At the forum Tuesday evening, the TTC poured cold water on my proposal to swap the CLRV and ALRV fleets between the 501 Queen and 504 King routes claiming that their studies showed that headways below 4 minutes could not be operated reliably in mixed traffic.
Others commented on the length of time it takes to get from The Beach or Long Branch to downtown, and as the evening wore on, comparisons became as bloated as the headways on the 501. One speaker claimed he could get from Buffalo to Toronto faster than a trip on the Queen car.
This post examines those two issues, and I will update this item if additional follow-up topics come to mind.
[Updated 4:20 pm, December 8: Bad links to charts corrected plus minor textual revisions.]
Headways Less Than 4 Minutes / CLRVs on Queen
Several have already commented on the dubious merit of the TTC’s claim, but to put it in context, let’s look at what would really happen.
Today, the peak headway on 501 Queen is 4’52” in the am, and 5’30” in the pm. Assuming we don’t change the capacity of the scheduled service and simply do a 3-for-2 swap to CLRVs, the headways would be 2/3 of today’s values. This gives us 3’15” in the am, and 3’40” in the pm.
On King, the peak headway on the 504 is 4’00” in the am, interleaved for about 45 minutes with a 4′ headway inbound through Parkdale giving a 2’00” combined service. If this were converted to ALRVs, the headway would be 3’00” although I could argue that Broadview Station deserves better than a car every 6 minutes. In the pm peak, the headway is 3’45” and assuming no improvement to scheduled capacity (something I don’t agree with, by the way), the ALRV headway would be almost 6 minutes.
The point here is that with CLRVs, the Queen headway would be comparable to what we now run on King with the shortest headways in the am peak when traffic congestion really is not a major factor on most of the route. The argument about minimum headways and mixed traffic operations is a very large, well-fed red herring.
I believe that the TTC’s real problem with the proposed swap of equipment is that they would need to continue running close to the same headways on King during the offpeak with ALRVs, but changing to CLRVs would increase the number of cars on Queen. Hello? Hello? That’s the whole idea!
The weekday offpeak headways on Queen would change to 3’40” midday, 4’55” early evening, 6’40” late evening, and to comparable values on weekends. Combine this with improved line management, and even a short turn at Woodbine would not put an immense gap in the service to Neville.
Another saving comes from elimination of the grossly excessive running time and huge layovers (two to four times what the TTC claims they should be) at Long Branch Loop. Recovering this time also increases the number of cars actually in service rather than taking a siesta while passengers wonder if their car will ever show up.
Is Buffalo Really Closer to Downtown Than Long Branch?
The travel time on any trip is made up of three components: stop access time (walking to the stop), waiting time (actual, not scheduled) and travel time. The access time has nothing to do with the service, with the possible exception that people will run a lot more for really bad service if it is in sight for fear of missing the last car of the day.
Waiting times depend on actual headways, and as we have seen, these are all over the map both in the Beach and on Lake Shore. Despite the advertised “frequent service”, waits up 20 minutes are not uncommon in the Beach, and they regularly get above half an hour on Lake Shore.
I have looked at travel times for individual segments of the route in previous posts, but here I have chosen to review times to Yonge Street from various points.
In each set of charts there are:
- Four pages each containing one week’s data for Monday to Friday. “Week 1” includes Friday Dec 1, and “Week 4” omits Christmas and Boxing Day. These charts include trend lines showing how the travel times vary by day and by time-of-day.
- One page consolidates all of the weekday times as a scatter chart without trend lines to show the degree of dispersion in the data.
- Two pages show the Saturday and Sunday times. Boxing Day is included with Saturdays, and Christmas Day is included with Sundays.
The charts show that weekday running times from Neville to Yonge lie mainly in a band centred on 33 minutes. This varies on some days and, for example, the early rush hour on December 22 is visible here as it has been on many other charts. Other notable points:
- The majority of the data lie within a 10-minute wide band.
- Weekday evening running times are consistently lower than midday times.
- Saturday times peak at about 1400 in a range close to that seen during weekday pm peaks. Boxing Day, included with the Saturday chart, shows the lowest running times.
- Sunday times behave similarly to Saturdays with Christmas being the best day of all.
I believe that it is reasonable to conclude that travel from Neville to Yonge will, on average, take at least 45 minutes of which 10 minutes is waiting time. However, due to variations in individual trips and headways, this will commonly rise to an hour comprised of a 20 minute wait and a 40 minute in-car travel time. Note that the unreliable headway contributes more to the long trip than the variation in the in-car time. This is extremely important in discussions about the relative merits of transit priority and headway regularity.
From the west, I have selected four points to illustrate various aspects of the problem. Long Branch is the terminus and Humber Loop is the location where half of the service is scheduled to short-turn.
Weekday time from Long Branch to Yonge are clustered around the 60-minute line, and most data lie within a 10-minute band of the trend values. During peak periods, these values spread out, especially in the pm. As with trips from the Beach, evening running times are about 10 minutes lower than the rest of the day.
Saturday times peak in the early evening reflecting congestion in the Entertainment District as I discussed in another post. Note that although there is a clear variation over the day, the actual values stay close to the trend lines. This indicates that congestion and loading delays are consistent, predictable events, not random delays. Sunday data behave differently as there is no congestion buildup in the evening, except on December 31. A few outlying values are due to CIS data errors and one of them, on December 10, causes the trend line to veer downward in the late evening.
When we move east to Humber Loop, we see that the times are starting to scatter and there is even a distinct pair of data sets, one above the other. These times include layovers at Humber Loop, primarily, but not exclusively, by the Humber short turn cars. What is clearly visible here is that the difference in times for this set of cars lies well above that of the through cars, especially on weekends.
The “recovery time” in the schedule is excessive. This exists, I believe, mainly to synchronise the two services (on paper) because the through and short-turn trips must differ by a multiple of the headway. If the route is split in two without integrated headways, the time wasted on meshing two services at Humber Loop can be reclaimed for use elsewhere on the routes.
Coming further east to Parkside Drive (east side of High Park), we see weekday times clustered around 30 minutes, slightly lower in the morning, higher in the afternoon, and with noticeably more scatter. This likely reflects variations in congestion through the whole Queen West area and the core as we have seen before on the individual link time charts. Saturday times rise from an average around 20 minutes in the early morning peaking at about 35 minutes by 1500 and then dropping off into the evening. Sunday times have a similar shape, but a lower peak, and as before show the effect of New Year’s eve.
As a side-observation, when the Long Branch car was a separate route, its scheduled running time from Long Branch to Humber was 25 minutes during most operating periods. What we see here is that trips from Long Branch to Yonge typically take about 30 minutes more than trips from Parkside, and this confirms that there has been no change in travel time from Parkside west where the line is on reserved lanes or a wide, uncongested street.
Finally, the chart for Wilson Park (just east of Roncesvalles) to Yonge shows similar behaviour to what we have seen at Parkside indicating, as we would expect, that the congestion and delays are all east of Roncesvalles. Travel times to Yonge lie around the 25-minute line, higher in the pm peak, with considerable scatter on some days probably reflecting unusual events.
We already know that headways west of Humber Loop are commonly above 20 minutes, and adding in the travel time we have the total trip to downtown easily at 90 minutes for many riders. Again, as in the Beach, the wait time is the greater source of variability.
The point in all of this is that a great deal of congestion lies outside of the core area and peak hours, and the larger proportion of the variability in travel times is due to unreliable headways. Any schemes that do not address these facts will do little to solve service reliability as seen by the TTC’s users.