In talking about transit planning I’m constantly stressing the need to think in terms of interconnected two-dimensional networks, not just the one-dimensional “corridors” that are the focus of so many transit studies. It’s a hard point to convey because (a) interconnectedness implies connections, also called “transfers,” which people supposedly hate, and (b) networks are complicated and abstract and hard to think about, which is why I’m always trying to create and promote tools for making them simpler.
What’s more, network effects are really hard to photograph. The closest you can come is a photo of a really smooth cross-platform connection, such as this one I observed in Vienna:
Here’s the whole network, with the red arrow indicating Laengenfeldgasse station, where the photo was taken:
The trains in the photo are a westbound U4 (green) and a southbound U6 (brown). So both are more-or-less outbound, but they’ve come from different parts of the inner city and are on their way to different outer suburbs. If you’re going from a part of the city on one line to an outer suburb that’s on the other, you just walk across the platform. What’s more, the trains actually pause here for about half a minute, so there’s plenty of time to make the move.
The trains make this timed connection even though they are both running every five minutes. In most North American and Australiasian systems, we wouldn’t worry about trying to time a connection when things are that frequent; we’d judge that a wait of less than five minutes for a connection wouldn’t matter much. But it still matters some, especially when lots of people are doing it, so the Austrians have taken the trouble to get it right. Even at high frequencies, connections can be points of stress for customers, who don’t know whether the connecting train’s about to leave. The customers on this connection can relax, because they know the trains will hold.
Notice too that becuase both trains hold, making the connection is as fast as staying on your own train; there’s no time penalty attached to the connection at all. This is how easy connecting or “transferring” can be!
I call this a photo of a network effect because people walking from left to right in this image are making a completely different kind of trip from those walking across the platform from right to left. In fact their paths of travel are almost perpendicular to each other, yet both find the same services useful because the connection is so easy to make. This connection effectively doubles the reach of passengers on both lines. That’s why good planning thinks about two-dimensional networks, even though so much of our transit planning is about one-dimensional “corridors.”
Note too that it would have been much cheaper to design this station so that U4 was all on one level and U6 all on another, because the lines obviously approach the station that way. In that case, if you arrived at the station on a U4, the train across the platform from you would be the U4 in the other direction, which is useless to you. Instead, they built a more expensive “weave,” separating the directions of the two lines and interweaving them so that the two outbound trains would be together at this platform, and the two inbound ones at another. There are many weaved stations such as this one in the better transit systems in the world, and they are always a sign that the system really has been conceived as a network, not just a bunch of corridors.
Of course, many ridership models would dismiss the possibility that connections can be this easy, becuase they would assign a heavy “transfer penalty” based on much more onerous connection conditions than these. This penalty, if overstated, will make the model biased against network benefits in general.
Connections can’t always be this easy, but good agencies are always trying to make them as easy as possible. Remember, if your transit agency makes it hard to make connections, they don’t really have a network. They just have a bunch of lines.
Jarrett- Vienna has a great subway system, I didn’t realize that they’d even timed the trains to make cross-platform transfers so seamless.
In Montréal, at Lionel-Groulx station, there’s a similar setup that also caters to ‘inbound’ and ‘outbound’ movements. It works well, the metro frequencies are good enough that there’s not much perceived hassle (though one can always ‘miss’ the train), but it’s definitely not timed like the Vienna example.
Check http://www.metrodemontreal.com/orange/lionelgroulx/index.html for a description/explanation. They clearly learned from the awkward transfers at the other (earlier) big hub station (Berri-UQÀM), which force big crowds up and down escalators and some narrow stairways.
I was thinking of Lionel Groulx. We also have at least one station like that here in Queens: Queensboro Plaza.
Having just ridden the new Canada Line in Vancouver, which has exactly zero connections with existing Skytrain stations, I am reminded of just how good the planning can be in other parts of the world.
There’s a less-well-thought-out example in Manhattan, at Seventh Avenue.
For almost any of the cross-platform transfers it would have made more sense to switch at the previous station. For example, if you’re on the D train going downtown and you want to be on Eighth Avenue, you would probably have changed for the A or C at 59th Street.
The transfers that make the most sense there, from the downtown D to the Queens-bound E and from the downtown E to the Bronx-bound D, require going upstairs both ways. I know; that was a regular part of my commute several years ago.
