Tails on Commuter Rail
An interesting discussion on Twitter came out of an alternatives analysis for Philadelphia commuter rail improvements. I don’t want to discuss the issue at hand for now (namely, forced transfers), but the discussion of Philadelphia leads to a broader question about tails. Commuter rail systems sometimes have low-frequency tails with through-service to the core system and sometimes don’t, and it’s useful to understand both approaches.
What is a tail?
For the purposes of this post, a tail is whenever there is a frequent line with trains infrequently continuing farther out. Frequency here is relative, so a subway line running every 2.5 minutes to a destination with every fourth train continuing onward is a tail even though the tail still has 10-minute frequency, and a commuter line running every 20 minutes with every third train continuing onward also has a tail, even though in the latter case the core frequency is lower than the tail frequency in the former case.
The key here is that the line serves two markets, one high-intensity and frequent and one lower-intensity warranting less service, with the outer travel market running through to the inner one. Usually the implication is that the inner segment can survive on its own and the contribution of the outer segment to ridership is not significant by itself. In contrast, it’s common enough on S-Bahn systems to have a very frequent trunk (as in Berlin, or Munich, or Paris) that fundamentally depends on through-service from many suburban segments farther out combining to support high frequency in the core; if ridership farther out is significant enough that without it frequency in the core would suffer, I would not call this a tail.
When are tails useful?
Tails are useful whenever there is a core line that happens to be along the same route as a lower-intensity suburban line. In that case, the suburban line behind can benefit from the strong service in the core by having direct through-service to it at a frequency that’s probably higher than it could support by itself. This is especially valuable as the ridership of the tail grows in proportion to that of the core segment – in the limiting case, it’s not even a tail, just outer branches that combine to support strong core frequency.
Tokyo makes extensive use of tails. The JR East commuter lines all have putative natural ends within the urban area. For example, most Chuo Rapid Line trains turn at Takao, at the western end of the built-up area of Tokyo – but some continue onward to the west, running as regional trains to Otsuki or as interregional or as intercity trains farther west to Shiojiri.
Munich and Zurich both use tails as well on their S-Bahns. In Munich, the base frequency of each of the seven main services is every 20 minutes, but some have tails running hourly, and all have tails running two trains per hour with awkward alternation of 20- and 40-minute gaps. In Zurich, the system is more complex, and some lines have tails (for example, S4) and some do not (for example, S3); S4 is not a portion of an intercity line the way the Chuo Line is, and yet its terminus only gets hourly trains, while most of the line gets a train every 20 minutes.
What are the drawbacks of tails?
A tail is a commitment to running similar service as in the core, just at lower frequency. In Philadelphia, the proposal to avoid tails and instead force what would be tails into off-peak shuttle trains with timed transfers to the core system is bundled into separate brands for inner and outer service and a desire to keep the outer stations underbuilt, without accessibility or high platforms. Branding is an exercise in futility in this context, but there are, in other places than Philadelphia, legitimate reasons to avoid tails, as in Paris and Berlin:
- Different construction standards – perhaps the core is electrified and an outer segment is not; historically, this was the reason Philadelphia ended commuter rail service past the limit of electrification, becoming the only all-electrified American commuter rail network. In Berlin, the electrification standards on the mainline and on the S-Bahn differ as the S-Bahn was electrified decades earlier and is run as an almost entirely self-contained system.
- Train size difference – sometimes the gap in demand is such that the tail needs not just lower frequency than the core but also shorter trains. In the United States, Trenton is a good example of this – New York-Trenton is a much higher-demand line than Trenton-Philadelphia and runs longer trains, which is one reason commuter trains do not run through.
- Extra tracks – if there are express tracks on the core segment, then it may be desirable to run a tail express, if it is part of an intercity line like the Chuo Line rather than an isolated regional line like S4 in Zurich, and not have it interface with the core commuter line at all to avoid timetabling complications. If there are no extra tracks, then the tail would have to terminate at the connection point with the core line, as is proposed in Philadelphia, and the forced transfer is a drawback that generally justifies running the tail.
Do the drawbacks justify curtailment?
Not really. On two-track lines, it’s useful to provide service into city center from the entire line, just maybe not at high frequency on outer segments. This can create situations in which intercity-scale lines run as commuter rail lines that keep going farther than typical, and this is fine – the JR East lines do this on their rapid track pairs and within the built-up area of Tokyo people use those longer-range trains in the same way they would an ordinary rapid commuter train.
