Updated February 16, 2013 at 11:00 am: The graphs of average trip times have been revised to show the data in two ways — grouped by calendar week, and grouped by day of the week. The commentary has been updated to reflect this change.
Updated February 14, 2013 at 6:40 pm: Graphs showing the average values of trip times in hourly blocks have been added to this article. This consolidates information from the scatter diagrams showing values for individual trips, and simplifies comparison of averages to the scheduled times.
Recently, I wrote about the reliability of headways on the Queen route during the month of November 2011. That article was intended both to confirm the erratic nature of service well-known to riders of the line, and to put in context the TTC’s own “punctuality” claims and targets.
A common rejoinder from the TTC is that services operated in mixed traffic are subject to unpredictable delays causing not only bunching, but also such a variation in running times that short turns are an inevitable result. Operators, on the other hand, complain that scheduled running times are less than what is needed for a car to actually traverse the city, and late running is built into the timetable.
This article reviews data from the service operated during November 2011.
Before I consider the route as a whole, it’s worth looking at four major segments:
- Neville Loop to Yonge Street
- Yonge Street to Roncesvalles Avenue
- Roncesvalles Avenue to Humber Loop
- Humber Loop to Long Branch Loop
The following sets of charts show the time taken for streetcars between each pair of points. These are plotted by time of day, and grouped by week and type of day (Saturdays and Sundays are separate).
During November 2011, there was a scheduled diversion between Shaw and Roncesvalles via King Street due to Metrolinx bridge construction at Dufferin. New schedules went into effect on November 20, but the diversion did not actually begin until the evening of Friday, November 25.
The “time” of each measurement is taken at the beginning of the segment for each set of charts. A 90 minute trip from Neville to Long Branch that appears as a data point at “1500″ represents a car that did not actually reach Humber Loop until after 1600.
Neville Loop to Yonge Street
In each chart set, the first five pages correspond to the five calendar weeks in November. Each day has its own colour for the data plots and trend lines.
For the westbound service between Neville and Yonge, all of the weekday data lie within a band about 10 minutes wide, although the centre of this band moves throughout the day. The week-to-week consistency is shown on the sixth page where all of the weekday data are consolidated. With only a few exceptions, the running time for this segment is predictable within that range.
Eastbound from Yonge to Neville is similarly well-behaved and, if anything, the band holding most of the data is a bit narrower than on the westbound chart. What is quite noticeable, however, is that the pattern of values eastbound has only one peak — in the afternoon — while the westbound data show distinct AM and PM peaks.
On Saturdays (page 7 of the charts), the peak is at about 1500 (3:00 pm) and it is more pronounced eastbound than westbound. On Sundays (page 8), the peak westbound comes slightly earlier than eastbound. Generally speaking, the data points stay within a 10-minute band although a greater amount of scatter is visible on some days.
Yonge Street to Roncesvalles
For the purpose of this pair of charts, Triller Avenue stands in for Roncesvalles. Triller is a short distance east of Roncesvalles on Queen, and using this point filters out trips that divert via King Street. There are no data points for this segment after the PM peak on November 25 when service began diverting via King.
Weeks 1 and 2 have consistent values in both directions, but by week 3, the congestion caused by some construction activities is evident.
Week 4, just before the diversion begins, is fairly well-behaved.
Weekends are a different story, with running times on Queen West almost doubling during the peak period mid-afternoon in both directions on Saturdays.
Sundays do not have as pronounced a peak except on November 20, the Santa Claus Parade, which looks more like a Saturday. Higher-than-average Sunday values appears east of Yonge as well on this date.
What is quite striking in both sets of charts here is the absence of congestion caused by fair weather traffic and events both in The Beach and on Queen West. Any additional activity that adds to traffic (auto or pedestrian) and transit demand will produce longer running times and, probably, less reliable service over the day. A schedule that may be adequate for “off season” operation will fail under the additional load of summer traffic.
Roncesvalles Avenue to Humber Loop
The segment covered here extends from Triller Ave. (east of Roncesvalles) to Lake Shore Blvd. at Humber Loop. Running times in both directions are quite consistent showing no sign of peak period effects, but despite the short distance they spread out over a band 5-6 minutes wide. Looking at the details, much of this variation occurs around the Roncesvalles & Queen intersection which has a long cycle time, and which is the west end crew change point.
Because this is a short segment and the trip from Roncesvalles west to Long Branch is rarely obstructed, any time lost here is not made up enroute and this contributes to widening the range of running times over the full route to Long Branch (see below).
As in the previous “Yonge to Roncesvalles” section, no data are shown after the PM peak on November 20 when service began to divert via King Street as no cars operated through the Triller & Queen reference point.
Humber Loop to Long Branch Loop
In these charts, “Humber Loop” is actually a point on the Lake Shore Blvd. just outside of the loop, and “Long Branch Loop” is at the loop’s entrance. These locations filter out any variation caused by layovers within the loops.
Westbound values on all days are consistently within a band of under 10 minutes with only slightly longer running times on weekdays as compared to weekends. The effects of the AM and PM peaks are quite small. Weekends show a slight peak in the afternoons.
