29 Dufferin is one of the TTC’s target routes for improved performance, and it also happens to be a route that I have been following with TTC vehicle tracking data from selected months going back to November 2011. This article reviews the line’s operation on weekdays.
This article is somewhat technical, but has been written for a general audience to the degree the subject allows.
Scheduled Service History
An important factor in reviewing service on the street is the schedule itself: the headways (time between vehicles) and the running times provided for each trip. Headways may look good on paper, but if service arrives unreliably, or if some of it never reaches the destination thanks to short turns, then the advertised service is a polite fiction. Running times also have an effect, especially if they are shorter than the typical time required to drive from one end of a route to the other. When an operator cannot make the scheduled trip, the bus runs late and is quite likely to short turn simply to get it back on time. In theory, this “restores” normal service, but if vehicles are chronically late, the process never ends. The “treatment” never cures the “patient”.
The months included in this article are:
- November 2011
- March 2012
- May 2013
- September 2013 (Dufferin Bridge at CNE closes)
- March 2014 (Diversion from College to Queen southbound for water main construction)
- November 2014 (Introduction of articulated buses)
- April 2015 (Major schedule revisions to reflect actual operating conditions)
This spreadsheet shows the scheduled headways and trip times for all of the periods covered by this article. In a few cases, there are two schedules shown for the same month because a change was implemented part way through.
The Dufferin Bus operates primarily between Dufferin Loop (at the Western Gate of the CNE ground) to Wilson Station. During peak periods there is a short turn at Tycos Drive where half of the service returns south. During certain periods (with seasonal variations), half of the service runs through the CNE grounds to the Princes’ Gate (eastern entrance).
In September 2013, the bridge on Dufferin at the rail corridor north of the CNE closed on very short notice for repairs. Service that was scheduled to operate to the Princes’ Gates turned back at Dufferin Loop. This resulted in half of the buses having more running time for their trips between Dufferin Loop and Wilson Station. Concurrently, a diversion for water main construction was operated southbound between College and Queen. These two offset each other, at least for the buses that were scheduled to run through to the Princes’ Gate. Extra running time (up to 10 minutes) was not added to the schedules until the end of March 2014 in anticipation of the re-opening of the Dufferin Street bridge.
In November 2014, the route officially switched to articulated bus operation Weekdays and Saturdays, although these vehicles had been present for some time before.
In April 2014, there was a major restructuring of the schedules: considerably more running time was provided to reflect actual conditions, and the split operation at the south end of the route was discontinued on weekends.
A Few Notes About The Charts
In the following sections, there is a standard format of chart which is used to plot four different sets of data: average headways at various points, average travel times between points, and the standard deviation values for each type. Within each week, the data are subdivided by hour. Standard deviations are shown in dotted lines on the same charts as the averages. The data are subdivided into four pages by time of day.
For the headways and travel times (called “link times” in the charts because they refer to a single link in a route rather than to the entire route), values are averaged for each week in the months under review. This can “flatten out” problems caused by major delays to a few vehicles, but it can also mask a situation where there is a wide spread in values that “on average” look good. (Weeks with only a few days – because, for example, the first of the month falls on Thursday – can have few data points within one hour, and if any of these are far from the average, a spike can result.)
The standard deviations measure how dispersed the values are. The SD will be low if the values are all close to each other, but it will be high if the values are spread out. For headways, a higher SD usually indicates bunching, especially in cases where the SD is as high as the average. This means that a majority of the values lie in a range from zero to two headways with many vehicles arriving in pairs. For link times, a low SD indicates that travel times from point “A” to “B” are fairly consistent within each hour. A high SD indicates that travel times are less reliable due to variations in conditions encountered by each vehicle.
In some cases, the values are quite “spiky” in spite of averaging over a full week. I will delve into some of the specifics in a separate article.
For the Headway Histories, the Y-axis is a constant value of 15 minutes. This allows direct comparison between different parts of the route. For the Link Histories, the Y-axis is 30 minutes except for the charts for the full route which use 60 minutes to accommodate the larger values.
One of the most common complaints about the Dufferin Bus is that the service is irregular. A passenger might wait much longer than the advertised headway, and then two or three buses will show up. The sets of charts linked here illustrate the behaviour of actual (as opposed to scheduled) headways at four locations and directions.
Service northbound at King, especially during periods when all vehicles originate at Dufferin Loop a few block to the south, should be reasonably well spaced. The charts show quite clearly that this is not the case. During the morning until fall 2014, the standard deviation values overlap the averages except for the first hour of service when headways are wider. Even so, the SD values rarely fall below 2 minutes. Starting in fall 2014, the schedules were changed for articulated buses, and the headways were widened. However, the SD values stay at roughly the same level.
Midday values continue the pattern with SD values persistently between 2 and 3 minutes.
