The Vexed DTX Tunnel

Just a techie tidbit note today regarding the DTX tunnel that is at the heart of the engineering side of the dispute between CHSRA and TJPA on whether the current design meets CHSRA's future needs. I know we've been over this too many times already, but somehow the issue needs to be put to bed such that everyone can save face and focus on maximizing California's share of the $9.5 billion in federal dollars already on the table. Protracted tiffs don't do much for the confidence of the general public, let alone potential private investors, in either TJPA or CHSRA.

My objective for this post is to suggest a tunnel sharing and operations timetable strategy that might help bridge the gap. First, a quick recap of the saga so far.

Issues Related to Planning and Funding

TJPA has been planning the entire Transbay Terminal since SF voters approved prop H in 1999, with the objective of bringing not just Caltrain but eventually also HSR under the same roof as a large number of bus services. The old building dates to 1939, has seen better days and is anyhow no longer up to seismic code. The only part that will be re-used are the bus ramps that once allowed electric trolley cars to run across the lower deck of the Bay Bridge. The underutilized neighborhood south of the new building will also be redeveloped.

The DTX tunnel is the bit between Caltrain's existing terminus at 4th & King Street and the basement of the new terminal building. As designed, it features three tracks under Townsend and 2nd Streets and widens into a curved throat section to reach six platforms tracks accessed via an underground concourse level and three full-length wide island platforms. Of these, two will feature the level boarding height to be chosen by CHSRA and the other the one chosen by Caltrain. The outside tunnel tracks are to be shared by both services and are permanent as inbound and outbound tracks, respectively. The center track will apparently alternate between inbound and outbound traffic.

CHSRA has recently raised a red flag regarding the design, claiming it couldn't support the 12 HSR trains per hour (tph, meaning that number each way) during peak periods claimed in the most recent ridership forecast for 2030. There seems to be fairly broad agreement among the readership of this blog that this is a new and excessive demand made public only after TJPA decided to bypass CHSRA in seeking a slice of the HSR dollars in the recent stimulus bill. Quentin Kopp is concerned this would set a precedent and cause CHSRA to lose control of project cost.

TJPA claims its plans are based on a previous estimate of just 4tph and a minimum dwell time of a full hour. The latter is itself a highly questionable requirement, given that terminal stations in Europe manage to turn HSR trains around in 6-10 minutes. They are treated as through stations that happen to require trains to reverse direction, generally with a change of driver. Old-fashioned terminus stations are grand buildings providing small town folk with access to a metropolis, reached after many hours or even days on the move. Thanks to high speed, even San Diego-SF-San Diego will amount to just a single eight-hour shift for employees, so in that sense SF could indeed be considered a through station along out-and-back routes originating in SoCal and Sacramento.

The Authority would be on much firmer ground if it stuck to AB3034, which point blank requires that the entire network must support headways of no less than 5 minutes. This refers to the time interval between the nose of one train and the next one running in the same direction on the same track. The technically required minimum is a function of speed and emergency braking distance at that speed. For moderate speeds, experienced operators of modern train control infrastructure can make do with as little as 2.5 minutes, so the legal target set by the bill is conservative. As will become evident shortly, the link between short headway capability and system throughput is not always straightforward.

Technical Issues and Solution Concepts

The beef that Clem/Richard Mlynarik, BruceMcF and yours truly have with the current design of the DTX tunnel is twofold:

a) at just 500 feet, the two curves in the tunnel will prevent at least the HSR trains from running them at much more than bicycle speed. In addition, we expect screeching noises as the long wheelbase trucks needed for high speed stability are dragged kicking and screaming around the corner. Passengers on board and especially, anyone present on the platforms would hear a sound reminiscent of fingernails on a blackboard, just much louder. Special lubricants might help keep both the noise and the wear and tear down, but no brand-new design should need them. The radii are also too tight for off-the-shelf Japanese shinkansen designs, which need ~925 feet. Even for European designs, 500 is the number the marketing department made the engineers sign up to, not one they'd recommend for brand-new track.

b) the slow speeds combined with the tunnel layout combined with train lengths of 660-1320 feet mean that during peak periods, inbound and outbound trains of both operators will block each others' paths to such an extent that the "throat" between the tunnel proper and the platform tracks becomes the throughput bottleneck. It's highly unlikely that HSR could achieve 5 minute headways during Caltrain's rush hour with that design.