Eastbound values are similar to westbound, although the effect of the peak periods is more visible for inbound trips where loading delays will have an effect.
There are a few anomalies in the trend lines notably on the morning of November 1 when streetcar service west of Kipling Loop was replaced by a bus shuttle and, therefore, there is no data for a “Humber to Long Branch” trip. This shuttle actually remained in operation until after the PM peak intermixed with streetcar service to downtown.
Full Route Westbound
Note that the charts in this section have a different scale to accommodate the larger values involved, and that the bottom of the charts is not at zero.
Although the Queen East and Queen West charts above both showed data over a band of roughly 10 minutes, these values did not combine to produce, generally speaking, a wider band when trips over the the full route from Neville to Humber (actually a point on the Queensway just before the loop) are considered. Much of the data stays within the 10 minute band, although week 3, more affected by construction than the others, shows more scatter and a higher peak value. Oddly enough, week 5, when the service was on diversion via Shaw and King, has lower off-peak running times than other weeks probably due to the lower level of congestion and busy, signalled intersections, on King than on Queen.
Running times from Neville to Long Branch do exhibit a wider range of values with the band smearing out to roughly 15 minutes, varying depending on which week’s data one looks at. As discussed above, delays at Queen and Roncesvalles add to the variation in trip times westbound from Neville to Humber or Long Branch.
Full Route Eastbound
Eastbound trips from either Long Branch or from Humber to Neville show a narrower range of running time values and stay roughly in a 10-minute band.
Scheduled Running Times
The values shown in this table give the scheduled running times and headways for the Queen route in November 2011. One set of values covers the first three weeks before running time was added for the diversion around the Dufferin bridge construction, and the other covers the last two weeks.
For each set, there are four values:
- RTT (Round Trip Time): This is the time allocated for a round trip between terminals, but not including “recovery time”.
- Rcvry (Recovery Time): The name implies that this is intended to give operators a break after a long drive across the city, but in fact the amount of time is dictated by whatever is needed to even out the trip to a multiple of the scheduled headway. There is no relationship between the length of recovery time and the time of day, day of week or length of trip.
- Total: The sum of the two values above.
- Hdway (Headway): The scheduled time between cars. As discussed in a previous article, TTC standards consider service operating within 3 minutes of this value to be “punctual”.
Scheduled running times to Long Branch on weekdays range from 90 to 95 minutes one-way, dipping to 80 minutes for late evening service. Weekend times are a bit shorter although Saturday afternoons have almost as much time allocated as the PM Peak period.
On the Humber branch, scheduled one-way times range from 65 to 69 minutes during weekday peak, daytime and early evenings, dropping to 58 late evenings. Weekend times are slightly lower as with the Long Branch trips.
Updated: The following graphs show the averages, by hour, of actual running time data. The degree of scatter in values can be seen in earlier plots showing individual points, while the graphs of averages show the mean of all values for each hour. The following text has been amended to refer to the averages.
Each set of graphs contain two pages. The first groups data for weekdays in each calendar week to show the variation, if any, over the course of the month. The second groups data for each day of the week to show differences between average trip times through the course of the week and on weekends.
Actual running times from Neville to Long Branch on weekdays lie above the scheduled times except in the late evening period. Averages between 1400 and 1800 (2:00 to 6:00 pm departures from Neville) are over 100 minutes when the scheduled running time is only 95 plus 7 minutes of recovery time. There is little variation between the data on a week to week basis except for a small increase in peak direction, peak period times for week 5 when the route diverted via Shaw and King.
Weekday averages stay close to each other except on Friday when running times are slightly longer midday and in the evening. This pushes them even higher than the scheduled times and leads to more ragged service and short turns.
A peak over 110 minutes occurs from 1400 to 1600 on Saturdays when the total scheduled time is 100 minutes. Sundays peak above the scheduled total time of 93 minutes, and this is not simply due to one day with the Santa Claus parade. Eastbound running times show a similar pattern, although some peaks do not last as long.
In these comparisons it is important to note that the times charted here are measured from outside of the loops at each end of the line, and so they slightly understate the total travel times.
The running times allocated to trips on the Queen route are shorter than the cars actually require during many operating periods. However, one challenge is that the “peaks”, especially on weekends, can be shorter than the scheduled period that includes them (such as “afternoon”). What might be adequate at noon or at 5:00 pm is too short during the height of Saturday traffic at 3:00 pm. As mentioned earlier, these data are from a month where most seasonal effects are not present such as tourism, shopping and street fair activity. Only the Santa Claus parade introduced a visible effect taking data values out of their ordinary range.
All of this suggests that a more finely-grained approach to allocating running times is required to bring scheduled provisions for service into line with actual conditions. This cannot simply be done by scheduling very large “recovery” times at terminals where there may not be capacity (as at Neville) to store cars taking long layovers. This also raises the issue of relief time for operators separate from time for their vehicles and of operating strategies that can preserve service quality.
I will turn to the issue of service management and short turns in the next article in this series.