Why we stopped building subways cheaply
We used to dig up roads to put trains underneath
The Linear No Threshold model says that there is no safe level of radiation exposure. There is overwhelming evidence it is false, yet it inspires the ALARA principle, which makes nuclear power unaffordable worldwide. Read the lead article from Issue 19 of Works in Progress.
Tunneling is one of the many technologies that make modern civilization possible. For one, a tunnel can dramatically reduce transportation costs by shortening travel times between two points. Prior to the construction of the Holland Tunnel beneath the Hudson River in New York, for instance, the only way across was via ferry, a journey that could take hours if the ferries were backed up. Tunnels are also needed to build large-scale infrastructure projects: hydroelectric dams require tunnels to divert water around the construction site and to feed water to the turbines. And tunneling can create new, valuable land beneath dense urban areas. By going underground, we can create the space for horizontal infrastructure such as subway lines without destroying existing buildings or disrupting the urban fabric. This makes tunneling an important technology for building cities that people like living in.
Historically, most subways were built using what’s known as ‘cut and cover’ excavation: digging an open trench, building the tunnel structure within it, and then covering the trench up. Cut and cover was used for the first London subway line, in 1860; it was used for the construction of New York’s first subway, in 1900; and for nearly a century it remained the preferred method of building subway tunnels. As late as the 1970s, most subway construction in the US was done using cut and cover.
For many types of underground construction, especially in undeveloped greenfield land, cut and cover is still widely used. And cut and cover is still the primary method of constructing underground train stations. But for urban subway tunnels, cut and cover has largely been supplanted by the use of tunnel-boring machines (TBMs), which tunnel horizontally beneath the ground without disturbing the surface. In a database compiled by Britain Remade of recent transit projects around the world, there are 80 projects listed as using TBMs, compared to just one being built with cut and cover.
Some transit experts believe that this transition was a misstep. Cut and cover is a much more disruptive construction method (since it tears up the street while construction is taking place), but it’s often much cheaper than using a TBM. During construction of the Canada Line in Vancouver, between 2005 and 2009, changing from bored tunnel to cut and cover saved more than $400 million in construction costs, 16 percent of the cost of the entire project. Alon Levy of the Transit Costs Project argues that ‘cut and cover is underrated’, and it should merit more consideration when deciding how a transit project should be built:
Regrettably, people don’t seem to even recognize it as a tradeoff, in which they spend more money to avoid surface disruption – some of our sources have told us that avoiding top-down cut and cover is an unalloyed good, a kind of modernity. Even more regrettably, this same thinking is common in much of the developing world, where subways tend to be bored.
But cut and cover has always been an unpopular method of construction, opposed by urban residents since the very first time it was used. Cut and cover may often be cheaper in terms of dollars, but as tolerance for the disruptive effects of construction has decreased, the political costs of using cut and cover have risen. And as other tunneling technology has improved, the relative advantage of using cut and cover has decreased.
A brief overview of tunneling technology
Understanding why tunnel-boring machines replaced cut and cover requires knowing a little bit about how tunnel construction works.
With cut and cover construction, the basic method – digging a trench and then covering the trench up – is simple. But there are a variety of ways that this can be done, depending on the specifics of the tunnel and where it’s being built. The most straightforward method is to dig a trench with gently sloping sides that require no additional support. Once you’ve dug down deep enough, you build your structure, and cover everything back up again.
Because the sides of the trench slope outward, this method occupies a lot of horizontal space. And the deeper the excavation, the more space is required. This can make it a challenge to use in urban areas, where that space is occupied by buildings and other infrastructure. To avoid this, cut and cover construction will instead often excavate straight downward, using support structures to prevent the walls from caving in.
These supports can be built in a variety of different ways. One common method is to use piles, large posts that are driven deep into the earth. Piles are typically made from either steel or concrete, and can either be spaced close enough together that they form a continuous wall (such as with secant piles or sheet piles), or spaced farther apart with infill structure between them, such as timber lagging (wood boards that span between piles) or shotcrete (sprayed concrete).
Cut and cover also uses different methods for building the tunnel structure itself. In the conventional method, known as bottom-up, the trench is fully excavated and the tunnel structure is built up starting from the bottom. With the top-down method, by contrast, the tunnel is excavated only partway down, and then the roof of the tunnel is built using the existing soil as a vertical support. Once the roof is in place, the rest of the tunnel is then excavated below it. With top-down construction, the surface can be completely restored after the roof has been built; with bottom-up, the top of the excavation will often be covered with temporary decking to allow use of the surface while tunnel construction is taking place.
