Down the Tubes with DTX!

DTX overview

San Francisco's Downtown Rail Extension project (DTX), officially known as the Transbay Transit Center Program Phase 2, is a two-mile tunneling project to extend the peninsula rail corridor from its existing terminus in the Mission Bay neighborhood to the purpose-built basement "train box" of the Salesforce Transit Center (SFC). The project is regionally important, as there are more jobs located within a half-mile radius of the SFC than within a half-mile radius of all Caltrain stops combined, from 4th and King all the way to Gilroy. The DTX is nearly shovel-ready, in the sense that environmental clearance is in hand and engineering is being advanced to award construction contracts the moment a key ingredient becomes available: money. Gobs and gobs of money.

Too Big To Fail

The last time the costs of the Phase 2 project were tallied in 2016, the total came to $3.9 billion in year-of-expenditure dollars assuming a 2025 opening. Due to delays, we can anticipate at least another five years of escalation at 5%, bringing us to $5 billion before any changes to the project scope. One can reasonably expect that Bay Area transit agencies' proven inability to deliver mega-projects on budget or on time is quite likely to blow up costs well beyond these figures. As a recent example, the Phase 1 project, completed in 2019, cost $2.4 billion (year-of-expenditure) or about 50% more than the $1.6 billion YOE budget of May 2010, adopted after the train box scope was added.

The DTX project's regional, state and national significance is certainly not lost on our Transportation Industrial Complex. To improve the chances of getting the Phase 2 project federally funded (after which any cost growth becomes easier to fund, following former SF mayor Willie Brown's "theory of holes"), the TJPA is undertaking a phasing study to make the project appear more thrifty. The various approaches include deferring or deleting components of the project, such as a pedestrian connector to BART, an intercity bus facility, and an extension of the basement train box. This nibbling around the edges amounts to $0.4 billion in 2027 dollars or about 8% of the total Phase 2 project cost, a drop in the bucket.

A $30 million project development study is now in the pipeline to get Phase 2 to the state of readiness required to apply for federal New Starts funding by August 2023.

PAX: The World's Most Expensive Grade Separation

If you thought the cost of grade separations is exploding, you really haven't seen anything yet: meet the Pennsylvania Avenue Extension (PAX) addendum to the DTX, a grade separation project that will approach $2 billion for two crossings, reaching the stratospheric cost of $1 billion per crossing.

Even after spending $5 billion (before inevitable cost overruns), the DTX project will leave two existing street crossings at grade, at Mission Bay Drive and 16th Street. Not to be outdone, the city and county of San Francisco has performed a methodical series of planning studies to conclude that a new grade separation project is needed. Rather than taking on the challenge of bending some design rules to keep it simple and make it fit, the favored paint-by-numbers engineering solution is a bored tunnel, which averts any conflict with a planned 27-foot sewer pipe and the sacrosanct pile foundations of the I-280 viaduct, each of which are under the jurisdiction of other agencies. The combined cost of DTX + PAX is estimated at $6.0 billion. Take away the latest (2016) $3.9 billion cost estimate of DTX and you get about $2 billion added for PAX.

Link21 Crashes the Party

Meanwhile, BART is in the early planning stages for beefing up its throughput capacity between the greater East Bay and San Francisco, with a second Transbay Tube. It's worth pausing for a moment to consider what an astonishing piece of infrastructure the first Transbay Tube already is: it carries almost twice as many people during rush hour as the entire ten-lane freeway that is the San Francisco - Oakland Bay Bridge, and at significantly faster speeds. Looking past the pandemic, long-term growth trends indicate that the region must plan for a second Transbay Tube.

Transbay Tube II is the centerpiece of an enormous regional rail program known as Link21, the scale and ambition of which dwarf the DTX. While there are many decisions yet to be made about the implementation details of Link21, perhaps the most critical decision centers on what technology to put in the tube: wide-gauge BART, standard-gauge regional rail, or both.

