A look at speed gains on the King Street transit pilot project

The City of Toronto recently began a pilot project to speed up transit traffic on King Street through the downtown core. The project entails changes along the busiest 2.6km stretch, whose main aim is to clear non-local auto-traffic from the transit-way. Parts of the street have been pedestrianized, many transit stops have been relocated to the far side of traffic lights, and stop bump-outs have been created and bollarded off from traffic. Probably the most important change however is the complex set of turn restrictions (which can be seen in detail here) which apply only to private cars (at all times) and taxis (except at night). These have the effect of clearing King Street of all but the most local, block-bound auto-traffic, while allowing transit users, cyclists, and pedestrians a more complete freedom of movement.

You can read about the whole pilot project on the City’s website. It’s really quite a bold move for a city these days, but one that is warranted by the fact that even with the sometimes remarkably slow transit service in this corridor, the city estimates that there are more transit users than private vehicles by a ratio of more than 3 to 1 (65k transit trips vs 20k daily vehicles).

The pilot project has been in place for a bit more than a week now, and we here in the SAUSy lab were wondering how it’s going. For my own (unrelated) dissertation work, I’ve been collecting GPS data from TTC’s entire surface fleet for almost a month now. This gives us a nice opportunity to have a little before/after comparison: have transit services actually sped up as a result of the changes?

The data we have are essentially GPS traces for each transit trip with associated timestamps. The sampling interval is about 20 seconds on average and the data comes from the NextBus API, which is what’s behind any real-time transit apps you may be used to using.

We wanted to measure travel times through the effected area, which is used by three different routes. The #504 and #514 are operated by streetcars, while the #304 is a night-only bus following the route of the #504. Below is a quick-and-dirty map of the overlaid GPS tracks for the three routes (red) and the limits (blue) of the pilot project area.

The bulge in the GPS traces through downtown is due to GPS signals bouncing off all of the tall buildings.Fortunately that cleans itself up a bit before the end of the pilot area.

There were about 14,000 trips in our dataset that traversed the study area, ~11,300 before the project implementation and ~3,500 post implementation. To estimate the travel time across the study area, we took an inverse distance weighted average of the timestamps on each trip that were within 150 meters of each of the study area boundaries. Or in simpler terms, we measured the time a vehicle entered and exited the study area shown above.

A plot of total travel times through the study area (below) shows 1) that we have a bit of missing data (my bad), 2) that there is a clear daily and maybe weekly periodicity in travel times and, 3) if you squint that there may be a regime change somewhere around Nov 12th, which indeed is when the pilot project officially went into effect (red line).

To take a closer look at the effects here, we’ll need to zoom in on the daily pattern. The following chart takes all the data from the above chart and overlays it on a single day.

Pre-pilot travel times are shown in black, post-pilot times in red. The smooth lines are loess regression curves for the pre and post sets.
Mean travel time Median
Pre-pilot 16.3 15.4
Post-pilot 13.8 13.5

The pre-pilot (black) shows a typical daily pattern with morning and afternoon peaks in travel time. The quickest average speed through the study area happens around 3-5 am with ~9-10 minutes needed to get from Bathurst to Jarvis. Presumably, that’s more or less the base time needed to get through all the lights and stop at most of the stops with minimal on-board crowding. During the pre-pilot period, the worst of the afternoon peak was at least twice that on average, though a different smoothing parameter could have resulted in a more peaky peak for the momentary average. Looking at the points themselves, I might be inclined to guess something more like 22-23 minutes as a maximum daily average.

The post-pilot times (red) however show a dramatic reduction in average travel times, kind of lopping the peaks off the rush hours. The same plot, but for the west-bound trips shows a somewhat different curve for both lines, but again, the big story seems to be the ~25% reduction in average travel times during the times when most people are riding.

Mean travel time Median
Pre-pilot 18.0 16.8
Post-pilot 14.7 14.2

I’d be willing to venture a guess that the variability in travel time is down substantially as well, but I’ll save that analysis until we have at least another week’s data. Really, we should wait until we have more data generally before drawing any real conclusions, but it looks like the pilot project is off to a good start, and immediately making a big difference in travel times for most people going through the area.

We’ll probably take another look at this issue as more data comes in, and I’m happy to share our data with anyone who requests it. I won’t link to here though since the dataset is still growing.

Also, I’d like to give thanks (happy American Thanksgiving!!) for Jeff Allen who did all of the charts and GIS analysis for this post! Jeff is finishing up a master’s degree this year, but with luck may be sticking around for another degree or two in the SAUSy Lab.

NACIS 2017 Posters

Who posts the posters?