To the same NYC point, the Hoyt-Schemerhorn (A-C-G) and Jay Street Stops (A-C-F) in Brooklyn achieve the same goal. In fact, there is a great running transfer between the F and the G lines, which you do by jumping off the F line at Jay Street, hopping on an A/C, and then picking up the G train at Hoyt, and it’s all cross-platform. People do it all the time. These and 7th avenue part of the IND Independent System, which involves plenty of weaves between distinct lines. It’s a network inside a network.
In Paris, there is a weave transfer at Chatelet-Les Halles (RER A & B) as well as at Gare du Nord (RER B & D).
As someone who often takes the last train in Montreal, having to transfer from Orange to Green, I can attest to how nice Lionel-Groulx is (as opposed to the megastation of Berri-UQAM. The station is very nicely designed, with ease of transfer maximized for something like 3/4 of the possible connecting passengers.
Berri-UQAM, on the other hand, is a nightmare to navigate. For the simple addition of one end point (the Yellow line terminates there), the station is a confusing behemoth compared to Lionel-Groulx.
Although they hold the last train at both stations so all possible connections can be made, I always wait until Lionel-Groulx because of the simplicity of the transfer.
The systems I’m familiar with do this only when the network is planned in advance; in the more common case, when the system is planned as one or two starter lines, then with extensions, transfers can be nasty.
For instance, Singapore: the two original lines, North-South and East-West, have their two transfer stations set up in such a configuration. In fact, the two transfer stations, which are adjacent, are set up with opposite configurations, so that you can transfer from any direction to any direction cross-platform, and only have to go one station further than necessary in one case out of four. However, transfers to the newer lines are not as convenient, and require walking hundreds of meters.
On the other hand, New York was literally planned line by line and strictly follows streets, and the transfers are so-so. Shanghai is even worse – it has a penchant for 200-meter transfers and out-of-system transfers at major stations.
New York was most certainly NOT planned line by line. There was a lot of good system-scale planning, and they even did some considerable modifications after the fact to simplify the route structure. The problem was that the subway was build as THREE entirely separate, non-connected systems. Each one might make sense individually, but taken together, you end up with a total mess in places like Downtown Brooklyn.
London’s Victoria Line is an example of careful attention to cross-platform transfers in the correct direction; and they went to tremendous trouble to do it, tunnelling around existing lines.
In contrast, London’s pre-nationalization Underground lines have fairly horrendous connections, because they were competing with each other. This is also the result in New York City; the IRT lines are smoothly connected, the IND lines are smoothly connected, the BMT lines are mostly smoothly connected, but switch between lines of different former companies, and it’s a headache.
In Switzerland, all the national railway network is planned as a whole grid. All lines have a 30-minute or 15-minute headway and arrive at transfer stations at the same time, so transfer time is always under 5 minutes. where cross-platform interchange is not available, an announce guides passengers to the next departing trains.
I’ll be in Switzerland next week, and look forward to seeing this in action!
I’m used to imagining a five-minute wait for a transfer being an example of a deluxe, specially-timed connection! It sure would be nice to have some of those! Amazing to consider that in other countries even that isn’t good enough, and they do much better.
Living in Zürich was a transportation planner’s dream. Here is a link to my web page on Zürich http://www.andynash.com/zrh-pages/anZRHhome.html. It has information about public transport, tours and also a public transport ‘tour’ that explains some of the more interesting aspects of the system.
There’s a great connections example in Berlin too. At night when the U-Bahn is operating at 10 minute headways, the U-6 and U-7 lines – in both directions – wait at the Mehringdamm Station for transfers, both cross platform and for those who need to go upstairs and cross over to the other platform. It brings tears to the eyes of transport planners (hope it’s still working that way!).
Thanks, Andy. Am in Berlin now, so will look for this.
In Vienna, you can see the differences in transfer philosophy based on when the station was built. Laengenfeldgasse was built in the past 20 years – with the full intention of a seamless U6-U4 transfer. Karlsplatz – with U2, U4, and U1 all intersecting has long walks between transfers.
Cross platform transfers was an item discussed in a class I took at the Technical University in Vienna.
Yes! The Mehringdamm weave is still there, and it’s beautiful!
I find this fascinating, because there’s nothing technologically advanced about the idea of a weave. A crucial weave at the centre of Sydney’s network was created in the 1930s, and I suspect many of the other examples cited in these comments are even older.