This question is already of great concern to TJPA, which writes in its August 20, 2021 Phasing Study:

BART and Capitol Corridor’s Link21 program is currently in the early stages of development and has not yet determined a preferred alignment, technology, or rail gauge options to meet their goals and objectives for a future transbay rail crossing. As expected at this stage of development, all options remain available for consideration. For example, Link21 may determine that a second transbay crossing best meets stakeholder needs if it provides additional capacity for the BART network only and does not provide a standard gauge rail crossing of the Bay. BART’s infrastructure and trainset design, however, are incompatible with Caltrain and CHSRA standards. Most significantly, BART operates on a wider track gauge with vehicles that may not meet collision requirements, and therefore a BART-only connection would not relieve congestion and conflicts on the DTX.

We can already see a problematic mindset emerging here, where "BART" is automatically conflated with "five-foot-six track gauge," setting up a false dichotomy of BART-or-standard-gauge.

Caltrain + BART: a Necessary Merger

The false dichotomy of BART-or-standard-gauge threatens to poison the debate around Link21 alternatives. The Transportation Industrial Complex has a vested interest in this incompatibility, as it ultimately forces multiple mega-projects to be built. Why build it right when you can build it twice and get paid twice? From the standpoint of scope and profit maximization, it would then make sense to keep DTX and Link21 as separate projects, despite their overlapping purpose and need to link the greater Bay Area megaregion together using high-capacity passenger rail infrastructure. Seamless integration is good for riders and taxpayers, but not so great for consultants and civil engineering mega-firms. That's why these firms have an interest in propagating the myth that BART and standard gauge rail will always be mutually exclusive.

Removing this false dichotomy is becoming a primary reason for merging Caltrain with BART to form a single Bay Area Rapid Transit system, although there are many other reasons. BART does not have to be synonymous with wide gauge; indeed, BART already operates a seamless standard-gauge extension between Pittsburg and Antioch, and provides day-to-day management of the standard-gauge Capitol Corridor. A new BART peninsula line, while indistinguishable from Caltrain's service vision, would suck the air out of the emerging pointless debate around the track gauge of the second BART transbay crossing. The Measure RR sales tax can serve as a dowry to integrate San Mateo and Santa Clara counties into a restructured BART district.

Link21, to its credit, places equity and inclusion at the forefront of its project development process. The contrast with DTX is jarring, as TJPA's Phase 2 project can easily be viewed as just another gold-plated white-collar rail project enabling nine-to-five technology and finance types to more conveniently access San Francisco's skyscrapers from the affluent suburbs to the south, without ever having to mix with the blue-collar working class. Bringing DTX under the Link21 umbrella, and merging Caltrain into BART, immediately defuses the classism and racism that underlies this anachronistic Mad Men commuter rail vibe.

Transbay Through Running

A stub-end terminal station suffers from fundamental throughput limits related to long turn times and the unavoidable crossing streams of inbound and outbound traffic in the station approach or "throat." For a given number of platform tracks, a through-running station configuration where all trains that come in one end of the station can exit the other end will always provide more throughput capacity, whether measured in trains per hour or passengers per hour. Trains don't have to dwell any longer than necessary at a platform, and don't foul opposing traffic on their way in or out.

With the DTX as it is, past operational analysis indicated that just 12 inbound and outbound trains per hour (8 Caltrain + 4 HSR) would push the limits of the terminus design, with near-saturated platform occupancy. If you uncork the other end of the train box (by having Caltrans clear some right of way i.e. dismantle and redevelop a couple of medium-rise buildings to the East) so that the DTX can connect directly to a new transbay crossing, everything changes. A lot of new capacity is created by virtue of not having to layover or turn trains right smack where your platforms and track real estate is the most expensive.

A recent through running operations analysis commissioned by the TJPA shows that the Salesforce Transit Center could handle up to 20 trains per hour per direction if no more than six of them turn at the station. Any more than six turning movements, and the excessively long platform re-occupancy times (as the study notes, due to the poor layout of the switches leading to tracks 1-4) will reduce throughput capacity to less than 20 trains/hour.

Broken Assumptions at Link21

The TJPA phasing study reports the following direct quote attributed to Link21 project team:

We have received briefings on the operational modeling for DTX and it would seem that even a three-bay DTX tunnel poses operational constraints. A robust service level through the transbay crossing is required to justify investment into Link21. Link21 is envisioning scenarios where not all trains that cross the Bay would continue to San Jose. At this point, there is no other location to turn trains around in the northern peninsula which makes flexibility in DTX important to the Link21 Program.

