When Different Capital Investments Compete and When They Don’t
Advocates for mass transit often have to confront the issue of competing priorities for investment. These include some long-term tensions: maintenance versus expansion, bus versus rail, tram versus subway and commuter rail, high-speed rail versus upgraded legacy rail, electronics versus concrete. In some cases, they genuinely compete in the sense that building one side of the debate makes the other side weaker. But in others, they don’t, and instead they reinforce each other: once one investment is done, the one that is said to compete with it becomes stronger through network effects.
Urban rail capacity
Capacity is an example of when priorities genuinely compete. If your trains are at capacity, then different ways to relieve crowding are in competition: once the worst crowding is relieved, capacity is no longer a pressing concern.
This competition can include different relief lines. Big cities often have different lines that can be used to provide service to a particular area, and smaller ones that have to build a new line can have different plausible alignments for it. If one line is built or extended, the case for parallel ones weakens; only the strongest travel markets can justify multiple parallel lines.
But it can also include the conflict between building relief lines and providing extra capacity by other means, such as better signaling. The combination of conventional fixed block signaling and conventional operations is capable of moving maybe 24 trains per hour at the peak, and some systems struggle even with less – Berlin moves 18 trains per hour on the Stadtbahn, and has to turn additional peak trains at Ostbahnhof and make passengers going toward city center transfer. Even more modern signals struggle in combination with too complex branching, as in New York and some London lines, capping throughput at the same 24 trains per hour. In contrast, top-of-line driverless train signaling on captive metro lines can squeeze 42 trains per hour in Paris; with drivers, the highest I know of is 39 in Moscow, 38 on M13 in Paris, and 36 in London. Put another way, near-best-practice signaling and operations are equivalent in capacity gain to building half a line for every existing line.
Reach and convenience
In contrast with questions of capacity, questions of system convenience, accessibility, reliability, and reach show complementarity rather than competition. A rail network that is faster, more reliable, more comfortable to ride, and easier to access will attract more riders – and this generates demand for extensions, because potential passengers would be likelier to ride in such case.
In that sense, systematic improvements in signaling, network design, and accessibility do not compete with physical system expansion in the long run. A subway system with an elevator at every station, platform edge doors, and modern (ideally driverless) signaling enabling reliable operations and high average speeds is one that people want to ride. The biggest drawback of such a system is that it doesn’t go everywhere, and therefore, expansion is valuable. Expansion is even more valuable if it’s done in multiple directions – just as two parallel lines compete, lines that cross (such as a radial and a circumferential) reinforce each other through network effects.
This is equally true of buses. Interventions like bus shelter interact negatively with higher frequency (if there’s bus shelter, then the impact of wait times on ridership is reduced), but interact positively with everything else by encouraging more people to ride the bus.
The interaction between bus and rail investments is positive as well, not negative. Buses and trains don’t really compete anywhere with even quarter-decent urban rail. Instead, in such cities, buses feed trains. Bus shelter means passengers are likelier to want to ride the bus to connect the train, and this increases the effective radius of a train station, making the case for rail extensions stronger. The same is true of other operating treatments for buses, such as bus lanes and all-door boarding – bus lanes can’t make the bus fast enough to replace the subway, but do make it fast enough to extend the subway’s range.
Mainline rail investments
The biggest question in mainline rail is whether to build high-speed lines connecting the largest cities on the French or Japanese model, or to invest in more medium-speed lines to smaller cities on the German or especially Swiss model. German rail advocates assert the superiority of Germany to France as a reason why high-speed rail would detract from investments in everywhere-to-everywhere rail transport.
But in fact, those two kinds of investment complement each other. The TGV network connects most secondary cities to Paris, and this makes regional rail investments feeding those train stations stronger – passengers have more places to get to, through network effects. Conversely, if there is a regional rail network connecting smaller cities to bigger ones, then speeding up the core links gives people in those smaller cities more places to get to within two, three, four, five hours.
