Auckland Council and the Government have recently signed an official agreement to jointly fund the City Rail Link (CRL) – a move that both had previously committed to in principle, but not on paper. This is good news for the city, as it gives us certainty about how CRL will progress. (It is also a fine example of the value of good analysis and patient persuasion – this government was initially very skeptical of the project but has gradually changed its tune.)

Given Auckland’s constrained geography and lack of future transport corridors, CRL probably won’t be the last major tunnelling project we investigate. If we want additional transport corridors, we’re going to have to reclaim land, build bridges, or dig tunnels.

Auckland development Corridors
Auckland: pinchpoints and natural corridors of demand

So it’s worth asking: Are tunnelling costs reasonable in Auckland? Could they be reduced? These are important questions. The cheaper tunnelling is, the more transport corridors we’ll be able to buy in the future.

To help answer this question, Alon Levy at Pedestrian Observations compiled a nice dataset of construction costs for rail tunnels (part 1, part 2). He gathered data for over 40 rail tunnel projects that have been completed or planned over the last decade or so. While the data is a bit imprecise – based on a mix of ex-ante cost estimates and ex-post contract costs, and converted between currencies using purchasing power parity exchange rates – it provides a useful basis for benchmarking CRL costs.

According to the full business case released in July 2016, the midpoint estimate for the cost to construct and commission the CRL is $2.5 billion, once adjusted up for expected future inflation. Here’s how the cost profile is expected to go:

CRL Business Case - CAPEX by year

Converting this back to 2010 US dollars is not an exact science, because we’ve got to adjust for recent and future inflation and purchasing power parities between NZ and the US, but as a rough estimate I would say that the CRL cost is equivalent to around $1.4bn in 2010 US dollars. As CRL is 3.4km long, this equates to costs of around US$410 million per km.

Here’s a chart showing how CRL costs compare to the costs for 42 other urban rail projects. (Note that a number of the projects on the lower-cost end of the scale had significant above-ground portions that tend to be cheaper to build.)


All in all, the CRL is ranked 11th on cost – it’s on par with the costs of the Amsterdam North-South Line or Budapest Metro Line 4. It’s nowhere near as expensive as recent underground rail projects in New York or London, which tend to cost more than $1 billion per kilometre. But nor is it as cheap as metro extensions in Spain or South Korea, which cost more like $100 million per kilometre.

For reference, here’s a subset of the data for 19 projects, including the City Rail Link. This shows a few important facts:

  • First, project costs vary more between countries than within countries – all of the projects in the US are ludicrously expensive, all of the projects in Japan are mid-pack, and all of the projects in Spain, Italy, and South Korea are cheap.
  • Second, there is basically no correlation between project scale and per-kilometre costs. It isn’t necessarily cheaper (or more expensive) per kilometre to build longer tunnels. However, there are likely to be other economies and diseconomies of scale that are harder to observe, such as crowding-out when trying to complete too many projects at the same time.
RankProjectCityTunnel lengthApproximate cost per kilometre (million 2010 USD)
1East Side AccessNew York (US)2km$4,000
2Second Avenue Subway Phase 1New York (US)3km$1,700
4CrossrailLondon (GB)22km$1,000
7Central SubwaySan Francisco (US)2.7km$500
8Singapore Downtown MRT LineSingapore (SG)42km$490
10Amsterdam North-South LineAmsterdam (NL)9.5km$410
11City Rail LinkAuckland (NZ)3.4km$410
12Budapest Metro Line 4Budapest (HU)7.4km$360
13Toei Oedo LineTokyo (JP)40.7km$350
14Nanakuma Line ExtensionFukuoka (JP)1.4km$320
20Paris Metro Line 14Paris (FR)9km$230
26Copenhagen Circle LineCopenhagen (DK)15.5km$170
27Barcelona L9/10Barcelona (ES)47.8km$170
33Naples Metro Line 6Naples (IT)5km$130
34Milan Metro Line 5Milan (IT)5.6km$110
39Seoul Sin-Bundang LineSeoul (KR)18km$90
41Helsinki WestmetroHelsinki (FI)13.5km$70
42Seoul Subway Line 9Seoul (KR)27km$40
43Barcelona Sants-La Sagrera TunnelBarcelona (ES)5.8km$40

Lastly, we should be asking: What can we do to be more like South Korea or Spain when it comes to tunnelling costs? Some of the differences between locations are likely to be impossible to change, as they depend upon geography. But others are possible to change, as they relate to construction methods, design standards, and processes.