You read that right: the Link21 team is thinking of turning Capitol Corridor trains at the Salesforce Transit Center, a completely American idea (copied straight from Penn Station New York) that is operationally insane if you think about it for even a minute. In a through-running configuration, all trains that cross the Bay should stop in downtown San Francisco and get out immediately. The California High Speed Rail Authority is planning a huge yard in Brisbane, a perfect place to clean, service and layover Capitol Corridor trains. These deadhead (non-revenue) moves are much less wasteful of infrastructure capacity than treating a through-running station as a terminal.

As was remarked in previous discussions regarding San Jose, the act of parking or laying over trains at a station platform is the railroad equivalent of parking an empty truck in the middle of a bustling loading zone, and then concluding that the loading zone fails to function adequately. Just stop it, don't even think of turning trains here!

The Bottom Line

Here are the pros and cons of merging DTX with Link21:

Pro	Con
Eliminates silly idea of a multi-gauge transbay tube project	Could further delay DTX, since Link21 is at an earlier stage of development
Increases SFC throughput capacity and bang-for-buck, making the enormous cost of DTX worth it	Exposes DTX to political re-prioritization
Provides faster Peninsula - East Bay connections than existing BART, and finally "Rings the Bay with BART"	Greatly reduces scope and profits for Transportation Industrial Complex
Makes more efficient use of taxpayer dollars by building  one project and building it right	Requires inter-agency coordination and mergers, which agencies abhor
Provides seamless regional rail connection from SJ and SF to Sacramento, if Capitol Corridor is electrified

Despite the obvious political and organizational obstacles, from the point of view of a rider and taxpayer, the pros vastly outweigh the cons. The answer is then obvious: the DTX should go down the tubes of a new standard gauge Link21 crossing, with Stadler bi-level EMUs operated by BART seamlessly connecting the peninsula corridor (a.k.a. the new BART Purple Line) directly to Oakland and points beyond. DTX should be built without delay and form the first building block of Link21.

August 2021 Timetable Review

Caltrain was recently returned to more or less full service, with a timetable that is supposedly simpler (with a claim of just five stopping patterns) and features 104 trains per weekday, the most ever. Let's take a closer look using our handy taktulator, which assigns a timetable a score based on frequency and connectivity. The formulation of the service quality metrics underlying the scores is described here.

Caltrain's 2021 peak hour timetable achieves a score of only 96, meaning the service is slightly worse than the taktulator's baseline, the 2011 peak hour timetable, which a decade ago earned our reference score of 100.

Why so mediocre?

It's mostly in the padding. Caltrain service planners have evidently given up on trying to run a tight timetable while also dealing with the debilitating variability in station dwell times inflicted by the lack of level boarding. The right way to solve this problem was and remains to plan for and implement level boarding, a system upgrade that (on a per dollar basis) has even greater service benefits than electrification. The lazy way is what we see here: about 20% of extra padding is baked into the station-to-station run times, allowing a train to easily make up time and arrive "on time" in the event of a dwell time delay along the way. In the absence of a delay, trains can dawdle along even more slowly than the ancient diesel fleet could manage, and just sit at stations until the clock says it's time to go.

Back in 2005, a baby bullet express with five intermediate stops was timetabled at 59 minutes. In 2015, it was up to 61 minutes. Today, the same express runs in 66 minutes. This follows a pattern noted by Alon Levy on the deterioration of speed.

Another factor that explains the lower score is one fewer train per peak hour, resulting in longer intervals between trains. This helps with fleet size, where only 16 train sets (+2 spares) are needed to operate the 2021 timetable where 18 (+2) were needed before. Two more train sets are freed up for maintenance downtime, a vital bit of breathing room as Caltrain's older diesel fleet is breaking down more often. The fleet is well past its expiration date due to the multi-year delays in the electrification project.

Is it optimal?

Can a timetable be devised that uses no more than 16 (+2) train sets and scores better than 96? Why yes it can. Here is a Silicon Valley Express timetable that uses just 14 (+2) train sets with four trains per peak hour per direction, and scores 102.

