With countdown clocks, it's easy to see how train bunching can impact wait times. (Photo by Kim Last)
I have been a big proponent of Transit’s PA/CIS initiative. In addition to vital behind-the-scenes communications upgrades, the new displays finally deliver countdown clocks to New York’s waiting masses. We now when our next train will arrive and how many minutes we must wish away while sitting in subway station on our way home.
With these new clocks, though, come some hidden glimpses into an infuriating part of the subway system: inconsistent scheduling. Take, for instance, the photo above my girlfriend snapped on the way home last night. She was waiting for a 1 train on the West Side to go down to Chambers St., and at 8:54 p.m., she had a nine-minute wait. To make matters worse, the train after hers was only two minutes further behind. That, of course, averages to a headway of around 5-6 minutes per train, but something clearly went wrong somewhere.
I too had a similar experience last night. I caught a movie at Court St. in Downtown Brooklyn last night, and while making my way over to the IRT local, I had to walk along the Manhattan-bound 4 platform at Borough Hall. The next 4 train at 9:45 p.m. was 14 minutes away, but the subsequent train was just six minutes behind that one.
I don’t have the hard evidence to back this up, but I’ve noticed this quite frequently. Downtown express trains on the 2 and 3 will arrive in bunches — an inexplicable happening considering the 3 makes only one stop after the Lenox Ave. terminal before meeting up with the 2 train at 135th St. After the bunches comes a very long wait, sometimes upwards of 10 minutes.
If anything, the countdown clocks allow us an anecdotal glimpse at the way trains are scheduled to arrive, and in my experiences, off-peak coordination is severely lacking. If trains are set to run every 6-8 minutes, a 3 train shouldn’t be two minutes behind a 2 train. Without better wait-assessment metrics and a more granular understanding of headways, all the promises in the world can’t make the actual schedules more reliable, and riders end up paying for the intricacies in variable and infuriating wait times.
23 comments
That’s not as frustrating as what happened to me late last night at DeKalb, when they were (apparently) switching the Q’s onto the N/D track because it was running in two sections. The announcers neglected to tell those on the platform which train was leaving first, so I got burned guessing wrong TWICE, first getting on the Q to watch the N take off, watching a D pull in behind it, and then moving over to the D since the Q had a red signal and may have been being held for other reasons, only to watch the doors close behind me without warning. Then I sat like an idiot on the D, waiting for them to throw the switch back so that we could move. Thanks, MTA.
Also, I’m not going to be satisfied until they put countdown clocks SOMEWHERE in the 34th St./Herald Sq. station, because if you need to get from there to Atlantic/Pacific in Brooklyn quickly, you have many options and no clue how to guess at the next train — not so bad during rush hour save for the crowds, but a hair-pulling tragedy late at night.
And at 14th St./Union Sq., it’s simply idiotic that they have clocks for the L down on its platform but don’t bother to run a wire upstairs to a panel on the main level. If someone is going to Brooklyn, it’s nice to be able to know what the options are instead of having to run down all those stairs to the platform.
We’ll be waiting for years for the countdown clocks to come to the BMT or the IND lines (except for the L). The MTA certainly should be working to install more clocks in more places off the train platforms, so places where you have multiple options, say Borough Hall in Brooklyn, you can effectively make a choice that will save you time.
side note: 9/11 on the 1 train countdown clock.
Depends on the MTA capital budget. If you search through the stuff on thier website the programs you are looking for is ATS-B to put it on the BMT, and CBTC on the IND. The original plans (back in the 90s) had all of it complete by now, but it takes longer then they thought. These things are very complex, and putting them on a running system even more so. And the signal system is saftey critical so a huge amount of additional effort is made check and recheck the designs and installations before it goes in service.
Frankly I give the people in charge credit – when ATS-A and CBTC had problems they didn’t charge on with the follow-on projects as scheduled and make a bigger mess. I’ve worked on stuff where that happened.
At Lexington Av./59th St., they placed a large countdown clock sign covering all IRT trains (4/5/6) on the staircase leading down to the BMT platform! Why would anyone be looking there to find out when the next uptown 6 is arriving???