A few years ago, Alan Davies (Crikey) identified a few of these factors:

Toronto transit advocate Steve Munro… analysed a report by transit agency Metrolinx comparing the cost of tunnelling for Toronto’s new 6.4 km Sheppard Subway with that for the new 40.5 km MetroSur line in Madrid. Madrid is a popular benchmark because it has literally built hundreds of kilometres of new heavy/light rail lines over the last 40 years.

After adjusting for differences in how land acquisition is costed, he says the respective costs of Sheppard and MetroSur were $142.5 million per km and $87.1 million per km. Both lines opened at the same time. The key differences Mr Munro identifies are:

  • No environmental assessment was conducted on MetroSur
  • The standards for fire safety are more stringent on Sheppard
  • Stations are 50% longer on Sheppard
  • Non-tunnel construction was undertaken 5×12 on Sheppard, compared to 7×24 on MetroSur
  • Sheppard was built with dual tunnels, MetroSur with a single tunnel. Mr Munro says “the trains in Madrid are smaller and require a smaller combined tunnel than would be the case in Toronto. Single tunnels eliminate the need for cut-and-cover box structures at crossovers and effectively reduce the scope of excavation at stations where these crossovers are located”.
  • The use of cut-and-cover tunnelling on Sheppard was confined to stations, but it was used for 30% of MetroSur
  • It was more expensive to tunnel through the glacial rocks and streams of Toronto than the compacted sand of Madrid
  • MetroSur provided greater economies of scale as it was one of a number of projects. “Construction activities simply moved from one project to another rather than being reconstituted for each expansion, and more of the design was done during construction.”
  • Sheppard has two large interchange stations over its 6.4 km, whereas MetroSur has five interchanges over its 40.5 km

Physical factors like geology explain a lot of the difference between Sheppard and MetroSur, but so too do standards. Community expectations on a range of variables – for example environmental standards, engineering and operating standards, safety standards, the level of citizen input – appear to be key drivers of higher costs in Toronto. It seems there is a very strong commitment in Madrid ‘to get on and get it done’.

In other words, if we’re going to build more tunnels in the future, we had best be prepared to commit to the concept and take a hard look at how we can get things done more cheaply.

What do you make of the data on tunnelling costs?

Share this


  1. Interesting that it’s so close to the Amsterdam N-S line – it’s run several years over and had a bunch of issues with properties ‘settling’ over it – it also passes under a river. I suppose they are digging through a swamp, and flat, which will simplify things.

  2. Seems like a lot of CRL’s cost is the stations. This is probably a fairly key factor to consider in the cost comparisons. Number of stations per km.

    1. Absolutely; tunnelling, as a rule, is pretty straightforward, always depending on the geology. Building buildings for thousands of of people underground, especially in the middle of cities, is an expensive and tricky business.

      On geology, it is no surprise that London had the first subsurface, then truly underground, railways, as that is pretty easy clay to work [Budapest was second!], and also no surprise that NY first tried an elevated solution; bedrock is hard. As anyone who follows the travails of the 2nd Ave extension will know.

      Another issue Peter mentions above is the building up of an industry and practice of doing it. This is something the Spanish and Italians have going for them; they don’t seem to disband the crews between projects, they are more or less always adding more tunnels and viaducts. In fairly different conditions; Italy is a very rugged topography and to build pretty much any transport link will require digging or bridging, and that is true in the north of Spain too. But in Madrid the difference is that the geology is very amenable to tunnelling, so they put a lot below ground, including many kilometre’s of highway along their river side, in order to return the riverbank to people and place.

      In both cultures the people expect a high standard of Transit provision and place quality in their cities, and expect their governments to fund both.

      1. In my experience (stormwater tunnels) the tunneling aspect is pretty straight forward. The time and cost of digging the shafts, would surprise people. It takes months to dig the hole, weeks to machine the tunnel. As RossR alludes to NZ is a small and distant country so the costs are going to be high. IF central government and Auckland came up with a long term pipeline for tunneling works to give a couple of contractors continuity of work i would say these costs would come down. And i do hope Auckland Council bring in experience to manage the construction because contractors are experts in variations (these can end up being considerable). If the contractors think this tunneling job will be their last they will try and make hay.

