Updated October 18 at 10:15 am: A few comments about system reliability during bad weather have been added as a postscript to this article.
In a recent post, I wrote about the TTC’s Capital Budget and the projects that are creeping into view as the true cost of adding capacity to the subway becomes evident.
Once upon a time, the TTC was really worried about the capacity of Bloor-Yonge Station, and came up with a scheme to add a third central platform on the upper (Yonge) level, and possibly a second, eastbound platform on the lower (Bloor) level. Interest in this project faded with the dwindling riding of the mid-1990s, but it never completely vanished. Plans such as a Richmond Hill extension raised concerns about YUS capacity even before recent ridership growth took back the “surplus” capacity available for many years to hide the problem.
Independently of the third platform proposal, the TTC came up with a plan to add to the number of trains on the line. If only they could convert to automatic train control (ATC), they could decrease the headway of trains and add to the line’s capacity. In practice, what happened was that the TTC had to replace the existing, worn out signal system anyhow, but really wanted other governments to buy into the project. At that point, ATC’s justification became not only the rejuvenation of the subway (a maintenance project), but a way to add capacity at lower cost than building a new line.
Of course, the trains the TTC was running, the H-series cars and the newer T1 fleet, are not equipped for ATC. A retrofit of the T1 fleet is possible but expensive, and this drives a “need” for a completely new fleet simply to make use of ATC on the Yonge line. In earlier fleet plans, the TTC treated the entire system as one pool and simply counted trains regardless of which type they might be. Now, however, they need a “YUS” fleet that can run ATC and a “BD” fleet that will run with conventional manual controls. (It is unclear what will happen if a BD train finds its way onto YUS trackage, say, for a diversion.)
With the recent, overdue arrival of the first TR train in Toronto, there were bold statements by the Mayor no less (although he was just parroting the TTC) about how these new cars would allow a 40% increase in subway capacity. Well, yes, maybe, but there’s a catch. Several catches, in fact.
Where Does the 40% Figure Come From?
That 40% is achieved in part (10-12%) from the additional space available for passengers on the TR trainsets compared with a 6-car T1 train. The rest comes from a 25% increase in the trains/hour that can operate with ATC. The compound effect of these changes is around 40%.
Current operating practices restrict the distance between two trains based on the block signal layout of the subway. Each block represents a safe stopping distance for a train at the presumed operating speed of the section of track it controls (that’s an oversimplification, but it will do for this discussion). When a line is running at capacity, especially near busy stations, trains tend to operate at lower speeds, and the “safe stopping distance” can be quite short. However, with a block signal system this cannot be easily implemented unless operators are given the ability to drive trains “on sight” almost up to the back of the preceding train. For various reasons, notably but not only the Russell Hill subway crash, the TTC does not allow this.
An ATC system uses a “moving block” scheme where the safe distance is a function of both the next train’s location and the speed of the following train. This allows trains to pull quite close together safely, and trains can follow each other through congested locations at a short separation. This would allow more time for loading at Bloor-Yonge, for example, time that is now “wasted” waiting for signals to allow the next train to enter the station.
If, however, trains are regularly going to run on closer headways, the entire line must be capable of handling this arrangement. It is not enough to fine-tune signals at one location, but leave bottlenecks elsewhere. Converting the entire line to ATC addresses many, but not all of the problems.
Today’s subway schedules bring one train every 140 seconds (2’20″), or just under 26/hour. It is possible to push beyond this by inserting trains occasionally, but this can create stresses elsewhere if the signal system cannot actually handle the closely bunched trains. If the headway comes down to 105 seconds (1’45″), this would give just over 35 trains/hour, a 35% increase in the trains/hour and, hence, the line’s capacity. In practice, the TTC does not need to get headways quite that close for the combined effect of larger train capacity and more trains/hour to yield a 40% improvement. A 110 second (1’50″) headway will do.
Dwell Times, Schedules and Terminals
The frequency of trains is determined by the rate at which they can flow through choke points on the system. The most obvious of these is Bloor Station where dwell times can reach 60 seconds and more when the platform is congested. Under these conditions, it is not possible to maintain a 140 second headway, let alone anything shorter, because the time for train “A” to clear the station and train “B” to arrive is constrained by signal block spacing.
ATC will shorten the approach time and, thereby, increase the possible throughput at such locations. This depends, however, on the arriving capacity (net of those who get off) being able to remove passengers faster than they arrive on the platform.
There is a similar issue, but over a much shorter distance, northbound from Union to Bloor in the PM peak where large volumes of passengers accumulate on platforms unable to board trains.