We do. These are the SAUSy lab entries to the NACIS 2017 map and poster gallery. The first is by the famous Jeff Allen, the second by my humble self, Nate Wessel. We’re looking forward to the first Canadian NACIS conference in 23 years, taking place October 10-13 in Montreal!

Paper published on retrospective routing with real-time GTFS data

The following paper has recently been accepted for publication by the Journal of Transport Geography:

Constructing a Routable Retrospective Transit Timetable from a Real-time Vehicle Location Feed and GTFS

Nate Wessel, Jeff Allen & Steven Farber

Abstract
We describe a method for retroactively improving the accuracy of a General Transit Feed Specification (GTFS) package by using a real-time vehicle location dataset provided by the transit agency. Once modified, the GTFS package contains the observed rather than the scheduled transit operations and can be used in research assessing network performance, reliability and accessibility. We offer a case study using data from the Toronto Transit Commission and find that substantial aggregate accessibility differences exist between scheduled and observed services. This ‘error’ in the scheduled GTFS data may have implications for many types of measurements commonly derived from GTFS data.

Preproof of the paper

New Paper on Dynamic Food Environments

New paper published in Applied Geography by SAUSy Lab members Widener, Farber, and Allen (and others):

How do changes in the daily food and transportation environments affect grocery store accessibility?

Abstract

A healthy food environment is an important component in helping people access and maintain healthy diets, which may reduce the prevalence of chronic disease. With few exceptions, studies on healthy food access in urban regions typically ignore how time of day impacts access to food. Similarly, most extant research ignores the complexities of accounting for the role of transportation in spatial access. Examining healthy food access is important, especially for populations whose day-to-day schedules do not align with a typical work schedule. This study profiles novel methods that can be used to examine the daily dynamics of food access in Toronto, Ontario, using grocery stores as a case study to examine the changing geographies of food access over a 24-h period, and the impact of a changing public transit schedule on food access. Walking and automobile travel times are also reported. Results indicate that access to grocery stores is severely diminished for large parts of the city in the late night and early morning, and that public transit travel times are higher and more variable in the early morning hours. Ultimately, this research demonstrates the need for further study on how residents with nonconventional schedules experience, and are affected by, the dynamic food and transportation environments. Future research should build upon the methods presented here to include a broader range of food retailers.

Photos from AAG 2017

A few photos of SAUSy Lab members from the 2017 AAG Annual Meeting below!

Misha Young, PhD Student – presenting his past work on transportation behaviours.
Dr. Michael Widener presenting his work on access to food and time use.
Tara Kamal-Ahmadi, PhD Student presenting her systematic review on using activity spaces in food environment research.
Boston

AAG 2017 Schedule

Check us out at the 2017 AAG! SAUSy affiliated presentations and abstracts in the links below:

Exploring the relationship between food shopping behaviour and transportation with time use data

A Systematic Review of Activity Spaces and Food-related Behaviours

Neighborhood context of aging-in-place: mapping the spatial patterns of aging in Canada

Transportation behaviours of single-person households in a Canadian context

Modelling the Effects of Space on Emergency Medical Transportation in Maryland, USA

 

Paper on urban cycling accepted!

Dwelling Type Matters: Untangling the Paradox of Dwelling Type and Mode Choice
(By Trudy Ledsham,  Steven Farber, and Nate Wessel) has just been accepted to the Transportation Research Record!

Abstract:
Urban intensification is believed to result in modal shift away from automobiles to more active forms of transportation. This work extends our understanding of bicycle mode choice and the influence of built form, through analysis of dwelling type, density and mode choice. Both apartment dwelling and active transportation are related to intensification, but our understanding of the impact of increased density on bicycling is muddied by lack of isolation of cycling from walking in many studies, and lack of controls for the confounding effects of dwelling type. This paper examines the relationship between dwelling type and mode choice in Toronto. Controlling for 25 variables, this study of 223,232 trips used multinomial logistic regression analysis to estimate relative risk ratios. Compared to driving, we found strong evidence that a trip originating from an apartment-based household was less than half as likely to be taken by bicycle as a similar trip originating in a house-based household in Toronto in 2011. Increased population density of the household location had a positive impact on the likelihood of a trip being taken by walking and a negligible and uncertain impact on the likelihood of it being taken by transit, but a
negative impact on bicycling. Further analysis found the negative impact of density does not seem to apply to those living in single detached housing, but rather it only negatively impacts the likeliness of cycling among apartment and townhouse dwellers. Further research is required to identify the exact barriers to cycling, apartment dwellers experience.

 

Unfortunately, we cannot provide a direct link to the paper at this time, but we’ll provide a link to TRR once the paper is published.