This is especially important to understand in the United States, which is poor in four-track approaches of the kind that the largest European cities have. I think both Paris and Berlin should be incorporating their regional lines into the core RER and S-Bahn as tails, but they make it work without this by running those trains on dedicated tracks shared with intercity service but not commuter rail. Boston, New York, and Philadelphia do not have this ability, because they lack the ability to segregate S-Bahn and RegionalBahn services. This means Boston should be running trains to Cape Cod, Manchester, and Springfield as tails of the core system, and New York should electrify its entire system and run trains to the Hamptons as LIRR tails, and Philadelphia should run tail trains to the entire reach of its commuter rail system.
Pedestrian Observations 
On tail with low frequency, like only 2-3 trains/hour off peak, to maximize convenience, wouldn’t it be desirable to run shorter and more frequent feeder trains, such that passengers don’t have to wait for a long time for their train to head home while also lowering the tail segment construction and operation cost?
In the abstract, “shorter and more frequent” is not generally a good idea. The savings for shorter trains get eaten up very quickly by the extra costs of running more trains, which are dominated by the per-train cost factor over the per-metre/per-car cost factor.
The case for shorter is when you are already at the lower limit of reasonable frequency anyway, it then makes sense to buy and run shorter trains. The principal savings are in the purchase cost.
This assumes that your outer tail is able to operate longer trains – in some cases the infrastructure would not allow longer trains, though the infrastructure cost (usually of moving switches or signals) is pretty modest. Expensive platform lengthening is only necessary at terminal stations; selective door opening should be sufficient at all other stations.
Shorter and more frequent vs. longer and less frequent is a discussion point only on how to create a certain line capacity. There are various philosophies, but IMHO, frequency does trump, despite the higher cost.
What about more trains costs more money? Can we make the trains fully automated and thus bring the costs down? Or is this about aerodynamic drag which is a function of train frontal area and length is a minimal addition?
The dominant factor in train operating costs is paying the people who work on them.
All the current full automation systems I’m aware of require staffed stations. If you’re running relatively infrequent trains, then even full automation requires more people than just operating unataffed stations.
The most cost-saving way to operate transit is do not provide any transit service. Offering very infrequent service is only a step away from this.
Sure. But I’m not proposing dropping frequency; I’m saying that the savings from ordering a few shorter trains for an outer tail are illusory, and better to just run the same number of full-length trains.
@ Richard Gadsden
What do those station attendants do? I’m not asking about large central stations where there will also be a lot of people and thus there is always something that needs done. (just cleaning bathrooms can be several full time jobs in a large station), but what about the farther out stations that only serve one train with a few passengers. The stops where only a handful of people are getting on/off each train? Stations that express train bypasses (if there is one)? What do they do that cannot be automated?
I really want to bring useful transit to everyone. Buses are expensive and so in less dense areas cannot be supported. Trains have high upfront costs, but those costs are amortized over a long time, and so if we can bring operating costs low enough, and carefully control construction costs they might cheaper in the long run thus letting us build them to less dense areas.
Fundamentally, getting passengers on and off trains is too dangerous to do without a human being present who is trained to do the right thing if something goes wrong (e.g. a leg gets stuck between train and platform). If there are no humans on the train, them there has to be one on the platform.
The result for infrequent service is that you put the person on the train; for frequent service, you put them on the station and automate the train.
Trains in areas of infrequent service generally just have a single operator, who manages the PTI (platform train interface) and usually drives the train.
Lille’s VAL is the oldest driverless metro and has always been without staff at stations (except major ones). Same for Lausanne M2. Paris Orlyval and M14 have staffed stations I think but most of them connect with other lines anyway. Boarding and unboarding is actually much safer since passengers can’t access directly to the tracks.
Vancouver’s sky train also does not have dedicated station staff. Just automatic intrusion systems and emergency stop buttons
> All the current full automation systems I’m aware of require staffed stations. If you’re running relatively infrequent trains, then even full automation requires more people than just operating unataffed stations.
Relatively infrequent train do not justify the extra cost of running automated system.
Also, Hong Kong’s South Island Line use remote monitoring for station platform from my memory, without having any staffs physically present on the platform.
But in term of passenger experience?
