maintenance cost per EMU set per year. Their estimate of $1.2 - 1.5M seems low in comparison. This could be due to high utilization, which would dilute the time-based component of maintenance cost.

Verdict: these EMU maintenance costs are not crazy -- they might even be too low.

Appendix: Gilroy branch

While we're at it, we can do a quick sanity check on the Gilroy branch, which will continue to operate with a reduced diesel fleet. There are four weekday round trips. Gilroy is 30 miles from SJ. The trains have five cars. This works out to 1200 vehicle revenue miles per weekday, or 306,000 vehicle revenue miles per year, or less than 3% of Caltrain's total. At four bucks a mile, that's $1.2M per year for the entire diesel fleet.

Caltrain's figures are closer to $7M per year (again, in 2023 dollars). That seems like a lot, but consider the extremely low utilization of the dedicated diesel fleet, less than 7000 miles per year per diesel locomotive, means that time-based vehicle maintenance costs will dominate, the opposite of the EMU fleet.

If the fiscal cliff is real and not some made-up crisis, then the under-utilized rail service on the Gilroy branch should be replaced by bus service and the entire diesel fleet that is dedicated to it, maintenance costs and all, should be unceremoniously dumped off Caltrain's balance sheet.

07 January 2024

New Year, New Risks

It's 2024, the year that Caltrain is supposed to go electric. All the wires are up and six trains are already on the property (see delivery spreadsheet), with more on the way shortly. After years of delays, will they pull it off?

Seems like a good time to review five risks facing the project.

1. PCEP schedule slips - while monthly reports of the Peninsula Corridor Electrification Project continue to assert that the project is on track for "Fall 2024," a nebulous date that could well be the last day of fall or December 20th, there are worrying slips in the project schedule. The November 2023 monthly report (from the January board meeting packet, PDF page 131) revealed a three-month slip in the critical path compared to the previous monthly report (from the December board meeting packet, PDF page 159). Completion of live runs on segments 1 and 2 between San Francisco and Menlo Park was pushed out from 12/17/2023 to 3/16/2024. Three-month slips this close to the finish line do not bode well for finishing on time.

Tree down on wires (KRON4)
2. Trees falling on tracks - as reliably as atmospheric river storms occur in the Bay Area, large trees will continue to fall across the tracks. Previously, a few hours of chainsawing was enough to clear the blockage and resume service. No longer: trees will now damage the overhead contact system (OCS), requiring repairs to high voltage equipment before service can resume. On January 5th, 2023, a large eucalyptus tree fell across the tracks and did just that. According to a news report, service was interrupted for most of the day to safely remove mangled poles and wires-- and this without any urgency to repair them before restoring diesel service.
Another one, February 2024

The craziest part of this story: it took until September 2023, a gestation period of nine months, to complete the OCS repairs due to long lead times to procure replacement parts. While there would have been more urgency had the OCS been needed to operate the service, this episode highlights a lack of preparedness for what will become a routine occurrence. It should not take more than a few hours to get temporary OCS repairs completed, and the winter months of 2024 will provide valuable opportunities for practice.

(UPDATE 17 February 2024: it happened again and we're at two weeks and counting for the repairs)

To mitigate this risk: hold negligent tree owners financially liable for damage and delays caused by their trees falling on Caltrain, and aggressively trim back vegetation. Establish a well-equipped rapid response team of "squirrels" (OCS maintainers) who can quickly deploy to an incident site to perform temporary repairs that allow service to resume quickly. Keep this crew sharp by regular practice of repair methods, and stock an ample and ready supply of spare parts.

3. Grade crossing collisions - crossing wrecks are another frequent occurrence that will continue into the electric era, even if the new trains have much more powerful brakes that can avoid some collisions. With old diesels, you could cut, bend and weld beefy steel parts, quickly returning equipment to service. With an EMU, a collision can do more damage: crumple zones will crumple, and the fiberglass front-end mask and cladding will be potentially costly and time consuming to replace.

To mitigate this risk: improve crossing safety equipment and lighting, and grade separate the busiest crossings. Keep enough spare parts (including entire front-end masks) locally, so repairs don't require long lead times or a trip back to the factory in Utah.

4. Wheel flat spots in wet weather - while the new EMUs have the latest in computer-controlled braking technology, their swift acceleration and braking will put greater demands on controlling friction at the interface between wheel and rail. Throw in some moisture and crushed eucalyptus leaves, and even the best computer won't always get it right. It doesn't take much sliding of a wheel to create a flat spot, making that loud whomp-whomp-whomp sound. BART found this out the hard way, having to pause delivery of their new fleet while software changes were made. Caltrain's plans for a 75-minute local require very aggressive acceleration and braking, increasing the risk of flat spots.

To mitigate this risk: do lots of wet weather testing to find the limits of the software, and set limits to prevent train crews from driving too aggressively. Get lots of practice truing EMU wheels on the lathe.

5. Copper theft - there has already been a problem with thefts of impedance bonds, devices that allow traction return current (at zero volts) to cross signal block boundaries. These bonds are easily accessible on the track, but European railways have also experienced copper theft of live components energized at 25 kV by thieves who know their way around high voltage.

To mitigate this risk: secure valuable inventory, use identifying markings to prevent stolen copper from being easily sold for scrap, and maintain a large supply of spares to rapidly restore service in case of theft. Another job for the "squirrel" rapid response team.