They actually have a BMT countdown clock system now, but it’s basically useless.
Case in point: A few weeks ago I was waiting for an R train at Atlantic, and I heard an announcement “There is an uptown [Manhattan-bound] express train one station away at 36th st.” Now if I were waiting for the D or the N, that announcement wouldn’t help me at all b/c
a. It can’t tell me which train it is
b. 36th st is so far away that there is quite possibly another train coming, but this system can only tell where trains are in terms of stations.
In other words, I end up with virtually no more knowledge than I had before the system was installed. A total waste of money.
The countdown clock examples you give also reflect delays rather than poor scheduling. I frequently use the iPhone app iTransNYC (one of your advertisers, coincidentally) to see the actual schedules.
The countdown clock image is a result of delays. Some the bunching — for example with the 2/3 — is a result of poor scheduling. It’s actually on the MTA’s schedules.
That’s the more important issue. On a stand alone line like the 1 it will always be evenly spaced. It gets tricker with the 2 and 3. You need them to arrive at the interlockings spaced out so they don’t wait for the switches to move. (This has to do with the safety locking of the switch.) But as soon as someone holds the doors for a bit at a couple of stations all the plans are in the trash. It takes less than you think. And if we make the 2 and 3 evenly spaced (in the theoretical schedule) what effect does that have on the 5, when it’s running to Flatbush. And affecting the 5 will affect the 4…and there is a line that can get really crowded with lots of door holding very quickly.
Now think about everything that comes together and separates down 4th Ave in Brooklyn – ouch!! (And think of it when the M went down there too!)
It’s an incredibly complex juggling act, and I really believe they have it about as good as it is likely to get.
Once the delays happen there are two basic strategies:
Fill from ahead: if you have a big gap the train sits somewhere and waits. These are those “held by the train dispatcher” announcements. Note that the crew doesn’t know the reason – they communicate by those orange lights you see near the conductors position. At night they may also hold if they are ahead of schedule.
Fill from behind: In a situation like the one in the photo the first train could be told to skip stops to put some space in between the two. That works, but it inconveniences passengers.
The other is Fill from the side: making an express local, but that only seems to happen if something really big is going on.
At the risk of going on too long I’ll tell a related anecdote: A few years back I went to a talk by the guy who had coordinated Northeast Corridor Schedules at Amtrak. The way he related it (of course I’m simplifying and I don’t remember details) it started with one of the MN branches in Ct. They would set the transfer time, which determines the MN schedules from NYC to NH. That sets the Amtrak between NY & Boston (since you put the Amtrak trains in available evenly spaced slots). Then since Amtrak is set the other commuter stuff (MBTA) can fill in the other slots. Some of those NY/BOS also continue (or originate) in DC. The other Amtrak trains (NYC/DC) are scheduled to provide even service and then all the other commuter lines to the south are filled in. I find it amusing to think of the whole NEC being determined by a branch in Ct.
There is also the issue of multiple at problematic crossings and merges. The IRT has 3 that can throw off on time performance on the 7th Ave and Lex Ave express lines. The grade crossing north of 135th St on the Lenox Ave line is an impediment to frequent and reliable service on 7th Ave express. The effects ripple throughout the the system on lines it shares tracks with. Getting it to work reliably at higher frequencies would require a higher degree of system monitoring and control that CBTC only begins to provide. There will be a need for trains to make up time with higher performance or physical fixes in the current line routing, merge, and crossing setups.
Bridge openings can be problematic for the A out to the Rockaways. JFK needs lots of fuel for those jets. You’ll also get the occasional recreational boat or bulk barge large enough to open the Jamaica Bay spans. That can affect the 8th Ave lines, and then the Queens Blvd Express.
As for other causes for delays, door holding leads to door malfunction. It can take 5 minutes for the crew to find and disable the balky door. The 1 also has the occasional Broadway Bridge opening to deal with. Off peak terminal utilization can also affect regularity during that time. If a line runs with as few trainsets as possible, there aren’t trains at the terminal waiting to go to buffer for late incoming trains.
