Cycling in Toronto: Route Choice Behaviour and Implications to Infrastructure Planning

1 This research investigates the route choice behavior of cyclists in the City of Toronto using data 2 collected from a smartphone application deployed to a large number of cyclists in the City. A 3 total 4556 cyclists registered for this study and logged over 20,000 commuting trips over a study 4 period of 9 months. In addition to the time-stamped second-by-second GPS readings of each 5 trip, information on age, gender and residence, work or school place is also collected on a 6 voluntary basis. Multi-nominal logit route choice models are calibrated for the commuting 7 cycling trips. The results have revealed the critical importance of cycling facilities such as 8 bicycle lanes, cycling paths and trails on cyclists’ route choice decisions, providing valuable 9 information for Toronto’s ongoing effort on bicycle network planning. 10 Li, Muresan, and Fu 3 INTRODUCTION AND BACKGROUND 1 Increased travel demand from population growth in many regions worldwide creates substantial 2 challenges for planners and engineers. In many regions, these travel demand increases lead to 3 increases in motor vehicle traffic, which in turn increases congestion, collisions, noise and 4 pollution. The Greater Toronto Area (GTA) in Ontario, Canada is one of Ontario’s fastest 5 growing regions, and its population is projected to increase by 40% by 2041 (1). Previous studies 6 have estimated that congestion in Toronto already costs commuters $3.3 billion in lost time and 7 additional vehicle maintenance, and may also cost the wider economy an additional $2.7 billion 8 (2). Additionally, it is estimated that transportation currently represents 24% of total greenhouse 9 gas (GHG) emissions (3). Regional governments therefore have all been taking initiatives to 10 encourage the use of alternative transportation modes. In particular, active modes of 11 transportation, such as walking and cycling, have been promoted as sustainable alternatives to 12 automobile travel. With a population of over 2.6 million Toronto is the fifth largest city in North 13 America. Cycling in Toronto has seen substantial growth over the past five years, with the 14 number of commuting trips made by residents increasing by 33% from 2001 to 2006. Despite 15 this, its popularity as a primary mode choice is still low, and represents only about 1.7% of all 16 commuting trips (4). 17 The City of Toronto is currently developing a new long term plan to revitalize its cycling 18 network to achieve the goal of increased cycling. A number of cycling network projects are 19 being proposed. These projects are to be evaluated on eight evaluation criteria, including 20 Connectivity, Coverage, Crossing Barriers, Current Demand, Potential Demand, Trip 21 Generators, Safety, and Population and Employment Density. One of the top requirements for 22 estimating the benefits of these projects is quantitative understanding of factors affecting traveler 23 preferences towards different modes as well as different bicycle facilities. 24 Although past research has investigated the effects of many different factors on cycling, few 25 have considered both trip operational and network factors. Also, of the past studies, most are 26 limited to the evidence of stated preferences with some hypothetical choice options or a small 27 sample of revealed preference data, which can only be used to examine the effects of factors 28 separately with little possibility of revealing hidden interactions and the relative importance of 29 each factor in cyclists’ route choice behavior. The objective of this study is therefore to 30 investigate the cyclists’ choice behavior using reveal preferences data from their real cycling 31 decisions. With a data set of over 20,000 commuting trips, the study has the opportunity to 32 examine the effect of multiple factors, including factors that have not been analyzed in previous 33 studies, such as types of bicycle facilities, energy consumption, and number of bus stops. 34 The remainder of this paper describes the literature on related topics, the methodology applied, 35 and the results of the analysis for the dataset. The next section presents a summary of the data 36 collected, including an overview of the processing required to prepare the data for input in a logit 37 model. Following this, the next section gives an overview of the model specification, including a 38 discussion on the rationale behind the variables included and overview of how the logit model is 39 applied to solve this problem. Finally, the last two sections review the results and provide the 40 conclusions and recommendations for future research. 41 Li, Muresan, and Fu 4