A few months ago, I received the TTC’s vehicle monitoring data (a.k.a. “CIS” data for “Communications & Information System”) for 501 Queen and related routes for the months of December 2008 and January 2009.  I have been whittling away at it for the past few months as time permitted, and now it’s in shape to begin publishing commentaries.

Of particular interest are the effects, such as they may be, of the new line management strategies implemented by the TTC.  Operations in the two months differ because of a change in the schedule.

• December 2008:  Drop back crewing was used at Connaught so that operators would leave westbound on their time while relief operators drove the vehicles to and from Neville.  The intent is to allow the operators to get on time without affecting the through service.
• January 2009:  A new schedule was implemented in which run numbers remained assigned either to the Humber or the Long Branch service.  The intent is to avoid short turns whose entire purpose is to sort out the relative order and location of each branch’s runs to make sure that they are in the proper sequence westbound.

For reasons best known to the TTC, relief crews were available during Christmas Week, but there was no extra line management.  Therefore, that week is in a way an example of a “do nothing” approach, although under less than the most strenuous circumstances.

Anyone who was in Toronto this past winter knows it was much worse than the previous few years and we had a particularly bad December.  This shows up in the service quality, but generally for the period needed to get the roads back in proper shape.

Because I now have data for December 2006, December 2007, January 2008, December 2008 and January 2009, we can review operations over three winters, a variety of weather conditions and different management strategies.

How The Data Were Analyzed

This is a very brief synopsis of a process that has evolved since early 2007 when I first started working on this type of data.

• Raw CIS data gives the location of each vehicle as estimated by the monitoring system.  In the original CIS system, this location was calculated by the offset from electronic signposts of each car, and translated to an intersection name.  For cars with GPS actively connected to the CIS unit, this location is calculated from the GPS position.
• None of the data up to January 2008 has GPS, and the implementation of GPS was in progress through December 2008 and January 2009.  Therefore some of the problems of locating cars (discussed in previous posts) remain even in the recent data.  I have made considerable effort to filter out these errors, but it’s hard to devise an algorithm that catches only bad data.
• From the location data, I plot the movement of cars through the city.  This is used to produce the service charts full of diagonal lines crisscrossing the page (you will see these for the current data in a future post) in a format invented by E.J. Marey in 1885.  This is not exactly a new technique.
• The location data also allows the reconstruction of an “as operated” schedule including the times cars pass various points of interest.
• When the “as operated” schedule is read “down” for one point, it gives the actual times of cars at that point and, thus, the actual headways.  When the schedule is read “across”, the difference between adjacent columns gives the link time from point to point.  Link times over longer distances can be calculated simply by choosing appropriate columns.
• If only the rows representing trips that crossed a standard point (say Yonge Street) and arrived at a specified location (say Long Branch) are considered, then we get the as operated headway outbound from downtown to the location of interest.  Note that this will not pick up vehicles that are short turned into the outbound stream further along the route, but examination of other data shows that this is rare.  Actual headways at the terminals are, of course, available in the headway charts (also coming in a future post).

The Charts

The charts presented here show the data for service outbound from Yonge Street to Neville, Humber and Long Branch for each of the five months for which I have data.  Each set contains six pages with averages and maxima, hour by hour, for weekdays, Saturdays and Sundays.  Note that the times shown are at Yonge and therefore the hour starting at “17:00” does not actually reach Long Branch until roughly an hour later.

In some cases, the value shown is zero.  This means that there was no datum from which a value could be calculated and, by implication, no service (at least originating at Yonge) during that period.  Some of these are really gaps at the outer ends of the line (especially on snowstorm days), and some are caused by major disruptions that caused diversions and/or split operations.  However, the overall pattern is clear and that’s what is of interest in comparing five months’ data.

(For the statisticians among you, no, I have not calculated standard deviations.  The primary reason for this is that “n” is quite small in many cases, and I think that the behaviour of the data (and the service) is sufficiently visible in the averages and maxima.  Possibly this could have more validity over longer intervals than one hour, although this might filter out some of the spikiness in the values.  I may play around with the data as a future exercise.)

To Neville

December 2006
December 2007
January 2008
December 2008
January 2009

To Humber

December 2006
December 2007
January 2008
December 2008
January 2009

To Long Branch

December 2006
December 2007
January 2008
December 2008
January 2009

In order to compare these charts, it may be useful for you to open all five for one location simultaneously and toggle back and forth to see the differences.

