This article continues the analysis of the King Street Pilot with May 2018 data. Part I is in a separate article.
Many topics are covered here, and the intended audience is those readers who want to look at the details of how routes operating on King Street behave. The sections include:
- Headway reliability on 504 King both in the downtown pilot area and at the terminals.
- Short turning on the 504 King car.
- Headway reliability on 514 Cherry and 503 Kingston Road.
- Headway behaviour for the consolidated 504, 514 and 503 services downtown.
- A detailed review of 504 King operation on certain days in May 2018.
A common factor through this article is that while travel times in the pilot area have improved thanks to the transit priority scheme, the headways on all affected services are erratic, especially on 514 Cherry and 503 Kingston Road.
The pilot transit project on King is intended to make service more attractive to riders, and this can occur in various ways:
- Travel times through downtown are shorter and more reliable than during the pre-pilot period.
- Route capacity has improved partly from schedule changes and recapture of excess travel time, but mainly through the replacement of the older streetcars used on King with larger ones, primarily the new Flexity cars.
- Headway reliability (consistent times between cars) can be improved if random congestion events are reduced or eliminated, and reliable travel times lessen the need for short turns.
No one change by itself “solves” King Street’s problems, and all three are needed to achieve benefits that will attract riders.
An important part of a transit journey is the time spent waiting for a vehicle. One reason the subway is so popular, aside from its speed, is that trains come frequently and reasonably reliably, although this has become a sore point in recent years. Surface route reliability has always been an issue, and although trips may be faster, there is still the issue of how reliably vehicles actually show up to carry riders.
The King Street Pilot has improved travel times, but problems remain with gaps and bunching.
In the following sets of charts, the headways (time between vehicles) is subdivided into four percentiles:
- The worst case, or maximum, headway at a location (100th percentile)
- The typical longest headway experienced by most riders (85th percentile)
- The median headway (50th percentile)
- The headway at or below which one quarter of the service operates (25th percentile)
A knock-on effect of bunching is crowding with more riders on a “gap car” and suffering poor conditions on board than overall average demand/capacity ratios would imply. When cars arrive in pairs (or worse), there is a good chance that more than half of the riders will be on the first car. This is not just a case of lemming-like behaviour, but a basic concern that the second car might be short-turned leaving riders to wait, yet again, in a gap.
In these charts, note that only 504 King cars are included. A separate set of charts with the 514 Cherry and 503 Kingston Road services merged in comes later in this article.
As an example, here are snapshots of the headway stats westbound at Yonge Street during the AM peak (8-9 am), midday (1-2 pm) and PM peak (5-6 pm).
For 8-9 am, the median headway lies in the 2-4 minute range as one would expect based on the scheduled service. However, the maxima consistently sit in the 10-minute range showing that a wider gap occurs at least once during this period every day. Similarly, one quarter of the cars operate at a headway of two minutes or less indicating that they are part of a bunch.
There is no visible difference in the values before and after the pilot’s implementation in mid-November 2017. This is unsurprising on two counts. First, the condition of service arriving westbound at Yonge is determined by the route’s operation largely outside of the pilot area (east of Jarvis). Second, from travel time stats, we know that the pilot itself has almost no effect on the AM peak period which did not show congestion effects before the pilot began.
By midday, the range of values has opened up with higher maxima, and a continuing presence of bunching as shown by low value of the 25th percentile. There is some improvement in shaving off the highest of the maxima and the median values are also slightly improved compared to pre-pilot values. However, note that early 2017 stats are also better than those for the period from May to October, and so an obvious question is whether the winter 2018 improvement is more a seasonal effect than a result of the pilot itself. Note the rise in the 50th and 85th percentile tracks beginning in May. Until more data come in, we will not know if this is a “blip” or a “trend” leading into the summer.
In the PM peak, the maxima are lower than at midday and some improvement is visible when the pilot begins in November. A further improvement show up in the 85th percentile with the mid-February schedule changes. However, the median and 25th percentile lines remain low showing that bunching remains a problem.
Service in the evening hours shows a continuing wide spread between the 25th and 50th (median) values and the 85th and 100th lines, and this has been worsening since early 2017. In the 8-9 pm period, the 85th percentile routinely lies on either side of the 10 minute line, and the maxima around 14 minutes. Riders will always encounter service that is a mix of short and long headways, and the service quality is noticeably less consistent than in early 2017.
The late evening situation continues the pattern.
The eastbound stats show similar patterns with a slight improvement in reliability during the PM peak, but deterioration in service quality off-peak especially in the evening periods.
The full chart sets as pdfs are linked below.
