In January 2016, the TTC implemented new schedules on route 501 Queen making the following changes:

• The route was split into two sections with an independent Humber to Long Branch service (in effect, a return of the long absent 507 Long Branch route), and
• Running times were increased so that scheduled and real-world travel times were better matched and short turns could be reduced.

The first article on this subject reviewed headway behaviour (the space between cars) with the new schedules. This article turns to running times (the time required to get from A to B) and service behaviour at terminals.

A Note on Reading the Charts

Many of the charts in this article are called “link histories” (my term) because they show the evolution of travel times over a segment (link) of a route over time. Data are presented from late 2013 to January 2016, although some months are missing because I was not tracking the 501 for those periods.

There are six pages in each set of charts, with each page containing data for a group of half-hour time periods starting at 0600 and ending at 2330 (11:30 pm). The solid lines track the average link times for their half hour with values averaged over a week (special days such as statutory holidays are omitted). The dotted lines track the standard deviations of the link time values within each week and half hour. Generally speaking when there is a spike in the SD values, this is caused by a delay within the link for the time period and week in question. It is the overall change, or lack of it, in values that shows how conditions evolve.

The times refer to the moment when a car enters a link and so, for example, the 0600 data include all cars entering the link between 0600 and 0629.

For a detailed explanation of the data analysis please refer to Methodology For Analysis of TTC’s Vehicle Tracking Data.

For the purpose of this analysis, locations close to, but not at terminals are used as screenlines. This avoids situations where a queue of cars forms within the terminal itself. The locations used here are:

• Silver Birch, one transit stop west of Neville Loop
• The Queensway west of the Humber River, but east of the entrance to Humber Loop
• Lake Shore Boulevard west of the exit from Humber Loop
• Lake Shore Boulevard east of Brown’s Line, before the entrance to Long Branch Loop

For the purpose of comparing travel times, vehicles are almost at their terminals when they cross these lines.

Between Neville Park and Yonge Street

The east end of 501 Queen is the less congested section, although traffic in The Beach can be troublesome during off-peak periods.

Westbound travel times are at their height during the morning peak (8:00 and 8:30 data on the first page), but values for midday (pages 2 and 3) are quite consistent. Values stay at roughly the same level through the PM peak (page 4) and begin to fall after 18:00 (page 5). By late evening, the link times are back to the same values as early in the morning.

Eastbound the pattern is similar, although the buildup for the AM peak is not as great as it was westbound.  During the PM peak, times are slightly higher than mid-afternoon, but not by much.

What is striking about both sets of charts is the consistency in values over time from fall 2013 to winter 2016. Although link times vary by time of day, the values at specific times are fairly constant for all months whose data are included here with few exceptions.

501_SilverBirchAve_YongeSt_LinkHistory

501_YongeSt_SilverBirchAve_LinkHistory

Between Yonge Street and Humber Loop

The west end of Queen is somewhat busier and more congested than the east end. This shows up in peak periods that persist a bit longer on that half of the route. Eastbound times are longer than westbound through the late morning, and the westbound times do not overtake eastbound until the early afternoon. By the evening, travel times both ways are roughly the same.

One effect worth noting occurs at the beginning of September 2014 when the TIFF-related diversion of King service onto Queen cause substantial congestion. This shows up as longer running times for the Queen cars.

As with the east end of the line, the times are fairly consistent since fall 2013 with a small seasonal variation (winter times are a bit better than for other seasons).

501_YongeSt_QueenswayHumber_LinkHistory

501_QueenswayHumber_YongeSt_LinkHistory

Between Neville Loop and Humber Loop

Looking at the full length of the main part of 501 Queen, we see how running times for cars operating between Neville and Humber behaved (short turns are not included because they did not make the full trip).

