Findings of the Cycling in Cities Study

Home , Route assignment

IMPROVING PUBLIC HEALTH THROUGH ACTIVE TRANSPORTATION: UNDERSTANDING THE INFLUENCE OF THE BUILT ENVIRONMENT ON DECISIONS TO TRAVEL BY BICYCLE

MEGHAN LESLEY WINTERS

M.Sc., University of British Columbia, 2006 B.Sc. (Honors), University of British Columbia, 1999

A THESIS SUBMITTED IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF

DOCTOR OF PHILOSOPHY in THE FACULTY OF GRADUATE STUDIES (Population and Public Health)

THE UNIVERSITY OF BRITISH COLUMBIA (Vancouver)

March 2011

ABSTRACT

Urban transportation is a public and environmental health issue. In North America, where urban environments have been shown to influence transportation decisions and physical activity, over two- thirds of adults are inactive. Consequently, there is growing interest in promoting active transportation. Cycling in particular offers one of the greatest opportunities for change. This dissertation aimed to understand how to design cities to support cycling, to improve public health through increased active transportation. It applied quantitative and qualitative methods to investigate the link between cycling and the built environment. The five studies that comprise this dissertation use data from the Cycling in Cities survey, which captured the opinions and travel behaviours of 2,149 current and potential cyclists across Metro Vancouver. The first study analyzed preferences for 16 types of cycling infrastructure, noting a clear desire for off-street and separated facilities, especially among women, people with children, and occasional and potential cyclists. The second study evaluated the relative importance of 73 potential motivators and deterrents. Environmental and engineering factors carried the strongest influence; specifically aspects related to scenery, topography, facility design, weather, and safety issues. The third and fourth studies mapped travel data to determine associations with measures of the built environment. The route choice analysis found that the majority of trips were less than 10% longer than the shortest distance route, and that bicycle trips detoured toward bicycle facilities and away from major roads, whereas car trips detoured toward highways and arterials. The mode choice analysis (bicycle versus car) made explicit consideration of the built environment around trip origin, destination and en route. Multi-level logistic modeling, adjusted for demographics and trip distance, showed significant associations with topography, cycling facilities, the road network and land use. The fifth study integrated these results with focus group findings to derive an evidence-based ―bikeability‖ measure. The utility of the index was demonstrated through its application as a planning tool. Taken collectively, these studies contribute to both data and methodological gaps in prior health, planning, and transportation research. This dissertation provides evidence on environments that support cycling and presents a tool to guide strategies to improve conditions.

PREFACE

Each research chapter of this dissertation (Chapters 2-6) was written as a stand-alone manuscript for publication in the peer-reviewed academic literature. Four have already been published (Chapters 2- 5). As primary author, I led each of these chapters. In this section I provide details of my contributions and those of my co-authors to each of the publications. Data collection for the Cycling in Cities survey was approved by the University of British Columbia‘s Behavioural Research Ethics Board (certificate: H08-80976). The collection of travel routes for Chapter 4 was approved by the Board in a separate application (certificate: H08-01045).

Chapter 2: Route preferences of ―near market‖ cyclists

For this paper, I conducted the literature review, completed the analysis, and wrote the manuscript. Kay Teschke designed and initiated the Cycling in Cities survey, provided guidance through the analysis and assisted with revision of the manuscript. My overall contribution: 90%.

A version of Chapter 2 has been published: Winters, M., Teschke, K. Route preferences among adults in the near market for bicycling: Findings of the Cycling in Cities Study. Am J Health Promot. 2010; 25 (1): 40.

Chapter 3: Motivators and deterrents of bicycling

Diana Kao conducted a review of the literature to develop the survey tool for the Cycling in Cities survey, and Kay Teschke designed and initiated the survey. I conducted the literature review for the manuscript, completed the analysis, and wrote the manuscript. Kay Teschke provided guidance through the analysis and assisted with revision of the manuscript. Gavin Davidson contributed to the interpretation of the results. My overall contribution: 90%.

A version of Chapter 3 has been published: Winters, M., Davidson, G., Kao, D., Teschke, K. Motivators and deterrents of bicycling: comparing influences on decisions to ride. Transportation. 2011; 38(1): 153-168.

Chapter 4: Built environment influences on route choice

Michael Brauer and I conceptualized this project. I designed the interview guide, conducted all interviews and data collection, mapped the travel routes, and completed all analyses. I prepared the manuscripts and completed revisions. Kay Teschke and Michael Brauer oversaw the analyses, and Eleanor Setton assisted with data issues and mapping. Michael Grant generated shortest path routes. All co-authors provided feedback for revision of the manuscript. My overall contribution: 95%.

A version of Chapter 4 is currently in press: Winters, M., Teschke, K., Grant, M., Setton, E.M., Brauer, M. How far out of the way will we travel? Built environment influences on route selection for bicycle and car travel. Transp Res Rec.

iii

Chapter 5: Built environment influences on mode choice

Kay Teschke and I devised the idea for this study. Eleanor Setton provided guidance on GIS methodology and compiling data. I conducted the mapping and statistical analysis, completed the literature review and prepared the manuscript. All co-authors assisted with interpretation of results and revisions to the manuscript. My overall contribution: 95%.

A version of Chapter 4 has been published: Winters, M., Brauer, M., Setton, E.M., Teschke, K. Built environment influences on healthy transportation choices: bicycling versus driving. J Urban Health. 2010; 87(6): 969-993.

Chapter 6: Defining and mapping bikeability

For this paper, I conceived of the study idea, derived the bikeability index, created the mapping tool, and prepared the manuscript. Michael Brauer, Eleanor Setton, and Kay Teschke provided feedback on early iterations of the index and tool, and helped to frame and improve the manuscript. My overall contribution: 95%.

A version of this paper will be submitted as: Winters, M., Brauer, M., Setton, E.M, Teschke, K. Mapping bikeability index: A tool for healthy travel.

TABLE OF CONTENTS

ABSTRACT ...... ii PREFACE ...... iii TABLE OF CONTENTS ...... v LIST OF TABLES ...... viii LIST OF FIGURES ...... x ACKNOWLEDGEMENTS ...... xi DEDICATION ...... xii

1. INTRODUCTION ...... 1 1.1. Literature review ...... 1 1.1.1. The link between transportation and health ...... 1 1.1.2. Cycling: a public health solution ...... 2 1.1.3. Health promotion models ...... 3 1.1.4. Targeting the built environment ...... 4 1.1.5. Existing research on cycling ...... 5 1.1.6. Gaps in research ...... 5 1.2. Rationale ...... 7 1.2.1. Overarching objectives...... 7 1.2.2. Dissertation structure ...... 7 1.2.3. Data sources and consistency...... 9 1.2.4. Study setting ...... 9 2. ROUTE PREFERENCES OF ―NEAR MARKET‖ CYCLISTS ...... 12 2.1. Synopsis ...... 12 2.2. Introduction ...... 13 2.3. Methods ...... 14 2.3.1. Location ...... 14 2.3.2. Design ...... 14 2.3.3. Measures ...... 15 2.3.4. Sample ...... 15 2.3.5. Analysis ...... 15 2.4. Results ...... 16 2.4.1. Demographics ...... 16 2.4.2. Preferences ...... 16 2.4.3. Current use patterns ...... 18 2.4.4. Current use versus preferences ...... 18 2.5. Discussion ...... 18 2.5.1. The near market for cycling ...... 18 2.5.2. Route preferences ...... 19 2.5.3. Current use patterns versus preferences ...... 20 2.5.4. Study strengths and limitations ...... 21 2.6. Conclusions ...... 21 3. MOTIVATORS AND DETERRENTS OF BICYCLING ...... 27

3.1. Synopsis ...... 27 3.2. Introduction ...... 28 3.3. Methods ...... 28 3.3.1. Survey design and administration ...... 28 3.3.2. Data summarization and analyses ...... 30 3.4. Results ...... 31 3.5. Discussion ...... 32 3.5.1. Survey strengths and limitations ...... 35 3.6. Conclusions ...... 36 4. BUILT ENVIRONMENT INFLUENCES ON ROUTE SELECTION...... 46 4.1. Synopsis ...... 46 4.2. Introduction ...... 47 4.3. Methods ...... 48 4.3.1. Setting ...... 48 4.3.2. Interview data collection ...... 48 4.3.3. GIS and statistical analysis ...... 48 4.4. Results ...... 51 4.4.1. Trip distance ...... 51 4.4.2. Built environment measures ...... 52 4.4.3. Road class analysis ...... 53 4.5. Discussion ...... 53 4.6. Conclusions ...... 58 5. BUILT ENVIRONMENT INFLUENCES ON MODE CHOICE ...... 69 5.1. Synopsis ...... 69 5.2. Background ...... 70 5.3. Methods ...... 71 5.3.1. Trip data ...... 71 5.3.2. Demographic variables ...... 72 5.3.3. Spatial analysis zones ...... 72 5.3.4. Built environment measures ...... 73 5.3.5. Statistical Analysis ...... 75 5.4. Results ...... 76 5.4.1. Bivariate comparison of built environment characteristics of bicycle versus car trips . 77 5.4.2. Multivariable models for built environment characteristics of bicycle versus car trips 78 5.5. Discussion ...... 79 5.5.1. Strengths and limitations...... 83 5.6. Conclusions ...... 85 6. DEFINING AND MAPPING BIKEABILITY ...... 95 6.1. Synopsis ...... 95 6.2. Introduction ...... 96 6.3. Components of bikeability ...... 97 6.3.1. Identifying built environment factors ...... 97 6.3.2. Developing the bikeability index ...... 99 6.4. GIS procedures ...... 100 6.4.1. Generating component raster files ...... 101 6.4.2. Scoring and combining component files ...... 101 6.5. Results ...... 102 6.6. Discussion ...... 103