This is also seen when it comes to reliability. When trains of different speed classes can use different sets of track, it’s less likely that fast trains will get stuck behind slow ones, improving reliability; already Germany has to pad the intercity lines 20-25% (France: 10-14%; Switzerland: 7%). A system of passenger-dedicated lines connecting the largest cities is not in conflict with investments in systemwide reliability, but rather reinforces such reliability by removing some of the worst timetable conflicts on a typical intercity rail system in which single-speed class trains never run so often as to saturate a line.
Recommendation: invest against type
The implication of complementarity between some investment types is that a system that has prioritized one kind of investment should give complements a serious look.
For example, Berlin has barely expanded the U-Bahn in the last 30 years, but has built orbital tramways, optimized timed connections (for example, at Wittenbergplatz), and installed elevators at nearly all stations. All of these investments are good and also make the case for U-Bahn expansion stronger to places like Märkisches Viertel and Tegel.
In intercity rail, Germany has invested in medium-speed and regional rail everywhere but built little high-speed rail, while France has done the opposite. Those two countries should swap planners, figuratively and perhaps even literally. Germany should complete its network of 300 km/h lines to enable all-high-speed trips between the major cities, while France should set up frequent clockface timetables on regional trains anchored by timed connections to the TGV.
In the 1970s (and presumably earlier) Market-Frankford in Philadelphia had 35 second headways at the 5 PM sharp peak. SEPTA also had combined 40 second headways on its subway surface lines, but didn’t signal in the stations. It also had variable platforms in the busiest stations underground. PATH trains had 30 second headways around 8:30 to 9 AM in the tunnels where there weren’t any stations. (also 70s). Culture is a big thing with headways. Passengers in Philly in those days didn’t run for trains and hold doors open. The need to handle sharp peaks has also been mitigated with most people not believing that a 5 PM quitting time is and absolute entitlement. (Leaving at 506 pm is quite OK nowadays or even later.)
I don’t think you understand correctly what a “headway” is. It’s the time elapsed between sequential train departures, or than between sequential train arrivals, not that between a departure and a following arrival.
I suppose you could get to 35 second headways, but that would be extremely difficult. Your doors can only be about for 15-20 seconds. (this is possible, but people will sometimes hold the door). You need track where each train is in great detail, with much smaller blocks than is typical, and ensure that all trains slow down enough that if the train in front has an emergency it can stop in time. This implies slower speeds, as in an emergency stop you can hurt people (who might be standing or have packages not secured) if you stop too fast from high speeds. Everything needs to be automated, and have redundant in many systems so that when something goes wrong it doesn’t result in collisions. If you have switches on the tracks they need to switch fast, and you better maintain them.
It would be fun to run such a system, and it is probably technically possible. However it would be rather expensive. Better to just make your stations longer and run bigger (longer and/or wider) trains with longer headway. Even in the largest cities, if you can’t keep up with common length trains at a 2 minute headways you can make a lot of people happy buy building a whole new track in a somewhat different direction that is a better route for their destination.
Please show me a timetable with 35 seconds headway!
From what I’ve been able to gather by studying its HSR system, it seems like Spain has been able to strike a happy medium between the approaches of the French and German systems: full utilization of dedicated, high-speed lines coupled with the option of short, intermediate stops to serve less populated regions in between. Although dominated to a clear extent by Madrid, there seems to reign in general a far more equitable planning philosophy there as opposed to France. When studying a Renfe map recently, I was surprised to see how well even the relatively thinly populated region of Galicia is served by high-speed lines.
Now look at a timetable for any of those lines. Unusable, most unused.
Spanish rail operations (outside of the metropolitan “S Bahn” cercanías) are a disaster.
So much sexy sexy infrastructure, so few trains, run so randomly, so infrequently, with so much airline-mimickry customer-hating baggage.
Just looking at a schedule for Madrid – Santiago De Compostela on a weekday next week, and there are fast connections every 1-2 hours — 10 trains all together. Seems pretty good? Or is that not typical?
So Santiago de compostela has a metro area population of 185k, and you can go onto Vigo and A Coruna which are like 300k each.
In contrast Plymouth (population 260k) where you can go onto Cornwall (the whole county has a population of 500k) gets an hourly service to London from 5am to 8pm of trains you’d want to catch – plus another slower service in the gap that you probably wouldn’t catch – and there’s also an evening sleeper train.