501_SilverBirchAve_QueenswayHumber_LinkHistory

501_QueenswayHumber_SilverBirchAve_LinkHistory

The trip each way peaks in the morning in a range of 70-80 minutes, with a steep drop for the lighter traffic at the end of December 2015 (that month is not included for previous years). For the afternoon peak, the range is 80-90 minutes. Allowing for seasonal fluctuation and construction effects such as various Gardiner Expressway ramp closings (these affected mainly King, but there was some spillover to Queen), the times have been remarkably uniform for over two years. The new scheduled running times (71 minutes AM, 82 minutes PM peak) will accommodate many but not all of these trips especially if other factors stretch out travel times. (Remember that the charted times are measured from about one minute away from the terminals, and do not include recovery time about which more later.)

Off peak scheduled times are generally in the range of actual travel times, and by late evening almost excessively so.

The more generous running times should show up as better service at the terminals, and that is what we saw with the headway charts in Part I of this article. Later, here, I will look at this from another point of view to show how much more service reached the termini in January 2016 than in 2015 with the old schedules.

Between Humber Loop and Long Branch Loop

For much of the day, Lake Shore is fairly uncongested by comparison with streets downtown. One area where problems have arisen recently has been the Humber Bay around Park Lawn. This shows up as additional travel time in the AM peak, particularly eastbound where trip times during fall 2015 and continuing into January 2016 exceed the 27 minutes allowed by the new schedule. It is, therefore, difficult for 501 Long Branch cars to make their scheduled trip times during the main part of the rush hour from 7:30 am onward.

501_LakeshoreHumber_LongBranchLoop_LinkHistory

501_LongBranchLoop_LakeshoreHumber_LinkHistory

Midday and evening services fare better with travel times and new schedules better suited to each other. As on the main part of the route, the allowed time is generous late in the evening when trips require less than 20 minutes.

A general point about all of the link time charts: The Standard Deviation values sit roughly in the 2-4 minute range for the shorter links (Neville-Yonge, Humber-Yonge, Humber-Long Branch) with larger values for the longer route (Neville-Humber). Values were even higher for Neville-Long Branch trips (not charted here) with SD values close to or above 10 minutes at times. This shows the cumulative effect of variations in conditions over a longer route when there is no chance to “reset” or recover to a schedule along the way.

Terminal Times

Scheduling problems can show up as inconsistent times taken by cars at their terminals. If there is consistently too much running time in the schedule, they will arrive early, and could even have an incentive to leave late knowing that they will be on time when they reach the other end of the route. This is especially true for short routes like 501 Long Branch where a substantial amount of time can be spent on terminal layovers rather than driving.

If schedules are too tight, two things can happen. One is that many cars are short turned and might well arrive early at their next terminal. Another is that operators desperate for a break, particularly on a long route such as Neville-Long Branch, will take a layover whether they are early or late if for no other reason than a call of nature and to relax after a long drive across town.

For the purpose of these charts, the “terminal time” is measured from the near-terminal screenlines used in the link time calculations. For example, at Neville, the time measured is for a car making a round trip from Silver Birch to Neville Loop and return. The shortest of these times will be for cars that took no layover and needed little or no time to serve passengers. They were probably on a running-in trip to Russell Carhouse, or were about to be relieved by a new operator at Connaught. Either of these is a powerful incentive not to dawdle at Neville.

The following charts plot the average terminal times by hour at three terminals – Neville, Humber and Long Branch – for November, December 2015 and January 2016. There are three pages per terminal with weekday, Saturday and Sunday values plotted.

501_201511201601_TerminalLinkHistory

At Neville Loop, the terminal time went up in January except for late evenings. The chart shows the average and SD values, but to get a sense of what the raw data looks like, here are charts showing all of the values for weekdays in January 2016.

501_201601_NevilleLoop_Terminal_MonthLinks

The lowest values observed, and there are not many, are about 3 minutes and these are for cars that arrived and left with no layover. Therefore it is reasonable to subtract about 3-4 minutes from the averages to obtain the time spent in Neville Loop. This gives us terminal layovers of about 5 minutes which are not unreasonable for a route of Queen’s length. However, these are just averages, and as the detailed data shows, the time spent east of Silver Birch often exceeds 15 minutes. Queues of cars are commonly seen at Neville (and at other terminals where generous schedules have been implemented).

The trade-off is to give enough extra time for most of the “difficult” days while not giving too much. It is worth noting that the link history charts show that January has lower travel times than the warmer parts of the year, and the situation at terminals may change as good weather arrives.