Both post authors and commenters have propsed a wide range of solution concepts on both this blog and Clem's. Here's a recap:

• terminating some or all HSR trains at 4th & King
  
• a central rail station at Market & 7th instead of the Transbay Terminal
  
• redefining the Transbay Terminal to include a narrow two-level heavy rail station under Mission Street, reached via a short underground pedestrian passage and featuring moving walkways along the 1/2 mile concourse level
  
• keeping the tracks in the basement, but redesigning the DTX tunnel as a one-way single-track loop through the building from east to west; any future second transbay link would then be BART down Mission Street
  
• redesigning it to run up 3rd Street to increase curve radius for just two - possibly individual track - tunnels
  
• minimally, some tweaking of the curve radii and the use of curved switches in the throat of the official three-track tunnel design
  

Dedicated Single Tunnel Tracks

In the vein of this last concept, I'd like to add one more suggestion to the list. Since HSR and Caltrain are supposed to get dedicated platforms and platform tracks anyhow, I figured why not give each of them a single, dedicated track in the DTX tunnel as well. That would at least get the operators out of each others' hair. A third track should be avoided to keep the tunnel engineering as straightforward as possible so the curve radii can be increased without breaking the bank.

The idea is really quite simple: each operator sets the signals in the tunnel track to outbound and runs a group trains out in quick succession. The timetable is arranged such that an equal number of inbound trains arrives at the mouth of the tunnel just as the last of the outbound trains clears it. The operator sets the signals in the opposite direction, the platforms tracks fill up again and the process repeats.

The advantage is that as long as traffic is guaranteed to be one-way, headways as small as three minutes are quite realistic. No outbound trains blocks any inbound one, no HSR train any Caltrain and vice versa. In other words, there is no throat in the classic sense, just two independent tunnel tracks that fan out to two resp. four platform tracks that are alternately used in one direction or the other for a well-defined period of time.

The disadvantage is that each operator's timetable would feature a number of trains in quick succession, followed by a period without service in that direction. Note that in each of these groups of trains, the fastest service classes (e.g. express) come first, followed by the slower ones (e.g. semi-express). This ensures headways do not decrease below three minutes in nominal operations. For two consecutive trains of a slower class, it may go back and forth from the minimum value to a higher one.

Spreadsheet Model

To get a better sense of the kind of timetable this would produce, I made a number of assumptions and plugged them into a spreadsheet. In addition to the three-minute headways, I assumed that the cumulative delay for an outbound group of trains would be no more than one minute. For inbound groups, I allowed a three-minute buffer and assumed a two-minute dwell time for trains making a stop between San Jose and SF. Finally, I assumed Caltrain's commuter EMU equipment would traverse the 1.3mi tunnel in two minutes (39mph average speed thanks to high acceleration and short wheelbase trucks), whereas HSR trains would take three (26mph average).

For service patterns based on half-length trains (660ft or less), the first inbound train would proceed to the end of the platform track. The second would stop before colliding with the first. However, since it's a terminal station, the later arrival would have to be the first one to leave, resulting in uneven dwell times. Also, I assumed a platform track would be cleared entirely before trains from the next inbound batch are admitted. This is how an operator would use the facility during rush hour. At off-peak periods, the ends of the platform tracks can be used for parking.

The results were as follows:

• For HSR service based entirely on single trainsets, I assumed a group size of four trains occupying one track on each island platform. The first two were express, the second two semi-express trains on the SJ-SF segment. This affects the length of time the track needs to be reserved for inbound traffic. Note that I did not consider service class impacts south of San Jose for this preliminary analysis. Instead, I assumed appropriate service groups would be created via wait states in San Jose.
  
  After allowing four trains to leave, four new inbound ones would be admitted, followed by the other four trains still in the station and four inbound ones to replace them. The period for this pattern of 8 trains each way works out to exactly 60 minutes, i.e. 8 tph. The minimum dwell times for the trains in each group worked out to 38, 44, 36 and 42 minutes, respectively. Because of the buffers built into the schedule, they could be up to four minutes longer.
  