6.6.1. Limitations ...... 105 6.7. Conclusions ...... 106 7. SYNTHESIS: CONTRIBUTIONS, IMPACTS, FUTURE DIRECTIONS ...... 115 7.1. Summary ...... 115 7.1.1. Objectives...... 115 7.1.2. Findings ...... 115 7.2. Unique contributions ...... 116 7.2.1. Interdisciplinarity ...... 116 7.2.2. A new population ...... 117 7.2.3. Integration of multiple methodologies ...... 117 7.3. Policy messages and impacts ...... 118 7.3.1. Policy implications ...... 119 7.3.2. Knowledge translation and exchange ...... 120 7.3.3. Impacts to date ...... 121 7.4. Limitations and future endeavors ...... 122 7.5. Conclusion...... 125 REFERENCES...... 126 APPENDIX 1: FOCUS GROUP REPORT ...... 138

vii

LIST OF TABLES

Table 2.1 Demographic characteristics of web/mail survey respondents in Metro Vancouver (n=1,402) according to cyclist segment ...... 23

Table 3.1 Demographic characteristics of respondents who completed the web/mail survey (n= 1,402) according to cyclist typea ...... 39

Table 3.2 Means and standard errors for each survey item, and results of factor analysis...... 40

Table 4.1 Interview guide (abbreviated) ...... 59

Table 4.2 Demographic characteristics of the study population (n=74 regular, infrequent and potential cyclists) ...... 60

Table 4.3 Distances of shortest and actual routes (km) ...... 61

Table 4.4 Comparison of built environment characteristicsa for each pair of actual and shortest routes ...... 62

Table 4.5 Multiple logistic regression of likelihood of detouring > 10% beyond shortest-distance route for bike trips (n=50), based on differences in the built environment features between the actual and shortest-distance routes, and reported safety perceptions ...... 63

Table 4.6 Comparison of distance traveled on each road class, by mode of travel ...... 64

Table 4.7 Examples of respondent‘s reasons for route choices on bicycle trips ...... 65

Table 5.1 Descriptive characteristics of trip takers (n=1902) (before imputation for missing data)... 86

Table 5.2 A priori hypotheses on measureable built environment measures that affect the likelihood of making a trip by bicycle, with hypothesized direction of influence, and relevant spatial zone ...... 87

Table 5.3 Trip distance and spatial analysis zone dimensions, by mode ...... 88

Table 5.4 Descriptive statistics for built environment measures for route zones, by trip mode ...... 89

Table 5.5 Results of multi-level logistic models for effect of built environment measures on the likelihood that a trip is made by bicycle instead of car ...... 90

Table 5.6 Cross-zonal multi-level logistic modelsa for effect of built environment on the likelihood that a trip is made by bicycle instead of car, for all trips, and for short trips only ...... 92

Table 6.1 Ranking of built environment factors in focus groups with current and potential cyclists (n=23) ...... 107

viii

Table 6.2 Developing an evidence-based bikeability index ...... 108

Table 6.3 Scoring the components of bikeability to create maps, using empirical data from Metro Vancouver ...... 109

Table 6.4 Concurrent projects using spatial data to capture ―bikeability‖ ...... 110

LIST OF FIGURES

Figure 1.1 Travel mode for work commute, by distance (Source: Statistics Canada)[21] ...... 11

Figure 2.1 Likelihood of choosing route type (if all route types were available) for 16 cycling route types among current and potential cyclists in Metro Vancouver (n=1,402) ...... 24

Figure 2.2 Mean preference score for 16 route types according to cyclist segment (A) and gender (B) ...... 25

Figure 2.3 Current use versus likelihood of choosing for 16 cycling route types ...... 26

Figure 3.1 Bicycle route and road network of Metro Vancouver...... 44

Figure 3.2 Top 10 motivating and top 10 deterring influences on cycling, of 73 items asked of survey respondents (n=1,402)a. Mean scores overall and as reported by each cyclist typeb ...... 45

Figure 4.1 Bike, car and shortest distance routes for participant who commutes using both modes . 66

Figure 4.2 The detour factor (Ratio of distance for actual trip: shortest trip) by modea ...... 67

Figure 4.3 Differences in road class usage between actual routes and shortest routes, by mode of travel ...... 68

Figure 5.1 Potential zones influencing decisions to cycle: route, origin, and destination zones ...... 93

Figure 5.2 Differential effect of large commercial and small commercial land use on likelihood that a trip is made by bicycle, instead of car ...... 94

Figure 6.1 Data sources and methodology for derivation of bikeability index ...... 111

Figure 6.2 Bikeability and component maps for Metro Vancouver ...... 112

Figure 6.3 Mean bikeability and component values for Metro Vancouver municipalities ...... 113

Figure 6.4 Mean bikeability and component values for Vancouver neighbourhoods ...... 114

ACKNOWLEDGEMENTS

The many people who have supported me and my research must be acknowledged here.

My first thank you is to my supervisor, Kay Teschke. I could not have hoped for a better mentor, both for her eternal support and encouragement, and as a model of rigorous research and teaching practice. I am grateful to Michael Brauer, both for the challenges offered and opportunities created. And to Eleanor Setton, who was always game to find technical solutions and provide patient explanations. I can‘t imagine many other students have had such an attentive and available committee.

I have many valued colleagues who facilitated this research and the knowledge translation activities. First to Gavin Davidson, whose collaboration in his role at Translink was invaluable in making this research both possible and useful. To Melissa Nunes and Michael Grant who provided essential technical contributions. It was a pleasure working with each of you. To Adam Cooper and Cam Pearce for their role in conducting focus groups, and to Christie Hurrell for continual assistance with communication and design. To Larry Frank, the UBC Bridge Program, and staff at the City of Vancouver, for providing input, guidance, and for taking interest all along the way.

Financial support for my training was generously provided by the Canadian Institutes of Health Research and the Michael Smith Foundation for Health Research. Financial support for the project was provided by the Moving on Sustainable Transportation Program, the Canadian Institutes of Health Research and the Heart and Stroke Foundation. I appreciate that these funding agencies recognized that cycling research is health research. The Cycling in Cities Survey was also funded by Transportation Program, Michael Smith Foundation for Health Research, TransLink, Metro Vancouver, and the cities of Langley, New Westminster, Richmond, Port Moody, Surrey, Vancouver, White Rock and the Township of Langley.

Thank you to my friends and family who have expressed endless confidence in me, and in the value of this work. To the eyes and ears of my fellow students Negar Elmieh, Glenys Webster, Anne Harris, Sarah Henderson, Conor Reynolds, and Perry Hystad, for their help along the way. To my parents, Louise, and my friends, for the many hours they contributed so that I could conduct this research. And finally, to Josh and the two young boys who came into our lives during this dissertation: your laughter and joy have brought me the balance I needed to get this done, smiling.

DEDICATION

For Nigel and Sebastien.

It’s my hope that research like this helps to make your world a healthier place, and a happier place to get around.

xii

1. INTRODUCTION

1.1. Literature review

1.1.1. The link between transportation and health Transportation is an essential component of society, providing access to services, education, employment, social opportunities, and goods and amenities. However, transportation is also a public and environmental health issue[1]. The existing auto-focused culture has negative health effects on individuals, both as a facilitator of physical inactivity and through injuries from traffic collisions. It also has negative health effects on the wider population, through air and noise pollution and the connection to respiratory and cardiovascular health[2-4]. Furthermore, there are more widespread environmental and societal issues: climate change, sprawl, highways that divide communities, and longer commute distances that reduce time for leisure and families. As a consequence of these negative health impacts, there has been growing interest within countries with high rates of automobile use (particularly in North America and Europe) in promoting public transit, walking and cycling as alternatives.

―Active transportation‖ refers mainly to walking and cycling for transportation. The 1996 US Surgeon General's Report[5] indicated that 30 minutes of moderate activity most days of the week was sufficient to achieve health benefits, even if activity was in 2-3 sessions over the day (e.g., a trip to and from work). This shift, from a previous focus on high intensity activities (e.g., jogging, aerobics), suggested that promotion of active transportation was a promising path forward to address the widespread levels of inactivity[6, 7].

Indeed, levels of physical inactivity are currently reaching epidemic proportions. In North America over two-thirds of adults are inactive[8, 9] and obesity levels have escalated in recent decades[10]. There are huge associated health care costs, with the annual economic cost of inactivity estimated at $2 billion in Canada[9], and obesity and inactivity together exceeding $70 billion annually in the United States[11]. The World Health Organization has estimated that physical inactivity costs a country €150–300 per person per year[12]. Inactivity is directly linked to obesity, and is a major risk factor for virtually all chronic diseases (e.g., diabetes, heart disease, and cancer) as well as mental

1 health[13]. With fewer than 10% of working Canadians currently commuting by active transportation modes[14] a shift in modal share could have a dramatic impact on population health.