Commentary

Scheduled Service

The scheduled service from Neville to Humber lies roughly in the following ranges for all dates of interest:

• AM Peak:  4.5 – 5 minutes
• Midday:  5.5 – 6 minutes
• PM Peak:  5.5 – 6 minutes
• Early Evening:  6.5 – 7.5 minutes
• Late Evening:  10 minutes
• SaturdayAfternoon:  4 – 5 minutes
• SaturdayEarly Evening:  6.5 minutes
• Saturday Late Evening:  9 minutes
• Sunday Afternoon:  5.5 – 7 minutes
• Sunday Early Evening:  9 – 10 minutes
• Sunday Late Evening:  9.5 to 11.5 minutes

Service to Long Branch is twice the headways shown above.

Note that the weekday schedules were changed with additional running time and wider headways in January 2009.  Saturday afternoon and Sunday services were improved concurrently to deal with overcrowding.

Service to Neville

For December 2006, the average service to Neville sits slightly above the scheduled value, but the real problem shows up in the maxima that lie in a cloud ranging between the 10 and 20 minutes.  Although many cars may reach Neville, they do not leave Yonge Street particularly well-spaced.

The averages in December 2007 are a bit worse than in 2006, and December 2008 is a real mess thanks to the weather.  The averages in January 2009 are a bit better overall than in 2008, and this likely reflects the impact of the changed line management for the east end of the route.

The maxima tell quite another story.  December 2007 is probably the worst of the lot, although 2008 isn’t much better.  Similarly, January 2008 is noticeably worse than 2009.  However, the consistent pattern even for the best case, January 2009, is that the maximum headways within any hour are rarely below 10 minutes and values regularly range to 20 minutes and beyond.  This is independent of the time of day, and shows that residents of the Beach will routinely encounter wide gaps in service regardless of when they travel.

Weekend headway averages are fairly reliable, although they regularly sit above the scheduled service levels.  Some storm disruptions, such as December 20 & 21, 2008, are clearly visible.  The maxima again show that gaps can be expected at all times of the day.  Given that the average headway is not double the scheduled value, these gaps are more likely to be due to bunching than to short turns.  Remember also that these data report the headways at Yonge Street and so bunching, if any, has already happened on the trips inbound from Humber.  I will turn to this problem in future posts.

Service to Humber

Weekday average headways behave similarly to those in the Beach, and January data tends to show less disruption than in December when shopping traffic and weather have generally had a greater effect.

The maxima for most months again show us that although the averages may look reasonable, the actual service contains large gaps at all hours.  January 2009 is the best of the lot, possibly due to the schedule and management changes, but the maxima rarely dip below 10 minutes and values lie in the band from 10 to 20.  As I mentioned above, this is the headway as seen outbound at Yonge.  This indicates some combination of the following effects:

• Cars are not arriving westbound at Yonge properly spaced.  This is a troubling situation considering the amount of supervision available at Russell Carhouse to dispatch cars inbound.
• A “regular” service may arrive at Yonge, but enough of it is short-turned to create gaps in the Humber service.

Weekend service shows the same sort of pattern we saw at Neville.

Service to Long Branch

Average headways to Long Branch show a different pattern from those elsewhere on the route in having more high peaks and more cases where there is no data for an hour or more.  This is a direct result of the combination of wider scheduled headways and short turns.  The averages are also much more disrupted compared to the schedule than at the east end of the line, and this suggests that less attention is paid to ensuring that service reaches the western reaches of the streetcar system.

Maximum headways are equally alarming with values well over half an hour common especially in the December data.

Saturday headways are better behaved, but I am particularly troubled to see average values in January 2009 that are routinely well above the scheduled level.  Much of the service that is supposed to go to Long Branch is not getting there.

Maxima over half an hour are not uncommon, and again January 2009 is troubling given that this is supposed to be the month with improved schedules and better management.

Sunday average headways are better for January 2009 showing, possibly, the effect of improved schedule headways and possibly less crowing delays.  All the same, maximum headways are still high showing bunching even with the improved service.

The TTC continues to resist splitting the west end of 501 Queen off as a separate route, but it is clear that service, if one can even use the word, is much worse than advertised west of Humber Loop, and the TTC has done an excellent job of destroying the credibility of streetcar service in southern Etobicoke.  That they contemplate a “rapid transit” corridor to the same community is rather sad.

In future posts I will examine individual days over the December-January 2008-09 period as well as headway and link time regularity along the route and over the months.