In the first article of this series, I reviewed the headways (time between vehicles) on route 501 Queen from August 2013 to May 2015. A pattern there is that headways on the outer parts of the route are consistently, at times extremely so, worse than the advertised frequencies. Even in the central part of the route, average headways are close to scheduled values, but the regularity of vehicle spacing is not – cars commonly run in pairs on a much wider combined headway than the schedule calls for.
With this route listed among those that will be part of the TTC’s “Ten Minute Network”, actually achieving that goal will be as important as the inevitable hype that will accompany the announcement. This is also a route destined for better service thanks to new loading off-peak loading standards.
This article turns to the question of running times – the period required for a typical vehicle to get from point “A” to “B” on a route. These are important for a variety of reasons:
- If the scheduled time is commonly less than the actual time needed, then vehicles will always be late, and there will be a strong incentive for cars to short turn.
- If the scheduled time is commonly more than what is needed, then vehicles will either dawdle along their routes, or take extended siestas at terminals.
- “Congestion” is a routinely cited reason for the TTC’s inability to operate reliable service, but it is not a consistent phenomenon across the route, by time of day or by day of the week. Some of the worst disruptions arise not from chronic congestion, but from events such as construction projects or diversions around festivals. The location of the delays is not confined to the core area.
There is a lot of material here, and I don’t expect that most readers will go into all of the detail. The first part looks at the route overall, and then I turn to individual segments. If there is any overall message, it is this: the operation of a long, busy route like Queen is affected by many factors. Some are institutional (schedules, procedures). Some are chronic (predictable congestion). Some are transient (accidents, illness). Some of the worst are from relatively short-lived events such as construction work or event diversions where the resulting service leaves much to be desired. There is no one “magic solution” that will fix all of them with minimal pain for either for the TTC or for other road users.
The underlying data for the analysis here are the same as used in the first article, but presented as travel times between points rather than as headways at points on the route. The format of the charts is substantially the same except that charts for weekday data are subdivided by half hour rather than by hour.
The Service History included in the first article contains schedule information for both headways and running times over the route.
Throughout this article, any subsection of the route (including the entire route) is referred to as a “link” by analogy to the standard terminology in which transit trips are described as having one or more links, each on a single vehicle or route.
Vehicle behaviour near terminals can be tricky to analyze because of layovers and queing. To keep behaviour on the main route separate from effects at terminals, for the purpose of this analysis the “terminal” locations are measured a short distance away:
- Silver Birch Avenue stands in for Neville Loop.
- A point east of Humber Loop on The Queensway stands in for the loop itself.
- A point east of Brown’s Line on Lake Shore stands in for Long Branch Loop.
The first part of this article gives an overview of the route as a whole – how long does it take to get from Long Branch to Neville – and the degree to which the schedule and actual operating conditions are out of whack.
The rest of the article looks at each segment of the route in turn for those readers who want the details. The purpose is also to show that the behaviour of service, including that bugbear “congestion”, varies from place to place and time to time. Any “fix” that tries a one-size-fits-all approach cannot deal with this variation, and may even work against acceptance of targeted changes.
The periods of data have been selected to avoid major disruptions of the route from diversions around track construction at Queen/Broadview (July 2014), Queen/Victoria (May 2014), Queen/York (July 2013).
Looking at the Entire Route
The charts in this section show running times for trips that operated from Neville Loop (Silver Birch) to Humber and to Long Branch. Yes, despite the short turns, some cars did make the entire journey one way or the other. Because these are long trips, remember that the times shown for each line on the chart represent the half hour (or hour in the case of weekends) when cars left Neville. Conditions will evolve over the length of the trip especially as traffic builds and ebbs, but from the point of view of schedules, running times are counted from the departure point.
- The westbound weekday trips build up in elapsed time from a range of 80-85 minutes at 6:00 am to highs of 100-110 minutes by the end of the peak. Time for the full route including terminal layovers would easily be 5-10 minutes more. The one-way AM peak travel+recovery time scheduled for this trip is 93 minutes, and so it is clear that that most cars will be late by the time they have made the Neville to Long Branch trip. The situation eastbound is similar.
- There are gaps in the brown (8:30 am) line indicating that no car leaving Neville in that half hour went all the way to Long Branch in the weeks where data are missing. (September-October 2014, when all cars turned back at Humber, is omitted on these charts.)
