Updated February 26 at 11:15 pm: The legend on the service chart has been corrected to reflect the actual location of the timepoints.
The 512 St. Clair route resumed streetcar operation from St. Clair West Station to Earlscourt Loop on Sunday, December 20, 2009. This was the test everyone had waited so long to see — would the right-of-way on the busiest part of St. Clair Avenue make a difference, and how would the line operate.
As we know from complaints that poured in to the TTC and to local Councillors, things did not go well during the first month. New scheduled were introduced in mid-February, and the decline in complaints indicates that riders are much happier. Nonetheless, it is worthwhile looking at January 2010 to see where the problems lay.
This series of posts uses TTC vehicle monitoring data to review the operation of the St. Clair route for that first month. I have requested the February data as well so that a “before and after” comparison will be possible with the new schedules.
This article is an introduction and, for those who have read these analyses before, a refresher on the methodology I have used and the format of the data presentation. In future articles, I will review the month as a whole, but here the data is from one day, January 1, to set the stage.
Historical Comparison — April 2007
In a previous series, I reviewed the service in April 2007 just before the western half of the line closed for construction. This was the last period when streetcars operated through to Keele Street.
Part 2: Headways and Link Times
These articles and accompanying charts give a comparison view of the line as it was before the right-of-way existed west of Vaughan Rd.
Technical Stuff
For those readers who wish to know the details of how the raw TTC data were converted to the charts used in these articles, please read this article.
Chart of All-Day Service
This chart shows the operation for New Year’s Day as a graphic timetable. Various effects are visible on the chart. (In this discussion I refer to cars by colour. The actual vehicle numbers are available in the underlying data.)
Slanted lines up to the right are cars moving westbound while those moving down to the right are eastbound.
Locations where the lines are horizontal show a car that is stationary. Small “notches” in the lines show places where a car stayed in one place long enough to report the same location over several 20-second intervals. These notches tend to appear in roughly the same vertical position throughout the day and show places on the route where cars are routinely delayed either by busy stops or unco-operative traffic signals.
At the bottom of the chart (St. Clair Station Loop), there are two separate locations where cars tend to sit. The higher location is the offloading platform and the lower one (usually zero) is the loading platform. Operators typically take their breaks at the unloading platform.
At the top of the chart (Earlscourt Loop), a similar pattern can be seen depending on where, exactly, a car lays over in the loop. These small scale changes show how accurately the GPS units are reporting vehicle positions.
The length of layovers operators are getting at each end of the line is clearly visible on the chart, and varies through the day.
At about 8:10 am, there is a “wiggle” in one car’s chart (brown line) near Lansdowne. This is a GPS error which is small enough that it is not rejected as being out of range. A similar wiggle occurs for this car at about 20:10.
Starting about 6:30 am, there is a car (blue line) that consistently has a wide gap in front of it. Shortly after 11:00 am, this car goes out of service at St. Clair West. It reappears at about 13:50. This could be either an equipment problem or a missing operator. At around 14:30, this car is short-turned eastbound at St. Clair West. Its gradual move westward within the loop is a side effect of the data mapping. Probably what happened is that it reported no data for some time, and then reappeared slightly west of its last known location. The mapping routine interpolated the missing time slots.
Starting at about 16:00, pairs of cars run together for extended periods. Note particularly the yellow and lighter blue cars which stay close together for a round trip, after which yellow goes out of service. At 19:00, yellow runs out of service. However, comparison to other Sunday/Holiday days and to the TTC schedule summary shows that there should still be 5 cars in service, not 4. This is probably a missing operator.
Midway through the evening, around 21:00, the pink car runs into troubles westbound from Oakwood. By the time it leaves Earlscourt Loop eastbound, brown has caught up, and pink gets a short turn at St. Clair West. Things get partly sorted out, but pink continues to carry a wider than usual headway. This problem gets worse at 22:45 when brown runs out of service. Looking at other Sundays and at the Schedule Summary, service is supposed to drop from 5 to 4 cars mid evening. However, one car (yellow) is already missing and only three cars stay in service. They run on their schedules rather than being spaced to a headway causing pink to carry a double headway for the rest of the evening.
Notable by its absence is any service west of Lansdowne. In the data supplied by the TTC, many days either have no shuttle buses at all, or appear to have some missing. I suspect that this may be due to the absence of GPS units on all buses assigned to the route. Bus data are included in the charts for what they are worth, but they are omitted from the headway and link time analyses.
New Year’s Day was comparatively uneventful, but Sundays through the month are quite a different story as we will see in a future article.
Headways
2010.01.01 Headways Westbound
2010.01.01 Headways Eastbound
The charts of headways show times for cars at Yonge, Tweedsmuir (just east of St. Clair West Station), Bathurst, Oakwood, Dufferin and Lansdowne. Times are not shown for the two terminal loops, but rather for nearby points. This avoids problems with small differences in the way each operator, even each trip, behaves right in the loop.
Clearly visible through the day is an erratic headway with wide variations to the point that the gap between cars can be as much as double the scheduled service. The evening periods show the effect of the one missing car when no effort was made to re-space the service and there were very long gaps.
From the service chart, we know that there was little short turning, and that most cars had at least a short layover at the terminals. The uneven headway is typical of what showed up in analysis of other routes — vehicles do not leave the terminals on a regular spacing and this causes uneven headways across the entire line. New Year’s Day had no congestion, no weather problems, no surge loads, and the service shows what happens with laissez-faire line management.
Link Times
2010.01.01 Links Westbound
2010.01.01 Links Eastbound
The link charts show the time required to travel between pairs of timepoints on the line. The charts are about as boring as they come, with the trend lines staying almost flat for every location, and very little scatter of data values around those lines. This shows that, at least on New Year’s Day, the travel time between points was quite consistent.