Additionally, TO and conductors aren’t always at their stations when the train is scheduled to go. Most of the time its a minute or two, but sometime its more. The later the train, the more crowded platforms can get. Larger passenger loads increase station dwell time, increases the likelihood a rider running for the train will get there in time to hold the doors, and further slow down the train. The decreased interval with the train after decrease passengers it has to pick up, reduce dwell time, and speeds its run. Thus the following train can close the interval on the late train. If line managers and dispatchers do nothing, this can magnify as the trains back and forth between stations.
Operator variability can also affect service regularity. An aggressive TO following a conservative TO can close the gap on a long run line with lots of stops. A TO that can accurately time the timed and grade signals and hug the reds can cut minutes out of his or her run.
Faster runs in the peak direction during the ramp up to peak hr service can leave larger intervals for the following train. The larger loads slows down the following train and leads to bunching, terminal capacity, and merge issues. Slower runs result in larger platform loads by end of run, and closing gap by following train. The effects ripple back up the line.
While like the clocks, I wish they’d only display the next 2 upcoming trains, like they do in every foreign city I’ve visited. There is no need to know that there is a train arriving in 28 minutes after the next 7 trains arrive, and there is no need to display the “if you see a suspicious package” announcement. It’s annoying when I need to know whether to take the train that just arrived or wait for the following one but instead of knowing when the following train will arrive I’m reminded to keep my belongings in sight at all times.
Yes, having messages (another message was asking people to let them know what they think of the signs!) other than the next arriving trains clutter up the signs, which are typically too small for all the info they display. Having signs with 2 lines of info displaying info about the arrival times of 6 different trains is confusing. Also, putting numbers in front of each line telling you the order of each train is just unnecessary.
Unbalanced scheduling and trains arriving in bunches is a direct result of the complex set-up of the NYC Subway. I can’t think of any other system in the world where train routes have to share tracks with so many other train routes. The system itself doesn’t allow for clean, evenly spaced headways, like the Paris metro does for example.
And with so much interweaving of trains going on, delays on one route are bound to affect the other routes that share (stretches of) the same tracks. So I’m afraid that’s something you’ll simply have to learn to live with.
Try any mainline rail rapid transit system, e.g. Tokyo’s JR East commuter rail network as well as many of the private networks, and anything in Europe named RER or S-Bahn.
…which have to deal with the same problem. At least, if you want to perceive it as a problem.
To a smaller extent – note how the lines have headways that are multiples of each other, for one.
But just to show you that it can be made much cleaner, look at the RER A timetable. Off-peak, three service patterns each repeats every 10 minutes and are offset 3 and 6 minutes from each other. It’s slightly less clean when one branches off from the other two, though.
In principle, New York could run the four IRT express trains perfectly cleanly, since the 2 only shares tracks with the 3 and 5 and the 4 only with the 5. There’s no one-line-branches-before-the-other-two issue as in Paris and Berlin. In practice, Rogers Avenue Junction mucks things up in the peak.
The 2/3 schedule is also annoying when the timing is bunched with the uptown 2 coming in, say, 7 minutes and the uptown 3 coming in 9 minutes. That 2 will be packed like a sardine can and the 3 will be empty. There are only 2 stops that are unique to the 3 so you’re getting all UWS/Harlem riders who can take either, along with all the Bronx people who need to take a 2. If bunching must occur due to scheduling they should reverse it so the 2 comes shortly after the 3. That way you pick off as many riders who can take the 3 as possible before the Bronx riders get on the 2.
I think they try to fix the problem by blowing through stops sometimes– the downtown 3 frequently blows through 125th (and the train will still be on the countdown clock!) but obviously this is just as frustrating and the solution is just to run the trains on time.
Of course off-peak coordination is terrible. I’m able bodied so I always choose to walk further to the final train I need instead of expecting to make connections unless it’s pouring rain or 20 degrees.
Seeveral times I’ve been burned lately waiting upwards of 10 minutes on the R platform at Jay Street/Metrotech…. ever since they opened the connection to the A/C/F, the R platform gets almost dangerously crowded at rush hour (that platform is so so slim). Perhaps the MTA didn’t anticipate this? Waiting down there with no Bay Ridge-bound R train in sight when more and more people fill it up is frustrating and maddening…. especially when R trains have such a long route, it’s no surprise that they tend to bunch together when headed to Brooklyn.
