Updated November 18, 2013 at 6:15am: Broken links/filenames corrected.
Route 501 Queen is the longest of the TTC’s streetcar routes, and among the longest in the entire system (54 Lawrence East is longer as a surface route, and the Yonge-University-Spadina subway is longer than both of them).
The Queen car is the subject of unending complaints about service quality. It suffers the compounding effects of its length, its passage through some very busy sections of Toronto, and the fact that it operates with two nominally interleaved services.
During 2013, this route was subject to a number of disruptions through one-day events and from long-running diversions, but the operating schedule for the route was not changed except recently when construction on Lake Shore Blvd. required that streetcars turn back from Humber Loop. This provided a series of views of operations on one route under various conditions.
The table linked here shows the headways and running times for various periods of operation for the 501 Queen car during 2013.
The only schedule changes during 2013 were:
- April 1: Midday, early evening and late evening services improved to address overcrowding.
- September 3: Humber-Connaught AM peak trippers changed to operate only to Parliament.
- October 15: Service split at Humber Loop with buses running west to Long Branch.
No adjustments were made to running times throughout the period to allow for the effects of construction and numerous special events affecting the route during 2013 including:
- May 1: A protest march required a diversion of service in the early evening via King between Church and Shaw
- May 5: Goodlife Marathon
- May 20: Victoria Day (fireworks in The Beach)
- June 29 to July 28: Reconstruction of the intersection at Queen & York. Queen service diverted via King between Church and Spadina.
- July 1: Canada Day (celebrations at Nathan Phillips Square)
- July 8: Major storm and flooding
- July 25-27: Beaches Jazz Festival
- August 6 to 20: Reconstruction of the intersection of King & Spadina. King service diverted onto Queen, and spillover road traffic from King affected Queen and other neighbouring streets.
- June 23 onward: Reconstruction of Kingston Road. Depending on the state of open and closed sections, Queen east of Woodbine suffered from extra traffic diverted south from Kingston Road.
- October 5: Nuit Blanche
- October 20: Waterfront Marathon
One-day events are handled with diversions, extra service and ad hoc management. Construction effects should result in schedule changes, but nothing was implemented.
In this article, I will review the headways actually provided at various points along the route. This information is derived from the TTC’s vehicle monitoring system (the same data that feed the NextBus system and other web applications) for the months of:
- May (as a “before” reference),
- July (service diverted off of Queen),
- August (service diverted onto Queen from King),
- September (return to quasi-normal), and
- October (split service at Humber starting on Thanksgiving weekend).
This is an unusually long article with many linked charts because I am covering a lot of territory and want to give readers who are interested lots of material to see what is happening on the route. What is evident is that 501 Queen does not suffer from occasional upsets, but that it provides chronically poor service under a wide variety of conditions.
This article focuses on headways. In a separate article I will turn to running times and the degree to which insufficient schedule time contributes to erratic service.
Background to the Analysis
The process by which these data are transformed from the raw TTC files into various charts is explained in detail in a separate article Methodology for Analysis of TTC’s Vehicle Tracking Data. That article is updated from time to time independently of the main sequence on this site to reflect changes in the processes used.
All of the charts in this article show headway information from various locations over the five months under consideration. There are four sets of charts for each location/direction corresponding to:
- Weekdays (Monday to Friday except holidays)
- Sundays & holidays
- Means, maxima, minima and standard deviations
For the first three sets, each point on any chart represents one vehicle passing a location on a route at the time shown on the x-axis (horizontal). The height above this axis gives the headway (spacing) between this vehicle and the one preceding it. For each day’s data, a trend line is interpolated between the points showing the overall pattern of values. (Note that at the ends of the day, this line may veer off as the interpolation tries to “predict” the direction it is heading. Wider headway at the beginning and end of service can confuse the calculation.)
If service is reliable, all of the values should lie fairly close to each other, certainly within the TTC’s own target of ±3 minutes of the scheduled headway. However, in the overwhelming majority of cases, this is not what the charts show. Regardless of the location, time of day or season, widely spread headway values well beyond the target band are common together with many very short headways showing that service is bunched.
Where the trend lines lie together on a chart (or on a series of charts), this indicates that overall the headways behave in a similar way from day to day, on average. The trend lines do not measure the dispersion of the values, and the same trend line could be produced by tightly clustered or widely spaced data values. The degree of dispersion is easily seen by looking at the charts and comparing values for various dates and periods. Simple visual inspection shows the degree to which many headways are quite outside the target range for TTC service and that this is a common situation.
A more abstract view of the data is shown in the fourth set of charts. These provide, in half-hourly intervals, the mean (arithmetic average), maximum, minimum and standard deviation of the values. The implication of each of these values is:
- Mean: This is the average headway of vehicles passing a point during a specific half-hour on the days in question for the month in question (e.g. from 9:00 to 9:30 on all weekdays in May). This gives no indication of the variation in values, and it collects together the effects of several days (compare to the first three sets of charts which give data for individual days). Headways of 5-5-5-5-5 produce the same mean as headways of 1-3-5-7-9, or even a very skewed set of values with many low numbers (bunched cars) followed by one large one as in 1-1-1-1-21.