1.1.2. Cycling: a public health solution

The promotion of cycling in particular offers one of the greatest opportunities for change. Cycling and walking have similar health benefits, but cycling offers the advantages of faster travel and extended trip distances, making it a more suitable substitute for car travel as trip distances exceed one kilometer[15]. Cycling is a widely accessible, cost-effective form of both transport and exercise, with minimal requirements for specialized equipment or training. Making commute trips by bicycle integrates physical activity into daily travel routines, and is a more sustainable means of meeting recommended levels of physical activity than structured activities (e.g., running or going to the gym)[16] with no reduction in health benefits. Research indicates that active commuters get more physical activity[17], have improved fitness[17, 18], lower all-cause mortality[19], lower cardiovascular disease[20], and less work absenteeism[21], compared with those who use motorized transportation. Moreover, current travel patterns show that a shift to cycling is feasible. Over one- third of Canadians have work commutes under 5 km[22] – distances that can be covered in 10-15 minutes by bicycle. At present, the vast majority of short commutes are made by car (Figure 1): about 70% of those between 1-4 km, and 80% of those between 5 and 15 km.

Additionally, cycling rates in Canadian cities are low compared to European regions with similar climates and demographics (2% of trips versus 15-30%, in Germany, Denmark and Netherlands)[23]. The differences in cycling rates between Canadian cities also suggest opportunities for growth. For example, Victoria has the highest share of work commute trips made by bicycle at 5.6%, versus Ottawa at 2.2%, Metro Vancouver at 1.7%, and Greater Toronto and Halifax at about 1%[24]. Comparisons in the demographics of those who cycle support perhaps even more important opportunities for growth. In Germany and the Netherlands, there is virtually no disparity in the proportion of cycling trips made throughout the age range from youth to post retirement, and both men and women cycle at similar rates[23]. In contrast, in Canada young men are approximately twice as likely to cycle as women and older men[24].

Is cycling really healthy? A public health lens must always look at risks associated with alleged ―healthy‖ behaviours. The evidence is clear that cyclists face significantly higher risks of fatality and injury, per time or per distance travelled, than people who use cars, bus or rail[23]. Luckily the

―safety in numbers‖ phenomenon[25] – the observed pattern that injury and fatality rates decrease as active transportation rates increase – offers hope that promoting cycling will in turn make cycling safer. Another health concern is the exposure to air pollution while cycling. A recent summary of this literature[26] suggests that cyclists are typically exposed to slightly lower concentrations than drivers, but may have higher inhaled doses due to increased breathing rates or longer travel times. However beyond experimental situations, many factors affect the relative exposure levels: the route selected, specific street configurations, car speed and conditions, trip duration, weather conditions, and more. Efforts to weigh the overall health impact of cycling have found that benefits greatly outweigh the risks (ratios of 9-20:1) despite uncertainties in the estimates[26, 27]. Models indicate that strategies to promote active transportation will have far more positive health effects for society than will technological improvements to motor vehicles[28].

1.1.3. Health promotion models This public health situation suggests that an increase in the uptake of cycling is a potentially useful strategy to improve physical activity. Traditional strategies that provide individually-tailored behaviour and lifestyle modification programs have proven effectiveness, but they are intensive and costly, and can reach only a small proportion of the population[16]. The discipline of health promotion offers a number of models of behaviour change that can guide future strategies to promote cycling. Health behaviour models can take a variety of perspectives: some theories attempt to explain health behaviours and health behaviour change by focusing on the individual, whereas others consider broader interpersonal, organizational or community influences on health. Two models in particular shed light on ways to achieve more dramatic changes in cycling rates at the population level.

The first is the Stages of Change Model[29, 30]. This psychological model considers an individual‘s readiness to make a change to health behaviour, and describes five stages that an individual moves through in adopting a behaviour change: precontemplation; contemplation; preparation; action; and maintenance. This model suggests that interventions will have the most impact by targeting key populations of those in contemplation (people intending to change in the next 6 months), or preparation (people intending to take action in the immediate future, measured within a month). Future cycling research should focus on understanding travel decision making in these populations.

The second is the Social Ecological Model[31]. This model suggests that behaviour is influenced by factors at multiple levels: interpersonal; intrapersonal (groups, families); institutional (workplaces,

3 schools); community factors (transportation network, design); and public policy factors (law enforcement, land use regulation). Proponents of this model argue that interventions are required at multiple levels (individual, social, physical environment, and policy) to achieve population-level changes in physical activity[7], and that ―the most powerful interventions should: (a) ensure safe, attractive, and convenient places for physical activity, (b) implement motivational and educational programs to encourage use of those places, and (c) use mass media and community organization to change social norms and culture‖. Future research needs to understand how to create environments that are safe, attractive, and convenient for cycling.

1.1.4. Targeting the built environment

Given that cycling rates vary between cities, and that theory suggests that urban environments can influence health behaviours, there is reason to focus on urban form as a possible determinant of healthy travel. Indeed, empirical evidence has shown that environment and policy approaches to health promotion, such as changes to the urban design of streets and communities, increase rates of walking and cycling[32-34]. In the urban planning and physical activity literature there is mounting research focused on healthy urban design, aimed to understand the ways that the ―built environment‖ promotes or inhibits physical activity. The built environment encompasses urban design (both at the community- and street-scale), land use (the distribution of activities) and transportation systems (the facilities and services that link one location to another). It is experienced by the entire population, and thus changes to urban form and transportation infrastructure have the potential to influence the health of large numbers of people. It is an appropriate focus from a public health perspective, as it removes the onus from the individual to initiate healthy choices and instead creates health-supportive environments for all. A number of recent reviews document the current knowledge and future directions for built environment research[35-39]. To date, the bulk of research has focused on the link between the built environment and walking or physical activity, but select studies have also considered outcomes of injuries, overweight and obesity rates, air pollution levels, and travel behaviours[40, 41].

Of specific relevance to this dissertation is that little of this built environment research explicitly considers cycling. In part this is attributable to data limitations. As cycling is a fringe transportation mode throughout much of North America, general population surveys have not provided adequate power for cycling-specific inquiries. Some studies have pooled cycling with walking (as ―active transportation‖) or used general physical activity outcomes. Yet walking and cycling are functionally

4 different in terms of travel speeds and distance, demographics of use (in North America), and legislation[23, 42]. Given this, evidence on how the built environment influences walking or overall physical activity may not be generalizable to cycling. In particular, the features of a ―walkable‖ neighbourhood (high density, access to shops, grid-based street networks, interesting street design) may not directly correspond to a ―bikeable‖ neighbourhood.

1.1.5. Existing research on cycling This section provides a very broad overview of knowledge on bicycle ridership. More detail on specific studies is discussed in each research chapter of the dissertation. The research originates from the public health, transportation, urban planning, and engineering domains, and not surprisingly, the diverse disciplines have applied a range of approaches and methodologies. Previous studies of cycling can be categorized as three common types:

 general population surveys documenting the number of people who cycle, the numbers of trips by bicycle, and the demographic characteristics of cyclists (e.g., the Canadian Census, or the NHTS in the US)[24, 43];  surveys and stated preference experiments which identify factors that might influence change in transportation mode based on people‘s opinions[44-46]; and  ecological studies that examine the influence of urban form (e.g., average population density, usual cycling infrastructure, average topography) on differences in observed cycling rates, using aggregated data to make comparisons between cities, or between countries[47-49].

These studies have documented who cycles, what proportion of trips are by bicycle, opinions on what may encourage cycling, and comparisons of cities with high and low cycling rates. Determinants of ridership include individual-level factors such as age, gender, income, education, ethnicity, cycling experience, and commute distance, as well as community-level factors such as safety, weather, traffic, topography, cycling infrastructure, proportion of students, and population density[44, 47-56].

1.1.6. Gaps in research Despite current knowledge, most North American cities are struggling to accomplish more than modest increases in cycling rates. Several methodological and substantive gaps on cycling and the role of infrastructure need to be addressed to move this agenda forward:

Population: Much of the research evidence to date has a focus on regular commuter cyclists as this is the population is the most easily accessed. National surveys capture ‗typical commute mode‘ (Canadian Census) or ‗number of trips per week‘ (US National Household Travel Survey), providing large samples to characterize regular cyclists, demographically and geographically[55, 56]. Other studies have used convenience sampling for recruitment via listserves or intercept surveys[50, 52, 54, 57], resulting in a bias (albeit acknowledged) toward regular cyclists. Much less is known about the population who currently cycle infrequently, or those who are contemplating cycling.

Opinion versus actual behaviour: Research has documented cyclists‘ opinions and self-reported travel behaviours[44, 52]. Stated preference experiments have provided evidence on preferred conditions given a number of hypothetical situations (e.g., route type, travel time, stop signs or car parking on route)[45, 58]. However, people‘s opinions do not necessarily reflect their true behaviour. Some research has tracked actual behaviour[57, 59, 60], but cost typically limits sample size or geographical scope. Additionally, the use of actual travel data cannot provide insight on new types of infrastructure, or types that do not exist in a given location. In reality, data on both opinion and actual behaviour are necessary; it is consistency in results across study designs that can lead to causal linkages.

Ecological versus individual data: Traditionally, a lack of individual-level data has hampered the ability to investigate a direct link between the environment and cycling. Census data can provide a broad characterization of the spatial variation in travel behaviour, but cannot disentangle the contributions of the physical environment and individual characteristics. Disaggregated and geographically located data is needed. A nascent body of research has tapped into survey and travel diary data[61, 62] or used intensive global positioning system (GPS) data collection[57] to generate individual-level data to examine the influence of the built environment on cycling.

Cycling-specific perspective: Much of the previous built environment research has used methodology tailored to walking or physical activity. A new lens is required that is cycling-specific, creating measures relevant to cycling (e.g., bike routes, bicycle-friendliness), and considering appropriate spatial zones (e.g., neighbourhoods, routes) to understand the relationship between the built environment factors and cycling behaviour. The pioneers in this area[57, 61, 62] have used differing methods and measures, making comparisons between studies challenging.

