The St. Clair streetcar route has seen its share of political battles over the years. Back in 1972, it was the heart of the fight to save the streetcar system from a plan that would have dismantled all routes by 1980 when, wait for it, the Queen subway would have opened. As a first step, trolley buses (remember them?) displaced from the North Toronto’s route 97 Yonge were to take over St. Clair with buses replacing streetcars on a 1:1 basis, a huge cut in the line’s capacity.
An ironic point about this plan shows how riding patterns can shift depending on other aspects of the network. Before the Bloor-Danforth subway opened in 1966, St. Clair had very frequent service (every 1’00” between Oakwood and Yonge including the Rogers Road cars), and the line carried many people east to the Yonge subway. Vaughan Loop was a major transfer point to the Bathurst cars which ran into downtown via Adelaide Street, returning on King.
After BD started running, many riders shifted to north-south routes to reach the new subway line and its connection to a route downtown via the uncrowded University subway. Demand and service on St. Clair declined. Years later, with the opening of the Spadina subway, many riders shifted back to the streetcar because it provided a direct link to the University line with one less transfer. More recently, the population along St. Clair is growing adding to demand on the line.
Unlike most streetcar routes in Toronto, 512 St. Clair operates on a street with more than four lanes. The TTC proposed conversion to a reserved right-of-way in the early 2000’s and this was approved by Council in 2005. The actual construction took forever thanks both to a legal attempt to block construction and to fouled-up co-ordination between various agencies and meddling by local Councillors in the timing of work. (You can read the whole sad story on Transit Toronto’s website.)
By spring 2007, the first segment from Yonge to St. Clair West Station was completed, and streetcar service returned, briefly between St. Clair Station and Keele Street. Summer 2007 saw the launch of work on the western portion of the line, and it did not fully reopen until summer 2010.
Recently, the TTC has been messing around with the schedules on this route and adding supervision in an attempt to provide reliable service, a sad situation given that the route is entirely on reserved lanes. The evolution of schedules from 2007 to 2015 is intriguing, and speaks to the failure of what should be a showcase route.
The information in this table is organized with the impending March 29, 2015 changes at the left and progressively older schedules moving to the right. Only the January schedules are shown for 2011-2013 to save on space during a relatively quiet period for service changes on this route.
- April 2007: The line operated over its full length, but with a right-of-way only on the eastern leg.
- June 2010: Service resumes over the full route with shorter running times, particularly on weekends, and more frequent service than in most periods in 2007.
- January 2011: Service during some periods on Saturday and Sunday has been improved by 2011 to handle demand on the route.
- January 2012: Weekday midday service has improved over 2011, but there are no other changes.
- January 2013: Peak service has improved slightly, offset by some weekend service cuts.
- January 2014: A 2012 service cut on Sunday afternoons was partly restored during 2013.
- July 2014: Slightly wider peak service (typical for summer) with improvement in Saturday early evening and Sunday morning service.
- October 2014: Running times substantially increased in response to a large number of transit priority signals being out of order (13 along the route)
- March 2015: Running times reduced (but not all the way to July 2014 levels) in response to repair of most (9) of the non-working TSP locations, and experience from a higher level of route supervision implemented in fall 2014.
It is worth noting that during almost all schedule periods, the allocated running times in October 2014 were equal to or longer than those used in April 2007 when most of the route ran in mixed traffic. Some of this was due to added recovery time (weekday schedules), and some to added travel time. In effect, the benefit of the right-of-way on scheduled speed was almost completely undone. This is partly, but not completely, corrected with the March 29, 2015 schedules, but there is still generally two minutes more running time for the route compared with most schedules from 2010 onward (presumably for the residual effect of non-working TSP locations).
A related problem on St. Clair has been irregular headways. The TTC’s stock response to complaints about this sort of thing is that “traffic congestion” is the root of all evil, and reliable service is impossible. In fact, as has been demonstrated by repeated analyses on this site, the real problem lies in uneven departures from terminals and from intermediate time points along the route. This cannot be explained by saying that operators are adjusting to known conditions because these irregular headways appear under all seasons, days of week and hours of the day.
