Analysis of Route 512 St. Clair — Part 3: When Things Go Wrong

In this, the final installment of the St. Clair analysis, I will look at the problem of short turning, as well as the details of operations on some days when service was disrupted. 

Previous installments included full-month charts of headway and link time behaviour.  To start off here, the chart linked below shows the destinations and spacings of cars westbound from Yonge Street.  As in previous route analyses, the spacing between the vertical bars allows quick identification of headway irregularities, and the length of the bars shows how far the cars actually went on the line.

Westbound Destinations from Yonge

A few caveats in reading the charts:

  • Only cars (and buses) leaving Yonge Street westbound are shown.  If a westbound trip originated at St. Clair West Station, it is not included.
  • When I did the analysis, there were reference points at Bathurst and at Dufferin.  The short turn charts cannot distinguish between cars going only to Vaughan (and thence to Roncesvalles Carhouse) and those going to Oakwood because both points are in between the same pair of reference locations.  Reasonable assumptions about which destination applies can be based on the time of day when a westbound carhouse trip is likely, or not.  (This is a problem of my own creation by the choice of reference points and has nothing to do with the TTC’s data.)
  • Starting at about 11am on Friday, April 20 through Sunday, April 22, streetcar service operated between Yonge and St. Clair West Station with buses going further west.  A few trips show up on the “short turn charts” which were operated by buses that came through to Yonge Street and therefore were picked up in the westbound analysis.

Stepping through these charts day by day, it’s important to look at the overall patterns and the way that service on each day is similar, or not, as the case may be.

  • The first page shows Sunday, April 1 when the service was fairly well behaved but with a few gaps and short turns.
  • The next four pages show Monday through Thursday, April 2 to 5.  The AM peak pattern with scheduled short turn service to Lansdowne is evident, as is the generally closer headways of the PM peak.  Midday service is reliable on some days, but spotty on others.
  • Friday, April 6 (Good Friday) has no data before 9 am because, I suspect, early morning service was operated on the overnight route number and was not reported in the CIS data for route 512.  There is a service disruption midafternoon.
  • Saturday, April 7 is fairly well-behaved except in the mid-afternoon.  From previous analyses, we already know that this is a period when running times are extended even beyond the values seen during peak periods.
  • Sunday, April 8 (Easter) has regular service as I discussed in part 1 of this series.  Note how the headways appear regular seen at the scale of this chart, but when seen at the detailed level in part 1 reveal the variation within the +/-3 minute scope accepted as “on time”.  It’s also worth remembering that the spread is smallest at the origin of the service (westbound at Yonge) and the service becomes progressively more bunched as it moves west along the route.
  • The week of April 9 to 13 shows the same general pattern but no two days are quite the same.  Only Monday and Wednesday get by with minor disruptions.
  • These themes continue through the month, and I will discuss individual incidents below.
  • A reader has pointed out that Hydro work started on St. Clair late in April 2007.  This explains the different behaviour of the line during the last week of the month.

The underlying point here is that service was irregular on St. Clair on many occasions through April 2007, and the obvious question is whether this situation will be improved with the implementation of a completely private right-of-way on the line.

Service on Wednesday, April 4

Through the morning, service behaves quite normally.  The regular pattern of Lansdowne short turns is visible, as well as the layovers of cars at all termini.  There is certainly no problem with inadequate running time.  At about 9:15, service is held eastbound at about Christie Street for about 20 minutes.  Note how 6 cars (!) arrive in a pack eastbound at Yonge around 9:55.  The gap is reflected across the route arriving at Gunns Loop as a half-hour gap between roughly 10:00 and 10:30.  Things really don’t get sorted out until after 11:00 am.  The most obvious problem here is the way a group of cars was left running as a pack well after the incident that created it in the first place.

(The zig-zag line at about 11:15 is what I have called a “teleportation event” where CIS becomes confused about the location and direction of a car and reports it at widely spaced points in quick succession.  Some of them are very hard to filter out of the data.  Once the TTC uses GPS based data, this effect will vanish.)