1.2. Rationale This dissertation aimed to understand how to design cities to support cycling in order to improve public health through active transportation. Addressing the gaps in the transportation, health, and urban planning literature mentioned above required a multi-faceted study. This research integrated qualitative and quantitative research, including opinions, actual travel behaviour, and focus groups, to investigate the association between the built environment and cycling in the Canadian context. It targeted the next wave of cyclists, or the ―near market‖, guided by the goal of getting a larger proportion of the population physically active. Additionally, since modal substitution – travel by bicycle instead of by car – will result in the greatest health benefits, the focus was on cycling for utilitarian purposes (for work, school, or errands) instead of recreational cycling.

1.2.1. Overarching objectives The following were the specific objectives of the dissertation:

1. To characterize the near market for cycling, according to demographic characteristics, stated cycling patterns and route preferences, and opinions on factors that influence their cycling. 2. To map built environment characteristics that may impede or facilitate bicycle travel. 3. To link built environment characteristics to actual travel data to determine which characteristics influence (a) decisions on route selection and (b) decisions on mode choice 4. To develop a measure for ―bikeability‖ based on evidence produced in this dissertation and associated projects, and apply it to Metro Vancouver. 5. To create outputs that promote the uptake and application of findings by governments and other agencies that plan urban areas, fund and build cycling infrastructure, or promote cycling.

1.2.2. Dissertation structure This dissertation consists of seven chapters: this introduction chapter, five research chapters that address the objectives above, and a concluding chapter. The rationale and primary objective of each research chapter are described below.

Chapter 2: Route preferences of “near market” cyclists Prior surveys on bicycle infrastructure have included a limited set of route types and used jargon that may not be widely understood by the general public (i.e., bicycle paths, bicycle routes). The study

7 describes opinions on 16 different route types. The opinion survey instrument used photos to improve the validity of responses. The objective of this chapter was to provide direction on how to design transportation infrastructure to increase cycling modal share by evaluating preference and usage patterns of the different route types, especially in potential cyclists and underrepresented subpopulations (women, families).

Chapter 3: Motivators and deterrents of bicycling The influences on decisions to bicycle are clearly multidimensional[63]. In order to compare the relative influence of factors related to engineering, education, encouragement, enforcement, and the environment, all aspects need to be included in a single study. The objective of this chapter was to identify key motivators and deterrents amongst an extensive list of 73 factors in order to inform strategic cycling promotion approaches.

Chapter 4: Built environment influences on route choice The knowledge gap on the travel patterns of cyclists has resulted in assumptions that the shortest (or fastest) route is selected, based on the notion that speed and energy are limiting factors. Such assumptions have brought inaccuracies to non-motorized transportation planning[64]. The objective of this chapter was to compare the travel patterns for bicycle trips and for car trips, and to determine how built environment influences route selection.

Chapter 5: Built environment influences on mode choice It is clear from the literature that there are many challenges in measuring the built environment, and that there have been only limited attempts to understand the influence of the built environment on cycling. The primary objective of this chapter was to apply a cycling specific lens on how the built environment influences the likelihood of cycling in different spatial zones (trip origin, destination, or route).

Chapter 6: Defining and mapping bikeability This chapter applied the findings of this thesis, synthesizing evidence from the previous chapters and enriching it with contextual data from focus group sessions. The primary objective was to develop an evidence-based construct for ―bikeability‖ and to use it to create a tool for use by planners, policy makers, and the public that can guide strategic interventions to improve conditions to support cycling.

1.2.3. Data sources and consistency The studies that comprise this dissertation reflect the perceptions and behaviours of one underlying population, enabling comparison and integration of the findings across chapters. The primary data source for this research was the Cycling in Cities survey. This was a population-based survey of Metro Vancouver adults in the ―near market‖ for cycling, defined as those who owned a bicycle and either currently cycled or were willing to cycle more in the future. The survey was initiated by researchers at the University of British Columbia, and was conducted in partnership with the regional transportation authority and local municipalities. It gathered opinions on preferences for route types and influences on decisions to cycle; it also recorded actual travel behaviours, including origin, destination, and travel mode for two common trips. Additionally, the Cycling in Cities survey population served as the sampling frame for two new data collection efforts required to complete these studies: gathering data on street-by-street travel routes to understand route selection; and recruiting participants for focus groups to contextualize what makes a neighbourhood bikeable. The chapters of this dissertation applied the data sources to a number of different study designs. Consistency in the results across chapters brings further strength to the link between the built environment and cycling.

1.2.4. Study setting The Metro Vancouver region was a suitable location to conduct this research. It is the third largest urban region in Canada and one with a climate that favours year-round cycling. It is comprised of 21 diverse municipalities, from the dense downtown core of the City of Vancouver, to suburban cities and rural farmlands. The widely varying neighbourhood characteristics, transportation infrastructure, and cycling policies provide good variability in the built environment from a research perspective. Compared with the premier cycling cities in Europe, the population density in the City of Vancouver (~5000 population/km2) is similar to Amsterdam and Copenhagen. However, despite the density, the cycling rates do not measure up. In Copenhagen and Amsterdam more than 25% of work trips are by bicycle. In Vancouver cycling rates are low, although regionally variable: only 1.7% of work trips are by bicycle across the Metro region, but mode share is 3.7% in the City of Vancouver and well over 10% in certain neighbourhoods. Portland, the US platinum-rated cycling city, has lower population density (~1,400 population/km2) but slightly higher mode share (3.9%)[65]. These density comparisons suggest that Vancouver could be a model city; the gap in

9 cycling rates indicates, at least to the optimist, a great opportunity. These conditions are fitting for the research conducted, and for its relevance to local policy action.

Figure 1.1 Travel mode for work commute, by distance (Source: Statistics Canada)[22]

walk/bike/transit

car

2. ROUTE PREFERENCES OF “NEAR MARKET” CYCLISTS 2.1. Synopsis This study aimed to determine the preferences and usage of specific types of cycling infrastructure, among current and potential cyclists. Data came from a population-based web-survey of 1,402 adults in Metro Vancouver, Canada, which used pictures to depict 16 different route types. Eligibility was restricted to the ―near market‖ for cycling, representing ~31% of the population or ~500,000 adults regionally. By broad route categories, most respondents were likely or very likely to choose to cycle on off-street paths (71 – 85% of respondents), followed by physically separated routes next to major roads (71%), then residential routes (48 - 65%). Rural roads (21 – 49%) and routes on major streets (16 – 52%) were least likely to be chosen. Within the broad categories, routes with traffic calming, bike lanes, paved surfaces, and no on-street parking were preferred, resulting in increases in likelihood of choosing the route from 12 to 37%. The results indicated a marked disparity between preferred cycling infrastructure and the route types that were available and commonly used. This study provides evidence to direct transportation engineering and urban planning on infrastructure design that could lead to an increase bicycle mode share.

2.2. Introduction The promotion of active transportation is a promising path to improve public health, addressing the widespread levels of inactivity in the population and simultaneously reducing air and noise pollution through the replacement of car trips by walking or cycling. Currently fewer than 10% of working Canadians commute by these active transportation modes. Cycling mode shares in Canadian cities (~2% of trips) are much lower than walking mode share (~10% of trips), and are very low compared to those in European regions with similar climates and demographics (15-30% of trips in Germany, Denmark and Netherlands). Therefore, increasing cycling for transportation offers one of the greatest opportunities for change, with potential corresponding benefits in physical activity levels and public health.

The goal of increasing population-wide physical activity levels calls for a shift from individual focused programs to widespread environmental and policy interventions[7]. Commuting by walking or cycling integrates physical activity into daily travel routines, providing a more sustainable means of meeting recommended levels of physical activity than tailored structured activity programs (e.g., going to the gym)[16, 17]. Creating physical environments for activity that are safe, convenient and attractive can have a positive influence that extends to the whole population[7, 39].

Recent literature has established the linkage between urban form and travel behaviour[36, 66, 67]. However, to effectively modify the design of the transportation network to induce a mode shift onto bicycles and out of cars there is a need for more detailed evidence on the specific types of infrastructure are preferred by potential users. One planning approach would be to adopt infrastructure types similar to what exists in regions with higher cycling mode share, such as the off-road or physically separated routes typical in European countries[68]. Alternatively, there are market research surveys that have elucidated preferences using very broad categories of infrastructure (e.g., ―lanes‖ vs. ―paths‖), but often without visual aids to ensure a common understanding of terminology[43, 44, 69, 70].

Studies on cycling infrastructure are heavily influenced by the population under study, since preferences for facilities may differ according to cycling experience or other personal traits[63, 71, 72]. Some surveys have sampled from the full adult population, including both cyclists and non- cyclists[43, 44, 69]. Other studies have focused on frequent cyclists, recruiting from bicycle-related clubs and list-serves, stopping cyclists on the road[45, 63], or tagging parked bicycles[73]. Others still

13 have studied specific populations such as university staff or students[74]. A strategic approach for increasing mode share is to survey the ―near market‖ for cycling, that is, members of the population most likely to be willing and interested to make changes in their travel behaviour. This group includes current cyclists who could cycle more frequently as well as non-cyclists who are willing to start cycling (―contemplators‖ [29]).

We conducted a population-based survey of adults in the near market for cycling in the Metro Vancouver region of Canada. We asked about their likelihood of choosing to cycle on 16 different route types, using a web or mailed survey instrument that included multiple photos of each of the infrastructure types. We also asked how frequently they currently use the same route types, to determine whether the routes used differed from the routes preferred. The goal was to provide evidence for urban cycling infrastructure development to promote a substantial increase in cycling mode share and attract a new wave of cyclists.