      2. reiterating others here, but have some current involvement in the planning of a tube extension in London – the tunneling is the easy part in urban areas. For the most part you don’t need to worry about utilities or property impacts. The big costs come in from property acquisition at portal/ station sites and station construction where a number of constraints are more pronounced like avoiding other underground structures/ foundations which all adds to complexity and time (cost)

      3. The reality of tunneling is there is a relatively small core of people working worldwide in this field. As such crews tend to stay together across projects in diverse geographic areas and companies employing them, with occasional changes in team members. The additional laboring crews, electricians, carpenters etc can get up to speed relatively quickly.

  3. At a certain point shouldn’t large cities that can forecast tunnel building requirements far into the future simply stand up a continuous tunneling operation. Rather than going project by project, try to standardize as much of the requirements and pace it out so you’re spending continuously $200m (for example) per year to complete out the tunnel networks. Similarly if there isn’t enough demand for this in Auckland certainly nationally there could be. I guess it would imply create a public works department though, so it’s probably a non-starter to take such an approach in New Zealand.

  4. A key factor has to be the large, experienced contractors based in Spain and Korea. It’s the home territory for these companies so they get the benefit of economy of scale and low mobilisation costs. Because there’s a constant stream of work in these countries (which nearly bankrupt Spain) there’s more competition.

  5. I think there probably is a second order effect between length and cost per km, simply because short tunnels tend to be done in very constrained situations, while long tunnels will logically have a greater probablility of sections being done under less constrained conditions.

    Case in point, the CRL is entirely built under the city centre with deep level stations every 1000m or so, in places that are quite awkward to build, in other words every metre of the 3km tunnel is difficult. However, if we were building a new metro line of 30km long, it’s probable that long stretches of that line would be technically simple, maybe only a third would be tricky work and two thirds fairly simple tunnel boring.

    Of course the real magic of the CRL is that 3km of tricky tunnel work is the last upgrade that buys us what amounts to 100km of metro system with 40 stations on it when it is finished.

  6. I wonder what the costs of a rail tunnel under the Tamaki Estuary into Pakuranga and eventually surfacing to connect the eastern rail line to a line down the middle of Te Irirangi Drive would be.

  7. “Second, there is basically no correlation between project scale and per-kilometre costs. It isn’t necessarily cheaper (or more expensive) per kilometre to build longer tunnels.”

    Correlation is not causation. In general there are mobilisation and demobilisation costs and other costs (such as the transportation and assembly of the TBM) that are independent of the length of the project. In the case of the 2 Seoul projects (the only data for which a vaguely “apples to apples” comparison can be made) the longer one has a considerably lower cost per km.

      1. Nasty digital cough you have there.

        I have discounted the Barcelona data because the L9/10 data is dodgy – the whole line is 47.8 km with the tunnelling being a subset of that and because work on the the La Sagrera Tunnel ceased in 2014 due to a corruption investigation of one of the contractors. Who knows what the cost will be if/when it restarts.

  8. Seems a no-brainer that once we have the tunnelling machine working, to keep it occupied for an extended time doing a sequence of tunnels throughout the city. It always seemed silly to me that after each tunnelling project, the machine would be packed up and shipped out of the country again.

    One issue however is the tension between different transport tunnel requirements, especially that between roads lobby and rail. Easy to standardise on a tunnel size for rail, but roads tend to come in different sizes depending on lane requirment etc. Suspect that love of road projects is also a barrier for some to support long term pipeline of rail tunnels, since that would tend to use up transport infrastructure budgets…

    1. The TBMs are all bespoke machines it seems. The engineers have to design them specifically for the rock structure, length of tunnel etc and I understand the cutting face is basically buggered at the end of it’s programmed life, while other parts are either salvaged by the TBM companies or just buried in a side tunnel.

      1. Exactly. The TBM itself is a small share of project cost. The TBM for the Waterview tunnel only cost around $50m out of a $1.4bn project budget.

        As others have pointed out, the big costs seem to come from:
        * Setting up the project and getting the tunnelling started – especially if there’s a need to re-route underground services prior to beginning
        * Building the interior services – ventilation can be expensive
        * The risk that the whole thing will get unexpectedly stuck due to bad geotech – a la Seattle’s Alaskan Way cockup

        1. TBM tunnelling systems are relatively capital intensive. It’s not just the TBM, it’s also the pre cast facility and all the services that feed the machine (grout, bentonite, conditioner, water and electrical supply, conveyors for muck, or pipelines and recycling system if slurry machines). TBM operations also generally require more expensive portal works for assembly etc. Aucklands geology is actually amenable to conventional tunnelling, I wouldn’t be surprised if the CRL goes this way.

