When Should HSR Serve New Urban Stations?

Greenfield high-speed rail lines frequently serve new stations rather than legacy stations; the TGV network is famous for this, and the discussion of whether to place intermediate stations in the city or on the outskirts has arisen in many reports and studies on the subject. What is less commonly discussed is what to do at the main urban stations. More often than not these are the legacy stations, but there are several exceptions.

Those are not the infamous beet-field stations in France, but something quite different – they’re in different neighborhoods from the older stations, but are still in dense urban landscape, often (but not always) as close to downtown as the older stations. Trains do not pass them at very high speed, so they’re chosen primarily to make through-service easier, in cities whose legacy stations are terminals or otherwise difficult to connect through. Indeed, I do not know of a single case in which such new stations are intended to serve as terminals – they are either through-stations from the start, or terminals intended to be used as through-stations with a later extension.

Example 1. Shin-Osaka is located just outside central Osaka, about 4 kilometers from Osaka Station. Osaka Station is a through-station, but there are sharp curves from it to the legacy Tokaido Main Line at both ends, and also there was not enough room to build additional tracks for dedicated Shinkansen use. Since the goal was subsequent through-service west of Osaka, it was easier to build a new station just outside the CBD, at the intersection of the Tokaido Shinkansen with the Tokaido Main Line (now Kyoto Line). The station is also connected to one subway line, which goes to Osaka Station and beyond. Although there has been development near the station, it is a secondary station, with far more traffic at Osaka; a transit-oriented CBD is too compact and dependent on a huge subway network to move so easily.

Example 2. Lyon Part-Dieu was built specifically for the TGV, since the old station, Perrache, was at a poor location for connection to the high-speed line. Part-Dieu is located in a busy neighborhood area of Lyon and has seen ample development, and the Lyon Metro, which is not much older than the LGV Sud-Est, serves it from multiple directions. In addition, commuter trains have been diverted to Part-Dieu from Perrache, so that now the station is France’s busiest mainline station outside Paris. Despite its use as a through-station, the construction of further LGV segments south of Lyon in the 1990s made it somewhat of a terminus for TGVs, while through-trains skip the city at full speed on a greenfield alignment to the east of the urban area or stop along the way, near the airport.

Example 3. Lille’s legacy train station, Lille-Flandres, is a terminus. This was unacceptable for TGV service, not least because the main draw of building a line to Lille was the onward connection to the Channel Tunnel, which was constructed at the same time. Thus, a new station was built a few hundred meters from Lille-Flandres, on the land of a former helicopter base; because of the city’s new position at the junction of high-speed lines to Paris, London, and Brussels, the station was named Lille-Europe. Like Shin-Osaka and unlike Part-Dieu, Lille-Europe is primarily a high-speed train station; Lille-Flandres is much busier (it is the second busiest provincial French station, after Part-Dieu). This is despite the fact that Lille has extensively redeveloped, using the TGV as a catalyst.

Example 4. Because of difficulty reaching Barcelona’s main station, Sagrera, the AVE is initially serving a terminal station a few kilometers to the west, Sants. A new track connection to Sagrera was built, in order to allow full through-service to points north and east of Barcelona. In this case, the choice of a new station was a temporary measure allowing the line to open earlier, rather than a change in alignment.

What all of these examples have in common is different from the usual conception of building new HSR stations, both in an outskirt setting and in a CBD setting. None of these stations has been about digging greenfield tunnels under city center – indeed Shin-Osaka was explicitly about avoiding such tunnels, and Sants was built as a way to allow service to open before such a tunnel were finished. None involves a station cavern; Lille-Europe is above ground, despite its CBD location. None is an urban prestige project.

Indeed, the decision to build a new underground station complex under Marseille’s terminal station, Saint-Charles, is one of many contributing to the very high projected cost of the LGV PACA project linking Marseille (and Paris) with Nice. A though-station very close to downtown exists, and an underground option there was judged slightly cheaper in an alternatives analysis, but all currently considered scenarios involve an underground station at Saint-Charles.

Another thing to observe is that neither Japan nor France compromised on station location in the capital, but at the same time neither built extensive infrastructure for it. None of the Paris RER lines or of the TGV projects has included any provision for building a single central Paris station for trains in all direction; such a station would require a large cavern with multiple tunnels, and the space and money for such tunnels is far more valuable for local transit use. Likewise, Japan has had no trouble cutting back legacy intercity and long-range commuter trains to bring the Shinkansen to Tokyo Station, but stops short of building a new cavern for it. The most it has done is reserve space at Shinjuku for a future tunnel for use by the Joetsu Shinkansen, requiring new subway lines that go nearby to be built deeper.