2.3. Methods

2.3.1. Location Metro Vancouver is comprised of 21 municipalities with widely varying neighbourhood characteristics and transportation infrastructure. The region is home to approximately 2.1 million people[75]. Bicycle mode share is estimated at 1.7% region-wide, and about 3.1% within the City of Vancouver[76]. The regional bicycle route network has over 1,350 kilometers of designated bicycle routes, ranging from paved off-road cycling paths to residential streets with signage only[76]. The climate is conducive to cycling year-round, with all monthly average low temperatures above freezing.

2.3.2. Design The survey instruments were developed after an extensive review of the literature, which identified 70 studies and 40 surveys related to the choice of cycling as a mode of transport. The literature provided a broad range of data elements used as the basis for designing two questionnaires that comprised this survey: a telephone interview; and a follow-up web or mail survey. The questionnaires were refined by a broad range of people interested in cycling and transportation (bicycle coordinators from the participating municipalities, members of cycling advocacy groups, and the regional transportation authority, TransLink) and in 6 focus groups (in two locations and of differing cycling segments).

The telephone interview filtered participants for eligibility, and collected demographics, travel patterns and mode choice. The self-administered follow-up survey asked about current use of and preference for 16 distinct route types using three photographs each to clearly identify different infrastructures.

2.3.3. Measures The 16 route types can be broadly classified as: (1) major streets, (2) residential streets, (3) rural roads and highways, (4) off-street paths, and (5) cycle paths next to major roads but physically separated from traffic; with additional detail covering road markings, bicycle lanes, traffic calming, route surfaces, and car parking. The current use question was ―How often do you currently cycle on this type of route?‖ and had a 5-point response scale: never; less than once a year; 1 to 10 times a year; 1 to 10 times a month; and 3 or more times a week. The preference question was ―If this and other route options were available, how likely are you to choose to cycle on this type of route?‖ with a 5-point response scale: very unlikely to choose; unlikely to choose; neutral; likely to choose; very likely to choose. Preference responses were scaled from -1 (very unlikely) to 0 (neutral) to +1 (very likely) for calculating and comparing mean scores. The questionnaire is available from the authors.

2.3.4. Sample The study was conducted in three waves distributed throughout 2006, with the focus on travel patterns in the previous 4 months. In each wave a random sample of names was selected from the telephone book and each was sent an introductory letter. In the second and third waves this was complemented by random digit dialing to increase recruitment. All study methods were reviewed and approved by the University of British Columbia‘s Behavioural Research Ethics Board.

In total, 31% of the individuals contacted were current and potential cyclists (those who had access to a bicycle and who had either cycled in the past year or were willing to consider cycling in the future), and were therefore eligible and invited to participate. 2,149 individuals completed telephone interviews (37% of those eligible). Of these, 1,402 completed the web/mail follow-up survey whose results are presented here. This subset did not differ demographically from those who completed the telephone interview, with the exception of potential cyclists (who comprised 19.5% of the telephone survey respondents, but only 13.8% of the follow-up).

2.3.5. Analysis Analyses included descriptive statistics and multiple linear regression using SAS Version 9.1 (Cary, NC) statistical package. Responses were weighted to reflect the age, gender and geographic

15 distributions of the region. Because of the large sample size, very small differences between groups were statistically significant. We have therefore not reported statistical significance, but instead focused our discussion on differences that are meaningful (i.e., % differences of at least 5%, mean score differences of at least 0.5, trends across categories).

The population was segmented into four subpopulations: regular cyclists (who cycled at least once a week, i.e., cycled ≥ 52 trips/year); frequent cyclists (at least monthly, i.e., 12-51 trips/year); occasional cyclists (at least yearly, i.e., 1-11 trips/year) and potential cyclists, who had not cycled in the previous year, but would consider cycling in the future. These annual trip frequencies were derived from responses for the number of one-way trips made by bicycle (for any trip purpose) in a typical week for 4 month season prior to that study wave. Individuals who reported zero bicycle trips in that season (67%) were also asked if they had made any trips in the year prior, which were also included in the annual trip frequency.

2.4. Results

2.4.1. Demographics Overall, 107 individuals were regular cyclists (weighted proportion=8.1%), 481 were frequent cyclists (34.6%), 617 were occasional cyclists (43.5%), and the remaining 197 individuals were potential cyclists who had not cycled in the past year (13.8%). Demographic characteristics of each segment are summarized in Table 2.1. Regular and frequent cyclists were more likely to be male (58.3% and 57.5%, respectively), and potential cyclists female (54.9%). Potential cyclists were older and more likely to be retired than the other segments. Most participants had access to a car (93.7%), although this was lower for regular cyclists (78.4%). Regular cyclists were more likely to live City of Vancouver (50.5%) than other groups (26.3%-38.1% in Vancouver).

2.4.2. Preferences There were clear differences in the desirability of route types (Figure 2.1). Off-street paths were the most preferred. Within this broad route type, nearly 85% of respondents said they would be likely or very likely to choose to ride on paved off-street paths for bikes only and 77% were likely to choose paved multi-use paths. About 71% of respondents were likely to choose unpaved multi-use paths, or cycle paths next to major roads, separated by a physical barrier. Residential streets were the next most preferred route type (48-65% likely to choose). Within this type, the presence of cycling facilities influenced preferences: residential streets designated as bike routes were preferred to those

16 not designated; and routes with traffic calming were preferred to those without. Rural roads were less preferred (21% to 49% likely to choose) but the value of cycling facilities was also observed within this classification; those with bike symbols were preferred to those with a paved shoulder, which in turn were preferred to those with no paved shoulder. Among the major street route types (16% to 52% likely to choose), routes with bike lanes were preferred to routes with only bike symbols, which were preferred to no markings at all. This was further modified by the presence of parking on-street; in all cases, the presence of on-street parking made a particular route type less favorable.

The two least preferred route types were major streets with no facilities, with or without parking (16% likely to choose). Only 79 respondents were ―very likely‖ to choose to ride on major streets with parked cars. They represented a unique subpopulation: 22.6% were regular cyclists (versus 8.1% in the overall sample); and they were disproportionately male (66.5%), aged 25-34, with a lower likelihood of having children (22.3% versus 46.8%).

Multiple linear regression models were run for the preferences scores for each of the 16 different route types, according to cyclist type, gender, age, education, household income and having children. Cyclist segment and gender were consistently significant predictors of route type preferences. For example, in the model of preference score for ―major city streets, with no facilities and no parked cars‖ (a low preference route type), regular cyclists gave an average score 0.55 greater (p< 0.0001) than potential cyclists, and 0.23 greater (p< 0.0001) than occasional cyclists. Females scored this route type 0.23 lower on average. Other variables were not significant. Since results were similar across many models the influences of demographics on preferences are best illustrated in Figures 2.2a and 2.2b.

Figure 2.2a shows preferences by cyclist segment. The rank order of preference for the route types varied little across cyclist type or other demographics, with one exception. Regular cyclists ranked unpaved off-street paths and residential streets without a cycling designation relatively lower than did other cyclist segments. The mean preference scores did vary by cyclist segment. For all route types, regular cyclists gave the highest scores, then frequent cyclists, then occasional cyclists, and finally potential cyclists. The largest differences were for the major street types.

Figure 2.2b shows preferences by sex. There were virtually no differences in mean scores between men and women for the six most preferred route types, but women scored the low preference

17 routes even lower than men. Similarly, respondents with and without children in their household scored the six preferred route types the same, but those with children scored the low preference routes lower than those without (data not shown).

2.4.3. Current use patterns For calculating the mean frequency of route use, each route use response category was assigned a value slightly below its mid-point (e.g., ―1-10 times per year‖= 4 times per year), under the assumption that the underlying within-category distributions were right skewed.

The three residential route types were the most commonly used types of infrastructure, with average usages of 40 times/year for unmarked residential streets, 33 times/year for streets marked as bike routes, 27 times/year for streets with traffic calming measures. Rural road route types and separated cycle paths next to major roads were the least commonly used (all <10 times/year). Usage patterns are included in Figure 2.3.

2.4.4. Current use versus preferences Figure 2.3 compares current use of the 16 route types to the route preferences. The crossing lines indicate a marked discrepancy between where people currently ride and where they would choose to ride, were all route types available. For example, the most commonly used route type in the current street network, residential streets without bicycling features, ranked seventh of the 16 route types in terms of preference. Major city streets with parked cars were the fourth most commonly used route type, but were the least desirable.

Many route types that respondents would be very likely to choose were not routes with high current use, likely because these route types are not widely available in Metro Vancouver. The most striking example is cycle paths next to a major street but separated by a barrier; these were third highest in preference, but the least used, because they are not commonly available in the region. As a route type group, off-street paths were the most likely to be chosen, but current usage was only moderate (ranked 5, 6, and 8 overall).

2.5. Discussion

2.5.1. The near market for cycling This study characterizes the cycling patterns and preferences of the near market for cycling in the Metro Vancouver region. This population, comprised of those individuals who reported having

18 cycled in the past year, or who were willing to cycle more in the future, represents those most likely to make a travel behaviour shift that could increase cycling mode share. In total 31% of those contacted fit into this ―near market‖. Projected to the adult population of the region[75], this represents about 500,000 adults; changing travel patterns in this population could have sizeable health and environmental impacts[77]. While 80% of survey respondents had cycled at least once in the past year, over half of these were only occasional cyclists, further indicating a great potential for change within this group.

