The Spadina Streetcar has operated now for just over a decade. For all the problems of getting the project approved and built, there is no question that it has transformed travel on Spadina. With such intensive scheduled service and a complete right-of-way, we should have the best possible quality. This series of posts will examine the actual service in December 2006.
As usual, I will start with Christmas Day to introduce the analysis under the best-behaved conditions, and then I will move on to regular weekdays, and to the month as a whole. Among the major points I have found are:
- The presence of a right of way greatly reduces the sort of variation seen over the course of the day on mixed-traffic routes, but does not eliminate it completely. As demand rises and falls, stop service times vary and this affects trip times.
- Although the averages are well-behaved, the degree of variation is quite substantial. Indeed, given the short distances travelled, the amount of variation is comparable to that seen on the much longer mixed-traffic Queen route. This variation undermines the benefit of the right-of-way.
- Some delays due to traffic signals are visible in the data, but the resolution of the TTC’s monitoring system (CIS) is such that I cannot report on this in detail.
- Short-turning is a chronic problem at Queen’s Quay, and much of the service destined for Union Station never actually gets there. At first sight, I was reminded of Queen Street in the Beach when I reviewed the charts. This is no recommendation for the benefits of exclusive right-of-way.
- Spadina is a route that demands management by headway, not by schedule. With extremely frequent service, the concept of being “on time” is meaningless to riders.
For those awaiting a review of a suburban bus route, please have patience. With all the discussions of service reliability on Queen, I thought it important to look at a route operating completely on reserved lanes to see just how it behaved.
For those who are new to this blog, I recommend reviewing the early posts about the King car in the Service Analysis category. The techniques used to distill the TTC’s data and present it for analysis are explained there in detail.
First up is the chart showing service for Christmas Day, 2006.
A few things are immediately obvious here:
- The short-turn service to King sits in the top half of the charts. Union Station is at the bottom, and the route via Queen’s Quay gets much less service than the heavily used Spadina segment.
- Horizontal lines visible at Queen Street are carhouse trips. I have plotted all locations to and from both carhouses at this co-ordinate so that they are visible, but the lines do not interfere with the regular service trips.
- Horizontal jogs in lines are visible at several locations. These are places where cars sat for long enough that CIS reported them twice, at least, in the same place. Although we might put this down to stop dwell time, the locations and times of day suggest that this is more likely due to interference by the so-called transit priority traffic signals.
- There are problems with resolution of vehicle movements at King. CIS “sees” a car passing the signpost at Queen and, based on the distance travelled, plots it south to King (not via Adelaide to Charlotte). At this point, the schedule says that the trip is supposed to go north, and CIS plots a northbound movement getting as far back as Queen Street. However, the car actually went around the block via Charlotte Street and eventually arrives at the King & Spadina signpost. CIS then corrects the car’s position. This type of error makes the data northbound from King to Queen meaningless because of the many phantom trips.
- Another CIS error shows up with Union Station trips that short-turn at Queen’s Quay. CIS thinks the car is still eastbound on Queen’s Quay until information from the electronic signposts corrects this, and CIS puts the car back where it should be on a northbound trip. This doesn’t appear on Christmas because short turns were rare, but we will see a lot of it on charts for weekdays.
- Departure times at Spadina Station are rarely clear. Instead, a car arrives at the loop, and then appears to take a leisurely trip to the surface where its location is picked up again in the data.
- Times a Union are unreliable in a different way. There is a signpost just before the portal on Queen’s Quay, and we get data for cars entering and leaving the tunnel. However, there is no layover at Union visible because, I suspect, the data for this trip is synthesized. Anyone who rides the 510 knows that cars take long layovers at Union, but this is not directly observable in the data.
