BART is a heavy rail transit line in the San Francisco Bay Area. Proposed in 1946, it opened in 1972 and seemed futuristic at the time. A much-delayed effort is planned to finally bring it to San Jose , which is the largest city in the Bay Area. It currently reaches the northern reaches of the town, but the extension would add about 6 miles of track, 5 underground, with 4 stations. The current estimate for this project is an astonishing $12.7B. Residents of Santa Clara County, which contains San Jose and Silicon Valley, is one of the richest counties in the world, have been paying a 0.5% extra sales tax on all their purchases for many years to fund this.

BART seems modern, but its core design is the same as the first commuter rail subways from the 1800s. At the dawn of 21st century transportation, it seems that modern alternatives could provide far better service to riders for a tiny fraction of the cost, even something closer to $1B, and much sooner as well.

Tunnels are to begin digs in 2027, but full completion is set for 2037, and based on the history of such projects, may be much later. With only 4 stations, only a small fraction of the valley will get to ride it. While envisioned as a circular line around the bay, BART has always been mostly in SF and the East Bay.

This is particularly important as BART is in crisis due to ridership decline, which never recovered from Covid, in part due to the adoption of work-from-home. The BART board has threatened it may need to close 15 stations and 2 branch lines, among other things, unless it gets a billion dollar annual injection it is submitting for referendum. The San Jose extension budget is managed by the Santa Clara Valley transit agency, known as the VTA.

BART is About as Expensive as it Gets

As transit lines go, BART’s very expensive, as the $12.7B cost of 6 miles should tell you. BART is non-standard wide rail gauge with 3rd-rail power. The non-standard width means BART needs all custom gear; the gauge is only used on BART and in India. 3rd-rail power requires all track be fully isolated from public spaces, so no at-grade crossings are allowed and lines must be isolated, usually underground or on heavy elevated viaducts. Like almost all rail it is a line (rather than a network) and stations are on-line (trains block the line when stopped) increasing headway. Unlike buses and cars, vehicles can’t pass one another except in limited ways.

Sadly, BART is an energy and money hog, using 2.8 times as much electricity per passenger per mile than an electric sedan like a Tesla. It costs $1.42 per passenger-mile just to operate and maintain (not including vehicles or infrastructure) compared to $0.11 for the electric sedan. (Values of 770M passenger-miles, 370GWh and $1.1B from BART web site .)

21st Century Transit Options

Last week I described 21st century transit called “Vansit” which operates on shared infrastructure (ie. roads.) There are also 21st century designs that use private right-of-way the way classic transit does. This is much less cost effective, but can have some virtues.

One slightly 21st century project, from a company called Glydways, is already partly underway to actually serve part of the BART extensions role, with service between SJC airport and San Jose’s Diridon/Caltrain station, west of downtown. BART’s 4 stops include that train station, another one in downtown Santa Clara on the back-side of the airport, as well as downtown San Jose and a less developed area where new development is hoped for after the transit comes.

While 1800s transit features large, heavy trains with on-line stations, 21st century transit uses much smaller vehicles, which turn out (against the intuition of many) to be more efficient as well as providing better and more frequent service. Journeys are station to station, with few or no stops at intermediate stations because vehicles can pass one another. Indeed, stations do not need to be that close to the “main line” because vehicles divert from the line to make their one stop, with ample time to exchange all or almost all passengers.

Vehicles are summoned and used like elevators in a 21st century building--there are no floor buttons inside these elevators. Passengers punch in their floor and are told “Take Car B” and everybody in car B is going to the same floor, or small group of nearby floors. It’s much faster and more efficient. This simple software change revolutionized tall buildings.

The lines of 21st century transport are not lines, they branch and loop as needed. The stations need not be right on the lines, and the lines can thus go wherever it’s most cost-effective to put them. The VTA BART plan requires the line go down one of San Jose’s main streets, which means a very expensive tunnel. The retailers on this street rebelled against the cheapest approach, known as “dig-and-cover.” With this method, you close and dig up the street, making a giant trench. You then build a roof over the trench and rebuild roadbed on top. The retailers would not accept long closure of their street for a subway almost none of them would benefit from. The latest proposal is for the most expensive approach, a giant single bore tunnel with 2 tracks. A massive machine has to dig deep underground, and the cost can get as high a $2B/mile when done.

With a deep tunnel, that also means deep stations. Riders are on the surface, they would rather board up there, but they must now descend through what is usually a very expensive station. It’s not at all uncommon for stations to cost more than the tunnels! San Francisco just spent $2B on a 2 mile, 4 station subway, most of that on stations.

For 21st century transit, you can dig tunnels, though these are much cheaper because the vehicles are smaller so the tunnels can be smaller. But the vehicles are also much, much lighter, so you can use low-cost lightweight elevated viaducts in areas which will tolerate those. And of course, you can use bare land if you can get it, with minimal construction cost--it’s just standard pavement, and at-grade crossings are tolerable with some loss of capacity.

The viaducts can go readily over freeways and most existing rail lines, absent tunnels and narrow underpasses. Because the vehicles tend to be much shorter, there’s a lot more places they can go, and if short tunnels are needed, they are small and lower cost. In many cases, existing highway medians can install guideways at surface level with no extra cost.

Cost estimates for low-weight elevated viaduct are extremely low, down to $10M/lane mile, but even if you boost that to $50M/mile it’s still vastly below BART’s cost. The cost is low because the land is already owned, the construction disruption is less, and the smaller vehicles (around 20 pax), at under 10,000lbs, require far less expensive track. Unlike road and rail overpasses which must expect immense loads to bear, no heavy vehicles are to be allowed on these tracks. The route will be longer than a direct tunnel but faster because trips are non-stop.