The afternoon values covering the PM peak show higher SD values until the introduction of the articulated buses in Fall 2014 and the revised schedules in April 2015. The peak in week 3 of November 2014 was caused by severe delays due to a storm. This shows up on charts for other parts of the route, especially those that became impassible. One might argue that the April 2015 data show the benefit of the new schedules, athough I would prefer to see another month’s data added to be certain. (This will be further complicated by construction on Dufferin for which there is no provision of added running time in May 2015.)
The evening page shows the trailing part of the PM peak which, on Dufferin, lasts well beyond 6:00 pm. As with the afternoon data, value do not “settle down” until fall 2014.
What shows up quite consistently is that the standard deviations rarely fall below 2 minutes and are commonly at higher values, and that these values seem to have little relation to the headway actually operating. This is almost certainly a direct result of the TTC’s acceptance of ±3 minutes as the target for “punctual” service. The problem with this level of variation leaving a terminal is that the wider headways tend to spread out as vehicles move along the route, and narrow headways collapse into bunched service. One can argue that keeping transit vehicles “on headway” at a level of two minutes or less is nearly impossible given the random effects of surge loads and traffic signals. The problem becomes one of maintaining the spacing as vehicles move along the route.
However, something quite striking happens between King and Bloor northbound – the service is actually better at Bloor than at King. The reason for this is that short turns southbound at College are common, and so there are more buses serving the stop at Bloor. I will return to this later in the article.
By the time the service reaches Eglinton, the average headways are roughly the same as at Bloor, but the SD values have grown. This shows the cumulative effect of headways becoming more uneven as vehicles travel along the route.
Southbound from Transit Road (just south of Wilson Station), the service is slightly better behaved with lower SD values in the morning and midday. However, by the afternoon and evening periods, the situation is not unlike what we see at King northbound. (Note that only half of the service operates to Wilson Station at peak periods. The averages for the 6:00 am slot are low because all of the service builds in southbound from Wilson Garage, but it does not all come back that far north, and wider headways from 7:00 onward result. There is a similar effect in the PM peak.)
By Eglinton Avenue, the Tycos short turns have merged into the service. The average headways actually operated are slightly longer than the scheduled values, notably in the PM peak where the SD values, as elsewhere on the route, are running roughly equal to the headway until the artic schedules go into operation, especially the April 2015 version. By the time the service reaches Bloor Street, the SD values are often greater than the averages, at least until the April 2015 schedules kick in. Although the service is advertised at a headway of 3-4 minutes, what actually shows up is a convoy of 2-3 buses much less often. This state of affairs has existed on Dufferin for years and is only now being addressed.
Running Time Reliability
An important aspect of maintaining regular headways is the reliability of running times. If trips routinely take much longer or shorter than average (not to mention the scheduled times), then some vehicles will always run early and others always late. Short turns will be common.
If the scheduled running time is reasonably close to actual requirements, then short turning should not be required, and regular vehicle spacing should be achievable with good line management.
The two charts above show the travel time from King to Transit Road in each direction. This segment is chosen to “lop off” the terminals at Dufferin Loop and Wilson Station where elapsed time has nothing to do with conditions along the route, and a lot to do with the layover an operator takes (not to mention efforts, if any, to dispatch buses from the terminals).
When comparing northbound and southbound times, remember that there was a southbound diversion between College and Queen Streets in September 2013.
During the morning period, southbound travel times are slightly longer reflecting the peak demand direction, even though the hills northbound can slow service. Midday times are slightly faster southbound.
During the afternoon period, northbound times are noticeably higher, and this is especially marked earlier in the peak period from 3:00 to 4:00 pm (the red line). This is caused by the fact that parking is permitted until 4:00 pm even though the street is already trying to sustain early peak period demand. Similar problems show up on some of the streetcar lines, and the City of Toronto has extended the hours of peak period restrictions to counteract this in some locations. Also notable here is that the running times are longer in the two periods with articulated buses, particularly northbound in April 2015.
By evening, the route settles down to fairly even times which decline hour by hour until 11 pm (23:00).
During the morning period, the trip between King and Transit Road is taking about 40 minutes northbound and 45 southbound, or 85 in total. This fits within the scheduled 112 minutes over the entire route from Princes’ Gate to Wilson Station and return. It is intriguing that the midday scheduled trip time, at 109 minutes, covers only Dufferin Loop to Wilson Station, a shorter distance, and so is actually more generous.
By the afternoon, peak, the new much extended running time is 143 minutes for the round trip from Princes’ Gate to Wilson Station. Roughly 115 minutes is consumed on the average trip, 55-60 northbound, 50-55 southbound. Buses should not have any problems staying within these running times which, if anything, may be too long for most conditions. TTC tends to give the same scheduled time to each direction of a trip, but this is a clear case where one direction should get more than the other. This particularly shows up with location and direction specific congestion as we will see later.