2.5.2. Route preferences This survey asked about preferences for 16 different route types, allowing for differentiation between broad classes of routes (off-street, major roads, residential routes, and rural roads) by characteristics of the facility (presence of road markings, signage, car parking, and traffic calming). In general, off-street and separated paths were the most favored route types, followed by residential routes, then major and rural roads. Within each route type, those with more cycling facilities were preferred to those without, and for each case considered, a route type without parking was preferred over one with parking. For off-street paths, paved routes were preferred over unpaved routes, especially among regular cyclists.

These findings add enhanced detail to available evidence on route preferences. A study of current cyclists in Edmonton, Canada (n=1,128) quantified the relative burden of 3 different route types[63] and found that cycling in mixed traffic (i.e., no facilities) was the least preferred: 1 minute in mixed traffic was equivalent to about 4 minutes on an on-street bike lane, or 3 minutes on a multi-use off- street bike path. A study of 167 university employee cyclists and non-cyclists found that they preferred off-road facilities to bike lanes, bike lanes to no bike lanes, and routes without parking to those with[78]. Our results differ from a US survey which found that frequent cyclists were more likely to want on-road bike lanes than off-street paths, whereas infrequent cyclists did not differentiate between route types[70]. The difference may be in the limited options provided to respondents in the US survey.

To increase cycling mode share one must motivate those who cycle least often to cycle more. This study found similar route preferences across frequent, occasional, and potential cyclists, making it straightforward to focus future infrastructure development. The top route types were: paved off- street paths; cycle paths next to major streets separated by a barrier, and residential streets marked as

19 bike routes, with traffic calming. Even among women and respondents with children in their household, subpopulations that were unlikely to choose many route types, these routes were highly ranked. This set of preferred route types would provide a variety of options to cyclists and transportation facility designers. The data on the least preferred route types also provide clear guidance to transportation planners on facility designs that are less likely to be effective additions to the bicycle network.

Some preferences of the regular cyclists differed from those of other cyclist types. Regular cyclists, on average, had a higher preference for nearly all route types. The one exception, unpaved multi-use paths, may be attributable to poor road surface conditions. It is notable that regular cyclists rated all major street route types substantially higher than other cyclists. This concurs with other findings that those with more experience tend to be less averse to cycling in mixed traffic[63]. These regular riders are a specialized group: they are avid cyclists, cycling at least once per week, and tended to be a younger, male population. Given their high preferences for nearly all route types, this group may not require special consideration in cycling network planning – they can be considered the ―first wave‖ of cyclists, those who will cycle regardless of conditions.

2.5.3. Current use patterns versus preferences The most commonly used routes were residential, followed by major streets and off-street paths, then rural roads. This order differs from the findings of Aultmann-Hall et al., who mapped 397 trips by commuters in Guelph, Ontario and found that 60% of total travel distance was on major roads, 35% on local streets, and only 5% on off-road paths or trails[59]. The contrast may be due to differences in the underlying transportation network and the availability of different route types between the two regions. In our study, the route type with the lowest reported current use was the physically separated cycle path next to major route. While the most highly used route types, residential streets and major streets without markings, are widely available, the physically separated cycle paths are very rare route types in Metro Vancouver (<500 m total at the time of this study) and indeed, much of North America. Variations of this type of facility appear in certain Canadian cities (e.g., Montreal) and are widely available in many European centres (e.g., Copenhagen, the Netherlands) where cycling modal shares are much higher.

There was great disparity between the route types that were in high use, and those that were preferred. The most extreme case was the physically separated cycle path next to major streets: it

20 was least commonly used, but just as desirable as unpaved off-street paths, or residential streets with bicycle facilities. This finding highlights one clear way to adapt the current road network so that it is more supportive of cyclists. Cyclists may perceive this route type as a safe way to access the many destinations located on major streets.

2.5.4. Study strengths and limitations Strengths of this survey were that it included 16 different route types and used photos clearly illustrating the infrastructure type. The eventual use of a given route will depend on the subtleties of its design and placement within the road network. The connectivity of routes is key: another study showed commuter cyclists deviate very little from the shortest route between the origin and destination and that for off-street paths, well-connected paths with good surfaces were used significantly more than others[59].

Our survey was conducted in three waves throughout the year to ensure route preferences and reported use were not influenced by the season of questioning. The design of the questionnaire may have resulted in some misclassification of cyclist segment. Individuals who had made zero trips in the four-month survey season were also asked if they had made trips in the past year, and this was used to derive annual trip frequency. However, those individuals who reported at least one bike trip in the survey season (n=712) were not asked about travel the past year, and their annual trip frequency was calculated solely from the number of trips in that season. This may lead to some misclassification of yearly, monthly, or weekly cyclists if this group of respondents cycled differently in the past 4 months than they had in the 8 prior to that.

Finally, this study surveyed the near market for cycling, i.e., the 31% of individuals contacted who were current cyclists or would be willing to cycle in the future, thus findings are not representative of the 69% of the population who did not have a bicycle or were unwilling to ride. As the latter individuals were not currently willing to change their travel behaviour to include cycling, there would be little immediate benefit in targeting interventions to them. We expect with future development of the cycling infrastructure, the climate and culture for cycling across the region will become more inviting, and a greater number of people may become open to cycling.

2.6. Conclusions In summary, these findings show that current and potential cyclists in Metro Vancouver express preferences for routes separated from traffic, in line with cycling infrastructure design in European

21 centres with high cycling mode shares. This survey included two additional pieces that will be presented in future articles: responses to survey questions about the potential influence of 73 potential motivators and deterrents on the decision to cycle or not; and an analysis of trip data reported in the telephone survey with objectively mapped features of the trip environment (population density, land use mix, elevation changes, proximity to bicycling routes, route density, trip distances, etc.) to determine which influence actual choice of cycling. Combined, this subjective and objective data will provide an important body of evidence that is crucial to guide policy decisions on bicycle infrastructure planning. We hope that continued research will build on our work, with the aim of quantifying how changes in cycling infrastructure impact cycling mode share and public health.

Table 2.1 Demographic characteristics of web/mail survey respondents in Metro Vancouver (n=1,402) according to cyclist segment

Cyclist typea Overall potential occasional frequent regular Weighted % of total (n) 13.8 (197) 43.5 (617) 34.6 (481) 8.1 (107) 100 (1402) Gender male 45.1 49.0 57.5 58.3 52.1 A ge 19 -24 6.3 10.4 5.1 12.3 8.2 25-34 19.1 20.1 21.0 21.7 20.4 35-44 31.0 26.7 30.8 22.8 28.4 45-54 22.8 23.1 23.1 21.0 22.9 55-64 12.0 14.6 13.9 14.8 14.0 65 & older 8.8 5.0 5.9 6.5 6.0 E ducation some high school or less 2.1 0.9 0.7 1.0 1.0 graduated high school or less 13.2 11.5 9.6 12.7 11.2 some post-secondary 82.9 86.4 88.5 85.5 86.6 E mployment full time 51.9 54.9 56.2 52.1 54.7 part time 11.2 12.3 11.7 11.5 11.9 self employed 8.7 11.3 11.9 16.5 11.6 student 2.1 5.8 6.1 4.5 5.3 retired 14.4 7.6 9.5 7.3 9.2 not employed 8.8 6.4 3.3 5.3 5.6 H ousehold incomeb <$30,000 10.6 7.1 8.5 14.1 8.6 $30-59,000 17.9 20.1 16.6 23.4 18.9 $60-89,000 21.6 19.5 23.8 12.3 20.7 >$90,000 30.3 32.7 30.0 33.2 31.5

Figure 2.1 Likelihood of choosing route type (if all route types were available) for 16 cycling route types among current and potential cyclists in Metro Vancouver (n=1,402)

Figure 2.2 Mean preference score for 16 route types according to cyclist segment (A) and gender (B)

Figure 2.3 Current use versus likelihood of choosing for 16 cycling route types

3. MOTIVATORS AND DETERRENTS OF BICYCLING 3.1. Synopsis In a survey of 1,402 current and potential cyclists in Metro Vancouver, 73 motivators and deterrents of cycling were evaluated. The top motivators, consistent among regular, frequent, occasional and potential cyclists, were: routes away from traffic noise and pollution; routes with beautiful scenery; and paths separated from traffic. In factor analysis, the 73 survey items were grouped into 15 factors. The following factors had the most influence on likelihood of cycling: safety; ease of cycling; weather conditions; route conditions; and interactions with motor vehicles. These results indicate the importance of the location and design of bicycle routes to promote cycling.

27 3.2. Introduction With the rising pressures of climate change and illnesses related to physical inactivity[79, 80], there is increasing interest in shifting the automobile-dominated culture toward active transportation modes[81, 82]. Incorporating bicycling into daily travel patterns offers benefits for both individual and environmental health, with negligible economic cost[77]. Furthermore, cycling is a feasible transportation mode: over 60% of Canadian adults have a bicycle, and more than 80% live within a reasonable cycling distance (<8 km) of at least one common destination[83]. Still, cycling mode share is low in North American cities, as compared to European centres[23, 84]. Evidence is needed on the types of initiatives that will create a supportive environment for cycling, and induce positive, long-term changes in travel behaviour and physical activity patterns[34].

The friendliness of a city toward cycling is a function of the policies, programs, and facilities in place[85, 86]. To determine which elements might be the most likely to influence cycling, we conducted an survey (―Cycling in Cities‖) of a population-based random sample of current and potential cyclists in Metro Vancouver, Canada. The focus was on those already cycling and those contemplating cycling (the ―near market‖), a strategic approach based on the Theory of Change used in many domains of health promotion[29, 30], and previously used in cycling research[72, 87, 88]. The survey encompassed 73 items, a broad array of potential motivators and deterrents identified from other surveys, qualitative and quantitative research, and input from local policy makers and advocates. The results provide evidence on the relative importance of a wide range of items that could motivate or deter cycling.