The upshot in the US is that the emphasis should be on functional train station locations, rather than on the most central locations. In particular, the Amtrak Vision‘s plan to bring intercity trains to Market East and Charles Center through new tunnels should be shelved in favor of the existing 30th Street and Baltimore Penn Station. In addition, a new track connection between Grand Central and Penn Station should be used only by commuter trains, which need it far more than intercity ones (it would also allow tighter curves, saving on expensive Midtown land acquisition), mirroring the fact that no TGVs serve Chatelet-Les Halles. If the example stations in this post are any guide, any Manhattan location south of 60th Street would work for New York’s primary train station, and Penn Station’s location is as good as any.

In California, what this means is not surprising: converting Los Angeles’s Union Station configuration from terminal to through-station is paramount. In addition, at the Bay Area end, it’s fine to end high-speed trains at the existing 4th and King terminal rather than Transbay, until future money for the final tunnel is committed. In the longer run, in San Diego, although the existing Santa Fe depot should be used if possible, another urban location would not be hurtful as long as it had some transit accessibility and was in a walkable location.


  1. jim

    I agree on the Baltimore and Philadelphia stations, but there’s a difference between them. All trains will stop at Philadelphia, so some slow zones near the station don’t matter very much: trains will be traveling slowly there anyway. But not all trains will stop at Baltimore (it’s like Wilmington) and the winding approaches will slow trains that need not be slowed. A Charles Center station only saves two or three minutes for a train that stops at Baltimore. But it saves nearly ten minutes (depending on allowable tunnel speed) for trains that don’t. So a tunnel along the PennDesign alignment (that was taken up by the Vision) but without a station cavern is still a good idea. It’s more expensive than the Wilmington bypass along the Shellpot Branch and presumably affects fewer trains, but it’s in the same effectiveness class.

    • Tom West

      Don’t under-estimate the impact of slow zones, just because they are near stations. The trackwork near Union Station, Toronto is being upgraded to allow speeds to increase from 25km/hr to 50-70km/hr. Doesn’t sound like much, but it will save about three minutes – which is significant when the typical journey time is 30-45 minutes.
      If you elimiate the slwo zone, the train can spend time accelerating instead of dawdling.

      • Richard Mlynarik

        This can’t be overstated! Ideally trains should enter the platforms at 60+kmh and do as much as physically possible of their braking along the platforms, not en route. Otherwise they’re not only wasting time, but blocking the approach tracks to major stations, whose interlockings are typically major choke points.

        Speeding up the slowest stuff is always (well, 90+%) the best thing to do. Nothing beats shortening dwell times, and then you go after the slowing to and restarting from a stop.

        I know of a few Euro-projects where apparently ridiculous amounts of money and effort were put into a concrete walls and pillars and wedging in higher speed turnouts and a critical crossover or two to make station throats clear tens of seconds faster.

        How refreshing (which is sad in itself) to hear that there is an North American example of understanding elementary arithmetic. (All you need is addition and division to work out that slowest = longest.)

      • Steve

        As I have pointed out before, the slow zone at 30th Street also happens to be Zoo Interlocking, the junction between the Northeast Corridor and the ex-Pennsylvania’s Main Line; today, it’s the junction between the aforementioned NEC and the Keystone Corridor. That is: two designated high-speed corridors.
        There is a Faustian choice to be made here: either you accept the slow zone of the interlocking in favor of existing (well-built) infrastructure and the connections a full high-speed junction allows, or you eliminate the slow zone, but this requires a ton of money for a new routing, and a second ton of money to figure out a new junction. The amount of money needed to remove the slow zone while retaining full connectivity is, in this case, any reasonable sum.

        One of 30th St.’s other problems is the fact that the Keystone Service has to do a reverse move. This stems from the fact that the station was built south of Zoo, while most New York-points west trains would use a station north of it. In the Golden Age of Passenger Rail, the PRR got around this by having the fastest New York-points west name trains stop at North Philadelphia (which still has the bones of a major transportation junction); this would, however, be as unpalatable today as proposing a Chicago station in Englewood.