Stations are even cheaper. Stations can be little more than parking lots with some passenger boarding areas painted in (or if you want to get fancy, made of concrete with some safety walls.) Anywhere you can grab some parking space you can have a station, including old surface or below-ground parking lots. And while non-stop travel is great, nothing stops a 21st century mini-bus from driving a few blocks on city streets to where the passengers actually prefer to have their station. Yes, there can be traffic, but it’s only for a few blocks until the van reaches the guideway. Managed lanes can also improve the traffic. Instead of hundreds of millions per station, one can cost millions. UI and fare collection is all digital on people’s phones or with a small kiosk.

A system would need to interface with BART. Passengers might arrive at the North San Jose BART station. There, ideally opposite their train on the platform, would be a series of vans, already waiting. They would know from their ticket (where they specified their destination) which van to get into, and what seat was reserved for them. If the load is particularly high, other platforms (parking lots) would exist for overflow, and as vans left, more would arrive. As there are only 4 or 5 stations, you’ll get into a van going non-stop to that destination. I say 5 stations because for those who have SJC airport at their destination, their station will now be directly in their terminal--the BART plan is a station on the far side of the airport, with transfer to a 10 minute shuttle bus which stops at both terminals, in time.

All those waiting vans of course brought passengers heading north on original BART, who wait for the arriving passengers to exit, and board it for their trip. A transfer, but a quick one that saved them stops. BART trains on this line arrive only every 10 minutes daytime.

The Glydways system uses 4 person cars, so its capacity is lower. It also is not able to drive on city streets, but today self-driving capability can be put into these sorts of vehicles. For now Glydways will serve only the rail station--frankly not a very popular destination for airport passengers--though one can get to the downtown and other locations on the light rail. (Another light rail also runs “near” the airport, with a free shuttle that is also very rarely used.)

With a 20 person “mini-bus” size and a high load factor (smaller vehicles with better service get higher load factor, and they simply don’t run when load factors are too low, yielding top efficiency) you can have 3,000 vehicles/hour on a line, which is over 40,000 seated passengers/hour, surpassing most heavy rail systems in standing-room-only mode. These vehicles whiz past stations, taking an off-ramp only when depositing their passengers.

You can also run all night, with fewer and even smaller vehicles. BART doesn’t run at night. Famously in San Francisco all the BART riders leave parties by 11pm to make sure they can get back home.

As noted though, BART’s in trouble and so is Caltrain. Sadly, this new system is much less useful with no BART to connect to. Unfortunately, most of BART is built in a way as to make off-line stations very difficult. New stations would need to be built, or at least new off-ramps since there’s no place for vehicles to stop off the line. However, since “stations” can be just a parking lot, if room for a ramp can be found, conversion of BART, or Caltrain to 21st century mode isn’t out of the question.

It’s also likely that if this extension could be build for a fraction of the $12.6B budget, it could easily save BART and Caltrain, though the money is meant for Santa Clara Valley improvements. The most popular journey among these stations without BART, namely downtown to the airport, is almost provided by Glydways if that should be completed.

The great thing about light vehicle guideways is you have lots of options on where to put them. Here’s a proposed map mostly using freeway right-of-way, which in some cases has room on the ground, and in other cases may need viaduct. Viaducts can’t go under underpasses easily, which can be remedied with a variety of tricks. Some creekbed is also used, that needs protection in flooding. Sadly, the Guadalupe River, which Glydways has planned for their track, has become a homeless camp that the city wants rid of, but that makes it amenable for construction that would never fly in other locations.

The yellow section, under 8 miles long, replaces BART, with better airport service and a better downtown location but no station in Little Portugal. The Little Portugal station does not serve any present need, it’s low density residential and industrial. VTA hopes that “transit oriented development” will arise. You could pick just about any similar place for such a station, and a spur line to allow a station there is also drawn, it doesn’t add a lot of cost-the freeway there has a wide median, though it has a lot of overpasses.

Tracks could also be placed over existing railway tracks with suitable contracts (if desired, overhead electric capacity could be added to those tracks. In many cases, elevated viaducts are unpopular in residential areas, though modern LCD windows could protect privacy and noise would be minimal. They are much more accepted on busy and wide city streets, or active railways or freeways. And there’s always tunnel, but much cheaper tunnel than for BART.

There are shorter paths to the airport and downtown but they don’t re-use track. It’s a money-vs-speed question.

In green are shown some optional tracks. Santana Row/Valley Fair is probably the top consumer destination outside downtown. And if the Santa Clara train station at the back of the airport needs a station, it’s doable. You can put a station almost anywhere it makes sense, particularly if you are ready to drive a few blocks on the surface.

Probably not. It’s too radical, and to be fair, too untested, for fans of 1800s transit to accept and consider. Some other systems probably need to be built elsewhere first. On the other hand, if BART does indeed have to scale back, it probably means that the San Jose might delay such a large expenditure to connect to a reduced system, giving time for a superior alternative.

In addition, this design still retains many features of classic mass transit, in particular dependence on private right-of-way, though it is specialized for lightweight minibus sized vehicles. It may make more sense in the future to use shared, managed right-of-way, where the transit vehicles have priority but share the extra capacity of the ROW with well-behaved vehicles such as robotaxis and private robovans. Current transit designs often run with headways of 5, 10 or even 15 minutes, leaving most of their capacity unused, both because they have on-line stations, and because they want to absolutely guarantee schedules which in the past could not be done with mixed traffic.

Private right-of-way is a very expensive approach (as the current BART cost shows.) Shared ROW is much cheaper but today faces congestion. Until congestion is mitigated with traffic management, like congestion charging, private ROW can be a costly solution. A good answer is a mixed system--use as much shared infrastructure as possible, but build private ROW in special areas where it is needed. This allows serving an extensive stop map with much less cost.