The service continues through the early afternoon, and some short-turning is still prevalent.

At about 15:50, there is a delay westbound from St. Clair Station that does not appear to clear until shortly after 16:00.  Many cars travel west in a pack, but this gets sorted out with a number of short turns at Lansdowne.  This is how short-turning is supposed to work.

The remainder of the day is fairly calm.  Note that most runs get good layovers at the termini right through the peak period.

Service on Saturday, April 7

The service is fairly orderly throughout the day which falls in Easter weekend.  Some variations in running time are visible between St. Clair West and Lansdowne, and these are likely congestion related.  (The problem also shows up on the link time charts in part 2 of this series.)

By about 14:00, runs 3 (yellow) and 6 (brown) are running as a pair.  They get layovers in all the usual places, and there is an attempt to sort things out by short-turning run 3 at Lansdowne.  It gets a nice siesta, but both cars wind up back at Yonge Street together and have a long layover.  On the next trip west, it’s run 6 who gets a short turn at Lansdowne.  This behaviour has nothing at all to do with congestion.

Service on Friday, April 13

Starting around noon, short-turning sets in at Lansdowne, but it is poorly managed.  Run 8 (blue) turns back, but rather than filling a gap coming eastbound, the car takes a layover and comes out in between two other runs 2 (pink) and 11 (light green).  This trio of cars continues together for one trip and then two of them vanish.  What has actually happened (looking at the underlying data) is that the two runs have swapped cars and they reappear later in the day once CIS starts reporting reliable data for them.

Meanwhile, just before 13:00, several cars bunch eastbound at Lansdowne.  They proceed east to Yonge as a pack arriving at about 13:30.  The pack spreads out a bit, but is still clustered for the next eastbound trip.  Meanwhile, there’s a lot of short-turning at least in part to deal with the disruption caused by the clusters.  It’s unclear how much of this is really necessary and whether it creates as many problems as it solves.

As the afternoon continues, service continues to be clustered at times, and the short-turns are rarely effective in filling gaps.  Commonly, cars sit in Earlscourt Loop only to emerge together with one or more eastbound cars from Keele.  Things don’t settle down until the evening, and even then some clustering of cars is visible.

Service on Thursday, April 19

CIS data for this date is a bit confused resulting in some “teleportation”, but under this, we have some interesting events.

Starting in the early afternoon, a cluster develops westbound to Keele at about 14:30.  These cars come east as a pack to Yonge, but on their westbound trip are caught westbound east of Oakwood in a long delay.  This large group of cars proceeds with some turning at Oakwood, some at Lansdowne and some going through to Keele.  The service is more or less sorted out within one trip.  Again, this is a good example of recovering from a major delay.

Service on Tuesday, April 24

Some congestion effects are evident here in the afternoon at the west end of the line, and Lansdowne short-turns to sort things out.  The major events come in the evening.

Westbound near Oakwood just after 9 pm a major delay causes a gap of over half an hour to Keele.  This is sorted out with some short-turning and service at Yonge after 10 pm is fairly regular.

Another delay just west of Lansdowne holds service and creates a 45-minute gap at about 10:45 pm.  The resulting parade of 5 cars comes east around 11:30.  Nothing was short-turned into this gap giving the impression that nobody was minding the store at CIS control.

Service on Friday, April 27

Service on this date gets worse and worse as the day goes on.  There is much short-turning, but the bunching of service gets more severe through the afternoon and early evening.

The impression I get is that whoever was managing the line was floundering making many short turns, but doing nothing to sort out service especially with bunches of cars accumulating at terminals.

Service on Sunday, April 29

Congestion effects are clearly visible at the west end of the route and there is a lot of short-turning in reaction.  However, in this case most of this accomplishes its purpose and cars come out of short-turn locations successfully filling gaps.

Conclusion

St. Clair does show effects of congestion on the part of the line which does not have private right-of-way, and this affects service during many periods such as afternoon and weekends, not just the peak.

As we have seen on other lines, some cars tend to cluster through a combination of operator practice and lack of attention to managing headways from terminals.  Short-turning is effectively used at times, while at others the result is at best neutral if not counterproductive.  This suggests variations in style and skill of various CIS controllers.

