In the previous article of this series, I examined headways on the Lawrence East 54 bus route for the months of November 2011, March 2012 and May 2013. The data revealed a route where staying close to the scheduled headway is a matter of chance, and happens far less commonly than “reliable” service demands.
If running times are fairly consistent, then the time taken from point “A” to “B” is predictable and headway maintenance should simply be a matter of short holds for the faster operators and encouragement to speed up to the slow ones. However, the headways are uneven right from the termini of the route and from an intermediate point (Lawrence East Station) where re-spacing service to a regular headway could easily be done.
A related issue with schedule adherence is the question of running times. Is the underlying problem that operators cannot make the assigned times in the schedules and therefore have no choice but to run at whatever chance headway occurs? I have looked at this previously on Queen and on Dufferin where schedules are a problem for some, but not all, specific time periods.
Finally, there is always the issue of traffic congestion, the bane of surface operations and a mantra to which the TTC often resorts when people complain about service.
Looking at May 1, 2013 in Detail
As an introduction to the topic, the first charts presents the times taken to cover each section of the route on one day, May 1, 2013. There is nothing special about this day, and it makes a good representative of the three months’ worth of data available.
Note: The route is subdivided into ten segments of my choice both for geographic reasons and to permit analysis of behaviour over specific sections. The locations have nothing to do with official “time points” in TTC internal schedules.
For reference, I also include the chart showing all vehicle movements on May 1, 2013. This maps location against time for each vehicle in a format seen in many previous articles here.
- The slope of the line indicates the vehicle’s speed with the more vertical lines indicating faster operation.
- Horizontal lines indicate vehicles that are not moving. This is commonly seen at termini and at Lawrence East Station. Short horizontal “steps” are usually for stop service and/or traffic signals, especially if the route is uncongested.
- Trips running into and out of service show up as lines that begin or end typically at Leslie & Eglinton, Lawrence & Birchmount or on Lawrence east of Markham Road. The route is based at Eglinton Garage near Pharmacy & Eglinton.
- Trips around the Orton Park loop appear to reverse direction. This is caused by the mapping which only considers east-west movement here. A bus running north on Scarborough Golf Club Road will also be running west somewhat before it turns east on Brimorton at the top of the loop.
When we look at actual data patterns, we see that “link times” (the time needed to traverse a “link” between two points on a route) vary much, much less than headways. Moreover, for many links, the times show little change over the day indicating limited peak-period effects of congestion or dwell times to load passengers.
On the westbound trip, the time from Starspray loop to Port Union is fairly consistent all day, but rises noticeably late in the evening. At off hours, the buses tend not to lay over at the terminal, but rather at points along the route and the time taken can show up in these charts as running time rather than terminal time. Layovers show up as long horizontal stretches on the service chart at or near Starspray.
From Port Union to Kingston Road, the times stay fairly close to the trend line with a few exceptions:
- At the end of the AM peak, there are five trips which lie below the main data. Four of these are buses running back to the garage. They show up in the data here because they are still “in service” as far as the vehicle tracking system is concerned.
- Early in the PM peak, some trips are affected by school traffic. See the service chart at about 16:10 where vehicles in both directions hold at the same location, a stop near a school.
Once we reach the segment west from Markham Road, the express operation of the 54E kicks in during peak periods. This is responsible for several data points well below the trend line. This shows up on the service chart as lines with different slopes and buses, on occasion, passing each other. However, the express portion of the line is comparatively short.
The next segment, from Midland to Kennedy, does not include the 54E buses because their trip ends at Lawrence East Station. Some of the longest trips over this segment happen in the evening suggesting layovers at Lawrence East Station. This can be confirmed from the service chart. Indeed, one bus at about 20:50 takes such a long layover that it later short turns without reaching Starspray.
Another factor quite visible on the service chart is the tendency of Orton Park 54 buses to enter service very close to a Starspray 54A bus. This sends a pair of buses west to Eglinton Station on double the advertised headway. One inbound pair, at about 21:20, is noteworthy for taking a layover together at Lawrence East Station. The official schedule shows that the two branches should interleave with even headways throughout the day.
Kennedy to Victoria Park shows little peak period effect. Some of the longer trips are associated with layovers at Warden.
From Victoria Park westward, peak period effects start to appear. They are small in the segments to Don Mills and to Leslie/Eglinton, but quite pronounced approaching Bayview in the AM peak (less so in the PM). From Bayview to Yonge, the trend line is flat although the values are scattered around it especially during the daytime. This may reflect the situation on this busy part of Eglinton shared by many routes where some buses are at the head of a gap and must handle more passengers at each stop.
Visible on the service chart are the common short turns westbound at Bayview and less-often at Laird. Note how most of these do not fill gaps eastbound, but appear intended more to get buses back on time than to even out service.
Eastbound trips from Eglinton Station show minimal effects of the peak period until we reach the section from Leslie/Eglinton to Don Mills with considerably longer link times in the PM peak. This shows up on the service chart as congestion between Leslie and Don Mills that is most pronounced between 1700 and 1800 (5 to 6pm) where the lines tracking the vehicles become more horizontal. This effect is eastbound only.