London is twice the destination that Madrid is, though.
England is a tiny little country though with a population density four times higher than Spain’s. The distance to Santiago is 60 percent farther from Madrid than Plymouth is from London. You also might want to take a look at the widely contrasting terrains.
Galicia is arguably one of the best run corridors. It had a big frequency bump this year when the final HSL section Pedralba-Taboadela opened, cutting travel times to a point where they are finally competitive with flying. It will probably have another frequency bump when the current shortage of rolling stock ends as the midlife refurbishment of Class 130 is finished and the new Class 106 finally enters service after at least a year of delay. Galicia is also, thanks in part to its polycentricity, the only region where Renfe does timed transfers in a minimally serious way.
But the rest of the country… well, Richard Mlynarik is painfully right about everything. The vast majority (like, over 80%) of transport infrastructure spending over the last three decades has been on highways and the rest on high-speed lines, basically leaving the rest of the rail network to rot — at least as far as not-Madrid-and-Barcelona is concerned. I could go on and on about this topic so I’ll restrict myself to a pair of examples that I think are illustrative lost opportunities in the context of network effects: Granada and Huesca.
Let’s start with Granada, a metro area of almost 550K which also houses more than 60,000 students as well as Spain’s most visited tourist site, the Alhambra. It lies some 90 km of mostly flat terrain east of the Madrid-Málaga mainline and the legacy line is kinda shitty, so it would have made sense for the government to build a high-speed branch and take advantage of network effects, right? Well, they sort of did… but they built a new line on the flat sections while reusing the shittiest part of the old line, putting in a third-rail mixed gauge thus making it even shittier. Oh and they also made most of it single-track so it is at capacity almost from day 1.
The current long-distance services are 3 trains per day to Madrid and 1 to Barcelona. There are also 3 to Sevilla (one of them offers a “timed” connection to Madrid that involves a 48-minute wait at Córdoba), and –wait for it– just 3 per day to Málaga, a metro area of nearly 1 million at 90 km distance which is a massive tourist hub and also increasingly Spain’s main tech industry hub. (There’s also Marbella, the city of 150K that sits 50 km southwest of Málaga and has no rail service whatsoever.)
Now let’s move on to Huesca, a province capital of just over 50K people, located some 70 km to the north of Zaragoza, a metro area of 770K which, besides being a major destination by itself, is also a major crossroads at the midpoint of the Madrid-Barcelona HSL, among others. So during the construction of the Madrid-Barcelona line, they decided to also spend € 250 million to refurbish the Zaragoza-Huesca branch by electrifying it with 25 kV, building a second track (standard gauge, 200 km/h) on one section and adapting the existing track to third-rail mixed gauge (160 km/h) on the other. So far so good, right? Sounds like a typical case of regional trains running every hour between Huesca and Zaragoza (maybe every half hour on the peak), stopping at several stations inside Zaragoza (to provide transfers to the local tramway and direct service to the university) and with timed transfers to AVEs and Alvias to forward destinations at the main station.
Haha, nope. There’s 5-6 regional trains per day along with 1-2 buses per hour running on basically the same route. Oh and did I mention that those trains run on diesel because 20 (!) years later Renfe has still failed to procure enough bi-current trains for regional service?
Oof. Thanks for the vivid examples. I guess I just got lucky with my dart throw and hit a region that’s more the exception than the rule. I still however wonder — although mired by its poor follow-through and lack of service to many areas — if Spain’s rail system doesn’t represent a middle ground between the German and French approaches, or is my judgement here also false? It just seems like, at least when I compare system maps (all I can do since I’ve yet to try out HSR in Spain), that Spain’s approach seems to mostly do the things Alon criticizes about the German and French systems.
I am sooo impressed that you always know what is best – regardless of the topic. Is it ok with you if I nominate you as the next Secretary General of the United Nations?
What he’s saying about Spanish rail operations isn’t too different from what Isabel Pardo admitted in an interview – that the timetables aren’t frequent enough. There’s a reason eve-of-corona AVE ridership was only 22 million.
