The City of Toronto is about to launch its review of how King Street “works”, and Chief Planner Jennifer Keesmaat has an op-ed in the Star, It’s time to reimagine Toronto’s streetcar ‘King’.
Both for personal interest and as part of my work for the TTC and City of Toronto on the behaviour of streetcar routes, I have been following the 504 King route for some time using TTC vehicle tracking data. This article updates the consolidated stats in anticipation of the King Street project with data to May 2016, and looks in detail at some of the current information from that month including vehicle speed profiles.
For information about the creation of these charts, please see Methodology for Analysis of TTC’s Vehicle Tracking Data.
A would-be rider cares foremost that a transit vehicle will show up soon (a short scheduled spacing, or headway, between vehicles), that it will show up reliably (consistent spacing between vehicles rather than gaps and bunching), and that it will have room for them to board and ride in tolerable loading conditions. These factors interact because inadequate service leads to longer waits and frustration with being unable to board.
Irregular service can transform what might in theory be adequate capacity into an inreliable mess with long waits followed by packed vehicles. If there is any empty space on the third of fourth vehicle to appear in a parade, it is of little use to people jammed into the first few cars, and that crowding produces its own self-perpetuating delay and a widening gap in service.
In the following charts, the average headways over each hour of operation, subdivided by week, are plotted from September 2013 to May 2016 (except for August 2014 and February 2016). Averages are shown in solid lines, and the standard deviation of the values is shown in a dotted line of the same colour. Typically, about 2/3 of the observations would lie within ±1 SD of the average, and the SD gives an indication of how tightly grouped the values are. If most service is running very close to the scheduled headway, then the individual headway values are close to each other and the SDs are small. If service is bunched and gapped, then there are many very short headways and corresponding long ones. That situation produces high SD values. When the SD and the average are close to each other, then much of the service is actually running as pairs and triplets of vehicles.
Headway values at terminals give an indication of the amount of short-turning that occurs on a route. This might be deliberate (a scheduled peak tripper operation, for example) or it might be simply the result of daily operations reflecting a different world from the official schedule.
Broadview Station is the eastern terminus of the King car. Headways here have been measured just south of Danforth Avenue so that congestion and queueing effects at the loop itself are omitted. (This is a particular problem northbound where streetcars can back up onto Broadview waiting to enter the station.)
During the AM peak, headways have been roughly the same for well over a year. A few items of note:
- The shorter average headway for the 7:00 to 8:00 am period is caused by bus trippers entering service southbound on Broadview. Their main purpose is to serve the central part of the route but they contribute to the total vehicle count and shorter headways from the eastern terminal during this hour.
- The spike in average headways in March 2016 is caused by the diversion around track construction at King & Charlotte (east of Spadina) which was operated with no additional scheduled running time. The result was an increase in short-turns to keep King cars on time.
During the midday, headways have been fairly consistent over the past year, and the improvements brought by revised schedules with added running time are evident in both the lower average and SD values thanks to fewer short turns. The spike for the Charlotte Street project appears here too as it will throughout the day.
PM peak service echoes the AM situation with the bus trippers entering service during the early part of the peak and producing shorter headways more typically found downtown.
Evening service, like the midday, shows an improvement over three years ago, but there is still a noticeable increase in SD values compared to earlier in the day.
By the time service reaches Yonge Street westbound, the average and SD values for headways are tracking almost equally during peak periods, and not far apart at other times. This indicates that peak service tends to arrive in bunches, and that off-peak service will be irregular with some bunching.
Late evening headways have improved thanks to a schedule change in late 2015.
An important issue with any transit service downtown is that the practical headway is a function of traffic signal timing. 80 second cycles are common, although the amount of green time King might receive varies from place to place. This effectively “dispatches” streetcars on 80 second headways whether that relates to the schedule or not. There is no transit priority at any signal between Church and University, and streetcars are at the mercy of signals designed primarily to aid with north-south flow of traffic.