In the fall of 2014, the TTC added route supervisors on St. Clair to act as dispatchers and regulate the service. This had some effect, but the level of on-street supervision cannot be afforded across the system. Indeed, there is no reason why dispatching on headways cannot be achieved centrally and at least in part automatically. This problem is not confined to streetcar routes, and it is a fundamental issue that TTC management must address particularly as the effects of larger vehicles and wider scheduled headways accentuate the problem on streetcar and articulated bus routes.
How has the actual service evolved over the years? For this we must turn to the TTC’s vehicle tracking data. As regular readers will know, I have been looking at routes on a selective basis since 2007. The discussion below includes data from April 2007 (pre-construction), July 2010 (full line re-opened) and September-November 2014 (pre/post implementation of longer running times and more aggressive supervision).
The purpose of this section is to establish a “before” behaviour against which post-construction operations can be compared.
The data here come from the era before the TTC implemented GPS on its vehicle tracking system. Instead, they used a set of location beacons scattered along routes together with mileage clocking from the vehicles to estimate a car’s location. This was subject to various errors including the effects of wheel slip and of unexpected short turns. Because the system assumed a car would behave as scheduled, the direction of travel was assumed to match that schedule.
The location data transmitted from vehicles was “signpost number x + distance” with no directional information. A car might be travelling opposite to its scheduled direction from a short turn, but the tracking system would not react to this until the car passed not one, but two signposts that confirmed the real location was somewhere other than the computed one. A car might be “tracked” as moving west when in fact it was going east. This produced no end of problems for the “CIS” (Communications & Information System) users including central route supervisors and operators of vehicles whose on time displays were meaningless.
The tracking data for St. Clair are reasonably well-behaved, but not at as granular a level as the GPS data available by the time of the 2010 and later operations discussed below.
The charts here show the headways westbound and eastbound respectively for Wednesday, April 4, 2007 as seen at various points along the route. Various typical behaviours are visible:
- AM peak headways leaving St. Clair Station stay in a tight band under 6 minutes because the scheduled headway is short (3′) and there is usually little to disrupt the service at this time of day.
- Headways at Caledonia and points west in the AM peak are wider because of the scheduled short turn at Earlscourt Loop.
- After the AM peak, headways become less reliable and this condition continues throughout the day. Note the many situations where a long headway is immediately followed by a value close to zero indicating cars travelling in pairs originating at the terminals.
To get a better sense of what is going on, here is a chart of vehicle locations by time.
This chart breaks the day into eight three-hour chunks and shows the movement of cars back and forth across the line. Because the source data come from the old signpost-based system, the locations of the cars are not reported as frequently as they are today with the GPS system (see later charts for examples), and so the lines tracking the cars are fairly smooth.
Time moves from left to right over the eight pages while vehicle positions move back and forth from west (top of the page) to east (bottom of the page)
Points to note here:
- Mainly on the second page (after 7:00 am), the short turning cars at Earlscourt Loop are clearly visible turning through the loop on a regular basis.
- At about 9:30 am, there is a delay eastbound at about Vaughan Road. Some cars disappear from view at Christie Street probably because they went out of service, while others have extended times between Christie and Bathurst as shown by the difference in the slope of the lines). There is a very large gap eastbound to Yonge between roughly 9:25 and 9:55 am (at Yonge). A parade of cars arrives eastbound just before 10:00 am. This parade travels back west more or less intact with only one of the cars short turning westbound at Lansdowne, and another eastbound at St. Clair West Station at roughly 10:20 and 10:55 respectively. The original pack of cars is not fully broken up until about 11:15 westbound from Yonge.
- Two of these quickly form into another pair eastbound from Keele at 11:50, and by the time they are westbound from St. Clair West at 12:45 the pack has four cars in it. This group continues for two full trips and is only disassembled by about 3:45 pm.
- However, by then a new pair has formed eastbound that eventually evolves into a parade of nine cars westbound at Bathurst just after 4:30 pm.
- It takes until about 8:00 pm for all of the pairing to settle down, if only because the scheduled headway is much wider.
Another way of looking at the headways is to chart all of a month’s operation into one collection.
These charts consolidate headways westbound at Yonge and eastbound at Keele for the month of April 2007 into one set of charts for each direction. The pages contain:
- Weekday data for each of four weeks.