“Long distance” British trains by contrast managed 139 million journeys pre-Covid by contrast – https://dataportal.orr.gov.uk/media/2064/passenger-rail-usage-jan-mar-2022.pdf. Corrected for population the Spanish should have managed ~97 million journeys.
(small nitpick: her surname is “Pardo de Vera”– it’s a compound surname; her second surname is Posada)
FWIW, ridership seems to have finally caught up to pre-corona levels this summer, and trains are murdering planes in the one corridor where liberalization has already gone into effect. (I know you were skeptical about liberalization, but I think the early signs are promising.) But yeah, the overall assessment is still correct. We haven’t yet exited the vicious circle of low ridership / low frequency / low rolling stock utilization / shortage of rolling stock, compounded by capacity bottlenecks that go unresolved for decades and by Talgo’s schedule slip in the certification of the Avril.
Pardo de Vera seems to be the only person with some power who understands the problem fully, but she’s been out of action for quite some time due to health issues. She’s now thankfully recovered and back on the job, but in the meantime the ministry has been effectively rudderless — the only major projects that were greenlit in her absence were highway widenings and rail soterramientos, that kind of project so customary in Spain we even have a specific word for it.
I express no opinions about things about which I’m not informed.
It’s a policy few adopt.
So what I’m hearing is an endorsement of Nuremberg tram expansion?
In the last fifty years, the tram network has shrunk on the net (there have been expansions, most recently in the north to
https://maps.app.goo.gl/apjYyCNXX1UhTRjh8 this busy stroad intersection seemingly in the middle of nowhere, but the shutdowns far outweigh those…) while billions have been poured into subway expansion.
While all current or planned endpoints of the subway (
https://maps.app.goo.gl/b6YXJefWSviU8wMD9 U2 South,
https://maps.app.goo.gl/3WReZa9ZTKBYGDNS8 U2 North,
https://maps.app.goo.gl/aJ2DdJGNpDE5NzoQ7 U3 South [planned],
https://maps.app.goo.gl/DGdA67QfqX4bo6mn9 U3 North, 54 Julius-Leber-Straße
https://maps.app.goo.gl/C8UpC9Knm1ySKnVA7 U1 East,
https://maps.app.goo.gl/6MZXTJNduFEqgQny6 U1 West [the actual endpoint is a bit to the north, but somehow I can’t share it. To the northwest of the metro station you can see a tram turning loop that was never used) might justify extension if it could be built for subway-surface costs, none of them are capable of expansion with current subway cost and funding. And Nuremberg U-Bahn was never intended as a subway-surface system – at least not past the point of rolling stock acquisition…
Meanwhile the tram network is perhaps one of the smallest in German cities of its size class (~500k) at less than 40 kilometers and it has a lot of forced transfers… It’s also about half the size of its prewar peak, but that has to be taken with the grain of salt of the U1 west corridor (central Nuremberg to central Fürth along Fürther Straße) was a tram line then and is a subway line now…
The service level capacity of a single track with stations mostly depends on train operating characteristics, station dwell time, minimum speed limits due to track conditions and terminal architecture. It’s nominally 40 tph (90 second headways) for the equipment, rights of way and 30 second dwell times for NYC. This link shows the service level capacity and actual peak service levels operated by the NYC BOT in 1949. It was included as part of their report
As can be seen, the Third Ave El operated 42 tph in the single track reverse direction. Just about every single line was operating in excess of 24 tph.
A similar map appeared in the NYC TA’s 1954 Annual report.
The removal of the Third Ave El’s downtown terminals limited its capacity. The map does show that the Flushing Line operated at 36 tph.
“Even more modern signals struggle in combination with too complex branching, as in New York and some London lines, capping throughput at the same 24 trains per hour…”
Our grandfathers knew that signalling poses a relatively modest decrease in service levels. The 1949 Third Ave El had branching at 129th St, Bronx Park and Gun Hill Rd. It managed.
What’s required are schedules that do not create merging conflicts and a train supervision system that keeps trains on the non-conflicting schedule. That’s different from a signal system’s function of Traffic Control, which is collision avoidance. NYC lacks adequate train supervision. It’s CBTC system has copied it’s non-existent train supervision so it provides no benefit in increasing service levels or avoiding merging delays.