Why is it better? First, it follows census patterns and puts the stops where they link the most residents and jobs, not where there is the most parking. It is more regular and has fewer gaps with long waits. While this does not figure into the service score, it makes far better utilization of the train fleet (83% of the time in revenue service, versus 70%). More efficient fleet utilization leads to fewer trains and fewer crews being needed to provide the same service, reducing labor and maintenance costs per passenger mile. All this is done without any magic: 15-minute equipment turns, comfy 40-second station dwells and a slightly less absurd padding level of 15%. There are zero overtakes, so fewer opportunities for cascading delays. Finally, this timetable is much simpler to understand for a rider, having just two service patterns.

One can only hope that despite this interim state of mediocrity, Caltrain will successfully implement its "moderate growth" service vision, which scores an impressive 240. Getting there will require reliable 30-second dwells for which level boarding is a must.

Credit to Richard Mlynarik who did the time-consuming part of this analysis.

The Exploding Cost of Grade Separations

Recently, the San Mateo County Transportation Authority prepared a grade separation program update, discussing past and future projects. What immediately jumps out of this document, and others published by Caltrain, is the exploding cost of grade separation projects. The project budgets are shooting through the roof, vastly outpacing inflation. Typical of this cost explosion is the Broadway grade separation in Burlingame, which will grade-separate a single intersection at the eye-watering cost of $327 million.

Cost Modeling of Historical Grade Separation Projects

With the SMCTA slides giving cost data for past and current projects along the Caltrain corridor, it is fairly straightforward to assemble a simple model of grade separation project component costs. All figures are inflated to 2020 dollars before fitting, and we break out unit quantities for each project of the following project components: fully elevated rail over road crossings, split (partially elevated) rail over road crossings, trenched road under rail crossings, pedestrian tunnels, stations, and the number of miles of corridor where the track elevation was changed. With all those quantities broken out for each project, we can fit a simple model that estimates the unit costs by (empirically, not rigorously) minimizing a least-squares fitting residual. The main result of this model is that projects from the mid-1990s through today consistently cost about $36 million per crossing, with not too much variation:

 

That brings us back to Broadway in Burlingame, which according to this model should cost only a third of the price tag of $327 million. That's right, even including two pedestrian tunnels and a new station, the entire Broadway project should cost no more than $100 million. This factor-of-three discrepancy raises some serious questions about how this project is being engineered, and whether it should even proceed in its current form. One could counter that the cost model presented here is too simplistic and doesn't reflect the unique local conditions of this project, but the model does okay with predicting the cost of every past grade separation project over the last 30 years. Is this a case of over-fitting the data, or have the engineers behind this project simply lost their senses?

With the most traffic of all grade crossings on the peninsula corridor and train-on-car collisions occurring on average once a year, the Broadway crossing is at the top of the state's priority list for grade separation, and we all know that you can't put a price on safety. That makes the Broadway grade separation project ripe for name-your-price taxpayer extortion.

Insane Costs are Baked in to the Caltrain Business Plan

A Caltrain business plan presentation from 2019 attempted to quantify the expense of partially grade-separating the corridor for each contemplated service scenario. The cost modeling for this was even cruder than the simple spreadsheet model described above: the costs for each project were either copied and pasted directly from each city's estimates (of wildly varying quality), or a standard grade separation unit cost of $255 - $355 million per crossing was adopted. This value is up to TEN TIMES the value estimated from past and present grade separation projects, and flies directly in the face of common sense. Despite the coarseness of this spreadsheet costing exercise, the resulting grade separation costs (on the order of ten billion dollars regardless of service scenario) were passed along into the regional Plan Bay Area 2050 exercise.

Why have costs exploded for a project like Broadway, which has proceeded far enough into detailed design to accurately estimate construction cost?