The point is that you might as well just run “the same length and more frequent” because the savings from shortening some trains are pretty minimal.
Shortening all trains is different; if you are having trains of different lengths, then you increase the costs of purchasing (by reducing the size of the orders) and you increase the cost of maintenance (by having more types of train). If you shorten all your trains, you save real money, but the context here is where you have a busy central core that needs full-length trains, so you don’t have the option of shortening those.
Let say you have a 200-km-long main line, and then three 10-km-long branch lines, toward different directions at both ends. So you have 60km tracks that have lower demand that passengers are going to merge into the main route, while there are 100km tracks that have high demand and is main route of operation. I cannot imagine the cost saving of ordering shorter trains for 3x2x10-km long lines is going to be less than the cost premium due to lack of commonality
Especially you can just buy shorter vs longer versions of same type of trains.
And then you also have station construction cost and land acquisition cost that can be reduced by using shorter vehicles to run separately on the line.
Also, long trains running into low demand stations make the train look empty and give reason for operator or anyone who fund the service to reduce the frequency.
Those aren’t tails, though! Those are branches.
Thie entire article is about tails, not branches.
If you organize through train operations on them then they become tail.
In general there are many tails running into one trunk. The frequency on each tail has to be low, not high, in order for all of them to fit into the trunk.
(Also, on above-ground lines the savings in running short trains are pretty minimal. You only get large savings when stations are underground and you get to make them shorter)
What I am asking is, would a forced transfer from all the tails into a trunk route with high frequency service on each of them using vehicles matching demand of each lines, be more preferable than a tail that get 1-seat ride into city center but the train only come every 20 or 30 minutes?
I agree with you. As long as it’s a forced transfer rather than clogging up all the slots on the trunk line. Alon seemed to define a “tail” as through-running so my previous comment was on that basis.
1 seat ride vs forced transfers are really about the goal. If this is the 1950’s and you want to get men who work downtown off the freeways: then a one seat ride is a bit faster and less thinking, so it will attract people who look at the schedule quick, see the train that stops by their house stops near work too. I said 1950’s male because in that sexist world your wife would be at home to take care of the kids if they get sick at school, and everyone worked the same shift all the time: they would be fired if they missed their train so a frequent train wouldn’t provide a useful option. (note that in 1950 the streetcar still existed in most cities providing frequent service)
A forced transfer means those people need to actually understand that transfers are possible. You can’t look at the trains that stop at your home station and your work station and figure out what train to ride. However if you put in some effort to understanding transfers you will discover not only can you get downtown, but you can also get to a job in the next suburb. Frequent mean if you kid a call that your kid is sick at school you just leave. Sure it is more difficult for people to figure out, but the forced transfers also enable a lot of trips that were not even possible before. If you work in the next suburb over this trip is possible, while with branches/tails it involves an unacceptable long wait in some station where you don’t want to be.
I framed the above in terms of going to work. However there are a lot of other actives people do that are out on a branch. These become possible with good transfers. There are a lot of niche hobbies that cannot support downtown rent, but attract people from all over the city (It is common for softball players to travel to different fields each week to play a different team, just as one example). Most people have family who have moved around the city. With a branch/tails system you have made the transfers to get to these things so bad that nobody will do it. However with frequent service and forced transfers these become at least reasonable to make (they do happen when there is no traffic, but you can still attract either because they drink alcohol, or they can save money not having a car.
Of course switching requires a station near those branches with enough platforms to handle all those trains. Building that station may be expensive (particularly if it is underground). The real world is messy that way. Still I would not plan a new system around tails, forced transfers and frequent service should get enough more riders to be worth it. Not to mention the construction costs of new rail is high enough that if you can’t run frequent service you can’t afford to build it in the first place.
With shuttles to branches your two seat ride, with a change at the main station, turns into a three or four seat ride and you don’t go.
^ Yes. In nearly all places, suburb to suburb rides will mostly involve buses (or at best trams).
Trains to the suburbs with good transfers and up zoning near the station will – over 20 years – build up higher demand centers around the major stations. Which is in itself a goal: get people on transit going to a place that isn’t far from home, while allowing longer trips once in a while (not daily) to the larger downtown as needed.
I agree the 4 seat rides won’t happen often, and 3 seat rides only work when all segments are short (but still travel a long distance).