Let’s consider a simple merge. It takes approximately 1 minute for a 600 ft. long train to enter and clear the interlocking. (Derived by observation and theory.) Let’s assume that a merged 30 tph (120 second headway) operation is desired. This means the interlocking is idle for 60 seconds for each 120 second cycle. A sufficient condition to avoid merging delays is that every train be within 30 seconds of its scheduled arrival time. The leader is 30 second late and the follower arrives 30 seconds early. The leader will have just cleared the interlocking when the early follower arrives. Similarly, trains must be within 15 seconds of their scheduled arrival for 40 tph (90 second headway) operation.
NYCT schedules have a precision of 30 seconds. This is true for both the static and run time GTFS schedules. They could have used an hourglass. They need realistic schedules with a 1 second precision and a feedback system to permit operators to make micro corrections at each station.
“with drivers, the highest [TPH] I know of is 39 in Moscow, 38 on M13 in Paris”
Both Moscow and Paris employ such train supervision systems to achieve these service levels.
Moscow has an up counting clock that records the number of seconds since the last train left each station. The driver and conductor know the headway they must maintain and consult this clock to adjust the station dwell time.
Paris uses a different system. There’s a clock with only the minutes and seconds showing. The clock setting is offset at each station by what the scheduled departure should be. If a train is scheduled to leave at 08:01:25, the crew is advised that is should leave each station along the line, when the station clock reads 01:25. There are several different clocks, to reflect different schedules throughout the day.
Do we know how reliable the trains were in terms of keeping to their overall schedule in 1949?
“Do we know how reliable the trains were in terms of keeping to their overall schedule in 1949?”
Yes. It’s in the same document.
The OTP for the BMT and IND was over 99%. The IRT lagged; it was 87.47%
Here’s a comparison with the MTA performance Dashboard:
The Dec 2019 OTP measure was Div A (IRT) was 84.8%. Div B (BMT-IND) was 76.7%.
Show some respect for our grandfathers. :=)
” It’s CBTC system has copied it’s non-existent train supervision so it provides no benefit in increasing service levels or avoiding merging delays.”
Nope. Any rail transit automatic train control (ATC) has a subsystem called automatic train supervision (ATS), be it CBTC or not. The well-known “formula”:
ATC (automatic train control) = ATP (automatic train protection) + ATO (automatic train operation) + ATS (automatic train supervision)
” Any rail transit automatic train control (ATC) has a subsystem called automatic train supervision (ATS), be it CBTC or not.”
NYC uses drivers and conductors. It does not have complete ATO. Train departures are subject to the whims of the operating personnel; they are not regulated by ATS to keep trains on schedule. The only dwell time regulation is by “holding lights” at a few selected stations. These holding lights are manually operated. The door closing reaction time when the holding lights are extinguished is unpredictable.
As I mentioned, their CBTC/ATS implementation copied their hourglass concept of train supervision. The GTFS-RT feed uses 30 second precision. Station dwell times are not incorporated into the GTFS-RT schedule.
There’s no reason to suspect that an ATS which mimics chaos will improve service levels. It has not in NYC. The current service levels for the two CBTC implementation are below that achieved in our grandfathers’ day.
“NYC uses drivers and conductors. It does not have complete ATO. Train departures are subject to the whims of the operating personnel”
It’s called semi-automated train operation (STO), or GoA2 automation. Such level of automation does not render ATS useless. Some of these systems have wayside starter signals, clocks, or LED screens that tell the operator time to go. But even without these wayside indicators, these information will be displayed on operator’s HMI, via the train-wayside communication system (TWC).
Moreover, the most powerful “weapon” the ATS has is regulating the “performance level” the train operates at in between stations. The schedules are always padded. So, if a train is severly delayed, the ATS will figure it out, and order the train’s onboard ATO to run at the maximum performance level (usually maximum speed allowed by the ATP, maximum acceleration and deceleration). If it’s early, the ATS will lower its performance level.