Layovers are commonly found at both ends of trips on this relatively short line, as well as moderately long times at St. Clair West Station.  Scheduled running times are quite adequate.  Indeed, the TTC needs to be sure, once the line finally does operate with its own right-of-way end to end, that running times are trimmed to suit the new conditions.  As we have seen on other routes, excessive running time leads to long layovers, irregular terminal departures and bunched service.

The right-of-way will contribute to service reliability, but improved operating practices and line management are essential to the success of the St. Clair project.

7 thoughts on “Analysis of Route 512 St. Clair — Part 3: When Things Go Wrong

  1. Just wondering whether the hydro work might have affected this study. I remember that the hydro upgrades west of Vaughan started sometime in April last year and ran all summer. There were some serious delays. I especially remember one day being in one of the streetcars that piled up at Oakwood where they were blocking the streetcar lane. Eventually they’d let them through in a group.

    Steve: That sounds reasonable. There is a distinct change in the line’s behaviour in the last week of April.

    Like

  2. Why do the short turns fail so badly, I think that’s part of the question? Is it that it takes so long to turn a car? Would it help if cars held (or perhaps gap cars waited) at a short-turn loop to jump right into any gaps that form?

    Steve: The problem is that nobody actually manages the re-entry of short-turns into the service so that they actually fill gaps. Some short-turns are intended only to get a car “on time”, although the length of the layover at Earlscourt Loop, plus the common layovers at Gunns Road, imply that they were not very late to start with.

    Like

  3. Line management is clearly the answer to this, and the other routes.

    Will the new GIS components for the CIS system allow for a simple display at TTC control visually showing where all the cars are? I’d think the ability to visually display such information in real-time is the first step in getting things under control.

    Steve: That is the intent. Additional information to be displayed includes the spacing (in projected headway) between the cars. Obviously this depends on prevailing running times because the same distance does not always represent the same headway.

    Like

  4. Is there a method or philosophy of setting running times?

    Obviously, transit riders want the vehicle to show up on time (whether by schedule or headway) and then be moved to their destination as expeditiously as possible.

    Some routes seem to have fairly generous running time, and operators often lollygag or sit for a couple of light cycles. On the other hand, there are routes such as Shorncliffe 123 where the bus is either accelerating flat-out, braking flat-out, or cornering flat-out. (There can still be enough layover at Long Branch for Shorncliffe buses to keep each other company for a bit.)

    Setting tight running times creates the problem that a slower operator can lag and create a gap ahead. Setting generous running times creates the problem of vehicles crawling along, and deliberately missing lights.

    At one time (so goes something I heard), Mississauga Transit set running times so that buses were forced to exceed the speed limit. And I do recall some fairly fast running in the PCC days. Not to mention high-rate operation on Bloor-Danforth.

    Does the TTC have an overall policy on setting running times? How do they come up with the times?

    Steve: That’s a good question for TTC planners, some of whom read this blog.

    There is always a challenge in setting running times for various reasons, but the main ones are:

    Traffic and passenger demand will vary from day to day, but the schedule has to work most of the time.
    Operators have come to expect layovers at terminals, but there is no convention on how long these should be at a minimum. Schedules have “recovery time”, but the amounts have more to do with making headways work out properly (especially on interlined routes like Queen) than with conditions on routes and time of day. This is a major issue for TTC labour relations that has not been addressed, or at least not successfully.
    Some operators are good at making up time with operating procedures such as leaving stops as quickly as possible, encouraging departing passengers to use rear doors, and driving somewhat aggressively.

    If the TTC moves to headway based rather than scheduled based line management, then the running times can be adjusted up and down through the day to reflect prevailing line conditions. Indeed, extra cars could be inserted into a line to allow a standard headway to be operated on a longer trip time without screwing up all of the crew scheduling.

    The important point is that even if a headway might be five minutes on a good day, but six or even seven on a bad one, it would at least be regular and short turns would not be needed to kep operators “on time”. Such a service would actually be better than three cars every fifteen minutes.