- The Standard Deviation (SD) values, are fairly consistent in the 5-10 minute range except where some event (visible by the spike in average travel times) pulls values away from the average.
- Over the two-year period, the averages and SD values have stayed fairly constant, and these data do not indicate, at least for this period, that AM peak congestion or stop service times have changed much.
- For the morning period from 9:00 to noon, average travel times both ways lie in a band from 90-100 minutes to which 5-10 minutes must be added for terminal time. This compares to scheduled travel+recovery of 98.5 minutes which is less than required for trips at the upper end of the range. Between 9:00 and 10:00 am, there are several weeks where no cars leaving Neville reach Long Branch due to short turns to get vehicles back on time.
- For the early afternoon from noon to 3:00 pm, average times shift upward into the 100-110 minute band (plus terminal time), well above the scheduled value. Worth noting are the lower values in winter 2014 which was particularly bitter, and when there would have been less than usual road traffic.
- The situation continues for 3:00 to 6:00 pm, although some trip times rise into the 120 minute band (plus terminal time) when the scheduled time is 102.5 minutes.
- By the early evening from 6:00 to 9:00 pm, average times settle down again within the 80-100 minute band (plus terminal time), the later periods having the shorter trip times. The scheduled time is 99 minutes and so eventually, on average, the service is able to operate within the schedule provided that there is no event to disrupt travel (about which more later).
- Late evening, from 9:00 pm to midnight, brings the average times down to a range of 70-85 minutes (plus terminal time) with lower values later in the evening. The scheduled trip time is 81 minutes.
Across almost the entire day’s operation, the average time for a streetcar to travel between Neville and Long Branch either way exceeds the time allowed by the schedule. Moreover, the fluctuation during some periods due to seasonal and/or construction effects is substantial. A single scheduled travel time cannot work for all conditions.
The next set of charts compare average weekday running times for 501 Queen Service between Neville (Silver Birch), Humber (east of the loop) and Long Branch (east of the loop) with the scheduled running times from May 2015. In each chart, the scheduled time is shown in red and it is generally lower than the actual average time. There are some caveats here:
- The times at which scheduled running times change do not land precisely on the hour, I have used the closest transition for these hourly plots.
- As the measurement points are slightly distant from each actual terminal, the average values would be a few minutes higher taken on the same basis as the scheduled times.
- Terminal layovers are not included in these figures. They are a separate issue discussed at the end of this article.
On Saturdays, by early afternoon the average westbound trip time to Long Branch is over 100 minutes, sometimes as high as 120, but the scheduled value (including terminal time) is only 100. By early evening, the trips are back to the 90 minute range, but this does not allow for terminal time and the schedule is still tighter than actual conditions. On Sundays, by midday the average travel times are above scheduled values and they stay that way right through to the late evening. The situation eastbound is the same with actual travel times above scheduled values for much of the day.
It is no wonder service on Queen is such a mess on weekends with schedules that are impossible to operate. New schedules are coming in September (the details have not yet been released), and we will see just how well they address problems shown here.
[A note about file naming: The prefixes “1xxyy” and “2xxyy” refer to westbound and eastbound respectively. This eliminates problems on routes which do not head in the same “direction” over their entire journey such as 504 King. The numbers “xx” and “yy” signify the points between which I have measured running times. As we move from east to west across the route, these numbers rise. The purpose of this convention is to allow filenames to sort automatically by direction and location independently of the street names. The “30” suffix in the weekday charts indicates that they are based on half-hourly data.]
The data for service to Humber includes both the through cars to Lake Shore and those that turn back from Humber. On the weekday schedule, the longest time allowed for a Neville-Humber trip is 74 minutes including 6 minutes recovery (midday schedules). However, the average travel time from Silver Birch to Queensway east of Humber Loop is above this level for much of the day guaranteeing that the Humber cars will always be late. The same applies, at times moreso, to eastbound trips.
Weekends see the same pattern as on the Long Branch service with trips taking longer than scheduled even before allowing for terminal time.
Averages and SD values give a general view of a route, especially the degree to which the schedule may be impractical regardless of finer details of its operation, but it is worthwhile looking a bit more deeply to see behaviour of individual weeks and days. I will explore this more in the third article of the series, but to begin with, here are the details for May 2015 corresponding to the charts above. The format is the same as in the monthly headway analysis in the first article.