- All weekday data plotted on one page.
- Saturday and Sunday data each on one page.
- Averages and standard deviations for weekday, Saturday and Sunday data
On the weekdays, the trend lines lie generally in the same part of the chart except for Friday, April 20 westbound from Yonge. What is notable about data for each weekday is the range of scatter of the data points on either side of the trend line (which roughly approximates the scheduled headway if almost all of the service is present). The TTC claims a goal of ±3 minutes to the scheduled headway, but it is clear many runs miss this target.
When all of the weekdays are combined, the data points are smeared over a band from 0-10 minutes with several points well above that value.
Weekends see much more chaotic service with many examples of very wide headways.
The statistics pages show the behaviour of the data reduced to averages (blue lines) and standard deviations (orange). The SD values are often close to the averages indicating periods when many cars are running as pairs, and the high SD values on weekends show that much of the service is operating over a wide range of headways. Ideally, the SD values should be low indicating values in a narrow band.
The situation for eastbound trips from Keele is similar to that shown for Yonge westbound.
The St. Clair line has just reopened from Yonge to Keele, and it operates entirely on reserved lanes. Also, by 2010, the CIS system is using GPS to track vehicles, and so their locations are much more precisely known. The data supplied by the TTC includes a record for each vehicle every 20 seconds rather than a record each time it passes an intersection. This permits a more finely-grained view of operations along the route.
As a sample of operations, I have chosen July 14, a Wednesday in a week uncomplicated by holidays, and a day without significant disruptions of the service. (In all of the following files, “Up” and “Down” refer to the direction of travel which, for 512 St. Clair, are “west” and “east” respectively.
The time-distance chart shows most trips operating the full length of the route. Of note:
- There is no major delay for the entire day.
- There are some short turns, mainly at Lansdowne. All service is scheduled to Keele unlike 2007 with scheduled short turns in the AM peak.
- Most trips have layovers at both ends of the line.
- GPS position resolution allows the chart to show short delays at intersections (both near and far side) related to signals and stop service. No location consistently appears as a source of delay.
The headway values are somewhat better for July 14, 2010 than on April 4, 2007, but mainly because there were no service disruptions to produce large gaps. What is common to both sets of charts is the regular pattern of very short headways followed by longer values indicating that cars tend to run in pairs. This behaviour originates at the terminals, not as a result of congestion encountered enroute.
These charts show the travel times over segments of the route, and they are best case numbers taken from an ideal weekday without major delays. Travel times are fairly consistent for all links and directions over the entire day. (One spike – westbound from Tweedsmuir to Bathurst at 10:25am – is caused by a car taking a layover at St. Clair West Station on the spare track. It then ran out of service.) What is particularly notable is that there is little sign of a “peak period” where running times deteriorate. Put another way, there is little “congestion”, that bugbear of TTC operations, and yet the headway reliability is still poor.
Comparing April 2007 to July 2010
An important difference between the “before” and “after” right-of-way measurements is the length of a trip from Yonge Street to Keele Street. (I have used this measure rather than attempting to decide when a vehicle actually “leaves” a terminal loop, a problem that is particularly challenging for 2007 with less fine-grained resolution of vehicle locations.)
Each of these files has a common layout. The first four or five pages (depending on the calendar) show weekday data followed by a page consolidating all weekdays on one sheet. This provides a day-to-day and week-to-week comparison of values so that “normal” behaviour is easily separated from unusual situations. Saturdays and Sunday/Holidays have their own page. (Street festivals and other events causing service replacement create oddly behaved charts on a few days, and these curves should be ignored.) Finally, the data are summarized with averages and standard deviations for each type of day.
For westbound service, comparing the weekday stats pages shows two important differences: the running times are lower by 3 to 5 minutes throughout the day, and the visible PM peak in 2007 is “shaved off” in 2010. Also notable is that Saturday and Sunday average times are reduced reflecting the congestion that plagued the route on weekends before it had a right-of-way. Nonetheless, the 2010 data points are dispersed over a band about 5 minutes wide showing the cumulative effect of small delays to some trips along the route. This is an important distinction for “average” versus “scheduled” running times, and the role of recovery time at terminals.