With a tunnel-boring machine, the basic method is different. Instead of digging downward, TBMs use large rotating cutting heads to excavate horizontally through the ground. Behind the rotating cutting head will be conveyors for carrying away excavated material (known as muck), hydraulic jacks for pushing the machine forward, and machines for installing the tunnel lining. A modern TBM is very much like a mobile factory that pushes its way through the earth and leaves a completely constructed tunnel behind it.
Like with cut and cover, TBMs comprise a variety of specific excavation technologies that vary depending on the project. At a high level, TBMs are categorized by whether they’re designed to tunnel through soil and soft ground or through rock (though today there are increasingly crossover machines that can do both).
Soft-ground TBMs evolved from unmechanized tunnel shields, large hollow structures that supported the sides of the tunnel while it was being excavated. The tunnel shield was invented by Marc Brunel (father of Isambard Kingdom Brunel) in 1806 for tunneling under the Neva River in Russia, and was first used to tunnel under the Thames in 1825.
Brunel’s shield consisted of a 21-foot-tall grid of iron frames, divided into 12 separate frames, each one consisting of three compartments stacked on top of one another. Within each compartment, the face of the tunnel would be supported by a series of boards called poling boards. A worker would remove a single board, dig away the soil behind it to a depth of around nine inches, and then replace the board and move on to the next one. After all boards had been dug out, the frame would advance forward with large mechanical jacks, and the process would repeat. Behind the shield, brick lining would be installed around the sides of the tunnel to form its structure. With Brunel’s shield, tunneling under the Thames proceeded at about eight feet per week on average.
Brunel’s shield was rectangular in shape, but most subsequent tunnel shields were circular. Early shields used workers with picks and shovels to do the actual excavation, but over time mechanical excavation equipment was added. In the early 1900s John Price developed a tunnel shield that had a large, rotating disc mounted to the front. Bucket-shaped cutters attached to the front of the disc would scrape away soil as it rotated and feed it into a conveyor for removal. Price’s mechanized shields were an immediate success, and over the next several decades were used to dig subways around the world, and are the ancestor of modern soft-ground TBMs.
The tunnel shield prevented the sides of the tunnel from collapsing while it was bored, but they still required some method to prevent the face of the tunnel from collapsing, and to prevent water from intruding when tunneling below the water table. By the late nineteenth century, the standard method was to use compressed air. By pressurizing the tunnel to several times atmospheric pressure, water would be kept out. Compressed air remained in use well into the twentieth century, and is still sometimes used today, but it has been largely supplanted by slurry machines and earth pressure balance machines, which respectively use a bentonite slurry and the excavated material itself to support the face of the tunnel. Today, earth pressure balance machines are the most common type of TBM for tunneling through soil.
Rock TBMs evolved separately from soil TBMs. In soil, the task of excavation was comparatively simple, and the primary challenge was finding a way to prevent the tunnel from collapsing while it was being dug. In rock, the tunnel could often support itself while it was being dug, and the primary difficulty was building a machine robust enough to carve through rock. This second task proved much more difficult, and successful rock-tunneling machines were developed much later than soil-tunneling machines.
Attempts to build rock-tunneling machines date back to the 1850s, but the first successes appeared in the 1950s, when James Robbins developed the disc cutter for the Oahe Dam project. Prior to this, most attempts at mechanical rock-tunneling machines used drag picks, sharp steel tools that scraped away bits of rock as the cutting head rotated. Robbins’s disc cutter, on the other hand, rolled freely over the surface of the rock like a wheel. As the tunneling machine pressed the disc cutter against the face of the rock, the rock cracked and flaked off.
Robbins’s disc cutter greatly increased how fast a rock-tunneling machine could tunnel. And disc cutters lasted much longer before needing to be replaced than drag picks, meaning the machines spent more time tunneling and less time down for maintenance. As a result, Robbins’s machine made it economical to mechanically tunnel through the rock for the first time. The Robbins Company remains a builder of all types of TBMs today, and the disc cutter continues to be the standard method for excavation on rock TBMs.
There are also other ways to bore a tunnel besides using a TBM. A roadheader uses a small, rotating cutter mounted to a boom arm that gets moved back and forth over the tunnel face (this is in contrast to a TBM, which excavates the entire face of the tunnel at once).