        2. Interesting comment Matthew “Aucklands geology is actually amenable to conventional tunnelling” – I’m keen to know more. By conventional, do you mean just hand / machine digging and then shotcrete concrete lining? Auckland being largely volcanic, presumably a lot of the digging is through lava flows and scoria, rather than harder rock or softer mud etc? Is that strong enough to hold up without lining? Presumably not, as I guess the water gets through the cracks, and so I presume you’re not talking about unlined tunnels?
          I know that building foundations in Queen St environs, they often hit lava flows there (the carpark next to Albert Park being the classic example), but what was the soil conditions like on the Waterview connection? Was it mud and clay or was it igneous?

          Presumably our light-weight volcanic stuff fractures easily and so is relatively simple to dig away, unlike tunnelling in Switzerland where everything is hard-as-nails granite, so hard in fact that they don’t line the tunnels but leave them exposed rock. I’m guessing we couldn’t do that in Auckland? In Wellington, of course, the local rock is greywacke and it is heavily fractured from all the faults, so we have to line the tunnels here with concrete.

        3. Yeah I meant excavating by an excavator / roaches derived and then supporting with shotcrete. The material most likely to be tunnelled through is a soft sedimentary rock under all the volcanic (think of the sea cliffs on the North Shore). Yes it would need to be lined. There are a few unlined tunnels in NZ but rare for a transport tunnel.

          The volcanics would normally be avoided as it is harder rock that would be difficult to excavate, particularly with a TBM if not designed for it.

        4. I believe the Homer Tunnel on the Milford Road is unlined as it runs through hard Fiordland granite. Used to get quite a soaking going through there with all the water running through cracks, but the seem to have managed to divert that in the last few years.

        5. TBMs operate well in the materials they are designed to cut through. If the design generally keeps the construction within the right material for the TBM (this can be anything from relatively soft soils all the way to rock that is multiple times the strength of concrete) the cutting can be significantly faster than conventional methods (road header, or drill and blast). This is why the geotechnical investigation upfront needs to be so intensive, and as a result relatively expensive.

  9. In order for this to happen, Auckland needs to intensify and grow its population (or atleast plan) to warrant the need for an expensive project. RIght now, I Auckland doesnt have the demand for tunneled train as it is not densed enough. Auckland is too spread out that station catchment is too low for projects like tunneling. CRL on the other hand is needed because the CBD is dense and a major destination. Suburbs on the other hands wont have underground until they get denser.

    1. Part of it is that what we feed will grow. We don’t have dense transit oriented already, but as the consistent 20% increase in transit ridership shows, when quality transit is provided, people will use it.

      So should we wait until densification happens, or should we build the bits we can in the current environment, so that we move towards transit orientated development?

    2. The case for tunnelling isn’t necessarily about density of demand around new PT stations, although that should certainly play a role, but about what we have to do in order to negotiate constraints in the built and natural environment. We’re running out of corridors…

      1. Very true. Some dense cities (Tokyo, for an excellent example, and suburban London, for another) still have lots of surface rail. Because generally the rail was built first before the city grew around it.

        The need to do high cost tunnelling isn’t because of density. It’s just the price we pay for not having planned ahead. Even if we didn’t build the rail in advance, we should have at least reserved rights-of-way.

        In Auckland’s case, the reverse is true of motorways – after about 1960, we built them first and then the city grew around them. So they were cheap, and they’re still relatively cheap to extend further out into the countryside. That was good planning, at least if you want motorways. But what is expensive is widening the chock-full bits we’ve already built, which is now all but impossible.

  10. IMHO, except for a rail based AWHC we are decades away from needing any other road or rail tunnels beyond what is committed now..

  11. Like with the CRL, its a no brainer to have a combined road/train tunnel from Auckland city to the North Shore as part of the State Hwy 1 network.

    The funding of this tunnel should be paid for by Central Government as a national strategic asset like the Auckland harbour bridge.

    Cost recovery for the tunnel road component comes from vehicle tolls similar to how the Auckland Harbour was paid for. With regards to cost recovery from the rail component would from those rail users like Kiwirail, AT, etc.

    China’s major government owned banks are keen to invest into New Zealand’s infrastructure, so the government should go to them for a loan to fund the construction costs.