Another important choke point is the terminals. At both Finch and at Downsview, it is physically impossible to operate a headway much shorter than the one we have today. The reason for this is the length of the crossovers which force a long cycle time on train movements. This can be avoided by short turning some service or by redesigning the terminal. Short turns are obviously not an option for Finch unless the line is extended.
Any operation of this type will require that operators work to the need of the trains, not the other way around, and this will likely cause random problems during peak periods wherever crew changes or breaks occur. The tighter the timetable, the less the margin for unscheduled extra time.
Finally, running times will have to be set shorter than peak trip times. The reason for this is that only a few trips actually require the scheduled time (complete with longer dwell times), and on the shoulders of the peak periods, trains regularly queue at terminals. This is a huge waste of equipment, manpower, and passengers’ time and patience. Attempting to operate the entire line at the minimum possible headway will guarantee many delays and long queues of trains.
The Platform Door “Business Case”
A rather strange report appeared on the September 30 TTC agenda. You won’t find the meat of it online, only a short introduction. This clearly shows that suicide prevention was the impetus for original work on this project, but as the presentation (not available online) shows clearly from its title, this has morphed into the “YUS Line Service Improvement Strategy”. The Commission approved this presentation’s two recommendations: (1) staff are to continue with planning for platform door installation and (2) this project is to be included in the 2011-2016 Capital Budget.
Note that “inclusion” does not mean a project is approved or funded, only that it is shown among the items competing for attention and money.
The details of the business case were not presented, only the summary. This is difficult territory because everything depends on the assumed monetary value of improvements, some of which are intangible or subject to interpretation. The major assumptions are:
- Benefits calculated over 30 years
- Ridership increases 1.5% per year
- Installation is completed over 6 years
- Capital costs are $9.8-million per station
The inclusion of ridership is intriguing because it does not, strictly speaking, affect the doors as much as the need for additional subway capacity. A 1.5% increase compounded for 30 years comes to about 56%, well above the claimed additional capacity planned for the YUS without even accounting for the current 10% shortfall.
The total cost of platform doors for the entire system is given as $511.6m with economic benefits over the 30-year span of $567.1m. When lines are reviewed individually, the greatest benefit comes on the YUS where demand is also highest. The presentation claims that the project economics remain favourable even when various factors are adjusted for a sensitivity analysis. Unfortunately, this entire analysis is not available.
Nonetheless, in the larger scheme, it is likely that many overall benefits of increased capacity have been included in the evaluation, and these would not be available as justification for other projects. In other words, if the economic benefit of carrying more passengers has been offset against the cost of platform doors, this same benefit cannot be used to justify more trains, a better signal system or station upgrades.
In brief, the analysis has omitted many related capital projects that all form part of the capacity upgrade, the nominal purpose of the study.
The Service Improvement Strategy states that the present capacity is 26 trains/hour and that this must increase to 35/hour to meet forecast requirements for 2030. (Note that 20 years at 1.5% gives about a 35% increase. Combined with the existing 10% shortfall, this would mean subway capacity would be exhausted sometime before 2030 even assuming we can reach the higher trains/hour factor and that no major new drivers such as the Richmond Hill extension added to load on the YUS.
As discussed earlier, running more trains requires a new signal system, better use of space inside trains and possibly longer trains. However, the other critical factor is reliability so that the design capacity is actually available a high proportion of the time when it is needed.
A review of TTC operations presented on June 2 (also not available online) identified three points based on international best practices:
“Toronto will not achieve the target level of reliability through automation alone.
“Even with TR [cars] and ATC, Toronto needs to reduce total incidents by 75% to achieve the target reliability level of 1 peak failure per week.
“Toronto needs to target key areas first and then evaluate every aspect of the subway in terms of reliability.”
In a review of incidents causing delays of greater than 5 minutes to service in 2008, Toronto ranks well (#5) among European and North American systems, but low related to new systems in Asia and South America. This is hardly surprising considering that the latter group of systems is generally quite new with technology and system design to match. Whether the new systems will retain their high reliability rates as they age is quite another matter, but we won’t know for a decade at least how these numbers will evolve.
Part of the improvement comes from technical changes on the trains themself. The TTC and Bombardier claim that the TRs will be much, much more reliable than older stock, but this must be achieved and sustained over the life of the cars. The last thing we need is a fleet that gradually deteriorates just as demand on the subway is building.
Part of the improvement comes from station management, and this includes platform edge doors. To what degree the doors contribute to the overall solution is not specified, but this is clearly only one part of a much larger systemic review of subway operations.