The S-4 in Zürich is a very bad example for what you want to show. A look at the map shows that beyond Langnau (that’s where the 10/20 minutes interval ends), the Sihl valley is pretty much no-mans land. The Sihlwald station (the operational end point) serves leisure passengers only (hiking into the protected forest etc.). Therefore, serving it just every hour is absolutely adequate.
München, OTOH does have tails, as described.
Sometimes, tails happen by branching (such as the ZVV S-12 goes to Schaffhausen or Wil). You will find that branching also in München, but there they actually split/join the trains, as the S-1 to Freising and Airport, or the S-2 to Petershausen and Altomünster. (FWIW, the latter used to be a branch with timed connections in Dachau, but that changed with the electrification and line upgrade).
Chinese metro planning tend to use a lot of forced transfers with smaller sized trains instead of through running lower frequency suburban tails into the core. The initial forced transfer between the Batong line and Line 1 of the Beijing Subway at Sihuidong station has been resolved with the signaling upgrades, but many still remain like Fangshan and Yanfang line having a forced transfer at Yancundong with Yanfang line running 4 car trains to Fangshan line’s 6 car trains.
The forced transfer between Line S1 which runs shorter 90m long maglev trains and Line 6 which runs 154m long 8 car B size trains, but the most egregious in Beijing is the forced transfer between Line 5 and Yizhuang Line at Songjiazhuang station despite both using the same 750V DC 3rd rail 6 car trains.
Shanghai has forced transfers between Line 17 and Line 2 at Hongqiao Railway Station , Line 5 and Line 1 at Xinzhuang, Pujiang line and Line 8 at Shendu Highway.
These are mostly Beijing or Shanghai specific problems rather than China-wide problems. Other Chinese cities like Chongqing, Chendu or Guangzhou have planned the quite opposite: running some or all suburban trains onto urban trunk lines. Some of these urban lines will even see dual-voltage trains of various consists.
Even for Beijing or Shanghai where such forced transfers do exist, these are not always “planning” problems. For example, Shanghai Line 5 was built at a time when rail transit expansion was overregulated. The local goverment built the line without acquiring required permissions by claiming it to be an “elevated trolleycar line”. It had no choice but to purchase cheapest commercial-off-the-shelf products from trains to signal systems, which did not guarantee interoperability with the existing and legitimately-built Line 1.
In Philadelphia…. desire to keep the outer stations underbuilt, without accessibility or high platforms.
Most of the inner stations don’t have level boarding either. It’s one of the reasons why SEPTA ridership is low. The trains take forever to make a stop.
If there are no extra tracks, then the tail would have to terminate at the connection point with the core line, as is proposed in Philadelphia, and the forced transfer is a drawback that generally justifies running the tail.
Philadelphia has four or more tracks through the core. There are four between Suburban and Wayne Junction and I’m not going to attempt to count them west of 30th Street. At least four in three cardinal directions.
Trenton is a good example of this – New York-Trenton is a much higher-demand line than Trenton-Philadelphia and runs longer trains, which is one reason commuter trains do not run through.
There are a lot more people west/north of Trenton. There’s a lot more New Jersey along the line. New Brunswick is roughly one third of the way between New York and Philadelphia and Trenton is two thirds of way.
It’s partly a much higher demand line because masochistic Pennsylvanians wend their way to Trenton or Hamiliton to take trains to Manhattan. Less masochistic Pennsylvanians drive to Cornwells Heights and take Amtrak. Or use 30th Street.
In a world with more capacity to Manhattan the SEPTA Trenton trains would originate at Suburban, run local to Trenton and then express to New York perhaps only stopping at Newark for transfer to PATH, for Wall Street. The West Trenton trains could originate at 30th Street, run local to West Trenton or whatever they are going to call the station near Trenton Airport and express to New York, again stopping at Newark for transfer to PATH…. it’s SEPTA’s job to move Pennsylvanians around, not NJTransit’s.
New York should electrify its entire system and run trains to the Hamptons as LIRR tails,
It would be great if they electrified the Hudson line to Saratoga Springs. West Trenton to Newark would be nice too. They could run trains between Boston and DC every hour or so.
On Long Island there is enough demand to run trains to the Hamptons as expresses. 50 or 60 days of the year. If there are whole trainloads of people, on summer weekends, they don’t have to stop anywhere in Queens except Jamaica and perhaps Babylon. The rest of the year, there aren’t many people out there, there isn’t much demand.