At 504 King’s Terminals
The TTC’s premise about service quality is that if service leaves the terminals on time (within reason), then headways along the route will look after themselves. This is a false premise on two counts:
- The service at termini does not stay within the target values even though these provide a generous six-minute window for being “on time”. For routes like 504 King where scheduled headways are short, it is possible for two or three vehicles to run as a pack and still achieve the “on time” goal.
- Uneven headways inevitably cause bunching as the gap car carries a heavier load and its follower(s) catch up. Headways along a route are never as good as at the termini.
For the purposes of this analysis, the headways inbound from terminals are measured just south of Danforth (at Broadview Station) and just south of Bloor (at Dundas West Station). This avoids problems with congestion at the terminals.
The patterns at both terminals are similar to what is seen at Yonge Street with 85th and 100th percentile values well above the scheduled headway during all periods, and matching low values in the 25th percentiles. This shows that riders will commonly encounter wide gaps and bunches of cars. The problem is worse in the evenings and the spread in values has worsened in recent months.
When the TTC publishes reliability statistics for its routes, the data are averaged over entire days. Hour-to-hour variations and the degree to which riders face ragged service are notpresented in detail. Even then, 504 King does not perform anywhere near the TTC’s target.
The TTC is quite proud of statistics showing that across the system as a whole, short-turning has been substantially reduced in recent years. The problem with their stats is that they are averaged over all routes and time periods, and do not show the actual behaviour of individual routes and locations.
The chart below shows for the month of May 2018 a measure of the degree of short-turning on the King car westbound. Data are accumulated for each week in the month, and weekends are lumped together (including Victoria Day). The values show the percentage of cars leaving Yonge Street within each hour that actually reached Dundas West Station. Many cars do not get there. In some cases, these trips are built into the schedule and this shows up where data for every week have similar values. Moreover, these trips occur primarily at the peak and shoulders, not on an all-day basis.
With all carhouse operations for 504 King now working from Russell and Leslie Barns, there are no trips scheduled to terminate at Roncesvalles as cars leave service at the ends of peak periods.
Note that although the situation is not quite as bad for weekend service, about a fifth of westbound trips do not reach Dundas West even on Saturdays and Sundays.
Measured at Roncesvalles, the percentage of trips is much higher showing that this is the common short turn location. (Other trips end at Dufferin, Bathurst or Spadina.)
The situation at Broadview is a more complex because of the way the route operates.
- Some eastbound cars short turn via Parliament and Dundas to Broadview, while others use the reverse looping. The latter cars will show up in eastbound counts at the Don River and King, but they do not actually serve Broadview Station.
- Some eastbound cars are scheduled to go out of service eastbound from Queen and Broadview, while others are scheduled to run in from Broadview Station. This has two effects. First, the cars making run-in trips from Dundas West to Leslie or Russell tend to run early for their entire trip across the city. Second, some cars that are scheduled to run in from Broadview Station actually leave service directly eastbound at Queen.
At the end of the AM peak, 12 of 39 cars in service on 504 King (May schedules) return to the carhouse, or just under one third of the service. This corresponds to the dip in the chart below, but that dip is considerbly deeper and wider than the removal of those cars would trigger, and this indicates that additional runs are being short turned.
There are also scheduled changes in service levels through the evening from 39 peak cars down to 26 in the early evening and then 14 in the late evening. Some of the scheduling effects are visible in the dips at 8 pm (20:00) and again at 10 pm (22:00), but again these are deeper than the change in service levels alone would produce.
Some of the King service is short turned at Parliament Street, and this shows up as a drop in cars reaching River Street (at the Don River) compared to those that left Yonge Street eastbound.
A further consideration about short turns is the degree to which cars re-enter service properly splitting gaps, or simply rejoining a parade as noted in Part I of this analysis. Even assuming that cars left terminals on a well-regulated basis, there is no guarantee that this would mesh with short turns joining the route along the way.
514 Cherry Service
Service on 514 Cherry has been extremely erratic ever since the route was created. Although it is comparatively well-behaved during the AM peak, headway reliability quickly deteriorates through the day and very wide gaps, many well over half an hour, are common at the two terminals. Anyone hoping to catch a car to or from the Distillery District has a good chance of finding themselves with a long wait for service. The degree of this problem goes very substantially beyond the standards the TTC sets for itself.
During the morning peak, the median has been well-behaved since the route began operating, but there are extremely large headway spikes, especially recently.
In the midday, maximum headways of over half an hour were common until the schedule change with added service in mid-February. However, the maxima and 85% lines have started to rise since mid-April.
During the PM peak, service is better than the midday period, but is not much different from values during the winter of 2017.
