December 1, 2006, was not a good day for transit operations.  As I have already discussed for the King route, it was probably the worst day of the month.  The weather was bad through the afternoon and early evening, and severe congestion problems affected many routes.

This is a contrast to Christmas Day, discussed in the previous post, where good weather coupled with little congestion or passenger surges made for ideal conditions.

Among the problems we will see for December 1 are:

• bunching of cars due to congestion
• pairs of cars running together over the entire route
• large gaps to the termini
• congestion, most severely in an area well away from downtown, and only in one direction

This shows what the line looks like under worst case conditions.  Even though the service is seriously disrupted, this data has important lessons about how the line is scheduled, managed and operated.

There are seven charts in PDF form here:

Queen 2006.12.01 Service Chart

Queen 2006.12.01 Headways Westbound

Queen 2006.12.01 Headways Eastbound

Queen 2006.12.01 Link Times Westbound

Queen 2006.12.01 Link Times Eastbound

Queen 2006.12.01 Destinations Westbound From Yonge

Queen 2006.12.01 Destinations Eastbound From Yonge

Compared to the Christmas Day post, I have added charts showing the destinations of cars plotted against their time leaving Yonge Street in each direction.  In effect, these are charts of short turns, scheduled or otherwise, and they reveal both the spacing and reliability of service in a straightforward way.  Each vertical line represents a car, and the length of the line shows how far it travelled.

On the service chart, you will see lines that wander in a way that doesn’t make sense.  As I discussed before, CIS can get confused about where a car is and the direction it is headed.  This is especially problematic when a car goes off route.  I look forward to the day when CIS will use GPS to locate cars, but meanwhile, we have to put up with the data we have.  One advantage of the charts is that looking at them, you can see the overall pattern and filter out the “noise” in the data.

Overall Operations

The Service Chart shows the line getting off to a reasonable start both in Etobicoke and the Beach.  Service westbound from Neville is decently spaced. 

The Saga of Run 16

One car, run 16 (you cannot see the run numbers in the PDF version), has a difficult time with equipment and schedule.  It starts off in Roncesvalles Yard with car 4201 (the salmon coloured line in the middle of the chart between 5:30 and 5:50 am), but this car never leaves the yard.  Run 16 shows up with car 4197 (a CLRV) at about 6:24 eastbound at Dufferin (the dark blue line begining at that point).  Since it is late, it short turns at Russell Carhouse.  Somewhere along the way westbound, it is changed off again to car 4075 (another CLRV, dashed blue line) and runs west to Long Branch.

On the return trip eastbound from Long Branch at about 8:30, it is late, and runs right behind run 3 (yellow line) to Yonge Street where, at 9:30, there is a long delay and many cars sit in a pack.  Eventually, it reaches Russell Carhouse where it short turns at 10:20 returning west to Roncesvalles.  On the eastbound trip at 11:20, it goes to Woodbine Loop.

The next westbound trip, leaving Woodbine at about 12:12, takes the car to Roncesvalles whence it returns eastbound at 13:14.  This time it gets all the way to Neville.

Leaving at 14:14, run 16 makes a round trip to Humber (the only one of the day).  The next westbound trip leaves Neville at 16:53 for a round trip, but only to Roncesvalles.  Finally, run 16 leaves Neville just after 19:00 for a trip to Humber and then in to Roncesvalles Carhouse for the night.

This car illustrates something I have noticed before on the King line, and will watch out for generally.  Some runs have an unusually high number of change-offs, usually at the same time of the day.  I am suspicious that some operators may take any excuse to get a change-off which, inevitably, will put them off schedule.  This creates gaps at predictable times and places on a route.  Another possibility is that “the cream of the crop” enters service fairly early in the morning, and the second wave gets the leftover cars whose reliability may be dubious.  Run 16 does not fit this pattern as it is a “regular” car expected to be out all day.  In any event, vehicle reliability is something I will have to review by digesting data from all routes to see which cars get the most use.