        Charles Center and Market East both seem to be an attempt to heal an ailing commercial center by shoving more transportation into it. I fear, however, Charles Center may be too far gone. Downtown Baltimore’s retail center is today the Inner Harbor; Charles Center is nowhere near it and has far less amenities or resources to fall back on. In addition, the rail problems in Baltimore go far beyond just the Penn Station approaches; neither Norfolk Southern or CSX have adequate double-stack routings all the way through the Northeast Corridor due to bottlenecks here. Some time ago the FRA proposed “Belt Line” tunnels replacing the Howard St. tunnel (CSX bottleneck) and the Baltimore & Potomac tunnels (Penn Station south approach, NEC bottleneck, highly dilapidated and needing replacement). Instead of treating the HSR and overall NEC needs separately–which is a major part of the NEC HSR cost bloat–as Amtrak is doing, HSR and non-HSR needs on the NEC need to be considered together, so that instead of proposing Belt Line and Charles Center (semi-)independently, what we get is an upgrade of the the NEC to HSR, where feasible, and a new intercity alignment, where not.

        • Steve

          Aha if I read the comments further the proposal I referenced would be “Great Circle”. In any event, though, the Great Circle proposal would provide every freight and passenger need Baltimore could possibly have for a century or more.

    • Alex B.

      There’s also a need for a complete re-working of all of Baltimore’s heavy rail infrastructure. CSX is working on double-stacked clearances south to DC and then to the West, but the connection up to the port of Baltimore is severely constrained both in capacity and height.

    • orulz

      If a bypass route were to be built for Baltimore, I’d like to think that a cheaper/better alternative could be found than one that goes directly under the center of downtown without stopping.

      If Penn Station is kept as the main passenger station, and freight is eventually rerouted from the Howard Street tunnel to the Penn Station alignment, new tunnels will have to be built both east and west of Baltimore Penn to augment the B&P tunnel and the Union tunnels. I would hold that the new tunnels could probably be built with sufficiently gradual curves in order to allow through passenger trains to go through fast enough, without the enormous additional expense of a new 7 mile tunnel under downtown Baltimore.

      I drew this idea out in Google Maps a couple years ago.

      • Eric

        A tunnel under Eager St and MLK Blvd would be about 3 miles long and with relatively low construction costs (very little tunneling under buildings, a not-too-dense neighborhood). From there the route could continue west in the median of Highway 40. Surely this option would be less expensive than the long tunnels under buildings you recommend. Not to mention that the new rail station would be half a mile south (closer to downtown) than at present.

      • Alon Levy

        The existing study by the FRA focuses on the B & P Tunnels, which are a bigger speed and capacity constraint than the Union Tunnels. The speed limit there is 30 mph, and the preferred Great Circle route would speed them up to 75 mph and save 1.5 minutes; in fact higher speeds seem plausible – the tunnel’s curvature is wide and its dynamic envelope is large in order to permit double-stacked trains. Allowing 125 mph trains, about the fastest that’s defensible within the city, would save about 30 more seconds.

        The Union Tunnels are limited to 45 mph, I believe (my track map is unclear on whether it is 45 or 60). But they’re fairly short, and allowing full speed through them won’t save too much time, especially since it’s probable that slightly higher speeds than 45 mph are in fact feasible.

        Charles Center might possibly save 10 minutes relative to today, but that also includes the effect of high run-through speed and better rolling stock. Relative to just getting better trains, building the Great Circle route, and maybe fixing the curves east of the Union Tunnels, it’s much less. Even the Market East tunnel, cutting off much more route-length in Philadelphia, would only save about 2 minutes relative to using the existing infrastructure in a way that’s optimal based purely on ROW geometry.

        • Adirondacker12800

          There’s no technical reason why Market East and Suburban don’t have service to DC, Harrisburg or New York. If there’s demand for it, they could run it. My copy of the 1956 Official Guide has a train or two an hour in each direction betwen Suburban and New York 20 hours a day with one in the dead of night.

          • Nathanael

            Basically, they’re full up with commuter trains, and the dead of night is used for track work.

            But yeah, there’s no reason you couldn’t run electric trains from those locations to DC, Harrisburg, or New York. You could even run the Keystone through without reversing ends, if you did a little work on the Reading lines to get the train to Trenton that way.