The full chart sets for both directions are below. Headways from Dufferin Loop are also irregular, but this is not as much of an issue because there is other transit service on Dufferin. The 514 service appears to operate with little attention to headway reliability and simply acts as a supplementary service on King Street as and when it appears.
503 Kingston Road Service
Since February 19, 2018, service to Bingham Loop is provided weekdays until the end of the PM peak by the 503 Kingston Road route. the western terminus was changed from York Street to the Spadina/Adelaide/Charlotte loop. (In late June, the 503 was consolidated into the 504 King service for the duration of various construction projects. In September, the 503 will revert to bus operation terminating at York Street.)
Prior to February 19, daytime weekday service was provided by 502 Downtowner operating as a bus from Bingham Loop to University Avenue. The 502 was changed to peak-only when the 503 became the primary route to add to service on King Street.
A long-standing problem on Kingston Road has been the reliability of service regardless of the route or mode serving this line. In spite of a scheduled headway of 10 minutes or more, it was common to see pairs of 502 Downtowner buses running close together, just as one saw pairs of 503 Kingston Road streetcars from February onward.
Headways leaving Bingham westbound are extremely erratic. The chart below shows the data for the third week of May 2018. This is typical. (Note that the extreme “tails” on the right end of the chart are a side effect of Excel attempting to fit data at the edge into a trend line.)
What is quite astounding here is the scatter of headway values over a half-hour wide interval with many points close to the horizontal axis, or zero. These are cars that left the terminal immediately after their leaders.
Here are the headways for Tuesday, May 15 split out to show the wide variation in service at Bingham, and this shows a complete absence of headway discipline on the route.
The situation westbound at Coxwell is only marginally better thanks to some short turns at Woodbine Loop, but wide swings in headways remain.
The headways outbound from downtown to Kingston Road are equally disorganized. Here is the chart for Week 3 of May 2018 eastbound at Yonge Street. As with the westbound service from Bingham, note the large number of headways over 20 minutes and close to the zero line.
When the full month’s statistics are combined, the worst of the highs and lows disappear into the average value, but the scatter shows up in very large values for the standard deviation.
When the SD value is close to the average (arithmetic mean) this indicates that well over half of the headways vary from that average by the headway value itself. In other words, Cars tend to run in pairs followed by gaps of almost a double headway. Even during the AM peak, the SD value indicates that headways are scattered over a range considerably above the TTC’s own “six minute rule” for “on time” performance.
Finally, this is not a new problem. Here are the headway percentiles from Bingham Loop inbound from 8-9 am for February to May 2018. As the detailed data above show, the morning peak is not too badly behaved, although it was badly disrupted during periods of cold weather and snow in late winter. In particular, there is little bunching and the 25th percentile line stays at around the 10 minute mark.
Midday is quite another matter, and the situation has been getting worse since mid-April (this is a common observation across much of the data for all routes reviewed here). The maximum headways are getting longer, and the 25th percentile line stubbornly remains well below the median. (Note that there is no data for early February because the 503 did not operate at mid-day until February 19.)
By the PM peak, the situation is even worse, and maxima well over 20 minutes are common.
The situation at Yonge Street eastbound during the PM peak is similar. Maximum waits of well over 20 minutes are common, although why anyone would bother is a mystery.
The full sets of charts are linked below as pdfs.
The Effect of Consolidated Routes on King Downtown
With three routes operating on King Street between Sumach and Spadina, the combined headway for that section is better than portions further out. Of course, the better service can only be fully used by riders travelling within that area. There is no co-ordination of the three routes and the degree to which they fill gaps in each other’s service is a matter of luck, not of planning or management.
Here are the headway statistics for the 504 King cars at Yonge Street westbound for May 2018.
And by comparison, the combined service of all three routes. In both cases, but especially with the combined service, the standard deviation values lie close to the averages indicating that service tends to run in pairs. For very close average headways, this is inevitable given the random interference of traffic, loading delays and signals, but it is not a good sign when this situation occurs outside peak periods when scheduled service is less frequent.
The eastbound service has a similar behaviour.
The pdfs linked below contain the full chart sets including weekend data which show a similar tendency for cars to run in pairs.
Note: In the charts below, only 504 King cars are shown. 514 Cherry and 503 Kingston Road cars are not included.
These charts show the position of vehicles over time. Distance runs east to west from bottom to top, and the line is charted as if it were a single straight line from Broadview Station to Dundas West Station. Cross streets are shown in scale to their location along the route. For the purpose of this analysis, the route is divided into 10m segments, and GPS vehicle positions are mapped onto these segments.
Time runs from left to right.
The effect is that each line represents a car moving back and forth across the route with the slope of the line indicating its speed (a horizontal line shows a stationary vehicle). On a few occasions, a car will appear to reverse direction briefly – this is caused by GPS errors that were not filtered out in the data cleansing that catches most of these events before the charts are produced.