  
• I also looked at HSR service based exclusively on full-length trains. With a group size of four, this also works out to 8 tph - but each of train would now have twice as many seats! The snag is that the minimum dwell times are down to just 11, 11, 9 and 9 minutes, respectively. As discussed above, that is considered enough for a terminal in Europe. However, if a large fraction of seats is actually occupied in SF, seat reservations and some pedestrian flow control would be highly advisable to avoid delays. Evidently, trainset utilization rates are much higher if full-length trains are used. Note that it's very easy to couple and uncouple HSR trainsets at stations, so operators could transparently switch to single trainsets during off-peak hours.
  
  
• HSR service based on full-length trains and a group size of just two yielded a throughput capacity of 7.06 tph, with minimum dwell times of 21 minutes.
  
  
• Separately, I looked at Caltrain local service based on 8-car trains that would be used during rush hour. With a train group size of four, throughput worked out to a very respectable 9.23 tph. Minimum dwell times were 6, 12, 6 and 12 minutes, respectively.
  

Fiddling with the parameters, it quickly became apparent that the minimum headway had the greatest effect on throughput, followed by the time required to traverse the DTX tunnel, dwell time at through stations down the line. Doubling the cumulative buffer period allowed for inbound Caltrain groups from 3 to 6 minutes resulted in a loss of only 1 tph in throughput.

Overnight Parking

Note that CHSRA still needs a solution for overnight parking of additional trainsets, as the first ones of the day will take some time to arrive from Merced and LA. The old Brisbane yard near Bayshore Caltrain would do nicely and could double as a transshipment center for High Speed Cargo trainsets that ride piggyback on single passenger trainsets during off-peak hours. Toward the end of the day, selected passenger trains would terminate at Millbrae/SFO.

Conclusion

Provided that inbound trains are grouped neatly and operators are willing to accept that consecutive trains in the same direction may run anywhere from 3 minutes to 13, 17 or even 21 minutes apart, aggregate throughput of over 17 tph is possible with this concept. For reference, a single loop track serving all six platforms with a minimum headway of 3 minutes would support 20 tph and provide more flexibility if CHSRA and Caltrain decide on a common platform height after all.

Cost Containment Opportunity

Just for kicks, I also looked at the possibility of saving some money in phase I by extending the dedicated single track for HSR beyond 4th & King to Bayshore. Caltrain already has four short tunnels in that stretch, CHSRA intends to bore new single track tunnels to either side of that. The CHSRA web site gives a time of 13 minutes for Transbay Terminal to Millbrae/SFO, so I figured it would take about 9 out to Bayshore. This longer single-track section caused HSR throughput to decrease from from 8 to 5.71 tph (4.14 tph for full-length trains with a group size of two but minimum dwell times of 33 minutes). For the single-trainset scenario with a group size of 4, minimum dwell times went up by 12 minutes. For full-length trains with the same group size and throughput number, they were unchanged.

In other words, it would be possible to shift some tunneling overheads south of 4th & King from phase I to phase II of the overall bullet train project, given that 4.14-5.71 tph will be enough for HSR operations for a while. Note that running northbound HSR trains east of Caltrain in that section would currently force UPRR trains to cross the HSR track. By the time HSR needs dual tracks between Bayshore and 4th & King, it's entirely possible the mighty Port of SF will no longer be served by UPRR's South City Switcher - a freight train running in streetcar mode.

Tight Squeeze At Transbay Terminal?

The San Francisco chronicle reports that CHSRA is now publicly complaining that the train box and tail tracks planned for the new Transbay Terminal Center in San Francisco (SFTT) may run out of capacity as early as 2030. For those who have followed the development of that project and its relationship with both HSR and Caltrain, this latest issue is in fact old news. To recap:

The design of the SFTT train box calls for six underground platform tracks, of which four will be reserved for HSR and two for Caltrain. Two tail tracks plus evacuation route will run down Main St. as far as the freeway supports. The tail tracks will enable offline cleaning/provisioning of HSR trains during the day and, provide additional overnight parking capacity for four trainsets. Caltrain would perform its cleaning in San Jose/Gilroy and use its existing station at 4th & King as an overflow overnight parking lot.