        • Steve

          They’re both the same link…just the first part of the study.

          I remember seeing the whole study, it was a thing to behold.

        • Nathanael

          If you remember that the Transcontinental Railroad was completed before anyone managed to build a way to get rails from one side of Baltimore to the other…

          …you won’t be surprised that Baltimore’s a mess to fix. The current alignments were actually driven by 19th-century NIMBYs.

          • Alon Levy

            NIMBYs, and uncooperative railroads. That’s why Baltimore has two substandard tunnels serving different stations, instead of a union station with one better tunnel. Boston did a little better – it did not build a single union station for the New Haven and the Boston and Maine, but when it was time to collect branches that fell under common ownership into one station, South Station included both the New Haven and the Boston and Albany.

  2. Pingback: Today’s Headlines | Streetsblog Capitol Hill
  3. Miguel

    I think you are wrong about the Sagrera Station in Barcelona. You seem to suggest that Sagrera is a working station and the other station, Sants, is new. It is the opposite: the train only stops at Sants, the old station, as a temporary measure because it is already there and it is easy. Sagrera will be the brand-new world-class train station of Barcelona but of course they needed to find a new place for that and also build a new tunnel though Barcelona. Sagrera will be the kind of new station a bit far from the city centre that you are talking about as the trend in the TGV.

    • Peter S

      Sants is also a significant transit hub in Barcelona. Two metro lines have a station there as well as many regional trains and long distance trains as well. While Sagrera is better served for HSR to France, Sants is better located in the city for urban transit as well as connections to Madrid and the airport.

    • Richard Mlynarik

      I bet Miguel knows infinitely more about La Sagrera than I, but just to add:

      The site of Bcn La Sagrera is an existing rail corridor, formerly (decades ago) freight yards, that is currently traversed by two pairs of regional rail (commuter rail, cercanías, rodalies) lines.

      The massive new station will serve those lines, as well as HS to the south and north. Nice overview here (13mb PDF). Two Metro lines (L4, L9) part of the complex, and four others within ~200m. Eight platform faces for regional trains (an island platform for each direction of each line, for short-headway overlapped dwells) and eight faces for HS above. North of the station there’s a big turnback and servicing facility — the majority of HS trains will run to the south, through Barcelona Sants, through the new connecting tunnel and terminate at Sagrera. Fewer HS trains will continue north on a new line towards Girona and France. All with huge taxi ranks, excellent bus facilities, a significant amount of parking, co-developed office buildings, and a park on the top.

      Once it is completed, all trains will continue to stop at the more central Sants station (Metro L3, L5, regional R1, R2, R3, R4, R7), but (as I understand) all will run though to Sagrera.

      All train/infrastructure enthusiasts, even those of us who don’t speak Catalan or Spanish are obliged to keep on eye on http://estovadeobras.wordpress.com (many must-see construction time-lapse movies.)

      So Sants today: an interim HS terminal and regional through station. In the near future: through station for HS trains.
      La Sagrera today: a massive construction site traversed by regional lines. In the near future: a massive HS/regional/metro/bus/parking/office/urban-redevelopment complex, with significant HS turnback capacity. Wow!

    • Alon Levy

      Ugh. I’ll update the post.

      My prior impression was that Sagrera was the existing main station and Sants was a preexisting secondary commuter station. Track configuration-wise, my impression was that Sants is better for service to Madrid and Sagrera for service to France, and the main point is to link them and let intranational trains serve Sagrera.

  4. Matthew

    I remember Shin-Osaka. We didn’t realize it was outside of the city, so we ended up with a lot less time to spend in Osaka than we thought.

    • Andrew in Ezo

      You must have budgeted very little time to see Osaka, given that Shin-Osaka is a mere five minute ride from Osaka Station, with 5 minute service headways.

  5. Michael John

    I saw a powerpoint of the Amtrak Proposal. They plan to dig a cavern under existing Moynihan/Penn Station, and build a new high speed rail station there. Why can’t they use the existing tracks and platforms there, as there won’t be any express trains passing through New York City, of all places. There would have to be upgrades made to the tracks and platforms, but it would be less costlier than building underneath the station. There still would have to be a new tunnels constructed for HSR under the Hudson and East Rivers though. The existing tunnels can be upgraded for commuter service Same goes for Boston South Station and Union Station, use the existing tracks, albeit upgraded and modified, to access the stations. Also, we don’t need to build another cavern under Grand Central Terminal!

