Quick Note: Amtrak’s Rolling Stock Shortage
It’s a commonplace that Amtrak can’t expand service frequency or even lengthen consists because it has a shortage of rolling stock. This is usually what is meant by “Amtrak is at capacity,” since there’s ample room to run longer trains. So I’ve been trying to investigate how much rolling stock constraint Amtrak actually has.
The sharpest shortage would be present in the Acela trains, since they have high seat utilization – about 60-65%, vs. 45-50% on the Regional. There are ten daily Acela roundtrips north of New York and fifteen south of New York, so at worst, fifteen consists are sufficient. The maximum frequency is hourly; the Boston-Washington trip time is just over 6.5 hours, so fourteen consists should be enough to provide more service than is available today, and with the current mix of hourly and two-hourly service currently used north of New York, thirteen are enough. There are twenty consists, so there are more than enough spares, and rolling stock does not actually limit capacity.
The New York-Washington trip time is 2:47-2:52, and the turnaround time is 8-13 minutes, which means that six trainsets could provide one extra hourly train. This implies Amtrak could do one of three things with the seven spares:
1. It could increase the frequency south of New York to half-hourly, except in the peak one hour in which the North River Tunnels are at capacity with current signaling.
2. It could couple two trainsets together. It could also mix this with option 1, depending on North River Tunnel capacity – i.e. couple two trains together just during rush hour and run every 30 minutes otherwise, and use the seventh spare to cover the mismatch in peak scheduling if necessary.
3. It could cannibalize the cars of some of the spares to lengthen the other consists from six to eight cars – or even ten if service to Boston is strictly two-hourly, which would require only ten consists.
For the most part, the platforms are long enough for the reconfigurations in options 2 and 3. All platforms are long enough for eight cars, and it’s fairly trivial to lengthen the few Acela platforms that are only eight-car long to ten cars except New London. All from New York south are long enough for twelve cars, used in the Pennsylvania Railroad days; Washington’s long platforms are low-floor, but it suffices to convert just one to high-floor. Option 2 really requires fourteen-car platforms – there are twelve cars but two power cars are in the middle – but the platforms are long enough at New York, Philadelphia, and Washington, and at the other stations people at end cars could walk to an adjacent car, a practice already used at the low-floor Regional stations in Connecticut and Rhode Island.
Note that in option 3, it’s in principle possible to make all service north of New York half-hourly, and either run all trains in double or cannibalize trains to create twelve-car consists. The problem with this is not platform length, but the complete lack of spares throughout the day.
There is, in other words, capacity for doubling Acela service south of New York, where the highest demand is. If Amtrak doesn’t provide this service, it could be an artificial shortage meant to keep prices high, or just insufficient demand for the quality of service. And if it cries capacity, it’s just after more money.
Pedestrian Observations 
“…thirteen are enough. There are twenty consists, so there are more than enough spares, and rolling stock does not actually limit capacity.”
There’s this little thing called “maintainence”, which means you can’t use all your rolling stock at once.
No, but you can set your maintenance cycles so that all of your rolling stock is available at the peak. SNCF’s proposals for US HSR mention that 80% of its rolling stock is available on each weekday and 98% is available on the weekend, at the peak of TGV travel. Using just 65% of your rolling stock and letting the other 35% sit at the yards collecting dust is not best industry practice.
It could be that the Acela is a hangar queen of course.
But, Alon, SNCF has a fleet of dozens and dozens of high-speed consists. Certain maintenance activities require multi-day parking of train consists. And there is unplanned maintenance and one-off events with trains consists. I guess the required spare fleet to cater for those situations follows more or less a Poisson distribution.
Fair enough, but it still doesn’t excuse 80% peak availability. I get that Amtrak can’t have 98% availability with 20 consists, but anything under 90% smells fishy to me.
2 in the shop is 90%. 2 in the shop and 2 spares gets you 80%. Want a spare at Boston, New York and DC, 75%. Use one of the spares and you have 3 in the shop….
Equipment in the shop is exactly what spares are. You shouldn’t do heavy maintenance in peak periods. Right now what Amtrak is doing is having lots of spares already – e.g. the schedule including turnaround time makes the NY-DC runs feasible with just 3 trainsets – as well as extra trains not used as spares or in-service.
Hard not to do heavy maintenance during peak periods when your union contract stipulates 8-hour shifts and time-and-a-half overtime (or similar things) and you have no political incentive to challenge the status quo.
I’m not sure that this is the reason they do heavy maintenance during peak periods, but I’ve heard similar things about the MTA’s maintenance force, and from what I understand the MTA has done more to drag its CBAs into the 21st century than Amtrak. Worth exploring, at least – anyone know any Amtrak maintenance workers? I don’t lurk around the railfan forums much, but I’m sure someone there knows the answer.