The last page of the charts shows the running time from Yonge to Lansdowne. The trend line is quite flat at about 18 minutes, although the data are scattered either side of it by a range of about 3 minutes either way.
On a “best case” day, running times can vary by a range of 6 minutes on a base of 18, quite a substantial margin. Some of this will be a question of individual operator style on a route with which many were probably unfamiliar. I will return to this question later in the analysis.
Steve Munro 
I believe that a major part of the problem was that the supervisors were short turning so much of the service to get it back on time that there were so few cars running the full route that they got soaked and became later. They would have been much better of just running a regular headway and making sure the operators got their break, which should have be at St. Clair West and not at at Earlscourt. I don’t know if this was the common practice but I did seem some breaks occurring there.
Steve: Although New Year’s Day has few short turns (mainly because the time cars took to handle the route was roughly what was on the schedule), other days in January were not so good as you will see in future articles.
LikeLike
These are very interesting graphs. For the Day Service Chart, I think that St Clair West is supposed to be above 200 rather than 180. I would have expected there to be more delays at Avenue Road.
How do you read the other two graphs?
Steve: Thanks for pointing that out. The legend is incorrect. Bathurst Street is at about 220, and the loop is around 200. Some cars’ layovers within the loop appear to linger either just above or just below the 200 line before returning whence they came. The reason for this is that the car stops reporting its location for a period and then reappears when it re-enters service. The mapping routine links the two points in the chart, and the slight movement of the car between the two known points creates a nearly horizontal line.
The headway and link time charts work in similar ways. Each point represents either the headway (time between two cars at the same point) or link time (time between two points for one car). Excel draws a line linking the points as well as a polynomial trend line which tends to smooth out the short term variations.
For headway charts, the constant up-and-down represents short headways followed by long one, in other words, cars running in bunches. If you step through charts for various points along the line, you will see a quasi-animated view where, generally speaking, the pattern just shifts along in time. This shows that the pattern of headways leaving the terminal is unaltered in the trip across the route, and no attempt is being mad by the operators or by supervisors to space the service evenly. What’s more, the amount of dispersion in the headways makes for extremely ragged service with some passengers having to wait up to twice as long as the advertised schedule even though there is no service missing from the route. The car in the large gap will tend to be the most heavily loaded, and the “typical” experience will be infrequent, crowded service. The empty space on and the short wait for the second and third cars in a pack isn’t seen by many customers.
For a headway chart, the ideal is to see points closely grouped around the trend line which, itself, will tend to match the scheduled headway provided that little service is short turned. Nothing prevents the TTC from dispatching vehicles on a regular headway from terminals except the will to do this. It is an endemic problem on every route I have studied.
For the link time charts, there will always be a small variation in times especially for points that are close together. Simply missing a traffic light adds close to a minute. On routes that have traffic congestion and/or variations in demand (and hence stop dwell time), this will show up as a rise and fall of the trend line. If the data points are widely scattered around this line, it means that the running time is quite unpredictable over this section of the route. If they fall in a narrow band, then most cars will traverse this section in roughly the same time, give or take a few minutes. This is important when comparing scheduled with actual running times.
The last chart in each set of link times gives the full trip from Yonge Street to just east of Lansdowne, and shows the cumulative effect of delays along the line. In the case of New Year’s Day, there is little change through the main part of the day showing consistent running times.
As for delays at Avenue Road, these show up in the charts for weekday operations. In the scale of these charts, Avenue Road is at 82. Even on New Year’s Day, you can see some notches at this location. Note that the eastbound notch (brown car at 10:35) is at a slightly different location (85) from the same car on its westbound trip (at 79). This shows the accuracy of the GPS data which resolves holds on opposite sides of the intersection. This is a traffic signal hold, not a passenger boarding delay.
LikeLike
When I see things like this (and I’ve seen things like this in person) it aggravates me. Two streetcars following one another.
I want to get angry at the driver – but – I don’t know if this is even a rule. Regardless it should be;
If you are right behind the next car, you should wait for a full cycle of the traffic light (red, green, red, then go) I know some buses do this when ahead of schedule, and on a RoW there is no reason a streetcar could not do the same. I know this would “slow” the route down, but as it stands we have short turns coming out right in front or behind another car. I think that this needs to be done to space out the vehicles at least a little bit better. I personally have seen this happen outside St.Clair West (I work in the area). Usually when I see two go in, I see two come out. At least the second car could/should wait at St.Clair west. They can then also do this at Keele (Lansdowne for now) and Yonge.
If drivers were told to / did this, and route supervisors stepped up and made short turns when needed, I think this would fix a lot of the problems on this route.
Am I on to something or am I way off?
LikeLike
A simple question with maybe a not so simple answer: How could the TTC solve the problem of streetcar pacing?
Also, should we expect more of the same from the Transit City LRTs?
Streetcars are like wolves. They’re hard to find, and they roam in packs.
Steve: Especially for Transit City, there is a good analogy on the subway where the signal system is used to enforce train spacing. This system is not ideal because it is schedule driven rather than headway driven, but in practice the subways tend to be operated more from the viewpoint of keeping operators on time without short turning trains. Similar operating strategies will be needed for Transit City.
On St. Clair, there are two related problems. First, the January schedules did not have enough running time to cope with actual conditions during some periods. Part of this was a problem with operator unfamiliarity on the route, part was reticence to drive at speed for fear of left-turning motorists, and part was due to improperly timed traffic signals. However, a big effect visible throughout the week 1 charts (which I have just published) is the absence of headway management. Short turning appears to mainly be in aid of keeping operators on time, while bunches of cars are allowed to form and remain together for extended periods. This requires a major rethink by the TTC in their approach to line management for surface routes.
LikeLike