From Victoria Park to Kennedy, the longer running times appears to be more related to time spend at stops/intersections than to operating speed between them. This includes stops at Warden which, given their length, are probably for crew changes.
Kennedy to Midland includes the stops at Lawrence East Station which can vary in length as discussed earlier.
Midland to Markham Road includes data for the express 54E buses and so the peak periods have some trips with lower running times than the average. These show up on the service chart as trips originating at Lawrence East and going to Starspray. Their faster speed often results in them overtaking a local 54 or 54A.
East of Markham Road, the data do not include the Orton Park short turns, and the 54E expresses are all running local. The running times are fairly consistent here.
East of Port Union to Starspray, the effect of varying layover locations causes spikes in the chart just as it did for westbound service.
Looking at an Entire Month and at Longer Segments
Although the preceding section has a lot of detail, it is almost too fine-grained. I included it as an example of what can be obtained from the data, but for a more general view we need to step back in two ways.
First, we need to see an entire month’s data consolidated to distinguish between constant patterns and one-day peculiarities. Second, we need to look at longer segments of the route to see whether operators can make up for short delays over longer distances. In other words, are the small variations we see in each route segment cumulative, or does the range of running times remain in a narrow band as the distance increases.
The following charts plot the link times for various segments of the route. The first five pages of each file show data for weekdays, and the sixth combines all weekdays together. The last two pages show data for Saturdays and Sundays.
Port Union is taken as a starting point because trip times east of here vary considerably due to the choice of layover points. I will deal with that segment of the route later in the article.
The weekday data for the segment from Port Union to Midland have a similar overall shape throughout the month. There is a “hump” in the trend lines at midday because express trips in peak periods pull the averages down for those hours. Just as on the May 1 chart, the express trips show up as a wider cluster of data points for the periods when they operate.
Saturdays are not unlike weekdays except there are fewer trips and no express operation. On Sundays, the midday peak is a bit lower than for other days.
Common to all of these charts is that much of the data lie within a 10-minute wide band except for the express trips on weekdays.
The segment from Midland to Kennedy includes Lawrence East Station. The data here are clustered in a roughly 6-minute band through the day, every day.
The segment from Kennedy to Yonge shows the peak period effects commonly seen on downtown routes with longer trip times in the peak. Week 5 data are more spread out because of extreme weather on some days. Weekend data show little variation except the usual lower values in early morning and evening periods. Again, the data may vary over the day, but the values stay within a 10-minute band a great deal of the time.
The question, then, is whether those variations from segment to segment add up over the course of journeys to produce a vary wide band of values for complete trips. The charts for Port Union to Yonge show that this does not happen and the data continue to lie within a 10-minute band much of the time. This implies that operators are usually able to compensate for time lost on one part of the route elsewhere in their trips, particularly outside of peak periods.
Eastbound trips from Yonge to Kennedy show the same peak period increases seen above for westbound trips. From Kennedy to Midland, the times are quite consistent all day long with occasional outliers in the data for buses that took layovers at Lawrence East Station. Between Midland and Port Union, there is the same pattern with more widely spaced peak period data due to express trips that is seen westbound.
The full trip east from Yonge to Port Union has a similarly shaped distribution of values to the westbound chart, but the AM peak times are shorter by about 10 minutes eastbound. PM peak times are roughly the same in both directions. (Compare the sixth page, the all weekdays cloud of data points, for the eastbound and westbound trips.)
As with westbound trips, although there is variation in each of the segments, these do not accumulate over the full journey. This indicated that operators are usually able to make up for lost time and make the overall trips within a roughly 10-minute band of times.
At the Terminals
These charts plot the round trip time into and out of terminal areas. Because layovers can vary substantially, treating these parts of routes separately eliminates this variation from analysis of running time on the main part of the route itself.
At Eglinton Station, the reference point is just east of Yonge Street. The measured round trip goes from there west to Duplex, south and through the station, and back out across Yonge. Traffic effects and dwell times will both contribute to the values plotted here.
On all days including weekends, the lowest values are about 5 minutes. This establishes what the shortest time is for this segment, but not a reasonable one for planning purposes. The centre of the data values lies at 8-10 minutes as one might expect for a reasonable amount of loading time and a break for the operator after a long trip.
Some buses, especially on weekends, get quite generous layovers at Eglinton Station. It would be interesting to correlate these with trip times to see if the longer layovers correspond to the faster trips. I suspect that they will, but that’s a job for another day.
At the east end of the line, as mentioned earlier, layovers are taken at various locations between Port Union and Starspray and there is no specific terminal location. Therefore, I have treated the entire segment as the “terminal”.
On weekdays, the values lie across a 10-minute band from 5-15 minutes much as they do at Eglinton Station. Unlike the western terminal, however, there is a marked drop in values at 2000 (8 pm). This almost certainly corresponds with the 15% reduction in scheduled running time for the “late evening” service. As at Eglinton Station, it will be interesting to compare the time spend east of Port Union with the time taken to get there from Yonge Street to see how much the extra time spent before 8 pm is a function of operating speed.