In closing, it is commonly accepted that electric trains are more reliable than diesels, as one would certainly hope given how often Caltrain's decrepit fleet breaks down. Mechanical problems cause an average of 47 minutes of train delay every day, coming in third position after delays due to construction and trespassers. While new electric trains should bring this number down, electrification itself exposes service quality to new risks that Caltrain must anticipate and mitigate. Failing to control these risks can quickly turn electric revenue service into a fiasco. 2024 is the time for robust contingency planning.

19 September 2023

First Look at Electric Service

Caltrain's proposed weekday peak service
Caltrain has started to pull back the curtain on their fall 2024 service pattern, when the electric fleet will (finally!) enter service.

The peak service pattern is depicted at right from a Caltrain slide, and looks like this in a string diagram from our trusty taktulator. The overall score for this timetable is a lukewarm 115 points relative to the benchmark score of 100 for the 2011 timetable, and it requires just 14 trains to operate, not counting spares.

The Good

  • Peak frequency will remain at four trains per hour per direction until ridership recovers more. This requires a waiver from the FTA, which originally made its funding contingent on operating six trains per peak hour per direction.
  • This enables operating with all-electric service in the portion of the corridor between San Francisco and San Jose, relegating the diesel fleet to where it belongs: not under the wire.
  • Stops are restored throughout the corridor, increasing service frequencies especially in Silicon Valley where the 2004 arrival of the Baby Bullet decimated ridership at skipped stops that nevertheless have healthy numbers of nearby jobs and residents.
  • Off-peak service is a subset of peak service, making for a regular and consistent half-hour "takt" throughout the entire day. This makes the timetable easy to use and memorize.
  • The Gilroy branch is served by a cross-platform transfer in San Jose, reflecting the results of a recent survey of its customers that revealed that a one-seat ride was their lowest priority.

The Bad

  • The Baby Bullet unfortunately survives as Express A, skipping perfectly worthy stops throughout Silicon Valley and tragically breaking the potential symmetry of the timetable. Express A should just be another Express B, giving evenly spaced 15-minute service from Redwood City to San Jose. This is very low hanging fruit that improves the timetable score from 115 to 118 points. The Baby Bullet pattern was a great idea for speeding up diesel runs, but in the new electric age, Caltrain should let it slip into history.
  • The "South County Connector" remains as diesel rail service. There is an opportunity for  enormous savings (and better service quality for riders) from selling off the entire diesel fleet and operating the Gilroy branch with much more frequent luxury buses. Operating and maintaining an excessively large electric + diesel fleet and wasting capital funds on the expensive but dubious idea of a BEMU distracts Caltrain from its core mission. If one believes in the idea of a fiscal cliff, then it's obvious that rubber tires can help Caltrain's balance sheet until the high-speed rail project electrifies the south county corridor.
The Implausible
  •  The proposed service pattern is built around a 75-minute all stops local trip between SF and SJ. This 75-minute figure is very impressive and would be a huge strength, if only it didn't border on the implausible. These are several issues with a 75-minute local run:
    • the trains have to accelerate very aggressively, something the equipment is admittedly capable of, but ouch, that electricity bill! Operating cost will depend not just on usage but also peak load, and high acceleration makes the peaks worse.
    • the timetable padding has to be shaved down significantly from today's comfortable margins
    • the station dwell time has to come down to a crisp BART-like duration of just 30 seconds. Without level boarding, this is unlikely to work reliably in daily service. The boarding step arrangement, door spacing and interior circulation of the new EMUs is identical to the Bombardier diesel sets, so it's hard to believe they would board any faster, before even considering the possibility of a crew interventions for wheelchair users.
    • There is nothing physically preventing a 75-minute trip, but if Caltrain can pull that off, what secret sauce of brief dwells have they been holding back from us all these years? Based on past operating practice, a more reasonable expectation would be an 86-minute trip. Let this serve as a prediction that Caltrain will find out the hard way that it really does have an urgent need for level boarding.

Taken together, there's a lot to like in this emerging service pattern, and a few tweaks can make it even more optimal. With freeway traffic getting steadily worse, the new service product will sell itself.

02 September 2023

Level Boarding: Still Not Getting It

The good news: Caltrain has initiated a small study effort to develop a level boarding roadmap, as part of its portfolio of capital projects.

The bad news: in the summary of this study, Caltrain shows no sign of grasping the purpose of level boarding. We might need to display it on a freeway billboard, like this:

Caltrain does a very nice job of explaining the benefits of electrification. Faster acceleration leads to shorter trip times, a strong message that they hammer often. On level boarding, however, the messaging is muddled. It's something-something about steps? Easier and more inclusive access?


It's about shorter trip times, just like electrification. While electrification saves time in motion, level boarding saves time at rest. The savings are big: cutting station dwell time from 45 seconds (typical for Caltrain today) to 30 seconds (typical for BART, which has level boarding) is worth almost as much "acceleration" as electrification. Electrifying without level boarding is half-baked, and Caltrain should be spending a lot more on level boarding than distractions like BEMUs. Please return to your core mission to quickly and efficiently get people where they are going.

About these numbers: the diesel trip time is for an all-stops local with 45 second station dwells and 15% padding. The EMU trip time has the same dwell and padding assumptions. The level boarding time assumes that station dwells drop to 30 seconds, and padding is cut down to 10%. The lower padding is appropriate because level boarding not only makes dwells shorter, but it makes dwells much more consistent and predictable, as discussed previously. If two wheelchair users need to board, it takes the same 30 seconds, not five minutes of staff assistance. Here are the detailed stop-by-stop stats and string diagrams if you want to tinker with assumptions.