If we try very hard, we can remember a time when the St. Clair line was not under construction. With last year’s project still unfinished, and this year’s barely underway, it will be a while before we see streetcars running all the way from Yonge to Keele.
For a brief period in 2007, the line was back in one piece, and as a “before” comparator of operating conditions, I asked the TTC for the vehicle monitoring data (CIS) for the month of April. We won’t be to an “after” condition until early in 2009 when this year’s project is completed and only the stretch from Caledonia westward remains to be rebuilt.
Rather than wait, I decided to spin through the April 2007 data to see what they revealed. What I found was disquieting especially considering all the hooplah around the construction of a dedicated right-of-way.
Although congestion does affect the line in some places and at some times, the overwhelming source of headway variation is the time spent sitting at the terminals and, to a lesser degree, at St. Clair West Station. If you have read my analyses of routes like Queen and King, you know what real congestion looks like on the charts with large changes in running times through segments of routes. None of this shows up in the St. Clair data.
First, I will review Easter Sunday, 2007, to show the basic data and review the way these analyses work for readers who are new to this. Next, I will look at the overall behaviour of the route for the entire month, and finally I will turn to specific days’ operation with events of interest.
April 8 started off with some overnight snow, but by dawn was “mostly cloudy” or “cloudy” with temperatures around freezing for the rest of the day and evening. Weather did not play a role in transit operations on that date.
The service chart shows the familiar zig-zag pattern of cars travelling back and forth across the route. St. Clair Station at the east end of the route is at the bottom, and Gunns Loop at the west is at the top. The scale is roughly 100 units to 1 km. Time progresses from left to right across multiple pages, and times after midnight count upward from 24 rather than wrapping around to 0.
Horizontal lines indicate two types of event.
- In the case of lines at value “195”, this is a car entering or leaving service. “200” is Bathurst St., and all points between Roncesvalles Carhouse and St. Clair are assigned a dummy value of “195”. This lets us see cars entering or leaving service, but the lines don’t interfere with headway and link time calculations or confuse graphs of car movement on the route itself.
- In all other cases, notably at values “0” (St. Clair Station), “180” (St. Clair West Station), “650” (Gunns Loop), these are layovers and stop service times. Note that due to location resolution problems with CIS data, I have charted all reported data for each location (loop entrance, various places in the loop, loop exit) as one value. Therefore, the times a car is at the “same” location include both layover (if any) and the time needed to serve the stop(s).
Segments that are nearly horizontal appear at “460” (Lansdowne). These are caused by longer than average times through that intersection which have two explanations:
- There is a CIS “signpost” just east of Lansdowne, and this is a location where the calculated location of the car (from odometer data) gets reset to the actual location.
- Lansdowne is the point where crew changes occur, and cars may sit for a few minutes here.
The CIS data for this date are fairly clean, and there is only one serious case where they appear to lose track of where a car actually was. Around 0900, there is a car whose trace makes U-turns at odd locations before settling into a westbound trip at about 0910.
There is no remarkable event throughout the day. Occasionally, cars may close up and run as pairs, but for the most part, the service is fairly well-behaved. How this translates into actual headways and travel times between points shows up in the next charts.
Headways are displayed for eight points along the route: St. Clair Station, Yonge Street, Tweedsmuir (east portal of St. Clair West Station), Bathurst (west portal), Dufferin, Caledonia, Keele and Gunns Loop.
Westbound headways at St. Clair Station are suspect because of data resolution problems at this location. There are few signpost reports here until about April 22 suggesting that the signpost was out of order. The values settle down on the second page, at Yonge Street.
Note that these headways are within a six-minute wide band for most of the day, and this meets the TTC’s criterion for “on time performance”. However, the service actually experienced by customers (along with any associated crowding) is of cars whose headways can vary by a factor of 2. This shows how the three-minute rule can mask what riders will perceive as ill-behaved service.
As we move west along the line, the peaks and valleys start to spread out. This is no surprise as the cars carrying short headways will spend less time at stops. By the time we reach the west end of the route, some cars are running in pairs.
Eastbound, the service leaves Gunns Loop within a six minute band of the headway, but as it moves east, the headways spread out and by the time we reach Bathurst, the band is almost 10 minutes wide. This continues east to Yonge Street.
The vital point here is that if the service is not closely regulated leaving the terminals, it will become even less reliable as it moves across the city. This is a short route compared to most, and yet we see the same effects of irregular headways through routine operating practices that have shown up on other, longer routes.
The link time charts show the running time between points on the route. If these are reasonably consistent, then we know that there are few random delays that affect the service. If they drift up and down at various times of the day, we know that there is a buildup and dropoff of congestion (or possibly of heavy passenger demand and stop service time).
The link times from St. Clair Station to Yonge Street are generally quite short. As I mentioned earlier, there were problems with the CIS signpost at St. Clair Station that caused departure times to be unreliable in some cases, and the spikes in this graph are from cars that reported leaving before they actually did. This caused the time to reach Yonge to be overstated.
From Yonge to Tweedsmuir, the running times lie in a band from 4 to 6 minutes for much of the day. The higher times before 9 am are a side-effect of cars sitting westbound at Yonge rather than in the station for early trips when the subway is closed. These layovers make the running time for this segment appear longer than it really was. This section of the route is entirely private right-of-way.
From Tweedsmuir to Bathurst, the times are consistent, although you can see two effects. One is the much lower time needed when the subway is not open and cars run straight through, and the other is the considerable time used to serve the station itself when it is open.
From Bathurst to Dufferin, the times lie in a band from 6 to 10 minutes and there is little sign of variation in speed over the day.
From Dufferin to Caledonia, the situation is similar. A few high points are probably crew changes westbound at Lansdowne. The final two segments of the route are similarly uniform over the day.
For the eastbound trip, some variation in running time shows up between Dufferin and Bathurst, mainly before noon. From Bathurst to Tweedsmuir, we see the effect of the subway opening at 0900 on running times, although eastbound cars appear to take less time going through the station than westbound ones. This may also be an artifact of the CIS data. From Tweedsmuir to Yonge eastbound, times are consistent although there is some variation late in the afternoon.
What this shows us is that there is not much change in running times over any of the route segments as we would hope to see on a quiet holiday. Almost none of the headway variation can be put down to “congestion”, but instead is simply the usual service permitted by the TTC’s three-minute rule for “on time” performance.
In the next article, I will look at the full month’s data and examine the route’s behaviour under a variety of conditions.