    Changing to such a system is not a trivial undertaking and will require a lot of work and trust between management and ATU 113 so that each party doesn’t assume the other will try to take advantage of the new arrangements for their own nefarious purposes.

    Like

  5. Thursday July 31, 10:05pm, St. Clair Station

    I ran from the top of the escalator to “catch” the streetcar that was parked well beyond the regular loading position. After 15 minutes of waiting for a driver I realised that the car was parked that way so that the rest of the cars on the route could fit in behind it. It was packed by the time we finally got going and we passed the last car coming east at Dear Park. The rest were in the loop (5 I think).

    I guess either the driver had a long break or was being relieved.

    A passing lane in the loop would be helpful but I don’t think we can afford that.

    Couldn’t the TTC establish a procedure so that the driver waiting behind a driverless car would take over this car and keep the service moving?

    Customers waited as much as 20 minutes for this car while the wait would have been more like 3 minutes if the 7(?) cars had been properly spaced.

    And everyone could have had a seat and been a little happier.

    Steve: And don’t forget that at this point the 512 is running completely on its own right of way. No traffic congestion, except that of the TTC’s own making, was involved.

    You are correct, there are 7 cars scheduled on that route on a 3-minute headway.

    Like

  6. Do you have any more info on the snappily-named “phase 3” construction between Vaughan and Westmount? When I made enquiries last week, I was told the contract is still being tendered … and the deadline is not until 08 August.

    As usual, this information was not on the TTC or the City’s transit websites.

    Maybe by Summer 2009 we will be able to take a streetcar west of St Clair West station …..

    Steve: As usual, the city seems to be taking forever on this, and there has not even been a public meeting to discuss phase 4, west of Caledonia. Many of the problems arise from trying to co-ordinate the TTC project with roadwork and hydro. If it were just a case of digging up and replacing the tracks, things would be straightforward.

    Like

  7. For the chart of Wednesday April 4th 2007,

    At 9:20, it looks like the purple car heading eastbound suddenly slows down, as seen by the change from steep slope to shallow slope. Pink and Teal catch up to Purple car somewhere west of St Clair West Subway station. Blue, Yellow, and Turquoise catch up to Purple, Pink, and Teal inside St Clair West Station.

    What exactly caused Purple to slow down at 9:20? Is that the same delay you were talking about at Christie Street at 9:15? Wouldn’t a holdup at Christie appear as a horizontal line instead of sloped?

    Steve: Back in 2007, the data quality was much worse than it is in 2010. The way things worked was that the central system (as described in the detailed article) calculated a car’s location from a number of factors and resolved this to an intersection location. One could go a long time without another report from the same car, and it’s next reported location could be some distance away. That’s what happened here.

    At 9:13 am, you can see “blue” disappear from the chart, and this means it signed off the system (out of service) at roughly location 260. In the co-ordinate system I used for 2007, this is the intersection of Christie and St. Clair. A “brown” car similarly disappears a short while later. Meanwhile, purple, pink and green, which were held by the event at Christie, next report their locations at about Bathurst Street. The shallow diagonals connect the two known points for the cars even though they didn’t actually behave that way. Such are the limitations of the data.

    Although you don’t see it, there was a lot of program code needed to filter out the worst problems in the source data from the TTC, and I was able to drop these routines with the much more accurate and detailed GPS-based data.

    Finally, in earlier non-GPS analyses, I was making up the internal scale fairly roughly (a) because the data wasn’t all that accurate and (b) I wasn’t attempting to measure to the metre the locations of each intersection. It didn’t really matter as vehicle locations reported by intersection name when they could actually be anywhere nearby within the same city block. In the 2007 charts, I chose a value of 650 for Gunn’s Loop. In fact, it is 6.7km from Ferndale to Gunn’s Road, and so I should have actually used at least 670. In 2010, I kept the same maximum value and scaled everything in between based on longitudes as St. Clair is a more or less straight line. This means that 1km is slightly less than 100 units.

    Like

Comments are closed.