Another problem this run shows is the difficulty of getting some cars back on time once they have gone off-schedule.

The Saga of Run 4

Some bunching is visible westbound from Neville just after 8 am.  This appears to be due to run 4 (turquoise).  This car runs late eastbound from Humber at 0658, and three other runs gradually catch up.  On the westbound trip, these three cars drop out by short turning, but we pick up another car (gray) that accompanies run 4 all the way to Long Branch.  They come back across the city as a pair until, just before 11 am, run 4 is turned eastbound at Church for a layover of about 35 minutes.  In the process, run 4 has created another parade of cars that arrives at Neville between 11:35 and 11:40.

Four cars (runs 21/red, 11/green, 23/gray, 26/light green) leave Neville and run across the city as a group.  There does not appear to be any attempt to space out this service.  Runs 26 and 11 go through to Long Branch, and remain married to each other for the eastbound trip until both are turned at Woodbine around 14:45.  Run 21 goes to Humber and comes east immediately followed by run 19/teal coming from Long Branch.  Run goes to Ronces and returns eastbound only to Russell.

Meanwhile, run 4 settles down to trips between Long Branch and Woodbine.

Again, I suspect that the behaviour of run 4 lasts more or less until a crew change given that it behaves so differently in the afternoon.  This shows two problems.  First, an operator can do a lot of damage to a line by running late, not just by running early.  The intent of this tactic is to get a short turn with a big fat layover, and it worked.  However, the more troubling issue is that nothing was done to space out cars running in packs across the city.

By the afternoon, the weather is turning bad and some incident on the Gardiner causes traffic to back up onto local streets.  We have already seen this for King on the same date.  This shows up on the Service Chart as a change in the slope of lines for westbound trips between Dufferin (500) and Roncesvalles (575) starting just before 16:00, worsening as the rush hour goes on.  This is also visible on the Link Time charts that I will discuss later.

Eastbound, there is congestion around Yonge Street starting at about 16:30.  This is visible in the change in slope of the lines on the Service Chart, and also on the Link Time charts.  Our friend run 4 runs turns eastbound at Woodbine around 17:30 to run in to Russell creating a gap of over 30 minutes to Neville.

Meanwhile in Etobicoke, the congestion in Parkdale plus what looks like a bad congestion delay on Lake Shore west of Humber around 17:00 combine to produce short turns and a 40-minute gap to Long Branch.

By 18:00, there is a long parade of cars eastbound with 10 cars crossing Yonge in a period of half an hour.  Things don’t settle out until mid-evening, but even then the service is irregular.

Headways

Westbound

The westbound headway charts echo the comments above regarding the service charts:  repeated events keep the service running in packs, and this continues right across the line all morning.  There are two very large gaps at Neville with cars leaving at 09:17, 09:47 and 10:20.  This results only partly from the eastbound delay at Yonge Street, and partly from bunching of cars described earlier.

These wide gaps are filled in by short turns so that when we see the headway at Greenwood, things have settled down.  All the same, alternating groups of short headways and long gaps continue throughout the morning, before the weather turned bad.  This has nothing to do with congestion, as we will see when we come to the link time charts.  The spread of headways is six minutes either way from average, double what the TTC calls “on time” performance.

What is quite fascinating is that, except for a few obvious major delays, the erratic, unmoderated headways continue all the way across the city, and it is hard to tell periods of genuine weather and traffic congestion from periods when conditions were fairer.

By the time we leave Humber Loop outbound on Lake Shore, the swings in headway are immense typically ranging from a few minutes up to 26.  Only for brief periods does the service operate within 3 minutes of the average headway.  (Note that the trend line is a fairly high order one.  A few points well off the scale pull the trend line away from the cluster of data points at times such as 18:00 to about 19:30.)

The service becomes even more ragged as we move west toward Long Branch.