CHSRA is planning ahead for the possibility of one day running as many as 12 trains per hour (tph) in each direction during peak periods. That corresponds to headways of 5 minutes, i.e. de facto maximum line throughput capacity. Whether this volume of HSR traffic will materialize on peak traffic days in 2030 is of course uncertain. Indeed, it may take far longer or, never happen at all. The point is that CHSRA does not want to build a very expensive network only to have throughput capacity limited by one of its anchor stations.

Most likely, there have been behind-the-scenes discussions on this for a while, but adding platforms would be far from trivial. However, now that TJPA intends to apply for a slice of the $8 billion reserved for HSR in the recent stimulus bill, CHSRA has decided to pull the rip cord by going public.

So, does Quentin Kopp have a valid point regarding capacity limitations at SFTT?

The math suggests that he does: 12tph each way means using two platform tracks for alighting and two for boarding. Just 9 minutes would be available for each operation, with one additional minute reserved for moving away from the platform. Cleaning/provisioning time would likewise be limited to just 9 minutes plus one minute for moving the train back out. Total turnaround time would therefore have to be limited to just 30 minutes, which is very aggressive.

A single single-level HSR trainset (8 EMU cars including 2 conductor cabins) can support around 400 seats, comparable to a Boeing 747. A bi-level trainset is more akin to an Airbus A380. Full-length trains consist of two trainsets - SNCF regularly operates 16-car TGV Duplex consists offering a total of 1090 seats on the busy Paris-Lyon-Marseille route. JR's largest shinkansen trainsets, the E4 Max, feature five seats abreast to support 817 passengers each, yielding a maximum of 1634 passengers per train - fully three Airbus A380s! Ideally, operators would like at least an hour of turnaround time for such monsters, but at full line capacity the SFTT design would require cutting that in half. Is that even remotely feasible?

Theoretically it is, but only just. Here's how:

1. A train, with up to 1200 passengers on board, pulls into an empty track at SFTT. The side of the wide platform is empty. Within 9 minutes, passengers alight through two doors per car and proceed to the building's concourse level. By the time the train is empty, a separate one will emerge clean and provisioned from the tail tracks and quickly proceed to the empty departure platform.
   
   Meanwhile, the first passengers booked on the outbound train 30 minutes later are already passing through security (if required) and sitting down in a waiting area that is sectioned by car number and row groups. Each passenger already holds an e-ticket with a mandatory seat reservation (cp. TGV in France) that specifies the departure time, train car, row number and seat identifier.
   
   
2. The now empty train quickly proceeds to an empty tail track, where a cleaning/provisioning crew of 4 (6 for bi-level trainsets) per car gets cracking for 9 minutes. For a full-length, bi-level train that comes to a brigade of 96 persons. Up to two trains must be cleaned in parallel and, there are two shifts per day. Accounting for 5% break time, staff out sick/on vacation/absent, this adds up to roughly 225 workers per shift or 450 personnel on the payroll. A few specialists would always be held in reserve to deal with unusual issues such as severely soiled seats or mechanical problems in the cafe car.
   
   Such a large operation will require a staff-only section at the east end of the SFTT concourse level. A walkway would get them to the north end of the tail tracks, where they would descend to a narrow, full-length island platform. A separate walkway at the south end of the tail tracks would serve as a secondary exit. Drivers of arriving trains would use it to reach grade level, where they would walk back the 1/4 mile to SFTT (or use a folding bike or shuttle service of some type). Walking the length of the narrow platform would be possible as well. Fresh drivers would board trains via the north access to the tail tracks.
   
   Meanwhile, departing passengers are permitted to descend to the boarding platform. One side of that is empty, as an earlier train has just completed its boarding process and proceeded into the DTX tunnel. After approximate sorting upstairs (cp. Southwest Airlines), passengers are now expected to queue up at their exact row number as marked on the floor and signposted. Note that staff would only enforce this procedure whenever HSR actually does operate at 12tph and, that 10 minutes are available to execute it.
   
   
3. When the clean, empty trains pulls up, passengers rapidly board through their designated doors. For bi-level cars, a maximum of 80 passengers would have to file through each door within 9 minutes. Because of all the queuing discipline, there should be minimal traffic jams inside the car - ideally, all passengers on the entire train would arrive at their reserved seat at almost the same time. As a matter of etiquette, passengers would be asked to avoid blocking the aisle. Large suitcases, folding bicycles etc. need to be stowed, but small items such as carry-on bags etc. can wait until everyone is on board.
   