I’m wondering if the MTA knows that this is a problem…. transferring to a Manhattan-bound R in the morning at DeKalb Avenue is a nightmarish sardine-packing exercise as well. Couldn’t they just bring back the rush-hour Nassau Street service to alleviate the crowding? They wouldn’t even have to run it all the way to Bay Parkway…. didn’t the rush hour Nassau M terminate at 9th Avenue in Brooklyn once upon a time?
I don’t think so. I know back between late 2001 and 2004, midday (M) trains terminated at 9th Avenue, but I don’t think they ever did it during rush hour.
This is why CBTC and ATO can’t come soon enough. Each interlocking in the IND is an island, but there is some coordination via the newer Master Towers in the BMT and IRT.
If a single computer system with redundancy was running the entire system, then appropriate trains could have priority through interlockings. Don’t put a late train first if there is another one that could pick off riders who just want the first train and ease the crowding on that late train. So much of the merging is done by human guesswork now, that would be eliminated by sophisticated computer simulation and algorithms. Have the computer test out what the conflicts are further down the line if one or the other goes first.
By having a computer running the trains, train speeds could be optimized for seamless merges without having to wait at red signals. Slow down a train if two are approaching a merge.
This could also maximize timed cross-platform interchanges, especially late night when headways are long.
Those technologies can’t solve a bad schedule. Look, I’m not even going to get into the 2/3 problems during peak service/peak hours. But when it comes to evenings and late nights, there is no excuse. Adjust the schedule. If the average is 10 minute headways then let’s make that the real deal. There is no reason a 2 or 3 train should even come close to sniffing each others rail dust if you balance everything out.
Now this still is a people business and when it comes to peak times, riders would need to cooperate. Holding train doors kills a schedule fast. I remember two decades ago when it was proposed to ticket door holding. I’m sorry but I’m all for it as long as it’s done with reasonable discretion. Some of it is purely blatant. Running up to a door that’s 5 inches from closure and prying it open to me is absolutely selfish and disturbing, but then again that’s the New Yorker’s entitlement attitude. I get to hold up a several hundreds of people on a train just because I want to get on this one instead of waiting for the next one.
Those technologies CAN solve a bad schedule! By doing away with the schedule!
Stop scheduling trains to depart the terminal at a certain time. The computer would figure out when the train needs to depart based on the rest of the merges/conflicts that train must deal with. Each terminal would have a count-down clock just like every other station so the Train operator would know when they needed to be ready to leave.
Adjust the speed of the train on the fly to overcome door-holder delays.
Big factor to throw in that no one is looking at. Each train operator performs the duty differently for one or many reasons depending on which area of the job you are looking at. Some operate fast because they want to beat the schedule with anticipation that something will kill the road causing them to be late, miss some of the lunch break or report late for clocking out. I can understand the mindset because you are stuck doing that job until the train reaches terminal regardless of if you are operating past clock out time. But what happens when the road cooperates? And what happens when you have a killer conductor working with that train operator? I have a friend who is now a TO but used to be a C/R and I just happen to be friends with another C/R and they both worked the 7 line as C/Rs. I was told how they knew they were on back to back trains and the one behind was operating the doors fast to try and get the front train to start pushing the envelope. Wild stuff eh?
Take a look at another factor. Ever notice how trains used to be held in station to adjust scheduling or meet a connection with another train across platform? Notice how that is now becoming rare on the IRT now that countdown clocks are in existence?
Let’s take it even further. Courtesy cross platform connections seem to be frowned upon on the Lexington Avenue Line where the discretion is left up to the conductor where as it’s more common on the 7th Avenue Line. For obvious reasons. One ill move on the Lex means you can destroy the road.
I’ll take it further. Any of you ever notice those countdown clocks next to the conductor’s board in Grand Central? Yep, basically you have a certain amount of time to get those doors closed before that clock reaches zero and starts counting the plus seconds you are delayed. Oh and you can thank those strategically placed stairs on that platform that promote overcrowding. Yes I know it’s an old station. A very old station.
So many things to consider when understanding why headways are screwed.