For trips over the entire route, the shape of the data is broadly similar from day to day and week to week, but there are noticeable differences between weeks and even individual days. Scheduling to match every circumstance is very difficult and this leads either to chronic problems with actual vs scheduled times (the current situation) or schedules that are padded to handle the bulk of situations. The latter approach has been used recently on 504 King, and similar changes are planned for 501 Queen. However, this brings its own challenges including running times that are so extended as to require operators to dawdle through their trips and queues of vehicles at terminals taking extended layovers. Padded schedules can also work against running “on time” because operators know they can always make up time later in their trip.
Note that there are a few anomalously long trips included in these data. Typically they are caused by cars that take a layover enroute and then continue in the same direction (for example, eastbound at Russell Yard). These have not been filtered out of the data.
In the following sections, I look at individual portions of the route to see whether the variations in travel time are uniform along the whole route, or concentrated at certain locations.
A point worth noting is that the Standard Deviation values are not cumulative over the route, although they are larger for the full trip than for an individual segment. In other words, although the averages for long distances are higher than for short ones, the degree of variation in values does not rise proportionately showing that some of the variation evens itself out in the progress of a journey across the city. Within any short segment, an SD value of 1-2 minutes is common, and a good deal of this can be attributed to small differences such as missed traffic lights and passenger loading times. Operators will usually make up for this on a small scale. When a car is seriously off-schedule, a short-turn might be required, but this does not show up in travel times.
Silver Birch to Coxwell
Westbound trips from Neville to Coxwell have had a fairly consistent set of times except for late summer and early fall of 2013 when something was clearly restricting traffic flow during peak and shoulder peak periods, hitting its maximum effect in late October. (Was there a construction project?) It is worth noting how running times build up from the half hour beginning at 7:30 to the one starting at 8:30. In the afternoon, the effect sets in at 17:30, peaks at 18:00 and then drops off again. There is a similar effect eastbound, but only in the afternoon.
Over the course of the day, running times rise gradually and then fall off through the evening, but except for cases where a disruption causes a spike, the values stay quite uniform. SD values stay at about 1 minute during these periods showing that although there is variation by time of day, within each half hour the values stay close together.
Weekend values are a bit more spread out and some values match those seen during the peak period, no surprise given the weekend traffic in The Beach.
Coxwell to Parliament
Travel times westbound from Coxwell to Parliament build up in the expected way from 0600 to 0830, although the change is quite substantial, from about 15 minutes to over 20, higher, proportionately, than in the segment through The Beach. Moreover, the SD values have many more spikes in them implying this segment of the route has less reliable journeys. From 0900 to 1430, the travel times are fairly consistent, although there continue to be many spikes in the SD values. This continues into the PM peak. Early evening brings two quite distinct sets of travel times for the period 1800 to 1930, and from 2000 onward.
Eastbound travel times build up in a similar way, but not with as much spread as the westbound values between 0600 and 0830 as one would expect for the counterpeak direction. However, something strange happens at 1130 beginning with the 2015 data – for only this half hour, the average running time is about 4 minutes higher, and the SD value jumps from 2 to 10 minutes. The balance of the day continues in a similar vein to the westbound data, including a marked drop-off in travel times from 1900 onward.
The data’s behaviour required a closer look to see just what is happening. Looking at the route at a more detailed level reveals that much of the irregularity is explained by activity at Russell Carhouse, most likely for crew changes.
For the segment from west of Coxwell to Greenwood, the “spikiness” of the SD values is particularly evident in the westbound data as is the abrupt change in running time after 1900. This is not a “traffic” phenomenon, but likely represents the end of a period when cars routinely take layovers at Connaught westbound. The situation eastbound is similar.
The oddity in data for the 1130 period in 2015 shows up in this segment. Examination of detailed data for the three months shows that on most days, one car arrives eastbound at Russell at almost the same time, lays over for a considerable period, and then departs to Woodbine Loop for a trip westbound. This skews the stats for the Greenwood-Coxwell segment, an example of the need to distinguish between “typical” behaviour and exceptions.
What is actually happening here is that the car in question is scheduled to run in to Russell Carhouse westbound from Neville Loop at 12:02 arriving at Connaught at 12:15. However, the car is often late, runs out of service eastbound, but does not log-off of the vehicle tracking system. This car then takes over for another run (which itself may also be late) and continues eastbound (the specific run varies from day to day). The result is that one car from Greenwood to Coxwell at the same time day takes much longer than the other trips, and it skews the stats. I have left this in the charts as an example of the need to investigate and understand how a route actually operates so that outliers like this in the data can be eliminated on an informed basis.
