The Queen streetcar is the subject of much discussion here, and I have been remiss in failing to post an analysis of the CIS (Communications & Information System) data for this route to substantiate many of my (and everyone else’s) observations.  Over the next few weeks, in preparation for a Rocket Riders’ meeting in early December, I will post a series of articles looking at the line’s operation in detail.

For those who are unfamiliar with the sort of analyses that will appear here, please read all of the articles about the King route filed under Service Analysis on this site.  As I write this, there are nine of them (with one more to come), and you should read them in order.  They include some of the background on how the CIS system works and the various ways I have sliced and diced that data.  I will not repeat this information here in the interest of brevity.

By way of introduction to the data, this post deals with Christmas Day, 2006.  This is important for a few basic reasons:

• Operating conditions on Christmas were as close to ideal as one could ask for.
• There was no traffic congestion.
• There was no inclement weather.
• Passenger loads were modest.

Collectively, this means that the observed behaviour of the line shows what happens when most of the sources of random delay are eliminated.

There are five charts (in PDF form) here.

Queen 2006.12.25 Service Chart

Queen 2006.12.25 Headways Westbound

Queen 2006.12.25 Headways Eastbound

Queen 2006.12.25 Link Times Westbound

Queen 2006.12.25 Link Times Eastbound

Overall View of Service

The Service Chart shows the travels of cars back and forth across the route in graphical format.  Time runs horizontally over the eight pages of the chart from 0400 to 2800 (for the sake of these charts, the clock does not roll over to 0000 at midnight).  I have not included run numbers on the charts, and in the discussion refer to cars by the colour of their chart lines.

Neville is at the bottom and Long Branch is at the top.  Major locations on the route have the following co-ordinates:

• 0 = Neville Loop
• 100 = Kingston Road
• 120 = Connaught (Russell Carhouse)
• 200 = Broadview
• 300 = Yonge
• 400 = Bathurst
• 500 = Dufferin
• 575 = Roncesvalles & Queen
• 600 = Parkside Drive
• 700 = Humber Loop
• 815 = Mimico Avenue & Lake Shore
• 980 = Kipling & Lake Shore
• 1080 = Long Branch Loop

This scale is roughly 100 = 2 km.

Moving through the chart, you can see quite clearly the Humber cars and the Long Branch cars and how they mesh (or don’t) through the day.  There are several artifacts of CIS and of the data digestion I have done that should be noted:

• Approaching Neville eastbound, CIS often “loses” a car until it leaves Neville Loop westbound.  The loss occurs roughly at Wineva Avenue where, irony of ironies, a senior TTC planner lives.  The net effect of this is that we don’t see the arrival time at Neville, although it can be inferred from the graphs.  Likewise, the layovers at Neville don’t show up explicitly as they do at Long Branch and Humber.
• A similar problem exists at Humber, but not with quite as serious an effect, where there is a “black hole” roughly from Windermere to Humber.  The data is missing either an arrival or a departure at Humber, and the layover does not appear as a straight horizontal line.  Visually, however, it is obvious what is going on.
• Every so often, there is a small blip in the lines caused by CIS re-adjusting its location for a vehicle to bring odometer readings into sync with an observed signpost.  (You can read all about signposts and their limitations in the King route writeup.) 
• On occasion, a car appears to move over very long distances instantly (teleportation, as I have called it), or to bounce back and forth across the city.  Both of these are due to CIS getting confused about where the car really is.
• CIS can get confused about the direction a car is travelling.  For example, at about 1115, a car that came eastbound from Long Branch is shown as returning westbound and eventually “teleporting” from Long Branch Loop to just west of Yonge Street.

A few points are immediately obvious on this chart:

• Layovers at Long Branch are quite generous on days when there is little traffic.  10 minutes is common, but much longer ones occur to the point that two cars are in the loop together even on a 15 minute headway.  Some cars have a layover eastbound at Humber in addition to their layover at Long Branch.
• The wiggles in the line indicate variations in speed across the city, but there is no actual service blockage that would show up as a gathering together of many lines at one time and location.
• Short delays for crew changes are visible at Russel Carhouse (120) and at Roncesvalles (575) where cars may sit for roughly 5 minutes.
• An example of how crew changes affect service can be seen in the car leaving Long Branch eastbound a little after 1000 (the dark blue line).  This run catches up to the preceding Humber car (dark green), and they are about 3 minutes apart by the time they reach the east end of the line.  The blue car leaves soon after the green one, but then sits westbound at Russell Carhouse for close to 10 minutes, almost certainly for a crew change.  Meanwhile, the following car (yellow) catches up and the two go across the city together.
• Other cases of cars travelling together as pairs because they leave the terminus at roughly the same times are visible through the day.
• Short turns are rare.  One car (dark pink line) lays over at Long Branch from 2054 to 2149.  On its eastbound trip it short turns at Russell.  That eastbound trip, by the way, is part of a parade of four cars carrying a gap that is 25 minutes wide by the time they arrive at Yonge.
  • The first (orange) leaves Humber eastbound after an unusually long layover just before 2200. 
  • The second (dark blue) came from Long Branch, but waited at Humber for “orange” to leave.  This car runs in to Russell Carhouse eastbound.
  • The third (light pink) came from Long Branch just before dark pink.
  • The fourth (dark pink) catches up at Humber.

Headways

The two sets of graphs show the headways at various points along the line.  Note that in both cases, Neville Loop is on page 1 and Long Branch is on page 15.  A trend line is interpolated through the data on each page to show the overall behaviour.  This line tends to stay close to the scheduled headway, more than we can say for the service.