Drill and blast involves drilling several holes in the face of the tunnel (typically using a mechanical drilling machine known as a drilling jumbo) and setting off explosives in them. Drill and blast was the primary method of excavating rock tunnels prior to the invention of rock TBMs, and is still widely used today.
The sequential excavation method (SEM), also known as the New Austrian tunneling method, excavates a tunnel in small ‘bites’ using mechanical excavators and other equipment, and supports the sides of the tunnel using shotcrete.
Changing technology, changing economics
For most of history, cut and cover was the cheapest way to build an urban tunnel, and boring was only done if cut and cover wasn’t an option. In the construction of New York’s first subway, in 1900, for instance, cut and cover was estimated to be just an eighth the cost of boring a tunnel, but it could only be used on about half the total length of the line.1 Because the ground of New York varies in elevation substantially, keeping the tracks straight required tunnels bored through rock, which were built using drill and blast.
But as tunneling machine technology continued to advance, this calculus changed. Brunel’s non-mechanized shield tunneled under the Thames at the glacial pace of eight feet per week. By the early 1900s, Price mechanized shields were achieving excavation rates of nearly 200 feet per week. And by the 1970s, TBMs were achieving rates of 1,400 feet per week in soft ground, and 1,900 feet per week in rock.
As TBMs got faster, they also got cheaper, and became increasingly competitive with cut and cover. When a TBM was used to bore some of the tunnels on the Bay Area Rapid Transit (BART) project in the 1960s, its costs were just 40 percent higher on average than the cut and cover sections, a far cry from the eight-times cost difference on the New York Subway. A 1994 study of French subway construction costs on over 90 miles of underground tunnel found that only in the most difficult underground conditions was tunnel boring more expensive on average than cut and cover.
Depending on the nature of the project and how disruptive surface construction would be, TBMs in some cases began to be cheaper than cut and cover. A 1980 environmental analysis for a rapid transit system for Los Angeles estimated that cut and cover construction would be more expensive than bored tunnel, due to needing to use eminent domain to buy and destroy homes along the roads. And when Seattle planned a tunnel to replace the Alaskan Way Viaduct in the early 2000s, the costs of a bored tunnel were projected to be comparable to cut and cover, but the disruptions to the city caused by cut and cover were projected to cost several additional billion dollars.2
Similarly, TBMs have high fixed costs (in the form of the time, effort, and expense to buy the machine and get it set up) but low operational costs: once they are up and running, the marginal cost of additional excavation is low. TBMs are thus often particularly economical on large tunneling projects where the fixed costs of the machine can be thinly spread. When Madrid built 60 miles of underground tunnel when constructing its metro in the late 1990s and early 2000s, it achieved a famously low cost of €42 million per kilometer (about $73 million per kilometer in 2023 dollars) using TBMs. And the recent extension of the L11 line in Madrid, which adds another 4.3 miles to the metro system, likewise found that excavation with TBMs would be cheaper than cut and cover.
Underground construction is high variance, and the costs of construction can vary greatly depending on the nature of the project and the conditions of the ground.3 The best construction technology for a given project will depend on the specifics of that project. As Alon Levy notes, cut and cover is still a useful arrow to have in a tunneler’s quiver, as per the $400 million savings it achieved on the Canada Line. Given the comparative labor intensity of cut and cover (TBMs are highly automated, and can operate with a very small number of workers), it is likely especially appropriate for countries in Asia and Africa with low wages. But as TBM technology has advanced, it’s become more and more attractive for urban tunneling.
Cut and cover gets harder
While tunnel-boring technology has gotten better and better, cut and cover has steadily gotten more difficult. The chief issue is the fact that cut and cover creates an enormous amount of disruption on the surface while excavation is taking place. The construction creates dirt, noise, and flooding, and can damage nearby properties as the ground is dug up; this has resulted in numerous lawsuits against transit authorities (tunnel builders will generally include a contingency to pay for buildings damaged as a result of construction for this reason).
Most importantly, in cut and cover construction the street gets torn up and portions of it become unusable, sometimes for years. Access to businesses is blocked, retail sales fall, and people complain. Disruption of traffic has been called ‘the plague’ of cut and cover construction:
Noise and dust receive their share of complaints, but these can be controlled to some extent to minimize nuisance. It is the day-to-day rerouted obstacle course of construction equipment, barricades, flagmen, and rattling deck beams that create an impression of confusion and personal affront to the daily commuter or casual visitor.
Traffic disruption from cut and cover is especially egregious because the nature of subway construction projects means that construction is likely to take place in the most heavily congested areas of the city, making the problem worse until construction is completed.