Note that this schedule contains absolutely no slack. If any part of the turnaround procedure takes extra time for any reason or, if trains fail to arrive on time, there are knock-on effects on at least several - perhaps many - subsequent trains. Such "brittle logistics" are highly undesirable, cp. the constrained pedestrian flow capacity at BART stations in downtown SF during rush hour.

Also note that doubling the number of tail tracks gives the cleaning/provisioning staff twice as much time per train. However, unless their numbers were also doubled to 900, each brigade would consist of half as many workers scrubbing just a s furiously as before - not much would be gained. Besides, decades of experience in assembly-line work have shown that workers quickly figure out how to get their tasks done as efficiently as possible.

On the other hand, additional tail tracks would increase the number of trainsets that could be parked right at the SFTT overnight. The question is if the additional expense for tunneling is worthwhile, compared to the overheads of an at-grade overflow yard for both Caltrain and HSR in Brisbane, the site of an old SP rail yard. This might double as a high speed cargo transshipment center, conveniently located for piggy-backing on single trainset non-express passenger trains during off-peak hours.

At SFTT, a far more serious bottleneck might be the very short interval available for boarding, mostly because orderly queuing is not nearly as common in American culture as it is in Japan or the UK. However, when things get crowded, it's human nature submit to sensible flow management procedures if they are properly explained on the ticket, supported by signs and helpful staff. Infrequent travelers, children traveling alone, the disabled and tourists (except tour groups) may need additional help, possibly even in foreign languages.

---

However, it's not all doom and gloom: CHSRA's scenario of 12tph is really quite extreme. First, as general ridership goes up, there will come a point at which it makes more sense operationally to switch to full-length and then to full-length bi-level trains to keep tph count from escalating. As discussed earlier, a full-length bi-level train with 16 cars and 4 seats abreast can offer ~1100 seats today. With distributed traction, the next generation AGV Duplex should support ~1200 passengers as well as top speeds of ~220mph. Alstom and SNCF are reportedly already collaborating on the design.

Taking that as a basis, consider the numbers: 2 x 12 x 1,200 = 28,800 passengers either boarding or alighting in a single hour. If the pace were maintained for a single 16-hour day of operations, there would be ultimate capacity for 460,800 passenger trips (assuming each seat is only occupied by one passenger for part or all of the route). Considering California's projected population of 50 million by 2030, that still a stupendously large number for intercity travel - even on Thanksgiving.

Side note: SNCF achieves 75% average seat capacity utilization on its TGV network and, tries to avoid letting any one line reach more than 85% to avoid customer satisfaction problems. Of course, individual trains are often sold out on peak travel dates.

Also, keep in mind that SF is not Paris and trains are not short-hop flights. Not every passenger will board at SFTT, there are other stations along the route to LA (at least San Jose Diridon). Therefore, the number of passengers alighting and boarding any given train at SFTT will generally be significantly less than the full complement of seats. Alternatively, it might make a lot of sense to terminate some northbound trains in San Jose on predictable peak traffic days. That would keep tph count at SFTT manageable, giving staff there more time to turn trains around. Some trains departing from SF would then run non-stop to the Central Valley and SoCal.

That means CHSRA's 12tph scenario and the strict turnaround procedure outlined above may not be all that realistic to begin with, even for peak traffic days in 2030. There's really no need to over-engineer the SFTT.

Nevertheless, CHSRA was right to express concerns now, rather than punt the problem to a future generation of planners and engineers. It's also worth pointing out that 12tph for HSR plus Caltrain traffic all day long would severely tax both the DTX tunnel capacity and the patience of peninsula residents whose property abuts the Caltrain right of way. Even if some cities did decide to retain some hardened grade crossings for now, they would have to close or grade separate them at a later date. If ridership on the two railroads really does reach such high levels, any above-ground solution would have to feature very effective sound and vibration damping to avoid real estate blight.

On the other hand, spending additional billions on a tunnel alignment through at least Atherton, Menlo Park and north Palo Alto up front - regardless of who pays for it - would be overkill if ridership grows much more slowly than CHSRA is projecting. Food for thought but again, there's a risk of very expensive over-engineering.