Eastbound data show a similar behaviour, although the running time saving is slightly less than westbound on weekdays. Again weekends benefit a great deal from the right-of-way.
By Fall 2014, service on St. Clair was still ragged, and running times had increased to the level of pre-right-of-way operations in 2007. This was due, according to the TTC, to the fact that 13 traffic signals along the route no longer provided transit priority – the adjustment of green times to favour transit, something that is particularly important for the farside stop placement on this route. As much as possible, streetcars should not have to stop twice – once for a red signal and again to serve passengers.
In mid-October, the TTC changed the schedules and instituted more aggressive supervision on St. Clair. Running times were increased, and supervisors at both terminals ensured that vehicles left on time on a regular headway.
The change in operations can be seen by comparing three days:
- Friday, October 10: The last weekday on the old schedules and with no supervision
- Tuesday, October 14: The first weekday of the new schedules and supervision
- Monday, October 20: The start of the second week
These days were generally free from major delays that would disrupt comparisons.
The westbound service leaving Yonge Street on October 10 shows the same pattern of unreliability we saw in earlier years’ data above. Plus ça change. Some wide headways westbound at Yonge (1200 to 1500) are shortened at Bathurst by short turns filling gaps at St. Clair West Station, but many other gaps travel untouched across the route. This can be seen easily by stepping from page to page of data and watching the pattern of long and short headways shift slightly in time at each point further west. This is a typical situation, and even with a completely private right-of-way, a short route, and a natural mid-point where service can be evenly spaced, it is clear that nobody is “minding the store”.
On the following Tuesday October 14 (Monday was Thanksgiving), the route is operating with new schedules and, more importantly, new supervision. The situation westbound at Yonge Street is completely changed with headways staying very close to each other during the peak periods. Middays are a bit more scattered and late evenings don’t show much change. By Monday October 20, things have improved again, notably for midday service. This shows that headways can be managed if someone makes the effort, but the story is not all as good as we might hope.
The service westbound at Oakwood, although better on October 20 than October 10, still shows a fair amount of scatter in the headways, and this appears to have developed between Bathurst and Oakwood (compare the behaviour of headways between the two adjacent charts). This suggests that problems with variations in travel time lie in this segment of the route. (I will return to that issue later.) Once the headways are back to the long-short pattern with pairs of cars running closely together, this is perpetuated westbound to Keele although the spikes are not as marked after Thanksgiving as before probably because the service started out in better shape from the eastern terminus and, to a slightly lesser extent, from St. Clair West Station.
For eastbound service there is a similar pattern starting at the west end of the line (which is at the end of the 10-page set of charts for each day). Service leaving Keele eastbound on October 10 has very wide gaps. Although these are filled in part by cars originating at Lansdowne (note the much lower peak headways from Lansdowne eastward), the values still swing widely with many close to zero (pairs of cars following each other).
On October 14, the service eastbound from Keele is quite well-behaved and the absence of short-turns at Lansdowne is evident because the chart at this location is so much like the one at Caledonia just to the west. This situation continues on October 20.
Again an evident problem is that the service leaves the terminal well-spaced, but by the time it works its way across the route, pairings of cars form placing actual headways beyond the TTC’s goal of ±3 minutes of scheduled values.
(In the preparation of this article, I reviewed data from September-November 2014. September behaves more or less like the sample day of October 10, and nothing would be added to this article by a detailed examination. Similarly, November behaves like the sample day of October 20. In other words, the change seems to have “stuck” at least while there is route supervision to enforce it.
The situation on weekends is not as pretty.
These charts are for Saturday and Sunday, October 25-26. Particularly on Saturday, the usual pattern of fluctuating departure times from St. Clair Station that was evident in the “before” data shown above is still very much here “after” the fact. The line is running with more running time, the common demand from operations staff, but the service is quite ragged. What is missing is the supervision to enforce the headways.
I am afraid that this does not say much for the general ethos of the operating personnel on this important route.
Another way to look at the request for more running time is to examine the time spent at terminals. For St. Clair I have examined the round trips from Yonge Street to St. Clair Station, as well as from Keele Street to Gunn’s Loop.
These charts are in the same format as the monthly link time charts shown earlier for the trips between Yonge and Keele, but they measure round trips at terminals to a nearby reference point.