Dundas West Station is the western terminus, and headways here are measure south of the station to avoid congestion effects at the loop (similarly to the situation at Broadview).
During the early AM peak, headways are comparatively short because trippers make one run up to Dundas West, but they do not return to the terminal with their service concentrated in the central part if the route.
As at Broadview Station, the off peak service has been substantially improved since new schedules were introduced in 2015. The effect of short-turns was especially severe on Roncesvalles Avenue with wider than scheduled headways and SD values close to the averages. Even the service that did operate to the terminal did so in bunches making for very wide gaps. The situation is somewhat improved now, although moreso for midday than for evening service.
Strachan Avenue is located at the boundary between Liberty Village and the Bathurst-Niagara neighbourhood. It is east of most of the common short-turn locations for King cars (although some turnbacks do occur at Bathurst and at Spadina). Headways here give an indication of the service provided inbound from the busy King West corridor.
As at Yonge westbound, the peak period average headway and SD values are roughly equal showing that service operates in bunches. Off-peak headways also echo other parts of the route with better-behaved service at midday than in the evening.
An important issue on this busy part of the route is vehicle capacity. Over time, King has seen a mixture or CLRVs (regular length streetcars), ALRVs (the longer articulated variety) and standard-sized buses. Many trips formerly operated with ALRVs are now served by CLRVs or buses, and some of the CLRV trippers have themself been converted to bus operation because of the shortage of streetcars. This has effectively reduced the route’s capacity, and the problem will only be solved by a return to larger vehicles (rebuilt ALRVs and/or the new Flexity cars).
The situation eastbound at Yonge is comparable to the westbound service with SD values showing that bunching is common particularly during peak periods.
Headway Details for May 2016
The following charts show the detailed data for May 2016 at the five locations used for the headway histories above. Several points are striking about all of these charts:
- The SD values track the headways closely throughout the day and even on weekends showing that bunching and gapping is the standard experience on 504 King.
- Headways exceeding the TTC’s target of a six-minute range bracketing the scheduled values are not uncommon, particularly on weekends.
- Headway reliability at terminals compared with other points on the route is somewhat better (lower SD values) on weekdays, but weekend service at terminals is quite unreliable (high SD values).
Looking at an even more finely grained view, here are the headways at various points along the King route for one day: Monday, May 16, 2016. These charts show quite clearly how service is bunched (a mix of low and high headways on either side of the expected headway) and that this problem begins at the terminal and continues at any point where trippers or short-turns may join into the route.
The service operated on May 16 was subject to only one major interruption, and that late in the evening when an incident at Jameson required two streetcars to divert eastbound via Queen and Shaw. Some congestion and other delay effects are visible:
- Up until about 7:10 am, eastbound cars are “delayed” at River Street. This is not a congestion problem or an accident, but something much more mundane: a Tim Horton’s coffee shop patronized by drivers on early trips.
- During the AM peak, congestion is visible most notably at Jameson Avenue (traffic bound for the Gardiner), and this actually gets worse after 9:00 am likely due to changes in parking restrictions. Some AM peak delays are evident downtown, but these are more related to longer dwell times near major intersections such as University than to pervasive congestion over several blocks.
- During the PM peak, congestion appears both ways at Bathurst Street, and to a lesser degree at Spadina and at University.
- During the evening, conditions are ideal with no congestion evident.
This is about as close to perfect as one can hope for on a weekday, and there were no major accidents, road closures, construction or a baseball game, not to mention congestion in the entertainment district which is affected by seasons, weather and the day of the week. Even under these circumstances, headways were less than ideal.
Travel Time Histories
Any review of changes to the operation of King as a street will want to see whether there was any actual effect. The following charts show the evolution of travel times from just east of the Roncesvalles/Queensway intersection and King/Parliament over the past three years. (Note that River was not used as the boundary because of the long-standing diversion via Queen around the closed bridge at King and River.)