Which weave are you thinking of? Is it the one I’m thinking of, which is the most amazing piece of rail engineering I’ve seen?
It’s visible from Google Maps south of Central Station in Sydney — the immense loom-like structure which takes a set of northbound lines on one side and a set of southbound lines on the other side and ties them into pairs of northbound-southbound shortly before they reach the terminal. Alternating tracks going south from the station go upwards or downwards, then all the upwards tracks merge to one side and all the downwards to the other, before regaining the same level.
That is one elegant and highly effective way to maximize turnaround capacity at a terminal station. I have not seen anything like it anywhere else in the world. It would certainly be suitable for other major cities which mainly have traffic coming in from one direction (and have the space to spare).
Yes, that’s the one, between Central and Redfern. It’s beautiful to ride
through, too, if you’re paying attention.
That looks quite remarkable on Google Earth. And that’s all electrified, too? (I’m used to electrified rail as the height of exotic, here in the US.)
Nathanael, subway systems around the world have the ability to turn trains at two-track terminals fast enough to maintain 2-minute headways. There’s no real need for expensive solutions like the one you describe for Sydney.
Nathanael, to clarify on the the railway flyovers at the southern end of Sydney Central Railway Station – the flyovers are for suburban train lines (Beige, yes, electric from 1926) which run north-south through Central and are an integral part of how the City Circle and Harbour Bridge lines work. Long distance trains finishing at the terminal platforms don’t use the flyovers.
For those not familiar, the Sydney suburban rail system is a heavy rail radial network where a number of suburban lines converge onto common tracks leading up to the City. The lines do not run independently of each other, and there is a high level of conflict between trains servicing different lines.
The railway flyovers at the southern end of Central are an invaluable asset in managing this type of network, setting up trains to allow logical cross-platform transfers at Central Station and allowing trains from any one line to move to any other line, providing great flexibility in routing trains through the system and especially when there are service disruptions. (Alon, this type of infrastructure wouldn’t be applicable to subway systems running independent lines.)
What is remarkable, in terms of the discussion here, is that the benefits of cross-platform transfers were well understood around the world more than 80 years ago (even in Australia!), and a lot of trouble has been taken in the past to achieve them. Here we are today, needing to relearn how to build transit systems, and even if the money is available, somehow still can’t manage to do it properly! (Your city/region/state may be different…)
The longstanding tragedy of many infrastructure projects, including transit–there is seldom enough money to do it right. Most transit projects, including ones which are well-run, are suboptimal in some way–whether its poor route configuration, grade crossings where there should be separation, ugly stations and platforms, etc. This is true in many other development activities–often times the stakeholder is interested in spending the smallest amount of $$$ necessary for acceptable functionality, which is well short of optimal.
And to be fair to stakeholders, it’s not common for planners and engineers, when given an unlimited budget, to use every penny of it on “gold-plating”. Such is a well-known risk of the engineering profession (I’m an engineer of the software variety, and we like to gold-plate even though we’re generally not constrained by materials costs).
This is mom and apple pie, of course, but a well-designed project (of any sort) will be a balance between user amenities and value, professional and technical requirements, time, and cost.
Ed. Thanks for filling in my readers on Bradfield’s remarkable weave in Sydney’s rail network. I have been meaning to blog about this, and will do more on it when I get home and have pictures.
Regarding the Sydney flyovers: effectively three track pairs are involved (actually four, but the fourth pair merges with the third at Central and is little used because there is no spare capacity beyond there).
The flyovers convert them from up/down up/down up/down at the suburban end to up/up down/down up/down at the city end. One track pair runs through the city to the harbour bridge, and two pairs turn into each other via a long city loop. So in practice there is little cross platform interchange of the sort described for Vienna as only a couple of stops can benefit from it. Nice idea anyway, dating from the 1920s.
Compare with the recent planning of the ‘CBD metro’ – an extraordinarily expensive, poorly justified, politically driven project which has since been scrapped. One of the government’s puffs for this was that it would be the ‘core’ of a larger metro network (no details of the larger network were ever given). A friend reported that at a community consultation meeting he raised the point ‘Will the design of city stations be done to allow cross platform interchange with the later additions?’ The planners had not even thought of it. Presumably because it would require planning the ultimate network intelligently in advance, which is not something one does in New South Wales.