If the train is dangling in the air Ivy City in the middle of it’s annual 40 work-hour inspection. it’s very difficult to have it replace the train in Boston that’s having a bad computer day and won’t run because they don’t all agree. Or the one where the air conditioner on car 4 has stopped working. Acela passengers would get really pissed off if something happens in Sunnyside and they trundle out something they borrowed from NJTransit.
Bad computer day? How unreliable are those trainsets, anyway?
How unreliable are the Acela trainsets? Well, I don’t really know, but they have a *bad* reputation. Apparently the Amfleet Is are more reliable. Does that help answer the question?
There’s a reason Amtrak is planning to replace them with completely dissimilar high-speed trains when they reach their scheduled end-of-life.
The Acelas have an MDBF of 20,000 miles, comparable to the New York City Subway back in the 1980s.
I think that if Amtrak could adopt a seat reservation system, option three would be more viable; seating passengers in cars that platform on the shorter platforms. A seat reservation System would also help with the ridiculous congestion encountered boarding trains in NY Penn, Philly and to a lesser extent, Boston. Lastly, optional seat reservations would generate a little more income (say $3 extra to reserve?) and allow quietcars to be added on the fly…
You don’t really need a seat reservation system – you can just tell them that the first half of the train stops in NYC, Philly and DC only, whereas the the others stop everywhere.
It’s similar how some hsr trains in Europe split up and branch into two at some point in their journey – you gotta be careful in which part you sit.
Oh, I know, though I suspect it couldn’t hurt. …I just find it irritating that so much is in place for seat reservations: car numbers, seat numbers and sections of the platform; yet this is a service that hasn’t yet been implemented. I’d think as a paid option it’d be very popular.
Seat reservations (train-specific) should be implement in a wider scale, together with a much more airline-style pricing, on Amtrak operations for a variety of reasons, but I don’t think this is one of them.
Amtrak’s ticketing system is already way too much like that of airlines. It’s stressful, inconvenient, and inflexible.
Greyhound tickets are much nicer.
I did some cat scratches on graph paper this morning which made it look like the current number of trips spaced an hour apart could be done with 10 or 11 trains, so 13 to 15 would probably allow for maintenance (and inspections, which take up a vast amount of time). Then again, the schedule I drew up might not be attractive to riders, but it could probably be tweaked.
Still, the simple math is that a single train can make 1.5 round-trips from Washington to Boston in 24 hours, with three hours left to possibly make a run to/from New York, yet they have twice as many trainsets as round-trips along the length of the corridor. That’s nuts — even worse than commuter services where there’s often a 1:1 ratio between daily round-trips and physical trains/buses.
I think Amtrak’s biggest issue may be one you haven’t really addressed yet: scheduling patterns. I say this because I ride the NEC frequently, both on Regionals and Acela, and have been doing so for the past two-plus years. I can’t remember the last time I was on a train that was only 60% full. That said, leaving DC or Boston, the trains are not yet full. The trains are generally full (or close to it) only on the central part of their trip–the last few stations leading into or departing from NYC. Clearly, Amtrak’s “capacity issue” is in that section of the NEC. My guess is that there is unmet demand for New York travel, but that Amtrak refuses to get creative about meeting it–running Philadelphia-Providence Acelas, for instance, or Baltimore-New Haven–in other words, trains that don’t serve the less crowded ends of the line. I wonder if Amtrak’s failure to provide sufficient service is a political issue–Amtrak wants to artificially inflate its occupancy numbers so that Congress will see fit to continue funding it.
Yes. Mean load factor (passenger-miles/seat-miles) is not necessarily a good indicator of capacity. A train sold out between Philadelphia and New York, half-full between Washington and Philadelphia and one-third full between New York and Boston has only a 51% average load factor. But isn’t likely to sell many more tickets.
I suspect that there are some political issues in running, say, Philadelphia-New Haven. But they’re not in artificially inflating occupancy. Acelas stop in every state between Washington and New York and those that extend to Boston stop in all three states north of New York. This isn’t accidental. State support is crucial for the NEC. The long distance trains can call on the support of all the Congressmen whose districts they cross. The NEC needs organized support.
“I suspect that there are some political issues in running, say, Philadelphia-New Haven.”
Interestingly, the Keystone corridor starts in Philadelphia, and the New Haven-Springfield line is currently being upgraded (but not yet electrified). It would probably be possible to always couple two Acelas between New Haven and Philadelphia, but then split some trains in Philadelphia to go along the Keystone corridor, and split some trains at New Haven to go to Springfield – this may require even more rolling stock, and electrification of the Springfield line, so cannot be done right now.