---

On the other hand: if the state's population really does keep growing as predicted and, HSR really does prove so wildly popular that SFTT introduces brittle logistics, there will be knock-on effects for transportation within the SF-Oakland region (and elsewhere). In particular, capacity and earthquake resilience considerations alone combined may prompt the construction of a second transbay tube after the HSR starter line is becomes operational in the 2018-2020 time frame (assuming no construction delays). CHSRA has already explored the option of running tracks up to Oakland, terminating in an intermodal station with West Oakland BART, to prepare for that possibility.

Unfortunately, like it or not, West Oakland isn't a significant draw for passengers and BART already occupies the available rights of way in downtown Oakland. However, a much more interesting option is emerging: the US Navy has turned over the former Alameda NAS to the city of Alameda, which has re-named it Point Alameda. Granted, the clean-up of this superfund site is ongoing. Meanwhile, most of the area has been designated a wildlife refuge for a little-known but endangered species of grey bird (the Least Tern) that has taken a shine to the old runways, because they help to hide nesting sites from predators.

For argument's sake, let's suppose the superfund cleanup can be completed in the next decade and the Least Tern colony recovers sufficiently to permit relocation to a permanent, perhaps smaller, refuge elsewhere in the Bay Area. I'm no ornithologist, but perhaps the uninhabited island just south of Richmond Harbor might be suitable. The issue would definitely require close scrutiny.

With these provisos, the city of Alameda could then consider careful, strictly transit-oriented development of the old Navy base, which is the largest patch of contiguous land available in the region. Readers may want to compare the suggestion below to what the existing community near Point Alameda is already working towards. Again, what type - if any - development happens at the point is up to the city (h/t to commenter Alameda?).

What I have in mind is primarily a large urban park / open space preserve, in addition to the already planned golf course across from Oakland harbor. Historic buildings worth preserving would be re-purposed, new construction would be limited to the southern and eastern edges of the area occupied by the old runways, plus the area just north of Seaplane Lagoon. Since virtually all of that is landfill (cp. Marina district in SF and Treasure Island), the focus would be on few high-rises separated by low-rise with green roofs. The model for the development in downtown Vancouver (BC) and its beautiful Stanley Park. On Point Alameda, trees and meadows (plus the golf course) would be irrigated with recycled water from the city of Alameda.

The catalyst for this effort could be a successful bid to host the 2028 Summer Olympics in the Bay Area.

Perhaps most importantly, motor vehicle access to the entire area occupied by the former Naval base would be restricted to permit holders. Alameda is famously averse to traffic, especially through the Webster and Posey street tubes across to Oakland. Therefore, the primary ways of getting to this new regional park and the new districts would be public transit. Local streetcars, buses, rental bikes (cp. Velib' in Paris), personal folding bicycles and bicycle rikshas (plus plain old walking) would all supplement heavy rail service.

Specifically, BART would be extended from downtown Oakland via a tube and run mostly underground out to the Point. The much longer second transbay tube would extend the SFTT tail tracks to new Caltrain/HSR stations on the Point, with an at-grade terminus/yard at Atlantic Ave. If desired, a parking lot there could be restricted to Alameda residents.

Here's a map of what the end result might look like:


View Larger Map

This concept of extending both broad and standard gauge systems into Point Alameda would give millions on both sides of the bay convenient access to a beautiful new urban park, open space preserve and golf course as well as new housing, office and commercial/nightlife facilities, including a large urban beach. Conveniently, it would also reduce the capacity constraints at SFTT, which would become a through station. Many passengers hailing from or destined for the East Bay would transfer to BART there rather than to buses or BART in SF. This would free up bus terminal capacity at SFTT for beefing up service to Marin county, connecting to SMART trains bound for Santa Rosa in San Rafael.

An optional feature would be an underground gauge change station for BART rolling stock retrofitted with variable gauge trucks. A regular standard-gauge locomotive operated by Caltrain would tow the BART train - possibly without any BART staff on board - across to SFTT and if desired, out to Millbrae/SFO on Caltrain tracks. It may sound a little funky, but variable gauge technology has been in commercial service in Europe for 40 years. FRA would have to approve the arrangement.