Weekends see higher round trip times from Port Union to Starspray, especially on Sundays. This is not traffic congestion, but a generous schedule that makes layovers possible. It is hard to reconcile the erratic headways during such periods with the amount of time available at terminals.
Actual vs Scheduled Times
From the monthly link time summaries, one can estimate a reasonable time one would expect most buses to spend on each segment. For mid-route segments, this will lie roughly in the middle of the observed times, possibly toward the upper one-third mark during peak periods. For terminal segments, a “reasonable” time would be the time needed for a minimal circuit through the terminal segment plus four or five minutes for recovery time.
The values for each segment can be totalled and compared to the scheduled times, as in the following table.
On the weekday schedules, the actual times are a bit higher than the scheduled ones. To some extent this could reflect operators “taking their time” and driving up travel times knowing that they will take a break at the terminal. However, it should be remembered that the terminal times in my table have already been adjusted to a “reasonable” allowance, not to the range of observed values.
Early evening times on all days are higher for the actual observations than the schedule. It is possible that the transition to the shorter scheduled times occurs later than the points I used as an “early evening” reference, and that there really is no problem with the schedule here.
Sundays consistently have adequate time, except in the early evening as noted. Indeed, the scheduled time actually is longer in the late evening so that the 54 Starspray service will mesh, at least in theory, with the 54A Orton Park buses.
If anything, I suspect my estimates of actual travel times are generous at least during periods of good weather (which May certainly would be). Winter months could be quite a different story.Even so, there is little here to explain the erratic headways and lack of attention to spacing of service so that the 54 and 54A branches blend properly and avoiding widely-spaced pairs during off-peak periods.
Evolution of Running Times
Out of curiosity, I also looked at the running times between Port Union and Yonge for the three months whose data I have been analyzing here.
These files each contain three pages, one for each subject month, with the “cloud” of weekday data points for each direction’s trip.
The westbound trips are slightly lower in March 2012 than in November 2011 or May 2013.
The eastbound trips are higher in the PM peak in November 2011 than the later dates. Midday off peak times are higher in May 2013.
Without an extensive set of data beyond what I have been requesting from TTC, it is hard to say whether these differences indicate any long term trends or simply seasonal variations.
In any event, there has not been much change in running times over the route since November 2011 and the current state of service cannot be put down to some recent phenomenon.
As we have already seen from the link time charts, parts of the route do not suffer from much congestion, and other parts see it only in the peak period when running times are extended.
Another way to display these data uses the vehicle positions as a sample every 20 seconds of where a bus might be found. Where buses run slower, or are stopped, they are more likely to be found and the number of observations will be higher. (This type of chart was introduced in a review of the King car earlier this year.) Because Lawrence East is a very long route, I have broken the display into two sections for the east and west halves to avoid crowding the information.
There are 19 charts in each set corresponding to the hours from 6am to 12mn. The location of each vertical bar is a position mapped from GPS data to a 10m resolution, and the height of the bar is the number of vehicles seen at a position during the hour in question. The scale along the x-axis is the internal distance measured from Starspray westbound with each 100 units being roughly 1km.
Locations of congestion show up as high points with a long “tail”. This corresponds to a stop where congestion forces buses to creep up gradually to the point where they serve passengers. (It could also be a location with a traffic signal, but no transit stop, but these are less common on the suburban routes than downtown.)
The charts summarize data for the first full week in May 2013, from the 6th to the 10th.
The most noticeable congestion effects are westbound between Leslie and Yonge in the AM peak, and to a lesser extent in the PM peak. Most locations on the route have a sharp vertical line corresponding to the stops. The area around Lawrence East station (right hand side of the “East” section) consistent shows higher counts because vehicles spend more time there. A similarly higher set of values shows up around Eglinton Station.
East of Scarborough Golf Club Road, the counts fall off because the 54A Orton Park service does not operate there.
Eastbound charts should be read right-to-left because they are built on the same template as the westbound ones, but vehicles are travelling in the opposite direction.
A few things show up here including Leslie and Eglinton as a point where buses hold eastbound, especially in the PM peak, to make the left turn east to north. There is a lightly used stop just around the corner, but the delays are on the approach to the intersection.
At the outer end of the line, east of Port Union, the varying locations used for layovers are evident.
These charts don’t look like much compared to the ones I generated for the King car because congestion is not as severe a problem on the 54 as it is on the 504. This is reflected in fairly consistent running times over large sections of the route that were shown earlier in the link time charts.
The flip side of congestion is the speed at which buses can move along the road. The following four sets of charts have the same format as the location charts above, but they plot average speed rather than counting vehicles at each position along the route.
The speeds are averaged for all vehicles seen at a point by taking the increment to the point where the vehicle appears 20 seconds later, scaling this to a value in km/h, and averaging these values for all observations at the point within the hour charted.
The variations in speed over parts of the route and by time of day are quite clear, and I will not burden readers who have come this far with a detailed commentary.
As with the location/congestion charts, these are best viewed by stepping through the pages to create a quasi-animation of the evolution of speeds by location through the day.
[Methodology: Background information on the techniques used in this analysis is available in a technical article: Methodology For Analysis of TTC’s Vehicle Tracking Data.]