The schedules changed on Sunday, October 12, and that is where we expect to see a difference in values.
At the east end, Yonge Street, for the first two weeks, the trend line is fairly straight with values in the 6-7 minute range. However, some data points lie near 3 minutes and these show the minimum necessary to speed into and out of the loop. Many data points lie considerably higher.
In weeks 3-5 the pattern changes completely with the shortest of times coming generally at the end of the peak periods. Even so, it is clear that cars have more time to spend at the eastern terminal.
On Saturdays there is much less change from the “before” (Oct. 4 & 11, red and yellow lines) to the “after” (Oct. 18 & 25, green and blue lines). The lower bound remains at about 3 minutes, but few values go above 10.
On Sundays, there is only one “before” day (Oct. 5, red) and the “after” days show generally longer times, but again, like Saturday, not as much variation as on the weekday operations. Thanksgiving Day (Oct. 13, green) stands apart from the Sundays showing the different conditions on this holiday.
The data at the west end, Gunn’s Loop, show some changes in the “after” conditions, but not as much as at Yonge. One might be tempted to think that, given the chance, operators would prefer to take their layovers at Yonge Street.
The additional running times, coupled with the already generous recovery times, have allowed most of the short-turning on St. Clair to be eliminated, but headway reliability still depends on supervision as we can see from weekend and evening operations.
Comparing Average Headways By Week and Time of Day
Another way to see the effect of the new supervisory regime on short turns is to look at the average headways at Yonge (east of the St. Clair West short turn), Bathurst (mid-route) and Caledonia (west of the Lansdowne short turn).
In these charts, the average headways are shown for each half hour of the day (separate horizontal lines) and for the months of April 2007, July 2010, and September-November 2014. Average headways are affected primarily by the scheduled values and the degree of short-turning. Although the operating environments in 2007, 2010 and early fall 2014 were different, the behaviour of the headway averages doesn’t change much until the implementation of new supervision in mid-October 2014.
At Yonge Street:
- Headways from 6:00 to 9:00 are fairly consistent. The jump at the end of April 2007 is caused by a major delay on one day in a short week. This shows how with a comparatively small number of data points, a single event can produce distortions that are not as evident when more data are consolidated. The blue line (6:00 to 6:30) stands higher than the others because scheduled service in the early part of the day is less frequent.
- Between 9:00 and 12:00, there is a marked change for the intervals beginning at 9:30, 10:00 and 10:30 indicating that more cars reached Yonge Street with the new schedules. This reflects a common problem on routes after the AM peak where much short turning occurs to recover from the peak period.
- Between 12:00 and 3:00, the averages become more consistent with the new schedules, and there is an improved headway in the 14:30 half hour.
- Between 3:00 and 6:00, the change after the new schedules are implemented is quite striking, and the improvement continues through into the mid-evening.
At Bathurst Street:
- There is much less change in average headways with the new schedules, although the averages through the PM peak are slightly better behaved.
At Caledonia Road:
- The data here show a similar pattern to the Yonge Street values with improved average headways for the weeks following the mid-October change in supervision. These changes are most notable after the two peak periods when under previous circumstances a great deal of short turning occurs to “fix” the effects of the peak.
Comparing Running Times Over The Route
This chart compares the weekday running times for five different months westbound from Yonge Street to Keele Street. The months are:
- April 2007: Before construction of the right-of-way west from Bathurst (red)
- July 2010: After re-opening of the line over its full length (yellow)
- September 2014: Current operations on the “pre change” schedules (green)
- October 2014: A mix of “pre/post change schedules (blue)
- November 2014: “Post change” schedules (purple)
Note the improvement for the line in July 2010, but that all of this has been lost again by fall 2014. The October schedule changes did not affect actual average running times which, for the three months in 2014, are almost identical.
The next change took place on March 29, 2015 with running times cut back in response to repaired signals. I will update this post when I have the April tracking data available. In advance of this, I will cull through the July 2010 data for comparison with 2007 and 2014 data to identify the route segments where the greatest time savings occurred in 2010, and where it was lost again by 2014. This will prove an interesting comparison to the April 2015 data where we should see the effect of all of those “fixed” transit priority signals.