(Why track vehicles and not runs? The reason is that passengers ride vehicles, and don’t care about the crewing practices “under the covers”. In early versions of my analysis software, I did track runs, and this produced many anomalies due to cars changing run numbers on the fly.)
Westbound from Greenwood to Broadview, we see the buildup of morning peak running times rolling off after 0900. This area contributes the most to the entire Coxwell-Parliament segment’s AM peak. In the PM, travel times rise, but not to the same extent, and the peak ends with a slight decline into the evening. Eastbound travel times over the same segment grow gradually over the day and then fall off into the evening, but without as exaggerated an AM peak as for westbound.
I have not included the Broadview-Parliament segment because it contains nothing unusual.
The charts below show the data for all weekdays for the Greenwood to Coxwell segment for March through May 2015. The regular appearance of the single car laying over (for varying periods) is quite evident.
Parliament to Yonge
This segment is fascinating westbound because there has been a major change during the AM peak between 2013 and 2015. Up to fall 2014, the period from 0730 to after 0900 shows a marked effect of added running time (and hence congestion or much longer stop service time). Although times in 2015 rise a small amount, similar to other parts of the route, the effect is nowhere near as pronounced as it was in past years.
I checked with the City’s Transportation Department to determine which changes on Queen might have produced so pronounced an improvement, and their sense is that it is primarily the increased level of lane clearance enforcement that was active by 2015. Another possibility is that completion of water main construction on Richmond Street would have an effect on westbound Queen traffic.
Weekends in this area are unremarkable.
This segment does show the effect of a diversion on another route. During August 2013, 504 King service diverted around the reconstruction at King/Spadina and this introduced much more service on Queen including turning 504 cars eastbound at Church. This shows up as longer running times eastbound on Queen for the affected weeks. Westbound, the diverting cars were present on Queen only west of Victoria, and they did not affect the 501 service westbound to Yonge.
Yonge to Bathurst
Between Yonge and Bathurst, the Queen service was affected by diversions of the King service for two events.
- August 2013 saw all 504 cars running on Queen from Shaw to Church (as well as a good deal of local traffic) because King/Spadina was closed for reconstruction.
- September 2014 saw all 504 cars running on Queen from Spadina to Church during the Toronto International Film Festival (TIFF) diversion. (This did not occur in 2013 because the King service remained on route with a transit mall through the TIFF district.)
The effects are particularly striking in the PM peak when there is more traffic attempting to fit onto available road space.
At a few other times, construction projects (including one west of Bathurst in fall 2014 that produced spillover effects to the east) affected travel times on Queen.
On weekends, the TIFF effect is quite noticeable through Saturday afternoon and evening, and a smaller effect on Sunday. It is worth noting that only one weekend of the four in September 2014 had TIFF diversions, and so the real scale of the effect is muted by averaging the “normal” weekends for the rest of the month. The real story is told by day-by-day charts of running times in this segment (below).
The detailed view of this segment for September 2014 shows both the effect of TIFF from Thursday, September 4 through Saturday, September 6, as well as some congestion effects on selected days later in the month after the TIFF period ended. These show how drastically a change implemented on another street and route can affect service nearby. These data also show the dangers inherent in averaging over longer periods because effects that last only a few days could be much more severe that the averages imply. Riders will remember one day that was completely screwed up far longer than several days that were merely disorganized at the expected level.
Bathurst to Roncesvalles
Running times between Bathurst and Roncesvalles are fairly consistent with the sort of hour-to-hour changes one would expect due to peak hour and direction effects, and on weekends the times build up through the afternoon.
However, in September 2014, there is a big jump in the average times westbound in Week 1 and eastbound in Week 2. Looking at the details, this shows up starting on Tuesday, September 2 for three days westbound, and on Friday, September 5 through to Thursday, September 11. (See the “MonthLinks” charts below.) At an even finer level of detail for two specific days, it is clear that the delay is westbound over the entire section from Dufferin to Lansdowne, and eastbound for various distances stretching west from Lansdowne to Roncesvalles. This is almost certainly due to some type of construction that occupied first the north side, then the south side of the street at or near Lansdowne reducing its capacity.