3.3. Methods

3.3.1. Survey design and administration The Cycling in Cities Survey was conducted in Metro Vancouver, British Columbia, an urban region of western Canada comprised of 21 municipalities that is home to approximately 2.1 million people[75]. The region has widely varying neighbourhood characteristics and transportation infrastructure, with bicycle mode share estimated at 1.7% region-wide, and at 3.1% within the City of Vancouver[76], a somewhat higher proportion of cyclists than other Canadian cities[56]. The climate is conducive to cycling year-round, with all monthly average low temperatures above freezing, although the region receives substantial rainfall (over 1200 mm rain/year)[89]. The regional bicycle route network has over 1,350 kilometers of designated bicycle routes (Figure 3.1), about 170

28 km of bike routes are off road, and the road network has over 400 km of freeways and highways, 1400 km of arterial roads, 1500 km of collector roads, and 6500 km of local roads.

The survey was funded by the Transport Canada Moving on Sustainable Transport program, TransLink (the regional transit authority) and local municipalities. All study methods were reviewed and approved by the University of British Columbia‘s Behavioural Research Ethics Board. The survey was conducted in three seasonal waves in 2006 (February-April, May-July, September- December). This prevented bias based on the season of the survey, since weather, available daylight, and personal schedules might impact respondents‘ opinions about cycling. Details of the survey development and administration have been published elsewhere[90]. The survey instruments are available from the corresponding author. In brief, there were two questionnaires: a telephone interview, and a self-administered survey (either via the web or mail). The telephone interview filtered participants for eligibility and willingness, and collected demographic information, travel patterns and transport mode. The self-administered follow-up survey asked about use of and preference for bicycle route types (see[90]) and about potential motivators and deterrents of cycling (the subject of this paper). In each survey period, a random sample of 4,000 names, addresses and telephone numbers was selected from the phone directory. Each selected household was mailed an introductory letter outlining the study purpose and the intended use of the information. In the second and third survey periods this was complemented by recruitment through random digit dialing to increase the numbers of participants. Given that the sampling frame was based on the telephone directory, the approximately 10% of the population who have only unlisted cellular telephone numbers would not be captured (representative of Telus, the landline carrier for the province of British Columbia, personal communication, unpublished).

The telephone survey had a contact rate for valid phone numbers of 40.3%. These respondents were screened for age, sex, regional representation and eligibility. To be eligible, respondents were required to be either current or potential cyclists, that is, they had to (1) have access to a bicycle and (2) either have cycled in the past year or be willing to consider cycling in the future. Of those contacted, 31% fit these criteria and were invited to participate in both portions of the survey. In total, 2,149 individuals completed phone interviews, and 65% of those (n=1,402) completed the follow-up questionnaire. The population was categorized according to reported cycling patterns to identify differences in motivators and deterrents based on frequency of cycling. The groups were: potential cyclists, who had not cycled in the previous year, but had access to a bicycle and would consider

29 cycling in the future; occasional cyclists (who cycled at least yearly, i.e., 1-11 trips/year); frequent cyclists (who cycled at least monthly, i.e., 12-51 trips/year); and regular cyclists (who cycled at least once a week, i.e., cycled ≥ 52 trips/year).

The extensive list of potential motivators and deterrents was compiled from transportation and health promotion literature on influences on bicycle use. The list was revised and refined by project partners (TransLink, the regional transportation authority, bicycle coordinators from the participating municipalities, members of cycling advocacy groups) and through pre-testing with focus groups. A total of 73 items were included in the follow-up questionnaire, presented under 13 general categories: vehicles; lane markings; intersections; distances, hills and connections; road surfaces and maintenance; aesthetics and access; coordination with transit; social interactions; safety; weather and darkness; legislation; and information and incentives. The survey questions asked ―how would [item X] influence your decision to cycle?‖ with responses on a 5-category behavioural intent scale: much less likely to cycle (influence score = -1); less likely to cycle (score = -0.5); no influence on decision to cycle (score = 0); more likely to cycle (score = 0.5); much more likely to cycle (score = 1). ‗Don‘t know‘ and refused responses (<4% for any item) were excluded from analyses.

3.3.2. Data summarization and analyses Data from the telephone interviews was entered using a computer-assisted telephone interview (CATI) system. Web-based questionnaires were entered to an electronic database and paper-based surveys were entered to match the format of the web-based questionnaires.

SAS version 9.1 (SAS Institute, Cary, NC) statistical software was used for data management and analysis. Analyses included descriptive statistics (counts or proportions for categorical variables; means, standard errors, ranges, and frequency distributions for continuous variables). Responses were weighted to achieve results representative of the age, gender and geographic distributions of the region. Because of the large sample size, very small differences were statistically significant. Therefore rather than reporting the many statistically significant differences, standard errors are reported, so readers can make comparisons of importance to themselves. The discussion focuses on differences that are meaningful (i.e., differences of at least 5% for demographic characteristics, mean score differences of at least 0.35, consistent trends across cyclist experience categories).

Exploratory factor analysis (EFA) was done to identify underlying constructs among the 73 motivators and deterrents. We employed Proc Factor in SAS with the maximum likelihood option

30 and orthogonal rotation to produce uncorrelated factors. Fifteen factors were extracted based on the scree plot and the Kaiser criterion (all factors with eigenvalues greater than one), and each was assigned a conceptual name based on the items that loaded on it.

3.4. Results Of the 1,402 respondents who completed the follow-up survey, 197 were potential cyclists (weighted % = 13.8%), 617 were occasional cyclists (43.5%), 481 were frequent cyclists (34.6%), and 107 were regular cyclists (8.1%). Table 3.1 describes the demographic characteristics of the population, by cyclist type. Frequent and regular cyclists were more likely to be male (57.5% and 58.3%, respectively), and regular cyclists were less likely to have access to a car than other groups (only 78.4 % had access, compared with 94% or more in other cyclist groups) and had more bicycles and fewer vehicles in their household, on average. As compared to the Metro Vancouver adult population, the ―near market‖ population tended to be somewhat younger (53% of respondents between ages 35 and 54 in comparison with 43% of adults in this age group across the region) and were more likely to have some post-secondary education (87% of the near market versus 67% of overall population).

The mean scores for the 73 survey items ranged from a high of 0.79 to a low of -0.86, on the scale of +1 to -1, where positive scores indicated motivators and negative scores deterrents. Figure 3.2 shows the top 10 motivators and top 10 deterrents overall, as well as the distributions of responses to these items according to cyclist type. The top 10 motivators had mean scores ranging from 0.49 to 0.79. The strongest motivators were routes that were away from traffic, aesthetically pleasing, and easy to cycle. The top ten deterrents had mean scores ranging from -0.57 to -0.86. The strongest deterrents were unsafe surfaces and interactions with motor vehicles.

Mean scores and rankings of the main motivators and deterrents were very similar across cyclist types (Figure 3.2), and indeed across all 73 items (data not shown). Where important differences existed, mean scores were highest for regular cyclists (most easily motivated) and lowest for potential cyclists (least easily motivated), so differences were greatest between these groups. The following five items had the greatest differences by cyclist type. ―The street is wide enough for motorists to safely pass cyclists‖ was perceived as a neutral influence for potential cyclists (mean score = -0.08), but a moderate motivator for regular cyclists (0.34). ―The distance to your destination is 10 to 20 km‖ was a strong deterrent for potential cyclists (-0.50) but had no influence on regular cyclists (- 0.10). ―I need to buy groceries‖ was a moderate deterrent for potential cyclists (-0.33), but no

31 influence for regular cyclists (0.07). ―It is raining‖ was a stronger deterrent for potential cyclists (- 0.74) than regular cyclists (-0.39), while ―The weather is hot & humid‖ was a moderate deterrent to potential cyclists (-0.33), but had no influence for regular cyclists (0.06). These are all safety and comfort related items that would be expected to vary with cycling experience level and equipment.

Table 3.2 summarizes the mean scores and standard errors for all 73 survey items. The standard errors (SE) range from 0.007 to 0.017, so differences in mean scores of more than 0.03 to 0.07 (2*1.96*SE) have confidence limits that do not overlap. Given that mean scores range from -0.86 to 0.79, many are significantly different from each other.

In Table 3.2, the survey items are grouped by the 15 factors identified in factor analysis and ranked according to their strength of influence (either motivation or deterrence). The grouping factors with the strongest potential influence on cycling were safety, ease of cycling, poor route conditions and weather, pleasant route conditions, and interactions with motor vehicles. These five factors included the top 7 motivators and the top 7 deterrents. The following factors were likely to have more moderate influences on likelihood of cycling: route surfaces; integration with transit; carrying loads; bike parking; end of trip facilities; education, information, and incentives; and lane marking and signage. Factors with very little influence on cycling were laws related to cycling, intersections and traffic signals, and the physical challenge of a trip.

3.5. Discussion This survey evaluated the relative importance of 73 potential motivators and deterrents of cycling among current and potential cyclists in the metropolitan area of Vancouver, Canada. This study contributes to the literature in the following ways: first, it captures opinions of a clearly delineated population-based sample, one most relevant to increasing modal share; second, it queries a wide array of characteristics not previously studied and compared in a single survey; and third, it illustrates that there are clear differences in the potential values of various policies, programs, and facilities for influencing cycling.