Non-CBTC ATC usually has fixed performance level settings (for example: 75% max speed, 75% acceleration, 75% deceleration). In case you don’t know, the Washington Metro’s ATC has eight distinct performace levels that can be automatically selected by the ATS. (https://en.wikipedia.org/wiki/Washington_Metro_signaling_and_operation)
CBTC gives trains a little more autonomy. Since CBTC enables broadband wireless communication between the train and wayside, the ATS will actually send the schedule to the train. So the train’s onboard ATO can do its own planning, with the help of pre-stored digital track maps.
My remarks have mainly concerned NYCT’s CBTC and ATS implementations. Any resemblance between these implementations and the capabilities of these technologies is purely accidental.
Here’s a case in point: this morning’s comparison between actual and scheduled terminal departure for the two CBTC/ATS systems currently operating.
Route: L; Direction: NORTH Terminal: Canarsie-Rockaway Pkwy; count: 27; avg seconds late – 40; std dev (sec) – 68
Route: L; Direction: NORTH Terminal: East 105 St; count: 9; avg seconds late – 65; std dev (sec) – 101
Route: L; Direction: NORTH Terminal: Myrtle-Wyckoff Avs; count: 2; avg seconds late – 13; std dev (sec) – 18
Route: L; Direction: SOUTH Terminal: 8 Av; count: 31; avg seconds late – 42; std dev (sec) – 75
Route: 7; Direction: NORTH Terminal: 111 St; count: 4; avg seconds late – 14; std dev (sec) – 16
Route: 7; Direction: NORTH Terminal: 34 St-Hudson Yards; count: 53; avg seconds late – 2; std dev (sec) – 8
Route: 7; Direction: SOUTH Terminal: 111 St; count: 8; avg seconds late – 17; std dev (sec) – 14
Route: 7; Direction: SOUTH Terminal: Flushing-Main St; count: 23; avg seconds late – 2; std dev (sec) – 7
Route: 7; Direction: SOUTH Terminal: Mets-Willets Point; count: 8; avg seconds late – 4; std dev (sec) – 12
Route: 7X; Direction: SOUTH Terminal: Flushing-Main St; count: 33; avg seconds late – 4; std dev (sec) – 14
Clearly, something other than technology has motivated the 7 Line crews more effectively than those on the L Line.
The Flushing line has better terminals than the Canarsie Line?
“The Flushing line has better terminals than the Canarsie Line?”
It’s a management issue. Main St is my home stop, so I witnessed CBTC’s growing pains.
CBTC made no operational difference, when it came on line. The Queensboro Plaza merging delays continued. I recorded some of these delays from the GTFS-RT data.
Byford evidently wasn’t satisfied. A week later there were supervisors on the Main St platform. Their task was to make sure that the operator and conductor were in the train and the brakes were charged, when departure time came. The merging delays disappeared.
I suspect they did it using the techniques you use in a car when you’re running late, i.e. speeding, harder acceleration and harder, later, braking.
They probably also opened and closed the doors on the train while it was moving.
The question is which of those techniques (if any) would be politically viable today.
“I suspect they did it using the techniques you use in a car when you’re running late…”
Usually by riding the yellow (signals) at full speed, by falsly assuming that the leading train won’t stop for some emergency. Should the leading train made an emergency stop, the following train would (at best) run a red signal or (at worse) end up in a collision.
I was on a train the other day when the driver forgot about my stop until they were a bit close to the station.
You could smell the brake dust when the train stopped but he made up 2 minutes and the ride quality was still better than a bus 🤷🏼♂️.
Maybe if we say that bus ride quality is an acceptable minimum there are corners the speed limits could be raised too by the same token. They could still target the current limit as the ideal in normal operations.
Schedules should include dwell time at every station.
Scheduled dwell time should add a few seconds to permit trains to make up time caused by operational variability. The number of seconds should be based on the standard deviation of the dwell time required to load/unload passengers. This way, there’s a relation between the amount of additional scheduled dwell time at each station and the probability of remaining on time throughout the trip.
I suppose that, instead of decreasing headways, if a train is delayed by a few seconds you can just delay every train behind it by the same amount and maintain the same frequent headways?