(The “Chart” files below are plots of the location of each car in the two dates shown here as examples. The more horizontal the line, the slower the car is travelling. These show clearly the slowdowns on either side of Lansdowne Avenue and the degree to which the effect backs up along the line. To complicate the situation, all westbound service diverted via Shaw and King from about during many afternoons when the eastbound service was severely delayed no doubt to avoid the delays at Lansdowne.)
This is an example of a short-lived disruption of transit service whose effects are felt miles away by riders who encounter more than the usual number of short turns simply because cars were tied up in traffic on the other side of the city. A clear question is to ask whether it is possible for the TTC to plan for such events and adopt an alternate service strategy to compensate for delays they should reasonably expect. There is little information in the eAlerts for the period in question to explain just what was going on.
That this disruption occurred at the same time as the TIFF delays further east shows how totally unconnected events can wreak havoc on one route.
Roncesvalles to Humber
Running times for this segment are quite consistent with a few notable exceptions:
- Eastbound times show more variation because they include the time spent at the eastbound stop at Roncesvalles where crew changes may occur.
- Westbound times jumped in April (actually the effect first shows up on March 31) because of a slow order from South Kingsway over the Humber River bridge.
Here are comparisons of the averages and SD values for this segment for March and April 2015 plotted together.
A detailed block-by-block review (which I won’t publish charts for here) shows that the increase all comes in the segment from Windermere to the Humber River. I asked the TTC if they have a project date for renewal of this track and removal of the slow order, and they advised that this work is planned for October 2015.
Humber River to Lake Shore
The segment including Humber Loop is subject to layovers within the loop, especially eastbound, and the data tend to bounce around a bit here. “Noise” from varying stop times in the loop is compounded by a traffic signal at the exit to Lake Shore Blvd. Initially this was tripped by a detector when a streetcar was present, but this was changed to a simple timer. The result was that traffic was held needlessly for the often-absent streetcar, and when a streetcar did appear, it had to wait for the signal timer to cycle. Eastbound service faces a similar problem.
Lake Shore West of Humber to Long Branch
Running times on Lake Shore are fairly consistent with minor changes over the course of the day. Two periods show the effect of long-lasting delays:
- October 2013: For the latter part of the month congestion backed up westbound service from Brown’s Line to about half way to Kipling (the extent varied day by day). This was a period of bus operation on the line during the reconstruction of safety islands.
- April-May 2015: From mid-April, westbound cars were delayed between Humber Loop and Park Lawn. This is related to condo construction-induced congestion at Park Lawn.
[Thanks to contacts in southern Etobicoke for verification of projects that affected service on Lake Shore Blvd.]
Part of the period covered by these charts involved bus operation on a Lake Shore shuttle due to construction further east blocking through streetcar service:
- Late April 2013
- September-October 2014
Bus running times were slightly below streetcar times only for the AM peak eastbound, except for the period in October 2013 when construction on Lake Shore extended travel times westbound. Given that times are similar during most periods, this could reflect differences in stop service time during the inbound peak. A more detailed analysis is required to tease that info out of the data. A related issue would be the comparatively more frequent bus service replacing the streetcars and, therefore, the lower number of boardings to be dealt with at each stop along the way.
Terminal Times at Loops
The running times over the full route were measured from points nearby the termini, but did not include the time spent at each terminus. These values are explored for the month of May 2015 in the following charts. For example, the “SilverBirch” charts show the round trip time from SilverBirch east to Neville Loop and return. The shortest of the times recorded indicate the basic running time needed for this journey with any additional being the layover at the terminal.
The times spent by vehicles at the termini vary substantially from hour to hour and day to day. Detailed examination reveals a variety of situations:
- Some very long layovers, notably at Neville, are caused by delays at that location.
- Service disruptions in the west end can cause cars to arrive quite early in the east resulting in long layovers.
- Similarly, short turns in the east end can result in cars arriving early at the west end.
Even allowing for these factors, there is a basic question of whether long layovers sometimes result simply from the length of the journey, and a fair certainly by operators that if push comes to shove they can get short-turned. With Queen being such a long route, and with its scheduled running time being inadequate for most trips, there is no real incentive to be “on time”. This also begs the question of whether crewing for long routes like this requires some form of scheduled “step back” mechanism to build in reasonable breaks without holding cars for long periods at terminals.
In the final section of this series later this year, I will review a few specific days’ operation in detail to look at the question of how service is managed, or not, when things go wrong.