But in the medium term it could result in better load distribution on the one hand, and increase ridership by giving more direct hsr services across the north east.
Well, yeah, a regular bus or train will get full in the middle of its route unless it’s a commuter-style service that dead-ends in/near the center of a city or region. I suppose I need to make a one-dimensional version of it somehow, but I did make this image a while back which shows how random trips in an area with flat density naturally cause a lot of traffic in the center of the area. Even though the same number of trips originate or end in the center of the diagram as anywhere else, the center shows heavier use.
Since many or most riders are going through New York rather than to the city, it might be useful to have some trains skip Penn Station entirely, though I think that would only partially solve the problem (and it might create other issues). My sense is that SNCF partially handles this type of load imbalance by making their TGVs operate more like aircraft than traditional trains, running in a much more point-to-point fashion with a small number of stops.
On the contrary, most riders do not ride through New York. On what would be the biggest through-markets – Philly-Boston and DC-Boston – Amtrak trip time is uncompetitive with flying. And to top it all off, trains dwell 15 minutes at Penn for a crew change. The largest Amtrak city pair not involving New York is Philly-DC according to the Master Plan.
The dwell time in New York has nothing to do with crew changes. Regional trains have a crew change in New Haven with a 2 minute scheduled dwell time. It’s about schedule recovery, to avoid propagating delays from Metro North to New Jersey Transit and vice versa. They chose NYC precisely because it’s at the middle of the line and relatively few people travel through it (I’d estimate 50-60% of passengers get off there).
Could Philly-Boston be made competitive with fewer intermediate stops?
Also, I’d be interested to know what percentage of the NE Corridor market uses the NYC suburban stations. I wonder if it would be possible to incentivize southbound passengers to transfer to Metro-North at Stamford or northbound passengers to transfer to PATH at Newark Penn. Doing that, theoretically, would lessen crowding at New York Penn and allow for reduced dwell times there.
No, it couldn’t – as it is, the Acela has few stops in high-speed territory. My understanding is that the aforementioned super-express Acela, stopping only at Boston, New York, Philly, and DC, was canceled because of insufficient ridership, not equipment availability. Too many people board at smaller stations.
It’s possible to get people to change trains, but not that many, and it would create more problems than it would solve. Amtrak’s long dwell time at Penn could be fixed very easily with more open boarding; the current practice of funneling all boarding passengers through one staircase so that a station agent could check their tickets is stupid. Conversely, direct service to Manhattan is one of Amtrak’s selling points over flying. Letting passengers transfer to Metro-North for East Side service with schedule and fare coordination would be great, but forcing a transfer is pointless.
For precise numbers of how many people use Newark and the other nearby stations, look at New Jersey Amtrak ridership numbers and change the URL the obvious way to get data for other states. The Connecticut numbers show that New London – the only Acela stop that can’t be easily lengthened – is the second least busy Amtrak stop at which Acelas call, next to Newark Airport.
Acela doesn’t stop at Newark Airport.
Yeah, you’re right – for some reason, I thought a few Acelas did stop at the airport, rather than just Regionals. Never mind. So New London is the least busy Amtrak station where Acelas stop.
There are 12 consists for the ten Boston-Washington roundtrips. Seven leave Boston before the second from Washington arrives. Two midday consists make only a one-way trip: they leave one terminus before a consist from the other arrives and arrive at the other too late to make a profitable return trip.
Four consists make the New York-Washington trips.
The remaining consists at any given point are in the maintenance portion of the rotation. The Acelas spend relatively short times in revenue service and relatively long times in maintenance, presumably because they’re overweight.
“The Acelas spend relatively short times in revenue service and relatively long times in maintenance, presumably because they’re overweight.”
Even our trains are overweight!
I think whatever number of consists Amtrak currently uses (16, I think) is the practical limit for reliable service. They tried using one more consist to provide a single “super express” round trip during the peak commuting hour (to NY in the morning, to DC in the evening, I think), but they ended up taking that train out of the timetable, because it ended up cancelled more often than not, due to lack of equipment. And yes, the Acela is somewhat of a hangar queen, but I think it’s more to do with complexity and uniqueness of its design than its weight per se. Perhaps the latter has an effect on things like the reliability of tilting mechanism.
It’s not only about weight, but also track conditions, regular elective preventive maintenance, weather etc.
Weather is certainly an issue in the Northeast. But if the Acela dealt with snow as well as the TGV does, there just won’t be any service for two months every year.