There is a lot of speculation about what the service patterns are going to be once East Side Access opens. The consensus is that the trains originating and terminating in Brooklyn will switch to Manhattan and service to Brooklyn, during peak periods, will be a shuttle between Jamaica and Brooklyn. Since it’s all the same trains, there may still be trains to and from Brooklyn when there is an event at Barclay’s Center.
Think bigger, decide how many trains an hour there will be to New England if Boston-New York is 90 or 100 minutes. And New York to DC is 90 or 100 minutes. Running trains from suburban origins to three destinations in Manhattan, Grand Central, Penn Station and Wall Street it too much capacity or very low frequency. And think about how many trains eastern Queens could fill. There could be a lot of changing trains in Jamaica, Newark or Secaucus and express service to the outer ends. …. like they do now.
Hey, this is something that Renfe does as general operating principle in most of its Cercanías Lines!
Barcelona´s R1, nothbound, for instance, has 6 tph off-peak to Mataró. 2 tph continue to Blanes, and one train an hour goes on to Maçanet and beyond. During peak, pattern changes to 7-8 tph to Mataró, 4 tph go on to Calella, 2 tph to Maçanet. Most of the Barcelona lines work the same way. R2 Nord run to Granollers, or Sant Celoni, Or Maçanet; R3 either to Vic, Ripoll or Puigcerdà; R4 to either Terrassa o Manresa. Similar patterns are in operation for points South, or in the Madrid, Valencia, or other systems.
Traditionally the “tail” trains were limited stop services closer to the core, so the Blanes trains skipped Ocata, Montsolís, Montgat and Sant Adria, for instance. As more service was added to the core, Renfe has tended to eliminate this kind of patterns (it is fairly hard to time overtakes o a 6 tph line), but it keeps them in quite a few places; in Madrid some of the “fast” trains at the Henares Corridor even use a dedicated bypass to get to the core (and hit Chamartin before Atocha).
I think it generally works well, with some caveats regarding travel times and how dang hard it is to keep scheduling reliable.
Note in reference to the statement that Philadelphia is the only all-electric American commuter rail system. Denver is all electric until the DenverBoulderLongmont line is built in the remote future.
Note about running through Trenton. In 1969 I rode a SEPTA Pioneer commuter train that was running through to NYP. It gave the return trip from NYP to Trenton a smaller train for the evening run back from Manhattan.
Oh, yes, sorry!
Wikipedia says there was through service between Reading Terminal and Newark until 1981. With a schedule from 1974, where they are still being called the Crusader and Wall Street.
Tails are something which larger Western urban areas need to learn. Or at least learn how to use it as a more explicit strategy to manage track and platform space rather than by accident. I get really annoyed at the UK ones in particular that have through-running stations which they use as terminal stations rather than building a number of terminal/padding platforms more cheaply in the city outskirts. This especially a problem for the North of England where you have urban areas overlapping each other.
Its also a way to handle a mismatch between branches either side of the trunk section, if you look at the Japanese cases in detail, JR West’s urban sections of the Tokaido-Sanyo services (1 line from Banshu-Ako/Himeji/Kakogawa etc till Kyoto where its splits into the Kosei and Tokaido “Biwako” lines) or JR East’s North-South regional rail (3 north vs 2 fit this pattern a bit more cleanly than the Chuo line which has that nasty two platform terminal at Tokyo station.
So the last bit of Nuremberg U-Bahn line U2 to the airport would count as a “tail” as frequency at peak goes down from 200 second headways to 600 second headways?
Yes.
Would it make sense to turn the last station on the core line (the buttock station?) into a mandatory transfer station for non-express tails? For example, take the line at 10 o’clock on this map: https://www.central-ppk.ru/new/passengers/schemas/schemas/Schm_RUS.jpg?v=15
There’s an S-Bahn/RER kind of line (D2, pink) that stops at Nahabino/Нахабино and a regional line (Ри, sky blue) that goes further out and has both regular and express service to central Moscow. What if the regular service was limited only to the tail of the line (from Nahabino/Нахабино to Shahovskaya/Шаховская), with passengers having to transfer to D2 at Nahabino/Нахабино? This way the core and the tail could have completely independent schedules, while express trains from major stations on the tail would still go to central Moscow.