Many rail transit operators have taken the “maintain the headway, not the schedule” approach given its obvious benefits. Most operators don’t. Upon a delay unsavable by the ATS, they’d rather skip some stops or terminate the delayed train and take it out of service, than delay every following train but ep the headway.
I think the “headway over schedule” approach is better for short, unbranched, frequent lines, like medium capacity transit systems and APMs. Frequency ensures that whenever you miss a train, you can just take the next. Trains are already slow, so a moderate delay don’t create problems.
Otherwise, “schedule over headway” is better. For longer lines, minor delays can accumulate to the extent that actually delays a passenger trip, leading to inconsistent trip time. For infrequent lines, missing a scheduled train has more consequences. Branched systems require “scheduled meet” that cannot be replaced by headway management.
I thought New York managed 30 TPH on the Queens Blvd express during rush hours. I could be wrong though.
It does in theory, but not in practice.
squinting at the E train timetable and the F train timetable, each runs every four minutes, towards Manhattan, at Union Turnpike, between 7:11 and 8:22.
“I thought New York managed 30 TPH on the Queens Blvd express during rush hours…It does in theory, but not in practice.”
One needs to investigate what causes the difference between theory and practice before concluding the theory is impossible to achieve.
If NYCT cannot muster a full complement of rolling stock and operating personnel, then no amount of sophistication in the signal and train supervision systems will provide full service. It isn’t a question that such operation cannot be achieved; it just has not been tried.
I became aware of NYCT’s inability to operate its full schedule several years ago. I created AWOL reports for the morning and evening rush hours, using their GTFS-RT feed.
The GTFS-RT schedules are listed 30 minutes before the scheduled terminal departure. I considered a trip as AWOL, if the GTFS-RT scheduled trip did not depart from its announced terminal. If a static scheduled departure never made the GTFS-RT schedule, it wasn’t counted.
They barely managed to send out 85% of their scheduled rush hour trains before Mr. Byford came on board. He managed to get that percentage up to near 95%, shortly before he left. It’s now back down to 85%.
This morning was a good day. Here are the results for the E and F for the morning rush hour:
AM AWOL Report for 23 Aug 2022
Totals for time period: 08/23/2022 06:00:00 to 08/23/2022 08:59:59 1603 scheduled; 1450 operated; 90%
Totals for time period: 08/23/2022 06:00:00 to 08/23/2022 08:59:59 by direction
Direction: NORTH: 785 scheduled; 696 operated; 89%
Direction: SOUTH: 818 scheduled; 754 operated; 92%
E: N: 31 sched, 28 dep, 90%; S: 36 sched, 33 dep, 92%
F: N: 37 sched, 26 dep, 70%; S: 39 sched, 33 dep, 85%
Adherence to schedules is also important. I’ve also kept track how timely trains left their terminals.
Route: E; Direction: NORTH Terminal: World Trade Center; count: 28; avg seconds late – 5; std dev (sec) – 14
Route: E; Direction: SOUTH Terminal: Jamaica Center-Parsons/Archer; count: 33; avg seconds late – 24; std dev (sec) – 36
Route: F; Direction: NORTH Terminal: Avenue X; count: 4; avg seconds late – 83; std dev (sec) – 97
Route: F; Direction: NORTH Terminal: Church Av; count: 1; avg seconds late – 0; std dev (sec) – 0
Route: F; Direction: NORTH Terminal: Coney Island-Stillwell Av; count: 16; avg seconds late – 11; std dev (sec) – 17
Route: F; Direction: NORTH Terminal: Kings Hwy; count: 2; avg seconds late – 82; std dev (sec) – 116
Route: F; Direction: SOUTH Terminal: Jamaica-179 St; count: 32; avg seconds late – 15; std dev (sec) – 21
I submit one must reserve judgement as to whether the theoretical 30 tph operation is achievable – I’m waiting for them to try it. :=)
If someone can find a 2009 schedule for the Nuremberg subway, we could confirm or debunk whether the shared U2/U3 trunk had 36 tph at peak during the year of mixed operations. It does today, fully automated.
Try web.archive.org (and its competitors) for the transit authority web site. They have timestamped versions.
Reblogged this on Calculus of Decay .