In The Great Society Subway, Zachary Schrag talks about the disruptions caused by cut and cover during the construction of the Washington, DC, Metro, stating that ‘cut and cover meant pain’:
Virginia Ali, whose Chili Bowl restaurant had served U Street since 1958, had endured riots and illicit drug markets, but subway construction was worse. With U Street itself blocked off, customers had to find their way through alleys. If construction workers hit a gas line, diners would have to evacuate, and frequently Ali found her restaurant’s floor inches deep in dirty water that ran off the wooden blanks that served as U Street’s decking. A block-long stretch of 7th Street turned into a twenty-foot deep garbage pit; residents fretted that children might climb through the shoddy fences and fall in. By the eve of completion, a neighborhood resident mourned ‘after five years of construction, the name of the game right now is survival’.
Cut and cover can also cause subsidence in the ground surrounding the excavation site, damaging surrounding buildings. TBMs will sometimes be chosen as the excavation method even when they’re more expensive purely to reduce this risk.
Because of the disruptions it causes, cut and cover transit construction has been unpopular since its inception. London’s first two underground railways were built using cut and cover, but public objections to the disruptions it caused, plus the fact that most of the remaining streets were too small to practically use, forced subsequent lines to be built using bored tunnel.
In the construction of New York’s first subway, cut and cover was described as ‘making life miserable for a few years’. Boston built its first subway (and the first subway in the US) using cut and cover, but when designing an extension to the Red Line in 1977 it opted for less disruptive tunnel boring in many locations, such as the segment between Harvard Square and Porter Square:
The deep bore tunneling method was chosen over the cut and cover method because it will lessen the impact during construction on the surrounding neighborhoods. Specifically, the deep bore method negates the need to tear up Massachusetts Avenue, which runs along most of the length of this section and which, if narrowed to half its width, would cause severe traffic congestion problems. Additionally, shops along this section of Massachusetts Avenue, which number between 40 and 50, would incur substantial economic losses due to a temporary loss of customer parking spaces, advertising exposure, and customer accessibility.
The disruptions caused by cut and cover often make using it difficult, even when it’s cheaper than other methods. In their 1981 textbook on tunneling, TM Megaw and JV Bartlett note that ‘The negotiations with those affected by a new [subway] line are likely to be far more difficult for cut and cover construction than for a deep tunnel. Objections sometimes are given disproportionate publicity; the promoters have to justify their proposals in much greater detail’.
During the construction of Atlanta’s subway, MARTA, in the 1970s, the planned use of cut and cover in the downtown area caused a major backlash, and caused the planners to switch to bored tunnel. When planning the BART extension to reach San Francisco Airport in 1995, transit advocates argued that BART’s preferred alternative required tunneling that would cost $135 million more than cut and cover, but if cut and cover was used the cities of South San Francisco and San Bruno would oppose the plan because of the construction impacts. When a cut and cover tunnel was proposed for the Alaskan Way Viaduct replacement in Seattle, it was rejected by Seattle voters.
The ability for citizens and advocacy groups to oppose disruptive cut and cover projects has likely been strengthened by environmental laws, such as the National Environmental Policy Act (NEPA), that require government agencies to investigate and disclose the environmental impacts of their projects. These laws first began to appear in the late 1960s, and provide an avenue for affected parties to oppose projects by arguing that agencies haven’t followed the proper administrative procedures. The previously mentioned Canada Line project, for instance, was sued by a group of citizens who objected to the use of cut and cover under Canada’s Environmental Assessment Act. They argued that the proper procedure had not been followed when disclosing the impacts of it. The appeals court noted that citizens objecting to public projects by ‘challenging not the substantive decision of government approving the project, but the process by which the government decision-makers informed the public’, had become common.
The nature of cut and cover construction means that we might expect it to get more expensive over time. A major expense of cut and cover construction is having to relocate below-ground utilities, and the more that have to be relocated, the greater the cost. As cities age, and accumulate more buried services, this will naturally make relocating them more expensive compared to simply tunneling beneath them.
With so much opposition to cut and cover, and the increasingly competitive costs of tunnel boring, it’s not surprising that developed countries have largely switched to the latter in built-up areas. And in fact, adopting a technology, then abandoning it later when its downsides are deemed to be too high, even if its replacement is more expensive, is a common arc of technological progression: until the 1940s, hydroelectric dams made up nearly a third of electricity generated in the US, but hydro plants stopped being built in the US in the 1970s, due largely to their disruption of river ecosystems and other negative environmental effects; aluminum wiring is cheaper than copper (which is why its used for long-distance transmission lines), but the risks of fire due to improper installation in homes caused it to be phased out in the 1970s; and, of course, there’s currently an enormous effort dedicated toward replacing carbon-emitting energy sources with more more environmentally friendly low-carbon ones.