- Maximum: This is the largest headway observed at the point during the half-hour on the days and month in question. One very large value in all of the days can make the maximum quite large even though they may represent few actual trips. However, the consistently high maxima, especially on weekends when there are fewer days and trips contributing to the data, suggest that large headways are not uncommon. This echoes the large number of high data points on the headway charts.
- Minimum: This is the smallest headway observed. Typically, the numbers will be quite small indicating that at least one pair of cars was running as a bunch during the times and dates in question.
- Standard deviation: This is a measure of the dispersion of the data. If the values were tightly clustered around the mean, then the standard deviation would be small. However, in many cases the mean and the standard deviation are similar indicating that there is a reasonably good chance that the headways lie between zero and 2H where “H” is the mean headway. In practice, this means that cars are running in pairs or closely grouped a great deal of the time. The TTC’s goal is that 70% of the service operates at ±3 minutes of the scheduled headway, or a band six minutes wide. With the shortest scheduled headway on 501 Queen at about 5 minutes, an SD equal to the mean headway shows that the band holding about 2/3 of the data is 10 minutes wide at least, and considerably wider during less frequent service. To put it another way, to achieve a goal of having 70% of the service within three minutes of schedule, the mean must match the scheduled headway, and the SD cannot be much more than 3 minutes. Neither of these is true much of the time.
I will review each section of the route separately to discuss local quality issues and the effect of various factors that changed the environment in which service operated.
With the exception of a few days, the trend lines for weekday headways westbound at Yonge Street are almost identical. What this shows is that the service provided at Yonge is, on average, about the same for any weekday we look at. However, the range of headway values generally lies in a band from zero to 20 minutes, and it is worse during the weeks of the King Diversion and, to a lesser extent, in August when King traffic was shifted north.
The eastbound charts are quite similar with consistent values on the trend lines, and headways scattered in a 20-minute band, worse during periods of diversions.
What is quite striking in all of this is the small number of cases where the trend is pulled away from its “normal” position. This indicates that the average level of service was consistent (in the sense that most of the trips operated every day), but that says nothing about the dispersion in the data values. For this, we turn to the statistics.
Consistently, the standard deviation of the headway is in the 4-5 minute range except in the morning. This means that the service is fairly well organized until late in the peak period, but that it never regains this level of headway regularity for the rest of the day.
On Saturdays, we see the same pattern as on weekdays, but the summer months and September are particularly bad with the “cloud” of data points expanding into the 20-30 minute range particularly in the evenings. From midday onward, the standard deviation tracks closely to the average headway showing how badly dispersed the data values are.
Sundays are even worse. Although daytime headways are scheduled to be no worse than 6’30”, many trips arrive in gaps of over 20 minutes in both directions. As on Saturdays, the standard deviation tracks the mean headway values during afternoons, particularly for July to September.
This is the service provided at the central point of the 501 Queen route, a point where, if anywhere, regularity and reliability are essential.
For both directions, headways are shown at two points:
- The charts for Woodbine show the service entering and leaving The Beach east of two common short turn locations: Russell Yard and Woodbine Loop (at Kingston Road).
- The charts for Greenwood show the service before the short turns.
The effect of short turns is quite evident in the difference between the location of the trend lines (which generally follow the average values) and the scheduled headway. At times the entire trend line lies more than three minutes away from the scheduled value showing that the level of service actually provided is, on average, outside of the target range. Also visible is the very wide range of headway values, notably after the AM peak when short turns are used in an attempt to get cars (and operators) back on time.
The same effect shows up on the statistics charts for weekdays. What is also quite clear here is that by the end of the AM peak, half of the service never reaches Neville loop and this situation stays constant through the day. It is quite clear that this practice is intended to get cars (and operators) back on time after one trip across the line during the peak. The turnbacks kick in sooner in months where diversions cause trips to be even longer than usual and cars/operators are running late earlier in the day.
Weekends are particularly bad with the trend line lying above the scheduled headway value throughout the day, and many headways at double, triple or more the scheduled values.
Standard deviations tend to be lower westbound early in the day because cars leave on an approximation of a regular headway, but once the short turns begin, there appears to be little effort to space service that originates from Neville and from Woodbine Loop. This causes bunching to begin westbound at the east end of the line and grow in severity as cars move west.
For eastbound service, the standard deviations are higher because these cars have already travelled the length of the city and bunching is quite common. We will see below what this looks like at the origin of the trips in the west end.
In the west end, headways are shown at various locations for each direction.
- Triller is just east of Roncesvalles, and shows the headways after the service has passed through the congested Parkdale neighbourhood but before most of the short turning occurs (some cars short turn further east, notably at Shaw, but this is a comparatively minor part of the service).
- Parkside Drive (the east side of High Park) is within the section still nominally served by both the Humber and Long Branch services, but it is beyond the Roncesvalles/Sunnyside short turns.
- “HumbLkShore” is the exit from Humber Loop onto Lake Shore Boulevard. This is the beginning of the route segment unique to the Long Branch service.