Eastbound

Headways at Kipling inbound bounce around, but things really get out of hand when congestion delays (and short turns) in Parkdale echo back eastbound with wide headways.  The worst case is a 52-minute gap inbound at about 18:00.  By late evening the line settles down, if we can call it that, to pairs of headways under 10 minutes and over 25.  There are periods during the day when teh headways at Kipling and at Royal York stay within 3 minutes of the average.

Inbound service at Parkside mirrors what we saw westbound, and all day long, cars run in clusters followed by wide gaps.  On the service chart, you can see that commonly the Humber cars leave eastbound just after a Long Branch has come through Humber Loop.  This is not a congestion problem.

The pattern continues all the way across to Greenwood, and the effect of short turns can be seen by the change at Woodbine.  Service east to Neville is ragged with wide gaps including at the end of the AM peak as described elsewhere.

Note that headways arriving at Neville are not reliable due to problems with CIS data.  Often a car is only reported when it leaves the loop, and we do not have an arrival time.

Link Times

Link times measure the travel time from one point to another on the route.  The locations I have chosen as “time points” have nothing at all to do with TTC schedules.  They were picked to bracket areas of interest and isolate effects in specific neighbourhoods.  Moreover, I avoided areas where CIS data is unreliable so that, for the most part, I have good confidence in the times from “A” to “B”.

Obviously, this time is made up both of travel time which can be affected by congestion and loading time which is affected by demand.  It is impossible to sort out the two effects, although in some cases common sense will tell you which component is most likely affecting any specific area.  One of the most important observations is that many links don’t change all that much.

Westbound

The westbound link times from Neville to Yonge do not vary much showing that the small differences are due to random effects such as traffic lights or slow loading at stops.  From Broadview to Yonge, the times build up early in the day and are more scattered, and then drop down again in the evening.

West from Yonge, things get a bit more interesting.  Times build up gradually from an average of 7 to about 9 minutes as the day goes on, and then down a bit in the evening for the segment from Yonge to Spadina.  The segments westward from Spadina to Gladstone/Dufferin have quite consistent times.  However, from Gladstone to Wilson Park (east of Roncesvalles), and from there on to Parkside, both show huge changes in link times with the afternoon peak rising to four times the normal levels for the rest of the day.  This shows the effect of a blockage on the Gardiner and of traffic diverted to local streets.

Times from Parkside to Lake Shore are consistent as one would expect for a private right-of-way situation.  On Lake Shore itself, some peak period congestion effects can be seen for both rush hours, and one big delay is quite obvious for the Humber to Royal York segment.  From there to Kipling has a slight variation as does Kipling to Long Branch.

Eastbound

Inbound link times from Long Branch to Royal York show the impact of two delays in the morning rush hour.

Link times eastbound through Parkdale show a slight rise through the day, but no effect anything like what we saw westbound.  Congestion effects finally start showing up in the Strachan to Spadina section, and especially Spadina to Yonge where PM peak times rise to more than double the usual value.

The Yonge to Broadview link shows the effect of the long delay just after 09:00, but otherwise this link is fairly stable.  Of some interest is the Greenwood to Woodbine segment which has some very long times sprinkled in.  These are again problms with CIS data and do not represent truly gigantic running times.

Woodbine to Neville times vary a great deal and rise gradually over the day.  Again, this is a side-effect of CIS data which usually reports only the departure times at Neville.  Any layovers are counted as part of the eastbound link time.

The important point about the link time charts, as we will see in looking at other days’ data, is that yes, there is congestion, but it doesn’t always happen downtown and it doesn’t happen the same way in each direction.  Any scheme that seeks only to provide transit priority on a limited segment, say in front of City Hall, will miss most of the locations where congestion occurs.

Short Turns

Finally, we have the destination or “short turn” charts.  These are plotted outbound from Yonge in both directions.  The spacing of the vertical lines shows the time between cars while the height of the lines shows where those cars ended their trips.