The terminal times at Neville show the same pattern on weekends as weekdays.

At Humber Loop, only cars that actually ended their trips at this location were included in the calculations. As at Neville, the average terminal time is longer in January, but the SD values are lower indicating that there was less scatter in the values compared to the fall 2015 months.

501_201601_Humber_Terminal_MonthLinks

As at Neville, there is a wide range of terminal times for cars whose trips terminate at Humber, and the low end of the range is at about 3 minutes indicating the travel time for a minimal in-and-out pass through the loop.

The January 2016 average values are particularly high on weekends showing that many cars get a quite generous layover at this location.

Out at Long Branch, we see the reverse situation. Where 2015 operations were at the end of a very long journey from Neville and lengthy layovers in the loop were common, January shows Long Branch operating as a separate route where immense recovery periods are not present. The average terminal time is lower in January.

Note that Saturday and Sunday/Holiday data for January 2016 include two days (Friday Jan 1 and Saturday Jan 2) which operated on the “old” schedules.

How Many Cars Reach Their Terminals?

Another way to look at these data is to compare the cars/hour at each terminal for the three months. This gives a sense of the proportion of short turning that occurred, on average.

501_201511201601_TerminalCounts

The values in these charts are calculated by taking the count of cars arriving within each hour for each type of day, and then dividing by the number of that day type in the month. This corrects for variations in the number of weekdays, Saturdays and Sunday/Holidays in the months.

Note that the late evening schedules were not changed in January.

At Neville, the cars/hour are consistently higher in January than in previous months indicating that more service reached the east end of Queen under the new schedules than the old. The AM peak in December was better than in November probably because with lighter traffic on many days, the number of short-turns required to “clean up” after the peak was lower than usual. The afternoon and evening values, however, are the same for both months.

The situation on Saturdays is particularly striking where the average cars/hour at Neville was well below the scheduled value all day in the 2015 data, and fell badly as the evenings wore on. This is a route that was supposedly part of the “10 minute network”. Sundays were even worse with fewer than four cars/hour all day in the November 2015 data.

Similar improvements are visible at Humber. The cars counted here are only those scheduled to reach Humber. Because there was a through service to Long Branch in 2015, cars could be short turned further east while some service remained further west, but many of the cars scheduled to reach Humber never got there.

At Long Branch, weekday counts show a slight improvement. What this masks for 2015 is the lack of reliability in headways as discussed in Part I. There might be five cars in an hour, but they would arrive and depart at widely varying intervals. The jump in values for January in the late evening is caused by the transition of the 501 service to a mix of through cars from downtown with some “Long Branch” cars that remain in service as local shuttles even if the downtown service is disrupted.

Conclusions

The new schedules on 501 Queen achieve their aim of getting most service out to the end of the line (Neville/Humber) and the separate Long Branch service is more reliable than the through service from Queen that it replaced. This is supposed to be a “temporary” arrangement, but the TTC should think very hard before reverting to the combined route.

Two problems remain. Cars still leave terminals on uneven headways, and this is allowed under TTC standards with a +1/-5 minute range for “on time” departure. When the headway is 5 minutes, this can cause bunched service to get a perfect score. The TTC needs to explicitly measure the degree to which vehicles operate on headways well below the scheduled values.

The other problem is the inevitable variation in running time as each car encounters different circumstances along the way – missing a traffic light, a surge load, a motorist who needs a lesson in parallel parking, not to mention differences in operator experience and aggressiveness. Transit vehicles will move apart even if they are dispatched exactly on time from terminals. The question is when does the variation become intolerable as a measure of service quality. For a short route, or one with a mid-point offstreet layover such as a subway connection, it is fairly easy to reset headways and dispatch vehicles “on time” for that point. This is more difficult for a long route without a natural layover point, and where knowledge of conditions enroute is essential. For example, under some circumstances, leaving Yonge Street early might be worthwhile if traffic in Queen West is known to be bad.

These are operating strategies beyond the scope of this analysis, but no analysis, none of the most up to date software in the world, can replace the basic ethos of providing as good service as possible.