Despite these limitations, we can still get a fair picture of how the line behaves. One important point about the chart from Christmas (and other quiet days) is that we see a fairly reliable service to Union Station. When I first reviewed the charts for weekdays, I thought that some problem with CIS must have prevented proper monitoring of trips into the tunnel. However, the fact that the data is well behaved on days when we would not expect much short turning shows that the problem lies with service management, not with the data.
As we have seen on other routes, headways may average what the schedule says they are, but there is some variation in actual operations.
For trips down from Spadina Station:
- The headways shown at Spadina Station are suspect because we don’t have reliable departure data. These values are more likely to represent variations in arrival times.
- The headways at Harbord are much more tightly grouped around the average and lie within a band of roughly 2 minutes either way. The very short headways in the evening are cars running in to the carhouse that follow immediately behind an in-service car.
- As we move south on Spadina, the headways start to spread out and by Dundas we see a range of 3 minutes either way. This falls within the TTC’s performance standards for “on time” service.
- North of King, the data includes trips that have actually gone via the Charlotte Street loop (see discussion of CIS plotting errors above).
- South of King, the average headway widens, but the spread in values stays mainly within the +/-3 minute standard. This starts to widen out as the service progresses east on Queen’s Quay.
- Arrival times at Union are, as I mentioned earlier, probably calculated from other data and represent the headway at surface locations.
For trips up from Union Station:
- Times leaving Union show a fair amount of scatter, just as we saw at Spadina Station, but this is reduced once we reach a surface reference point at York. Even so, headways can reach close to 10 minutes on a holiday service with no source of interruption. This is probably due to variations in dwell times at Union Loop.
- Once we are north of King, the short turn service cuts in and the average headway goes down. The effect of the “ghost” trips from short turns described earlier can be seen by comparing the charts for south and north of Queen. South of Queen, the ghost trips are present, while north of Queen they are not. The valid headway information northbound on Spadina from King actually starts with the “North of Queen” chart.
- Headways from Queen north stay within a six minute band, although they are slightly more dispersed once we get past Dundas Street probably due to variation in dwell time at the northbound stop.
Southbound from Harbord to Queen
Southbound/Eastbound from Queen to York
York to Union Station and Return
Westbound/Northbound from York to Queen
Northbound fom Queen to Harbord
Harbord to Spadina Station and Return
Compared with previous route analyses, I have defined more “time points” for Spadina in an attempt to capture the dwell times at traffic signals. The result of this is not yet conclusive, and I need to dig into the raw data to understand what is happening at each location. I have not presented detailed link by link charts here yet for this reason, although some idea of the granularity I am attempting is visible from the number of different locations in the headway charts discussed above.
However, we can look at the link times over substantial parts of the route to see how well-behaved they are on Christmas Day.
For both northbound and southbound trips, the time from North of Harbord to South of Queen holds a steady average of 8-9 minutes with a small rise northbound later in the day. Note, however, that the times are scattered in a band about 5 minutes wide. On a base of less than 10, this is a considerable variation.
Similarly, for trips between east of York to south of Queen, the trip times average around 10 minutes, but there is a spread of 5 minutes or more. We will see more scatter in the data when we turn to the all-month charts in a later post.
At the terminals, as we have seen on other routes, there are resolution problems that preclude accurate reporting of dwell times. However, we can look at the round trip times to a nearby point where the data is reliable and infer from this the variation in station times at the terminals.
The round trip from York to Union consistently averages about 7 minutes on Christmas Day, although there is a spread of about 6 minutes around the average. Similarly, at Spadina Station the round trip from Harbord averages 6-7 minutes within a band of about 6 minutes.
What is troubling about the link times, especially along the route, is the range of variation relative to the average. The data is scattered all day long and even in the detailed data (not presented here) there is no clear time of day, location or direction on Spadina that contributes to the overall total except between Dundas and College, mainly northbound. (This is no surprise given the demand at the northbound stop.)
The situation on the southern part of the route is even less reliable than on the northern part. I cannot help concluding that the variation arises primarily from traffic signal delays, and will explore this in a future post.