Cost Drivers

Utility relocation. Whenever you dig, surprises happen where utilities buried underground are found elsewhere than expected. The more and deeper you dig, the more surprises you will find. Every new discovery delays or even stops construction work, running up costs. Almost every digging project undertaken by Caltrain runs into this situation. Just in the last couple of years: in South San Francisco, construction of a grade-separated pedestrian access tunnel was delayed for 17 months, at an additional cost of $10 million (and still counting!) due to utility relocation issues. In San Mateo, the 25th Avenue grade separation project, where several new crossings were dug, was delayed by over 500 days due to negotiations with Union Pacific over the relocation of fiber optic cables. Pacific Gas & Electric also had to be paid to move a high pressure gas line. The budget for utility relocation almost tripled, from $12 million to $32 million, not counting the cost of construction delays. Meanwhile, corridor-wide, Caltrain's electrification program (while not a grade separation) is continually digging up their own brand new train control fiber optic cables, which were buried in places that don't match what the contractor said they did. This is causing many months of delay to foundation installation. The matter is now tied up in court, as one of several smoldering side sagas in the big bonfire of litigation over the CBOSS project, still building up to a climactic jury trial in 2022.

Vertical curves made for freight trains. Changing the vertical profile of the tracks, whether up or down, is subject to design constraints on the radius of vertical curves, or how quickly (and over what distance) the slope of the tracks is allowed to change. You might think this issue primarily affects faster passenger trains, but amazingly, the biggest culprit is heavy freight. Freight cars maintained to the bare-minimum standards practiced in the United States can derail at the slightest provocation, so industry track design standards are set extremely conservatively. The maximum vertical acceleration allowable for freight cars is 0.1 ft/s^2, six times less than for passenger trains. At equivalent speeds, the grade change (for example from level track to a one percent slope) must then take place over a distance six times longer than for passenger trains. If you wanted to design the vertical profile of a grade separation to the most aggressive vertical radii and shortest structure lengths allowable for passenger trains, the freight trains would have to be slowed down to 1/sqrt(6) of the passenger train speed to stay under their six times lower vertical acceleration limit. On the peninsula corridor, where we design for 110 mph passenger trains, short grade separations require that the freight trains can't go any faster than 45 mph. Unfortunately, new grade separations such as Broadway in Burlingame or downtown Redwood City are being engineered for 60 mph freight speed, which makes all the vertical curves (and bridges, embankments, trenches, etc.) almost 80% longer than they need to be for 110 mph passenger trains.

Vertical curves that can't overlap bridge spans. Recent preliminary design drawings, such as for the Redwood City grade separations, reveal a new design constraint has been applied that does not appear in older Caltrain engineering standards. The vertical alignments are configured such that wherever the track crosses over a bridge span (as for a grade separation) there is no vertical curvature. To understand how wasteful and silly this is, ask any engineer--never mind, ask any kid: is a train bridge supposed to look like design A or design B, where this  constraint has been applied so no vertical curvature exists where the tracks pass above the under-crossing? Anyone can see this design rule will blow up structure height, length, cubic yards of concrete, and of course cost. And yet, that's what we see in all the profile drawings.

Paint-by-Numbers Structure Depth. There are well-worn preliminary engineering rules for how thick a bridge deck needs to be relative to the loads it must support and the width of a span. Blind application of these rules during preliminary engineering, when the vertical track profile is often decided, results in bridge decks that are comically deep, as measured from soffit (bottom surface of the bridge) to top-of-rail. These massive bridges result in a much higher track profile, needlessly increasing the length, height, visual impact, and cost of grade separation projects. Bridge structural forms exist that minimize structure depth, and it is often possible to shorten spans by adding support columns.

How to build affordable grade separations

Here are some golden rules for designing affordable grade separations. These are rules that are clearly not being followed for Broadway, or for the Menlo Park plans, or for the downtown Redwood City plans, and directly contribute to stratospheric cost estimates for these projects.

• dig as little as possible. Wherever possible, go up and over.
  
• limit freight train speeds to no more than 45 mph.
• allow bridge decks and vertical curves to co-mingle.
  
• from the very beginning, aggressively minimize structure depths.

Another important consideration, in view of the large number of grade separation projects that will be required to advance the decadal process of grade separating the peninsula rail corridor, is to standardize designs. There ought to be a small set of bridge designs that can be repeatedly adapted to each situation, using standard prefabricated structural elements. Not every project needs to be a special snowflake.