When I started the analysis of the TTC’s vehicle monitoring data some years ago, my great hope was to be able to analyze behaviour at stops and intersections. Alas, the quality of data on the “old” version of the TTC’s system was utterly incapable of this type of use for two reasons:
- Although vehicle position was polled every 20 seconds, vehicle location was resolved to a nearby intersection, and reported only sporadically. This made fine grained location analysis impossible.
- A location, such as a major stop, might not be reported at all for a vehicle’s trip, only the vehicle’s appearance at two nearby locations on either side of the point of interest.
With the new GPS-based data, position information is available for almost all vehicles every 20 seconds. Anything that causes a vehicle to stop for some amount of time will be registered by a failure to change location in two or more succeeding data points. (I say “almost” because some GPS units are a tad unreliable, and they report rogue data points causing missing samples in otherwise clean data.)
A great debate about St. Clair (and other routes with the same arrangement for left turns and farside stops) focuses on the following sequence of events:
- Streetcar approaches an intersection and is caught and held by a red signal.
- Streetcar waits for the red phase, and the following left-only phase to complete.
- Streetcar moves through the intersection, serves a farside stop, and then departs.
In some cases, there is no stop associated with a traffic signal, but streetcar service may be held by it in any event. In theory, approaching streetcars are supposed to extend their own green time, or to shorten the cross-street’s red, but empirical observation suggests that this does not always happen.
This article reviews several intersections on St. Clair to determine how transit-friendly the traffic signals actually are, and whether any change in their behaviour was visible over the month of January 2010 while many aspects of the line’s operation were still being adjusted.
An obvious and troubling observation from these data is that in general, streetcars spend more time waiting for green lights than they do serving the farside stops at the same locations. This shows the frustration and inefficiency from an arrangement where cars must stop twice because they do not get the so-called transit priority they are supposed to receive at major, and even some minor, locations.
Readers familiar with these analyses will know that one element of them is “link time”, the time taken to get from one point to another on a route. To review intersections which included stops, I created three time points quite close together which would define two links.
- Just east of a farside eastbound stop
- Middle of the intersection
- Just west of a farside westbound stop
In cases such as Oakwood where a nearside stop exists (e.g. Oakwood westbound), the time point is just west of the intersection.
This creates two links which viewed from the direction of travel are:
- Near an intersection to its midpoint. If a streetcar has a clear signal, it will traverse this link quickly. However, if it has a red signal, it will not reach the middle of the intersection until after it is able to proceed on its next green phase.
- Midpoint of an intersection to the point just beyond a farside stop. This link measures the time needed to come to rest at a stop, serve passengers, and pull away. If nobody wants to use the stop, the link time will be quite short.
For intersections which do not have associated stops (e.g. Via Italia and Alberta), only one link, defined from just west to just east of the intersection is needed. If a streetcar is not held by the traffic signal, it will cross this link in a very short time. Otherwise, the link time will demonstrate the traffic signal delay.
I have included one intersection from the original part of the 512 route, Avenue Road, as this shows up both in the service charts as a delay point, and has been mentioned by some comments to earlier posts.
Limitations of the Data
For the most part, there is a report of a vehicle’s location every 20 seconds, and the GPS values are usually very accurate. (Those of you who have watched the animated version of these data will know that the location of vehicles on the route is quite well-behaved.) All the same, the 20-second cycle requires certain assumptions.
- If two successive reports show a vehicle on either side of a timepoint, it is assumed to be at the timepoint haldway between the two data. For example, if at 12:00:00 a car has not yet reached a timepoint, but at 12:00:20 it has gone past it, it is assumed to be at the timepoint at 12:00:10.
- For closely spaced timepoints such as are used here, it is possible for a car to pass several of them in one interval. In these cases, the link times will be calculated as zero because the car was “at” the timepoints all at the same instant within the accuracy of the data. I prefer to use “zero” which is recognizably from this type of situation rather than interpolating values over a 20-second range.
Avenue Road & St. Clair
The “Nearside” charts show the time taken from a point just beyond the end of the farside stop in the opposite direction to the middle of the intersection. Westbound, the trend line sits at about 30 seconds although the data are scattered up to the 2 minute line. Eastbound, the trend lines are lower, at about 20 seconds, but several data points still lie in the 1 to 2 minute band. Clearly this is an intersection where the streetcars are often held in the interest of flow on the north-south street.