Unlike the headway charts, the travel times are reported in half-hour intervals, but otherwise the format is the same. Travel times (aka “link times”) are shown in solid lines with the corresponding standard deviations in dotted lines of the same colour. The time intervals refer to the point where a vehicle crosses the beginning point (Parliament westbound, or east of Queensway eastbound). For example, data for “10:00 am” refers to the time taken for vehicles beginning their trip across the city between 10:00 and 10:29.
Notable in both sets of charts are the downward notches at New Year’s, although the effect was less pronounced in 2015/16 than in earlier years likely due to the better weather in the past winter.
Peaks upward are visible for the recent track construction diversion, and during earlier periods when Gardiner Expressway work diverted substantial traffic onto King, or triggered changes in access patterns with resulting delays at intersections leading to on ramps.
Although there have been some changes in traffic regulations on King, it is worth remembering that these affected only the shoulder-peak periods from 9:00 to 10:00 am, 3:00 to 4:00 pm and 6:00 to 7:00 pm. Any individual change would have to address an area of quite severe delay to existing service for it to have much effect on overall travel times across the route.
Some of the worst disruptions in travel times co-incide with construction projects either constraining road space and capacity, or forcing a service diversion. Although there are time-of-day variations in the trip times, they have been quite consistent during periods when such major upheavals were not underway. This begs the question of whether “increased congestion” is really a factor, or whether it is the prevalence of construction work (including the Gardiner, but also the many lane occupancies from condo construction) and increased activity by delivery vehicles that are the culprits in motorists’ perception of greater congestion. How much of that “congestion” is a series of events that more frequently are the norm that in years past, and how much of this is controllable?
Finally, there is the question of just where transit vehicles are delayed along King Street.
The following two charts are published as spreadsheets rather than in pdf format because of resolution problems with the exported pdfs from Excel.
These charts show the average speed by time and location along the King route for weekdays in the last half of May 2016. The westbound charts are read left-to-right while the eastbound charts read right-to-left. There are separate tabs in the spreadsheet for each hour of the day, and you can step through the tabs as a sort of “flip chart animation” to watch how speeds and congestion vary through the day. Where the values are low (notches in the chart), vehicles are not travelling very fast, on average, at a point. These correspond to any location where vehicles will slow, typically for a signal or transit stop. The width of the notch shows the degree to which vehicles slowed in advance of a stop typically due to congestion that prevents them from pulling righth up to the location with minimal delay.
The effect is particularly striking at major congestion locations such as downtown, eastbound to Jameson, and westbound to The Queensway, but it is visible also at other locations as an indicator of how much the speeds vary over the day (or not).
Of interest in coming weeks will be the removal of stops at some locations to, in theory, speed travel along the route. For example, stops will be removed both ways at both Simcoe and at York. The degree to which these affect travel times today at various periods can be seen in the charts. The effect is likely to be greater at times when the stops are actually used and streetcars fall out of step with the traffic wave on which signal timings are based, but also at Simcoe there is transit priority to extend green time, a function not found at York.
Two other locations of interest are on Broadview with the stops both ways at Mountstephen (between Gerrard and Dundas), and farside northbound at Queen (a stop introduced some years back for the Don Mills night bus). Both of these are at locations where there is no traffic signal, and so the delay, if any, is due to stopping for passengers. The stops at Mountstephen will remain (thanks to pressure from residents and the local Councillor) while the stop at Queen will disappear. On the speed charts, there is a drop in average speed for the farside stop at Queen, and to a lesser degree for the stops at Mountstephen, but these are not as severe as the drops at major stops such as Gerrard and Dundas where there are also traffic signals with transit priority.
When the June data become available, it will be interesting to compare the speed profiles before and after the changes planned for June 19.
The important point about the speed profiles is that delays both from traffic and from passenger service occur all along the route. “Fixing” the King car requires more than tweaking at a limited number of times and locations.