    • ComradeFrana

      “There still would have to be a new tunnels constructed for HSR under the Hudson and East Rivers though.”

      The East River tunnels should have more than enough capacity for HSR once ESA opens.

      • Adirondacker12800

        If they had opened it on schedule in the early 80s. I can’t find an authoritative source – the trains that currently go to Brooklyn will be redirected to Grand Central. The trains that currently go to Penn Station will continue to go to Penn Station. It’s why the MTA is busy building a new terminal in Jamaica for the shuttle that will replace service on trains that currently go to Brooklyn.

        • jim

          the trains that currently go to Brooklyn will be redirected to Grand Central

          Will that be enough? ESA Grand Central will have eight tracks vice (effectively) five at Atlantic Terminal and will have tail tracks to help manage them. I don’t get the sense that Atlantic Terminal is saturated, either. Just redirecting Brooklyn-bound trains to GC won’t fill the station.

          I’d expect LIRR to either run more trains in total or to divert some from Penn to GC.

          • Adirondacker12800

            I don’t have authoritative sources. They will be running more trains in total. There’s not going to be less trains to Penn Station. Trains that run to Brooklyn will be redirected to Grand Central. There’s going to be service added – from the Port Washington Branch for instance. Port Washington branch riders are going to introduced to the concept of changing at Woodside for Penn Station. ( Off peak anyway ) The opening itself will induce some demand. Less crowding on the trains to Penn Station will also induce demand. If they had completed it in the 80s there would have been a few less trains. There won’t be in 2017, all it will do is relieve overcrowding on the existing trains.

          • Adirondacker12800

            And short term-2030, 2040 or so, instead of terminating trains in New York and turning them around in Sunnyside they could terminate them in New Haven, Springfield or Boston.

      • Alon Levy

        The East River tunnels have more than enough capacity for HSR today. Total LIRR traffic at the peak is a little less than 40 tph, and four tracks allow 48 tph quite easily. If there’s enough demand for 8 tph north of New York, there’s enough money in operating profits for anything.

        • Adirondacker12800

          Add 5 or 6 trains an hour from New Haven and PFFT there goes your slack.

          • Alon Levy

            Ah, but with a transfer at Sunnyside, trains from New Haven can be substituted for LIRR trains to Penn Station one-to-one, to improve service to Downtown Brooklyn.

          • Adirondacker12800

            The trains won’t be able to go to Brooklyn. From

            Click to access 2010-14questions_and_answers_v2.pdf

            Phase 1 provides for the construction of a new platform at Jamaica
            Station along with associated track, signal and switches. This new
            infrastructure will create a dedicated location within Jamaica Station to
            operate the new Cross-Borough Scoot service between Jamaica and Flatbush
            Avenue, to be initiated as part of the East Side Access operating plan.

          • Adirondacker12800

            …. and trains that go through Sunnyside, assuming they put the station at Queens Blvd. will only be going to and from Penn Station. You can see the tunnel being dug under the west side of Sunnyside Yard in current satellite images. It would be under 41st Ave if 41st Ave ran across the yard.
            No big deal from the point of view of a Metro North rider – they can get to Grand Central on a train thats going to Grand Central. Same thing for LIRR riders except those that are walking to Sunnyside… but if they are walking to Sunnyside they can just as easily walk to a station on the 7.

          • Alon Levy

            The main point of Sunnyside done my way (i.e. at the yards rather than at the intersection with QB) is to increase frequency to each Manhattan station from each outlying station. A secondary benefit is to let people from Co-op City get to Grand Central slightly more easily than they can today.

          • Adirondacker12800

            Putting it at Queens Blvd means it’s a long but reasonable walk to the subway. Putting at Queens Blvd means they could even put a subway stop at the station. They’d probably have to move the 7 slightly or even redo the whole bridge. Put it anyplace else and it’s Secaucus East without the parking.

  6. D. P. Lubic

    Off topic, but for those who are interested and perhaps haven’t checked the site real recently, The Infrastructurist announced it was suspending operations as of Jan. 6, 2011, apparently at least in part because its writers and editors found other work. You may want to check what’s on it before it goes away entirely.

Leave a Reply to Adirondacker12800 Cancel reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

This site uses Akismet to reduce spam. Learn how your comment data is processed.