The SD values for trips between King and Transit Road generally stay below 5 minutes, but this does give a ±5 minute window that is substantially wider than the scheduled headway. Without adequate recovery time, the long trips have no hope of leaving on a return journey on the scheduled headway.
Between Bloor and King, northbound running times are slightly longer than southbound except in September 2013 when southbound service was diverted. The SD values stay at or below 2 minutes indicating little variation in travel time.
The segment from St. Clair to Bloor is unremarkable and shows quite consistent travel times with rises in the peak and falls in the off peak. One point stands out: the hour from 3:00 to 4:00 pm (red line, afternoon page) northbound. This location has shown up in previous analyses of 29 Dufferin as one where the lack of parking restrictions until 4:00 pm regularly delays service between Bloor and Dupont Streets.
Northbound travel times between Lawrence and Wilson are quite consistent until mid-morning, and they build up substantially into the PM peak (the degree of this effect is much more marked in recent weeks than in past years) with the primary delays occurring around Yorkdale Mall’s parking entrance. This effect does not show up in the southbound data, and it is responsible for the difference in running times over the route for each direction noted above.
The Effect of Articulated Buses
In November 2014, headways and running times were adjusted for the larger vehicles. This was offset by a roughly 50% increase in vehicle capacity, although a major problem here, as on other routes scheduled for articulated buses and streetcars, is that not every vehicle on the route actually operates with the larger artics.
In April 2015, the headways were further widened, and this completely negated the capacity bump from the original artic-based schedule. The situation in the PM peak is even worse with capacity now scheduled on Dufferin is actually less than was present in the pre-artic schedules.
AM Peak PM Peak Headway Buses/Hr Capacity Headway Buses/Hr Capacity March 2014 2'37" 22.9 1237 2'45" 21.8 1177 November 2014 3'30" 17.1 1317 3'38" 16.5 1271 April 2015 4'00" 15.0 1155 4'45" 12.6 970 Vehicle capacities: Standard bus 54, Artic bus 77.
The TTC would argue that the April 2015 schedules with substantially expanded running time will actually provide better service because buses will stay on time, short-turns will be minimized and service will be more reliable yielding the same effective capacity. That, however, depends on reliable service, not to mention having artic vehicles assigned to all runs on the route.
One obvious question is whether there is a difference in running times between “standard” 12m buses and the 18m “articulated” vehicles. This can be influenced by grades, engine performance and loading delays depending on whether all-door boarding is used at major stops.
For November 2014 and April 2015, although the route was officially converted to “artics”, some 12m buses did appear in service. Their average travel times from King to Transit Road have been separated in the following charts.
Given that there are few 12m buses during both periods, there are fewer “sightings” to nail down the 12m average running times and some variation is inevitable. However, the values for both vehicle type track fairly well together through the day in both months.
A perennial issue on Dufferin especially at the south end of the route is short turns. The evolution of the route’s behaviour can be seen in vehicle counts for various locations, notably at Bloor and at King Street.
These charts show the count of vehicles in each hour for weekdays in three separate months. The data are displayed in three formats:
- The first page shows the ratio of vehicles northbound at Bloor to those at King. A value of 1 indicates that the same number of vehicles passed both points within the hour. A value less than 1 indicates that some vehicles joined the line north of King.
- The next three pages show the counts for several points on the line in line format.
- The last three pages show the same counts in bar chart format.
In March 2014 the counts for King are below those for Bloor through much of the day showing how much of the service at Bloor originated north of King. The short turns may sort out running time problems further north on the line, but they can harm service at the south end.
By November 2014, the spread is reduced, and this continues in April 2015 with the new schedules showing that the use of short turns had declined, but not completely vanished. Note that the short turns (lower counts for King) tend to fall after the peak periods when service is being sorted out after the rush hours. This pattern shows up on other routes where sorting out the service after the peak can play havoc with the shoulder peak period.
(Note that the counts at Lawrence and Wilson are lower than for other points during the peak periods when half of the service short-turns at Tycos Dr.)
The TTC has at long last made attempts to improve service on the Dufferin Bus, and some of this is demonstrable in the vehicle tracking statistics. However, some of the early benefits of added peak capacity, particularly in the PM, have been lost to the wider headways and extended running times to deal with “real world” driving conditions and constraints on the fleet size.
Short turns have been reduced at the south end of the route, although not totally eliminated.
Headways continue to be uneven, although for a good part of the day much of the service falls within the TTC’s ±3 minute target relative to schedule. The fundamental problem with such a target is that for shorter headways, bunched service is considered “within target”. During periods of wider scheduled headways, uneven gaps between vehicles persist. One might ask whether the laissez-faire attitude during periods of shorter headways carries over to the route as a whole.
If there is any active management of dispatch times at terminals, this does not appear to be reflected in actual operations. Buses continue to run in groups, although on the now-wider headways for artics, they are not always nose-to-tail.
In the next article in this series, I will delve into the specifics of certain dates and locations to understand some of the “spikes” in the data presented above.