The “Farside” charts show the time from the middle of the intersection to the far end of the stop, where the car would have served its passengers and begun to move off. Note that the westbound values show PM peak effects probably from longer stop service time (there is also some extension of times on Saturdays at midday). A comparable effect is not seen eastbound.
Bathurst to Vaughan
Anyone who rides the 512 knows that it spends an eternity getting through the intersections at Bathurst and at Vaughan. Again, this is a clear sign that the traffic signals make little or no provision for “transit priority”.
Westbound approaching Bathurst, the trend line lies around the 1-minute line, but there is a great deal of scatter in the data. This shows that a delay westbound here is quite likely, and some of these will be well over a minute long. Some of the particularly long delays may be due to conversations with Route Supervisors (a comparable effect is visible eastbound farside).
Crossing Bathurst and serving the carstop shows better-behaved data, and the values are fairly consistent throughout the month.
Approaching Vaughan, the trend lines lie at the 30-second level, but many data points lie above 1 minute especially on weekdays. The signal westbound at Vaughan often catches streetcars. Crossing Vaughan and serving the carstop shows better times, although they are more scattered on weekends.
Eastbound approaching Vaughan shows a similar pattern to westbound with many times in the 1-2 minute range, while the time needed to handle the stop itself are shorter and less scattered.
Eastbound approaching Bathurst shows the familiar story with many times between 60 and 90 seconds, but the real surprise is on the east side of the intersection. This is not a stop where very long dwell times should be common, and I can only conclude that it is a place where cars are held up for chats with a Route Supervisor.
Alberta is one block west of the Winona Drive Stop, and had a traffic signal added as part of the St. Clair project. This signal spends a great deal of time whenever I have seen it green for traffic into a shopping mall. The signal is green, but there is often little or no traffic, and the signal delays westbound streetcars.
This is a good example of the problem with closely spaced signals and the inability of operators to request a green wave when they need it.
The trendline for Alberta westbound lies at about 30 seconds for weeks 1-3 of the month, but drops to 20 seconds for week 4 suggesting that some adjustments were made here. However, there is still a fair number of data points at the 1-minute line indicating that this signal interferes with streetcars quite regularly.
From riding the line, the typical pattern is:
- Streetcar pulls westbound through Winona drive on a green phase shared with other traffic.
- While the streetcar is serving the westbound stop, the signal at Alberta, probably timed to fit in with the gap that will come west from Winona (or a wave eastbound from Oakwood), turns red in front of the streetcar just as it is leaving the stop.
- The streetcar waits for the red phase plus the eastbound left turn (even when there is no traffic waiting in the turn queue).
By contrast, the eastbound times at Alberta indicate that holds for the signal are quite rare.
The westbound stop at Oakwood is nearside, and times for this link show the combined effect of stop service and traffic signal delays. A small amount of peak period effect can be seen on weekdays, while Saturdays tend to have longer times at midday. The farside times are very short as there is no stop beyond the middle of the intersection.
Eastbound, there are short delays at the traffic signal, but generally cars are held here only briefly. Most of the delay here is consumed by stop service time. This is the transfer point between eastbound St. Clair shuttle buses and the streetcars.
Westbound cars on Dufferin spend longer waiting for traffic signals than serving the farside stop. This is primarily an offloading stop because cars are near the end of the line, and that activity is faster than boarding passengers eastbound. Eastbound cars on Dufferin spend roughly similar times waiting for the traffic signal and serving the stop.
The length of some signal delays show that this intersection is organized to avoid interference with the north-south flow on Dufferin.
Via Italia has a traffic signal between the stops at Dufferin and Earlscourt. The trend lines both ways lie in the 15-20 second range, but some values up to 1 minute are seen. Westbound times are slightly longer.
Although St. Clair is a “transit priority” route, many of its traffic signals do not give preferential treatment to transit vehicles. This is quite noticeable where the streetcar route crosses major north-south streets.
The farside stops compound rather than reduce delays at such locations. The stop configuration is designed to allow room for a nearside left turn bay, and this brings a “catch 22” situation for transit rights-of-way as they are implemented in Toronto.
Unfortunately, the “before” data for St. Clair comes from the “old” monitoring system where location resolution for vehicles does not allow this type of fine grained analysis.