When cars go off-route for diversions and short turns, they disappear where a car leaves the route and reappear when they rejoin it.
Wednesday, May 1
There were two delays during the PM peak, one of which was within the pilot area.
A westbound delay west of Dufferin triggered a diversion via Shaw and Queen (westbound cars can be seen “disappearing” west of Strachan and reappearing at Roncesvalles. The more interesting effect, however, is with the parade of cars after the delay clears. Visible in the chart are capacity backlogs both at Dufferin, and a much greater problem at The Queensway. This is a challenge for any increase in service through Parkdale – the degree to which frequent service cannot get through the Queensway intersection. This would be worsened if there were any attempt to add a Harbourfront West service via King and The Queensway.
A similar, but less severe, problem can be seen eastbound at Spadina after a delay just west of there clears. Cars approaching the intersection do not get through quickly when several of them are present. (The situation is much worse during diversions if cars have to turn rather than running straight through.)
A related issue is the location of farside stops which, with the larger Flexitys, can be blocked by cars sitting in the stop zone. Although the stop service time may be brief, this can create just enough of a delay that the following car is held nearside by the traffic signal.
Monday, May 7
The early midday period was not disturbed by any major delays. Westbound bunches from Broadview Station are visible in the chart below, although they were better spaced out coming back from Dundas West Station and with a few short turns at Queen & Roncesvalles.
During the afternoon, the pattern with some bunching continues. One visible problem begins westbound from Broadview Station just after 1 pm when a car that is on a reasonable headway runs progressively later, especially through downtown, and other cars catch up forming a bunch. Another pack of cars forms at about 1:30 westbound. Some of these are short-turned at Roncesvalles, but they go back east as a bunch rather than filling in the gap coming from Dundas West Station. Another gap and bunch form in part from a minor delay westbound at Jarvis just after 3 pm.
By the time the westbound parade reaches Dundas West, some cars have been short turned at Roncesvalles, but only one of them actually fills the gap while others simply rejoin the parade. Some of the eastbound trips will short turn at Parliament causing a large gap to Broadview Station at about 5 pm.
The situation is complicated by a delay east of Bathurst that caused a diversion via Shaw, Queen and Spadina between about 5:00 and 5:40 pm, although the gap caused by this was, ironically, partly offset by delays to streetcars eastbound at University that spaced out a parade.
Many cars short turned westbound at Roncesvalles between 5:40 and 6:20 pm.
Through the early evening, some parades of cars remained, and a few large gaps to terminals were created by short turns. This has the appearance of trying to keep streetcars (or operators) on time without regard for actually maintaining reliable service. Although one could argue that if cars stay on time, the service quality will look after itself, but that is not what happened here. A close look will reveal short turns at many locations (westbound at Roncesvalles, Dufferin and Bathurst; eastbound at Church, Parliament and Broadview/Dundas), but cars emerging from the short turns do not always “split” the gaps and can simply contribute to bunching in the opposite direction.
Layover times at terminals, particularly at Dundas West Station, were generous indicating that the service was not running chronically late.
Tuesday, May 8
A delay eastbound at John during the 1-2 pm period contributed to a spike in the travel times through the pilot area, but this was only one event among other gaps and bunching in service.
Wednesday, May 16
A delay after 8:30 am eastbound near John trigered a diversion via Spadina, Queen and Church …
… but the effects are sorted out after 10 am. Many cars short turn eastbound at Church between 11:30 and 12:30.
During the early afternoon, a delay eastbound near Yonge creates a small bunch in the service, but this does not last long. One car short turns at Broadview/Dundas and appears shortly later westbound from Parliament (the pink line just after 2 pm).
The service continues to be fairly regular through the afternoon peak. Note the large number of cars short turning at Queen/Roncesvalles westbound and at Parliament or Broadview/Dundas eastbound, but comparatively little bunching.
Wednesday, May 23
A delay westbound at Spadina at noon triggered a diversion via Queen and Shaw, but the service was already bunched coming east from Dundas West Station.
Through the early afternoon, gaps remained in the service, although a large one was created at Dundas West from about 2:15 to 2:45 by westbound short turns.
By the afternoon peak, the worst of the gaps had been ironed out, but bunching remained a problem. Note the backlogs of cars westbound to The Queensway, and a delay at Queen/Broadview at about 5:50 pm triggering a diversion westbound via Dundas and Parliament (cars disappear at Dundas and reappear at Parliament westbound).
Bunching and gapping continued into the early evening.
Many different issues show up in these charts, but what is generally not evident is severe traffic congestion, especially downtown.