Various notes about these charts:

• Westbound leaving Neville, headways lie in a band of plus-or-minus four minutes for a great deal of time.  This has nothing to do with congestion, but the irregular departures (as we already saw on King) are the direct cause of uneven service across the line.  Around 0920, there is a small convention of cars at Neville and again at about 1023.  (This can also be seen on the service chart.)  This sort of thing happens from time to time with pairs of cars leaving close together.  This happens too often to be put down to something like equipment problems.  Possibly the operators are exchanging Christmas gifts.
• The irregular headways continue across the route and by the time we reach Yonge, they mainly lie within a 6-minute band.  This is actually “on time” performance by TTC standards which treats 3 minutes, one way or another, as acceptable.  The impact on the headways seen by riders is obvious.
• By mid evening, headways are getting ragged due to cars running as pairs.  There is nothing in the service chart to give any reason why this is so.
• Once we reach Parkdale, headways start to open up as cars that are on short headways catchup to those in the lead.  The band of headways is roughly eight minute wide and this exceeds even the TTC’s generous plus-or-minus three goal.
• By the time we reach Humber, headways over 20 minutes are common because Humber cars tend to catch up with Long Branch runs.
• Service on Lake Shore is all over the place, and the variation is well beyond the TTC’s goal.
• Eastbound departures from Long Branch are not reliable due to problems with CIS data, but the information at Kipling gives a good picture of inbound service in Etobicoke.  The service stays within the six minute band most of the time except around noon.  Looking at the service chart, there is some disruption in CIS data quality at this point.
• The very long gap eastbound just after 2100 is a combined CIS artifact (one car that clearly went to Long Branch was never reported as getting there) and a service disruption caused by the long layover for our friend “pink” discussed earlier.
• Once we reach Parkside Drive, the service settles down into the desired 6-minute band for most, but not all of the day.  The four-car parade described earlier shows up as a long gap followed by three short ones at Parkside just after 2200.
• As on the westbound trips, once cars pass Yonge eastbound, the headways tend to spread out with cars on short headways catching up to their leaders.  By the time we reach Woodbine, swings of ten minutes are common.
• Arrival headways at Neville are not reliable due to limitations in CIS data and this chart should be ignored.

Link Times

This part of the analysis has not yet been published for King (I plan to push that out the door soon).  What these charts show is the time taken for cars to travel between various points on the route over the course of the day.

Traffic congestion will show up as a gradual rise and fall in times for a segment, while delays will show up as a short-lived spike.  These charts clearly show that there was no congestion on Christmas day because the running times between points stayed fairly constant all day long.  (Where times go to zero, a CIS teleportation is in progress.)

Westbound

• Neville to Woodbine:  This stays quite reliably at about 6 minutes with a few exceptions.
• Woodbine to Greenwood:  Consistently around 5 minutes with some jumps that are probably due to crew changes at Russell Carhouse.
• Greenwood to Broadview:  Consistently around 7 minutes.
• Broadview to Yonge:  Around 9 minutes, although with a bit more variation than we saw further east.
• Yonge to Spadina:  6-7 minutes.  Note the slight rise in the trend line through the daytime and the falloff late at night.
• Spadina to Strachan, Strachan to Gladstone, Gladstone to Wilson Park:  These are all quite consistent through the day.
• Wilson Park to Parkside:  There are some peaks in this chart due to crew changes at Roncesvalles.
• Parkside to Queensway-Humber:  The large peaks in this chart are due to CIS errors where Humber cars are not “seen” until they leave on their eastbound trips.  Therefore, their layovers are included in the westbound running imes by the charting program.
• Queensway-Humber to Lake Shore-Humber:  These are the Long Branch cars coming through Humber Loop.  Some cars take a break westbound here.
• From here to Long Branch, running times stay quite consistent all day long.

Eastbound

• From Long Branch to Humber, times are fairly well behaved.
• Through Humber Loop, times vary somewhat due to layovers.
• From Queensway-Humber to Parkside, times are unreliable because departure times for Humber cars are mis-reported by CIS as described earlier.
• From Parkside to Spadina, times stay quite regular.
• From Spadina to Yonge, we see the same slight increase in travel times from 6 up to an average of 8 minutes peaking in the early evening.
• East from Yonge, running times are consistent with the exception of the Greenwood-Woodbine segment that is affected by crew changes eastbound at Russell Carhouse.
• Woodbine to Neville times are not reliable because arrivals at Neville are usually not reported in the CIS data.

Summary

There’s a lot to look at here, but it gives an impression of how the line behaves under ideal circumstances.  A major problem lies with headway regulation causing erratic service from the rider’s point of view.  Congestion, at least on Christmas Day, has nothing to do with it.

In future posts, I will look at other data to show:

• Examples of “good” and “bad” days on the line and we will see how commonly various effects cause irregular service on the line.
• Service at a few sample points over the course of the month to show how reliably some runs arrive every day and others show up at random.
• Link times over the month to show how they vary by week and time of day at different points along the route.  This is vital for understanding the behaviour of congestion and the degree to which it is predictable and should be provided for in schedules.  This will also show where congestion commonly occurs and whether any sort of transit priority will be beneficial in various neighbourhoods.
• Short turn patterns and the quality of service as measured by where cars actually end up.
• Vehicle assignments and the degree to which 50-foot CLRVs are doing the work of 75-foot ALRVs.

I will also look at operations on the 502 Downtowner car both to examine the quality of service on Kingston Road, and the degree to which the 502s contribute to smoothing out headways on Queen Street.