Analysis of 504 King: Part IX – Headway Reliability and the Two Minute Wave

So far, we have been looking mainly at data one day at a time over the route as a whole.  In this discussion, we will look at service at specific points for every weekday in December 2006.  This allows us to compare the behaviour of service, the experience of a rider who tries to use the system for a regular work trip.

On King, the TTC claims to run a two-minute headway and this is true, but only for part of the line and only in the morning rush.  For roughly an hour, from 7:30 to 8:30 am, there is a “wave” of two-minute service scheduled to come east through Parkdale and the Bathurst/Niagara district.  Riders in these neighbourhoods complain of erratic service and overcrowded cars.  What is actually going on?

There are four charts linked from this post representing four locations on the western part of the line.

  • Jameson Eastbound displays the departure times eastbound from Jameson Ave. in Parkdale.
  • Crawford Eastbound displays the same wave of service as the previous chart, but as seen slightly to the east.  This includes cars short-turned at Dufferin.
  • Marion Southbound displays the departure times southbound from Marion, one stop north of Queen, giving a view of the service as it is seen inbound on Roncesvalles.
  • Marion Northbound displays the departure times northbound from Marion to show the condition of the service as it heads north to Dundas West on Roncesvalles.

Notable points in this data are:

  • There is a wide variation in departure times compared to the schedule with ±3 minutes quite common, although some runs have even wider variations.
  • Some runs do not operate on each day causing gaps in the scheduled 2-minute headway.
  • The schedules operated on December 22, 27, 28 and 29 are different from those on other weekdays and should be viewed separately although they are presented together with the other weekdays.

(When reading this chart, note that at the edges (around 7:00 am and 9:00 am), some runs appear just after 7:00 on some days and just before 9:00 on others.  This is a side effect of the data selection process where the last appearance of a car within the two hour period is the one reported.  If this varies across the 9:00 am selection boundary, an earlier trip will be shown on some days, and a later trip on others.  My focus is on the period from 7:30 to 8:30 and, therefore, this is not an issue here.)

Missing runs present a major problem in the amount of service actually provided during the two-minute wave.  I suspect that these are unmanned runs that are crewed from the spare board or by volunteers, or that there was no equipment available to operate them.  A related problem is that these runs may not enter service as close to schedule as crewed runs, and therefore they do not appear reliably at the same time each day and/or may be short-turned producing gaps in the through service.

Of particular interest is run 67 which was either off-schedule, short-turned or not operated on 11 of the 15 regular weekdays between December 1 and 21.  This run is scheduled to come eastbound through Parkdale just after 8:00 am at the height of the inbound demand.  This predictably causes disruption to the following runs and a chronic irregularity of service.

The Marion northbound chart shows that the service is bunched and possibly out of order before it even begins the heavy inbound trip.  Many cars appear reliably at the same time, but some vary quite a bit and some are often missing either because they were not operated, or because they short-turned before reaching this part of the line.  (Be sure to ignore the lines zigzagging across the chart for runs that sometimes manage a second appearance before 9:00 am.)

The Marion southbound chart is, intriguingly, more regular than the northbound one indicating that cars tended to leave Dundas West fairly predictably and on a regular headway.  This pattern breaks down later in the peak due both to arrival of cars that have traveled across the city, but also to the fact that many runs whose on-time entry to service and regular appearance at Dundas West for this trip are dubious.  A delay involving run 5 at Dundas West on December 11 is quite clear in the chart.

The Crawford eastbound chart shows the same service wave as the Jameson chart, but the reliability of times is not quite as good as at Jameson.  This shows the effect of heavy loading and also the cumulative effect of any gaps in service on arrival times of following vehicles.

Following the Jameson chart, there is a table describing the behaviour of each run in detail.  Note that the scheduled times are rough estimates based on the actual times through the month.

What is clear from these charts is that irregular service on this part of the route during the am peak is chronic, and that it is mainly caused not by congestion, but by poor management of the service.  Many cars enter service late or not at all, and they may be short turned without serving the very times and neighbourhoods for which they are scehduled.

Although vehicle assignments are a subject for a later post, an important point here is that most of the extra runs in the “wave” do not reliably operate with the larger ALRVs, and the capacity one would expect from the scheduled service is, in fact, not present on the line even without the unpredictable operation.

504 King Eastbound at Jameson

504 King Eastbound at Crawford

504 King Southbound at Marion

504 King Northbound at Marion

2 thoughts on “Analysis of 504 King: Part IX – Headway Reliability and the Two Minute Wave

  1. These results aren’t surprising for anyone who waits for the streetcar at the same stop and about the same time for more than a few days. On the 501 by Long Branch, there are morning runs which are reliable and other runs that are not. The clever rider will try to catch the reliable runs (which is comparatively easy because they show up the same time every day).

    Steve: I will be turning my attention to Queen soon, and have seen exactly the pattern you describe on Long Branch. I believe it has a similar cause — cars entering service randomly possibly due to the way that the schedules are constructed and staffed.


  2. Another thing I thought of: is there a board change in your data? It would be interesting to see if reliable runs turned unreliable, or vice-versa, after a board change.

    Of course, the TTC should not be relying on exceptional operators to keep service running smoothly. Operator ability and experience varies; the line management has to be able to deal with these variations.

    Steve: The data I have are for the month of December 2006, and the Board Period changed in mid-November. Special schedules are in operation for the week between Christmas and New Year’s, but this is not the same as seeing the effect of a new period, and new crews, on the stability of service for the first few days. Also, of course, Christmas Week is very, ver quiet for traffic and riding.


Comments are closed.