    1. I’m not sure the road aspect is a no-brainer (depending on your definition of that term).
      I would say a light rail bridge would be significantly cheaper and deliver most of the same benefits.
      Does anyone know what the toll would need to be to collect $6 billion over say 30 years?

      1. Agreed, a PT bridge would be much cheaper than a PT tunnel. Some people are assuming that a PT crossing needs to be tunneled because others concluded a motorway crossing needs to be tunnelled.

        A motorway crossing is very different to a PT crossing. The AWHC needs to join up to the VPT. So it needs to cross Westhaven. If a bridge it would scythe through/over Westhaven and the Fanshawe St area.

        The PT crossing (if LRT or bus) could land at the tip of silo park and continue at grade from there. Much less disruption.

      2. The exact figure would depend a bit on the interest rate you were being charged on the loan to finance construction, as well as the ‘ramp up’ in demand, but as a starting position I’d say somewhere in the range of $400-500 million in toll revenues per annum.

        At present, somewhere on the order of 200,000 people travel over the harbour bridge on an average day, including people in buses. That’s about 75 million trips a year. So based on those levels of demand you’d need a toll in the range of $5-7 per trip, levied on the existing and new crossings and in both directions.

        Even if demands immediately *doubled* following development of a road tunnel, you’d still need a toll in the range of $2.50-$3.50. This is still quite high.

        1. We’d be better off to instigate the toll and see if demand stays high enough to justify the tunnel.

        2. Yep I very much doubt demand would double after applying a 2.50-$3.50 toll. Would be lucky not to go down.

        3. Half or a third of a cup of coffee? People will spend $2.50 or $3.50 without the blink of an eye if it is going down their throat. I’m sure they wouldn’t take a different route to work just to save that?

        4. You assume that

          a) most motorists are travelling to work.
          b) most motorists must drive for their current trip.
          c) most motorists must take every trip.

          All of these are false

        5. That toll guestimate also assumes it costs nothing to collect the tolls, but even with the electronic system on the northern tollway, a chunk of the toll goes into administration costs and GST. The administration cost has decreased from about a $1 per vehicle to 55c or so from 2009 to 2014. But still, at the current volumes of vehicles crossing the harbour your estimate of $5-7 would need to be $6.30 – $8.60 once you include GST and a 55c/transaction admin cost.

        6. “How much is the bike charge going to be for Skypath?” Charging money to ride a bike while making it free to drive your car is one of the daftest things I have come across in transport.

        7. @mfwic While we all know how hard it’s been/long it’s taken to get Skypath going & the rationale behind the toll, I definitely agree it sends the wrong signals to have free cars/paid bikes. I’m still surprised that John Key didn’t declare it a bikeway of national significance and pay for it outright from the national cycling fund of his.

    2. It’s not quite a no-brainer that road and rail should be combined. When you really look ahead and if you were free of sensitivities about increasing density in certain suburbs that seem to be off-limits then rail via Devonport makes more sense on a number of dimensions, which doesn’t align to the SH1 corridor.

  12. Thanks for this analysis Peter. Very useful data for those of us (maybe just me!) who believe that Wellington needs some sort of tunnelled rail extension to “complete” its regional rail system.

  13. The one missing is the Perth tunnel under the city that would be an interesting comparison as its similar in length and twin tunnels under all buildings with underground stations. Its was built about 10 years ago. Cheers

  14. Taking the Sunset Rd. / Constellation Hill on the Northern motorway as an example; How long would it take take to pay off the extra costs of a tunnel through fuel savings compared with the existing set up? Do the economics change for EVs with their regenerative baking? Tunnel would also mean more rateable land.

    1. +1 interesting thought game. Saved time and fuel over that hill would be huge and could build apts over the top.

      1. When the Busway is converted to rail what would be the case for tunnels vs the three hills en route to Silverdale?

      1. Well that one had me thoroughly dumbfounded. Amazed that I had never heard of it before. Very tempted to order one for myself, but I don’t think they serve to Brooklyn (Wellington), just to Brooklyn (New York). Pity really – I quite fancy a burrito…

  15. What really surprised me in the graph, was the cost of the Sha Tin station in HK… The tunnels were already in place, what they did was dig down and build station infrastructure. That’s not to belittle the task – They took up a massive block of land to build the station, which considering the price of land in HK was most likely the main contributor to cost.

    This also raises another question – What is the total land purchase component of the CRL build?

Leave a Reply

Your email address will not be published. Required fields are marked *