In fact, the existence of subways at all is arguably a result of this sort of progression. Prior to the subway-construction era in the US in the early twentieth century, many cities had rapid transit systems based on elevated trains. New York built the first elevated train line in the world in 1868, and by 1890 its elevated trains gave New York the largest mass transit system in the world, ten years before it began construction on its subway. By 1900, elevated urban trains had been built in Kansas City, Sioux City, Chicago, and Boston.
But, not unlike cut and cover construction, building an elevated train came at the cost of incredible disruption to the surrounding area (and unlike an elevated train, that disruption remained after construction was complete). The structures reduced sunlight to the street below, and steam-powered elevateds on steel frames were incredibly noisy for surroundings, lowering the quality of life and property values in their immediate vicinity. Urban elevated trains were thus frequently opposed by residents.
When Los Angeles was considering a mass transit system in the 1920s, it was opposed by citizen groups such as the Taxpayers’ Anti-Elevated League, and an LA reporter who researched elevated trains in other US cities came away with the conclusion that ‘an elevated is a many-legged and roaring steel serpent and should be shunned by all cities for the machination of the devil that it is’. Cities began to prefer subways to elevateds, despite the fact that they were two to four times as expensive to build as elevated trains. Since 1908, the US has only built one new urban elevated railway, in Miami, a city where the high water table makes underground construction difficult.
The US is partly an exception: dozens of elevated railways have been built across the developing and middle-income worlds in the past few decades. A handful have been built in developed countries, including the Docklands Light Railway in London, in the late 1980s, and the Yurikamome which connects Tokyo with Odaiba artificial island, in 1995. But both of these were built into virgin terrain as part of redevelopments, and residents came along later with their impacts already ‘priced in’. Both are also automated and electric, and run on concrete supports, making them much quieter than early 1900s elevateds (though the DLR can screech when it turns sharp corners).
Cut and cover will sometimes be the best choice from a basic cost-benefit perspective, as will elevated rail, and both are useful to have in the tunnel construction toolbox for that reason. But technological deployment decisions are rarely governed purely by economics. There’s always a broader calculus at work with any technology: a set of shifting norms, assumptions, culture, and institutions that governs what methods are considered acceptable for solving problems. Political costs and civic opposition will often make a technology unviable regardless of how the dollars and cents add up. Perhaps a policy will overcome this problem – if not, then cut and cover seems to be a casualty of this broader calculus, like so many technologies that have come before.
This article originally appeared in Issue 14 of Works in Progress. Ben Southwood, one of our editors, wrote a response to this piece on his Substack.
Brian Potter is the author of the newsletter Construction Physics.
722 Miles: The Building of the Subways and How They Transformed New York, Clifton Hood 2004, p. 86
Tracking down the costs of tunnel boring versus cut and cover for this project is difficult, but they appear to be similar. The initial, six-lane cut and cover tunnel was projected to cost $4.6 billion. That was later cut down to a four-lane tunnel that would cost $3.4 billion. The ultimate budget of the bored tunnel, which used what at the time was the largest tunnel-boring machine in the world, was $3.3 billion. Environmental impact documents from the project also seem to show bored tunnel as having been cheaper than cut and cover, though making a direct comparison here is difficult since the projects have somewhat different scope.
The previously mentioned French study found that in the most difficult ground conditions, tunneling costs per mile varied from ~250 million francs per mile to ~1,200 million francs per mile.
 | A guest post by
|
Great overview of construction methods and associated direct and indirect costs. Only very few companies make modern TBMs and usually they are not moved from one city to another but made for a specific job and partly assembled on site. This adds cost. The article doesn’t speak about the cut and cover hole needed to launch or retrieve the TBM. It also doesn’t mention how Musk’s Boring company’s TBM achieves a higher rate of speed than conventional TBM’s, albeit so far only for smaller tunnel sections not useable for subway, LRT or trains.
Brian, that was a great post. Both the history and the social effectors that drive design were nuggets of information that certainly backfilled my understanding of the transit and tunneling.
When the MASS dig project opened, you could feel the collective sigh for those driving in the area. Thank you.