To get a sense of the types of delays, I pulled together a list of all of the subway delay notices issued by TTC’s E-Alert system since July 1, 2010. (I have them going back much further, but there’s a limit to my patience in transcribing this stuff.) What is quite clear from this list is that the overwhelming majority of problems occurs on the YUS, and that certain stations, at least for periods of time, appear to be “hot spots” for events. This suggests that some local factors are at work that could be (and may already have been) addressed without the complexity of a Platform Door system. Some delays, of course, have nothing to do with PDs.
The question this begs is whether the TTC has examined the cause of all of its delays in detail to find which type of delay contributes the most, and which of various countermeasures will have the greatest benefit. Instead, the entire exercise seems bent on “proving” that we need platform doors without establishing a real case for them.
How Many Trains Are Needed?
If the TTC is going to run more trains/hour, then it needs more trains unless it can offset the shorter headways with faster operation. The number of trains needed for a line is quite simple:
N = RTT / Headway
“RTT” is the round trip time including any layovers at terminals or along the way, and the “Headway” is the frequency of service. It is self evident that if headway goes down, then “N” goes up unless there is some offsetting change in “RTT”.
The AM peak service in YUS requires 44 trains, not including 3 “gap trains” that are on standby for use when there is a service delay or to slot added capacity into the crowded section from St. Clair south (these trains originate at Davisville Yard). I am cheating a bit here because part of the service short-turns at St. Clair West and “RTT” is shorter for these trains, but for the sake of argument, this will do.
If there are to be 25% more trains/hour, then N must rise from 44 to 55 absent any other change on the line. If the short turn point is extended north from St. Clair to Glencairn or Downsview (with the Spadina extension), then more trains are needed for the existing line.
Any extension brings additional equipment needs, and these go up if the planned headway gets shorter. For example, if the base headway is 2’20″, then beyond any turnback it would be 4’40″. However, if the base goes down to 1’50″, then beyond the turnback it would be 3’40″. That service improvement requires more trains.
If the line can be operated with a shorter round trip time, this will reduce equipment requirements. Although some time might be saved with faster station stops, the real savings would come in “high rate” operation on hilly sections and portions of the route with widely-spaced stops. A few of us remember what this felt like on the BD line when “high rate” was used, but it’s a distant memory.
The TTC stopped using “high rate” (which has faster acceleration and a slightly higher top speed, notably on hills) due to problems with motor wear on older subway cars (now retired), as well as concerns about increased track maintenance and power costs, and the need to retime signals for the faster trains. If the TTC intends to save on fleet size and give faster rides using high rate, then it must address and accept the tradeoffs this will bring.
The round trip time from Finch to Downsview is 118 minutes. From my own observations on trains that have been operating in high rate (shhh, don’t tell anyone), about 10 minutes might be shaved off of this. Therefore, the likely improvement in fleet size due to high rate operation would be a saving of about 8%. A slightly higher value might apply once the line is extended given the wider-than-average station spacings.
Finally, there is the matter of spare equipment. The TTC has justified the premature retirement of its older cars on the basis that they are unreliable, and that newer cars built with more robust subsystems will not need as many spares to provide the same availability for service. We shall see. The TTC has a long history of operating most of its fleet (subway and surface) with excess equipment. New vehicles come on stream, but old ones remain available while the kinks are worked out.
H4 subway cars, retained against the possibility of extra service needed for a Toronto Olympic bid, are still in operation and they provide a cushion for the lower-than-ideal reliability of other parts of the fleet.
This happened with the PCCs and CLRVs, and the continued presence of GMC New Look buses is a testament both to their longevity and to the reliability problems we face with the new hybrid bus fleet. There will be a long period of co-existence of old and new subway fleets, and I doubt we will ever actually see the spare factor screwed down tightly, at least not until only TR trains can operate on the YUS because only they will have ATC equipment.
A reasonable target spare factor is about 15%, although this number has crept up in the transit industry as vehicles become more complex. If a new fleet actually achieves high reliability, the 15% factor might be achieved, but this might not be sustained once the cars age to the point of major overhauls (requiring a pool to be available for service) or to the point where things wear enough to be less reliable.
More cars, of course, means more carhouse space, or at least space that can be used to store trains when they are not in use. Transitional periods between old and new fleets push this requirement even higher while new trains undergo acceptance testing, and old trains await decommissioning.
Today’s fleet consists of 112 trains plus 6 odd cars:
- 44 H4s (7 trains + 2 cars)
- 136 H5s (22 trains + 4 cars)
- 126 H6s and (21 trains)
- 372 T1s (62 trains)
Once the current orders for TRs are delivered and all of the H series retired, the fleet will contain 132 trains (including the cars for the Spadina Subway extension and some growth in demand):
- 372 T1s (62 trains)
- 420 TRs (70 trains)
A further order for TR cars is included in the Capital Budget along with storage expansion at Finch Station.