- The LongBranch charts show headways just east of Long Branch Loop with the effect of short turns at Kipling Loop. Note that the scale on statistics charts for locations west of Humber Loop goes to 60 minutes rather than 40 to make room for (most of) the longer headway values observed here.
- Royal York is midway along the Long Branch section of the route east of the Kipling Loop short turn.
- “QsyHumber” is the exit from Humber Loop onto The Queensway, and it is a point where the inbound Humber and Long Branch services should be blended into an even headway.
- “Triller” is just east of Roncesvalles, and headways measured here pick up cars entering service from the carhouse or from a short turn at Sunnyside Loop.
Westbound service at Triller shows the same pattern as in the east where bunching is common as cars make their way across the city. The minima on the stats charts are usually close to zero, and the maxima are considerably higher than the mean headway. The mean stays fairly close to the scheduled headway (i.e. most service actually gets this far) except in July when delays caused by rerouting onto King Street caused a lot of short turns east of this point.
Saturdays and Sundays show the common pattern of a low standard deviation early in the day, but by mid-morning, the SD is up to the same level as the mean showing chronic bunching.
By the time we reach Parkside Drive, some of the service has short turned at Roncesvalles. The mean headway is noticeably above the scheduled value, especially in July when, on weekdays, it sits at double the expected value. The SD values are not quite as close to the mean as at Triller because some pairs of cars have been broken apart with a short turn (usually a “Humber” car).
The mean values on weekends also show the effect of short turns when these charts are compared to those at Triller.
West of Humber Loop, service can bear little resemblance to the schedule. On a nominal headways of about 10′ peak and 12′ midday, the means stay fairly close to expected values in May. In July, however, the means rise again due to the diversion via King. This shows that some of the scheduled Long Branch service did not even get past Humber Loop. The statistics settle down again later in the year. Note that October’s stats contain a mixture of through streetcar service and local bus shuttles. Even with the buses, however, headways are not particularly reliable as is evident on the headway charts.
At Long Branch Loop, the effect of short-turning is clear. Even in May when there were few disturbances, the midday means climb above the scheduled headway showing the effect of short-turns at Kipling. This is particularly severe in July. By October, when the stats are dominated by bus service, the weekday numbers are back down to expected levels, although the SD value is still higher than appropriate for a relatively wide headway.
Weekends are particularly bad with extremely wide mean headways and large SD values. Service to the outer end of the line is considerably worse than advertised, and what’s more, it does not arrive on anything approaching a reliable basis. In spite of the wider headways, minima of zero to a few minutes are not uncommon showing that cars travel to Long Branch in pairs with wide gaps between them.
The well-behaved weekend service with Long Branch operating as a separate route shows just how bad the situation is thanks to the through-routing of service with 501 Queen.
With the streetcar service cut back to Humber, weekend afternoons see a good deal of short turning at Roncesvalles (compare mean headways at Triller and at Parkside for October).
Eastbound at Royal York on weekdays in May, mean headways and SD values are fairly well-behaved although the SDs do grow later in the day. The contrast with July is quite striking when all of the lines (mean, SD and maxima) shift upward showing that there was considerably less service on Lake Shore during the King Street diversion thanks to short turns. What service did operate did so with an SD value of about 10 minutes (i.e. two-thirds of the service was on headways ranging 10 minutes either way from the mean). This is not the sort of reliability the TTC advertises. The situation improves as the summer wears on, and by October, the effect of the local bus shuttle show up.
Weekend service in May shows quite reasonable stats until after 9am with the means, minima and maxima close together and low SD values. However, as the day wears on, the headways become less reliable. Similar patterns show up in succeeding months.
Eastbound from Humber Loop to The Queensway, the mean headways are greater than the scheduled values showing the effect of short turns further east. Also, at times, the SD values are close to the headways implying bunching. In other words, cars from Humber and cars from Long Branch tend to leave together. I will explore this problem (and a similar issue at Woodbine Loop) in a later article.
At Triller eastbound, the mean headways are close to the scheduled value except in July when there are also some very high maxima and SD values. Weekends echo what we have seen elsewhere on the route.
Conclusions and Next Articles
The following conclusions arise from the data presented here:
- Even during a “normal” month, service on 501 Queen becomes erratic late in the AM peak (or on weekends in the late morning) and stays that way for the balance of the day.
- Chronic short turns reduce service on the outer ends of the line, and this is particularly so when major events such as a diversion for which there was no scheduled provision make it impossible for cars to keep on time. Headways beyond the short turn point are not reliable.
- Cars run in pairs commonly on this route. They do so inbound from outer parts of the line, and the bunching is accentuated as they travel across the city.
- Periods with wider headways, notably weekend evenings, have more erratic service than those with shorter headways.
- The “three minute standard” is not achieved by most of the service.
In future articles I will review the actual running times required for services to operate under various conditions, and the degree to which these differ from the scheduled times. I will also review the operation of scheduled and unscheduled turnback points to see the degree to which services blend and short turns fill gaps, as opposed to simply creating new pairs of vehicles that travel together.