Alon, your suggestion of using trains in platforms shorter than them is hazardous and would require additional, double fail-safe devices to prevent door opening of the “wrong” train cars, plus a revamped protocol for evacuation in station, probably involving adopting evacuation on track which would mean minor disruptions could require the shut-down of the entire sector.
This practice is outdated, dangerous and should be forbidden by FRA, not encouraged. On top of safety concerns for passengers, it poses challenges for certain mobility-impaired passengers and, more importantly, it greatly increases the stop times of trains as one would have to account for the time taken for a lot of passengers embarking/disembarking queuing up and clogging the last train entrances.
If that were not the case, train stations would all be super short and passengers would board the first 2/3 cars and always find their way moving between cars, instead of having the whole system directing them to the “correct” car position in platforms.
Unfortunately, it’s not always financially feasible to expand all the stations to the necessary length, and putting in such a requirement would mean that many existing stations would simply have to close, because nobody has the money to lengthen the platforms or run special short trains just for those stations. And selective door opening is a fairly widespread practice worldwide, with examples in the UK among other places. Finally, when I last rode the MBTA commuter rail system, most of the trains didn’t have automatic doors, and the doors were left unlocked between stations, and frequently just left open entirely.
Selective door opening has been banned or eliminated in countries like Italy, France, Switzerland and Netherlands out of safety and comfort issues. In some cases, they extended the platforms, in other, stations not justifying the investment are now served only by shorter regional trains and people change for faster trains in stations with longer platforms.
For clarity, the UK uses SDO in a very specific manner, spotting trains so that every car has at least one door platforming, and with the other door double-locked.
That said, on longer-distance services which have longer dwell times in any case, there is really no problem with stopping at short platforms, regardless of the train.
Selective door opening is common on Metro-North, where there is a mix of 4, 6, 8, 10, and 12 car platforms. And I don’t think another bogus prohibition from the FRA is the way we want to go if we’re trying to improve our only HSR.
The conductor just throws a switch that prevents train-lining of the doors, and that’s that. You ever hear of people falling on the tracks at busy stations such as Fairfield or Stratford? No, because it’s fairly easy to open the right cars. Proper announcements and, in Amtrak’s case, signage, could make it easier to run longer consists with selective door opening.
Amtrak ran 14 and 16 car trains on the NEC into the 80s. Wikipedia says the train that was in the accident in Chase Maryland had 16 cars. I vaguely remember the foamers on Railroad.net say the major stations have some platforms that are 18 cars long. You’d probably have to do some rehab in places like Newark NJ but they are long enough for 16 car trains.
Whenever I’ve heard “equipment shortage” in reference to Amtrak, it wasn’t in reference to the NEC — which currently is running under capacity and with ridership slacking — but with reference to the cars for LD trains and the off-NEC corridor trains.
Wait, that’s not true. Amtrak was complaining about a shortage of functioning electric engines for Northeast Regionals too. I guess they just bought some more though.
Amtrak is *definitely* short on single-level LD cars, painfully so, to the point where it’s funding new cars out of its own budget. It’s also hitting its limits on functional bilevel coaches (it’s now trying to get the states to buy new ones).
And Amtrak’s single-level corridor fleet is hampered by being full of Horizon cars, which nobody likes and which freeze in the winter. 😛
a) Amtrak owns 20 Acelas, and operates a schedule that uses 17 on any given weekday.
b) 8-13″ turnaround time? You have to clean and restock a train before you refill it with people. You cannot clean and dump the toilets of the train, restock the galleys, and remove the rubbish in 8-13″. That’s a laughably stupid suggestion.
c) The trains are serviced in 4 annual cycles of 13 weeks each, in which the 90 day PM cycles are broken down into hunks that can be performed in the off hours. The 3 that are not in service are having more intensive work done on them. If this makes them hangar queens, so be it, but Amtrak bought what it essentially had to given the politics of the 90s HSR purchase. Most in the firm wanted the X2000, but ABB refused to build them in the US, so they lost out.
d) You do know Amtrak is buying more Acela coaches, right?
The Shinkansen trains turn in 10-12 minutes. Apologies for assuming that a schedule in which a train arrives at Union Station at 2:47 and the next train leaves at 3:00 has a 13-minute turnaround rather than a 1:13-turnaround.
By the way, Acela disaster or not, I’m glad that you commented and I think what the board did in 2005 was a travesty.
d) I’ve seen this talked about and saw a report that Mr. Boardman had made that decision. Has an RFP actually hit the street? I’m an ex-bureaucrat and tend to disbelieve in the existence of things that aren’t on paper.