However, the BD subway only requires 43 trains for service, and the Sheppard line consumes a further 16 cars or roughly 3 trains. When spares are added at 15%, this brings the total need for a T1 fleet to about 316 cars, about 56 short of the actual fleet size. Even doubling the service on Sheppard would only add 16 to the peak requirement plus a few spares.
TTC fleet planning was, until recently, based on running T1s on the entire system, and early planning for what became the TRs assumed full interchangeability. However, the switchover of the YUS to ATC changes this and requires that each line have its own fleet.
The result is that we have more T1s than we need unless there is a major increase in service on BD and Sheppard. Such an increase on BD is impossible because of the same constraints in the signal system and in terminal configurations that afflict the YUS.
The current peak level of YUS service requires 47 trains, and allowing for spares this would bring the total to 54.
For quite some time, the TTC will have more subway cars than it actually needs to operate service, and it will be difficult to get a handle on the actual spare factor with which the system could, if pressed, be operated. All those extra cars require storage space and running maintenance.
Having lots of spares allows for a maintenance policy that can let things slide, leave problem trains on the sideline because there is no real need to make them available for service. Whether the generous size of the subway fleet will lead to such practices remains to be seen. The problem, if this happens, is that when the cars are actually required, they may not be available, or there may be pressure to buy more new cars rather than to tighten maintenance practices and expect the reliability claims for the once-new fleet to be achieved.
The Downtown Relief Line
The TTC and Metrolinx are currently studying transit service and capacity into the core area. A status report was on the recent TTC agenda, but it received little comment at the meeting due to the long list of more contentious issues.
Although the TTC focuses much effort on possible ways to add capacity on the existing YUS, the true cost and limitations of this approach are becoming evident. Within the TTC, there is a strong pro-YUS upgrade faction who would just like the DRL to wither away, Sadly, because the DRL is seen as serving only as a relief valve south of the BD line (Dundas West to Pape as a notional routing), the benefits of new rapid transit service north of the BD line is not considered.
If we are to build new capacity into downtown, it has to go somewhere in its own right, not simply act as a bleeder valve for the worst of subway congestion. Trips must be diverted off of existing lines on an all-day basis for this route to have value in the larger scheme of the network.
By analogy, imagine that people who might otherwise ride the Spadina Subway were told that they don’t really need a line, and that everyone would be fitted in on Yonge. That’s the sort of attitude some at the TTC bring to this debate.
During the election campaign, we have heard a lot of bilge about possible new subway lines, and the TTC has made its own contribution with a flawed presentation of the capacity question for the Yonge subway. The new Council, Mayor and TTC Chair will, no doubt, be strongly lobbied to accept the official view of our problems. Newcomers to the TTC tend to have stars in their eyes, be blinded by the complexity of it all, and be utterly unwilling to ask hard questions about alternatives.
Those questions need to be asked in an era of spending restraint that will exist at the City and Provincial levels no matter who is in office. Whether the debate will be well-informed is quite another matter.
Postscript: A Small Question of Weather
System reliability is vital to achieving the hoped-for additional capacity on the subway network, but this is doubly important during periods of bad weather. When it snows, even when there is heavy rain, more people commute by transit rather than face conditions on the roads. It is at precisely such times that many weather-related problems afflict the TTC — frozen switches, ice in the brake lines, door problems brought on by even more people trying to crush on board.
Reliability stats are usually reported on an averaged basis with the peaks blended into the offpeak, and the good weather blended with the bad. One of the goals cited by the review of TTC service was to achieve no worse than 1 peak delay of 5 minutes or more per week. From past experience, when weather is bad, they would be hard pressed to achieve better than 1 per hour, let alone 1 per week.
This affects two parts of subway operations. One is the physical reliability of systems during poor weather. If trains regularly stall because of iced-up third rails or signal systems that see “ghost trains” thanks to track circuit failures, the system’s capacity will be far below what we hope for.
More importantly, the system under normal circumstances must have enough headroom to absorb the overload conditions of bad days. If the YUS is so close to complete breakdown because trains cannot handle the passengers, then the last thing we want is a bump in ridership when weather is bad. Those who plan subway capacity would do well to leave room for this type of demand rather than assuming that just enough for normal days is acceptable.
The best possible advertisement for transit is service that works under bad weather conditions when driving is not the pleasant romp down the expressway some make it out to be. If transit is at its worst on those days, riders will flee to their cars the moment the snow stops falling.