Transportation planners have categorized four ―E‘s‖ of cycling: engineering, education, encouragement, and enforcement[91]. This framework has been commonly adopted in municipal and regional bicycle planning in Canadian and US cities (e.g., Vancouver, BC, Boulder, CO and Madison, WI). Most of the 15 factors identified in our analysis can be mapped onto one of the four Es. Engineering would encompass ―interactions with motor vehicles‖, ―route surfaces‖, ―lanes

32 markings and signage‖, and ―intersections and traffic signals‖. Encouragement would encompass ―integration with transit‖, ―bike parking‖, and ―end of trip facilities‖. Education would map to the factor ―education, information and incentives‖, and enforcement would map to ―laws related to cycling‖.

However, many of the factors with the most perceived influence in our analysis do not fit well in the 4E scheme. A fifth E – environment – would address the physical, safety and social environments that are often reported to influence physical activity, and cycling specifically[63, 92]. The factors from our analysis that would fit here include ―safety‖, ―ease of cycling‖, ―weather conditions and darkness‖, ―pleasant route conditions‖, ―carrying loads‖, and ―physical challenge of trip‖. Many of these factors would be best addressed in the initial phases of route planning, when selecting the location of a cycling route, for example, locating routes away from noise and air pollution, along minimum grades, and in areas with aesthetically pleasing scenery. The importance of environmental conditions concurs with findings of the National Active Transportation Survey of Canadian adults, where better weather and increased safety in traffic were the 1st and 2nd ranked items (of 14) that would help them cycle more often[83].

Key engineering items were related to bicycle facility location (away from high traffic speeds and volumes), design (i.e., bike paths separated from traffic; bike signage, pavement markings & bike activated signals on residential streets) and maintenance (smooth surfaces). These themes corroborate the results of the other part of our Cycling in Cities Survey on route types[90], where the most preferred route types were off-street paths, cycle tracks along major roads but separated from traffic by a barrier, and designated bike routes on residential streets with traffic calming. The influence of traffic was also apparent in a recent stated preference survey on route choice trade-offs of Texan cyclists, where motorized traffic volumes and speeds were the most influential variables, when compared to on-street parking, bicycle route characteristics, and physical characteristics (hills and stops)[93]. Route features were also mentioned in the National Active Transportation Survey[83], where better designed on-road routes was the 4th ranked item, continuous routes to key locations was the 5th ranked item, and maintenance was the 7th ranked item. Our findings add to previous research by providing much more detail on the subtleties of route design. While particular engineering items were very strong motivators (―the route has bicycle paths separated from traffic for the entire distance‖, mean score = 0.69), others had little influence (―the bike lane has a solid white line painted between moving cars and the bike lane‖, mean score = 0.16).

33 The Cycling in Cities Survey had the benefit of a large sample size representative of the regional population of both current and potential cyclists. Sampling techniques for surveys have often focused on specialized populations such as students[74], on recruitment from bicycle list-serves or clubs[50, 52, 94], or on regular cyclists (by stopping commuters or tagging locked bicycles)[73, 95]. A focus on students or those who cycle most often may not capture the motivators and deterrents important to women, adults with children, and older men, all of which are demographics underrepresented among North American cyclists[23, 84, 96]. This survey excluded individuals without access to a bicycle, or who were not willing to cycle (69% of those contacted) but was representative of the remaining 31% of the population. The sampling approach was population-based, but focused resources on understanding the opinions of those currently willing to change travel behaviours; the ―contemplating‖ population, or those with the greatest potential to increase mode share in the short to medium term. We included both current cyclists and potential cyclists as the ―near-market‖ for cycling. Individuals in the regular cyclist group are avid commuters, cycling at least once per week. However, the other cyclist groups show great potential to increase their cycling patterns: frequent cyclists, cycling only monthly, and occasional cyclists, cycling only yearly, could certainly be encouraged to make more bike trips given the right conditions; and potential cyclists, who currently don‘t cycle but are willing to cycle more in the future, represent the greatest latent demand for cycling.

The reported influence of the 73 items included in the survey was similar across the regular, frequent, occasional, and potential cyclist groups. Any differences in influence scores varied most between the two extremes of cyclist type – the potential and the regular cyclists – but still the difference was rarely more than 0.2 on the scale of -1 to +1. Others have reported that cyclists are a non-homogenous group in terms of route choices and preferences. In an analysis of actual route choices, Aultman and colleague‘s[59] suggested that there were two types of cyclists, those who are not averse to cycling on direct routes that may be major roads with traffic, and those who are willing to travel longer distances for better route options. In our population, long distances, proximity to motor vehicles, hot or rainy weather, and needing to shop had less deterrent effect for regular cyclists, than for the other groups. This corresponded to differences in route type preferences observed in the other part of this survey[90]: regular cyclists scored routes along major roads higher than potential cyclists did (though both groups ranked paved off-street paths for cyclists only as their top route choice). The fact that there were so few differences between regular and potential

34 cyclists is encouraging for planning: improvements related to the strongest motivators and deterrents have the potential to increase cycling of the whole near market population, not only a select subset.

3.5.1. Survey strengths and limitations This survey of motivators and deterrents of cycling included a list of 73 potentially influential items developed based on previous literature and expert opinion. This broad range of items complements results from stated preference surveys on cycling, which quantify trade-offs between options (i.e., willingness to travel longer for a more desirable route type) but are limited in the number of comparisons they can make[45, 58, 93]. By capturing opinions on each item individually, components can be combined to optimize design. For example, off-street paths provide the desired separation from traffic, noise, and air pollution, but other strongly influential items should also be considered in the design; ideally such paths should also be flat, lit, and provide direct routes between common destinations. Still, some items were excluded during survey development in the interest of minimizing respondent burden. In many cases, these were subtle design features of interest to transportation engineers, but difficult to describe in a self-administered survey. In a survey of such scope, using pictures or videos of environmental or infrastructural items, or detailed descriptions for programs or incentives was not feasible. The questionnaire was pretested with cyclists, advocates, and the general public to ensure items were clearly written and understood, but respondents‘ responses were still subject to their own interpretation of each item.

We used mean scores to summarize study findings, based on a scale from +1 to -1, where positive scores were motivators and negative scores were deterrents. Mean scores averaged the ordinal scale responses (much more likely to cycle, more likely to cycle, neutral, less likely to cycle, much less likely to cycle). This approach assumes that the difference between any two adjacent categories is identical when in fact it may not be. In addition, a summary mean does not provide a picture of the underlying variability in responses. In the extreme case, a neutral mean score could result either if all individuals had a neutral response, or alternatively if half the population felt it was strong deterrent and the other half a strong motivator. Such extremes of opinion were not observed in this survey; only 1 item (―the street is wide enough for motorists to safely pass cyclists‖) had a bi-modal distribution.

This study was conducted in a single geographical location, but one that is comparable to many other cities in the western world. The Metro Vancouver region has a temperate climate that is similar

35 to many European and North American centres influenced by coastal weather patterns. Its population density (~1000/people per km2 in the region as a whole and ~5000/km2 in the City itself) spans the range of most North American and European cities with at least 100,000 population[75, 97].

Finally, these findings are based on behavioural intent, not actual behaviour. In other concurrent manuscripts, we have used GIS to conduct analyses of the actual travel behaviours made by this same study population, to determine how urban form affects mode choice between cycling and driving[98], and also how it affects route choice for cyclists or drivers[99]. Such analyses, as well as those underway elsewhere[100] will provide other sources of evidence about the types of infrastructure and environments that influence choice of cycling. Outcomes of these travel behaviour studies can be used to test whether opinions are translated into actions, though they cannot provide relative rankings of the broad scope of items included in this survey.

3.6. Conclusions We quantified the relative impact of a wide-ranging set of items that could influence cycling behaviour. The top motivators were: routes away from traffic noise and pollution; routes with beautiful scenery; and routes separated from traffic. The top deterrents were: ice and snow; streets with a lot of traffic; streets with glass/debris; streets with high speed traffic; and risk from motorists. These findings can direct transportation planners and public health officials to select urban planning and maintenance items that are likely to increase cycling rates. Certain items reported to be strongly influential, such as flat terrain, darkness, or poor weather, may not seem modifiable except via an individual‘s responsibility to have appropriate gear and a minimum fitness level. However, transportation engineers can make decisions that will mediate these for the entire population. Examples are to locate bicycle routes to minimize gradients, ensure that bicycle facilities are lit, and ensure that they are plowed and salted in winter weather.

While a comprehensive, multifaceted approach to bicycle promotion is likely to prove most effective, this study provides evidence on where to focus efforts to maximize changes in cycling mode share. Factors related to the built environment for cycling (separation from motor vehicles, ease of cycling, and pleasant route conditions) were the strongest perceived motivators and therefore most likely to make a dramatic impact on mode share. These factors could be addressed by building direct bicycle routes that are physically separated from motorized traffic and minimize gradients.

36 Such designs may appear to be costly and reduce road space for motor vehicles, but one way to accomplish them with minimal impact on the road network is to develop cycle paths along old rail corridors. Such policies have been widely used in some North American cities but have not yet been fully exploited in many regions, including Metro Vancouver. Another approach that has met with wide success in Portland, Oregon is ―bike boulevards‖, or bicycle routes along traffic-calmed neighbourhood streets. Detailed design guidelines are now available for implementation in other cities[101]. In Vancouver these constitute a significant portion of the bicycle network and are referred to as ―neighbourhood bikeways‖; they are a highly ranked infrastructure by all types of cyclists[90]. Locally, the results of this survey have been used by Vancouver city planners to gain council approval for a $25-million plan for cycling including network improvements such as separated bicycle paths along arterial streets[102, 103] and to guide the development of an online bicycle route planner[104].

In a handful of cities where the built environment for cycling is already of high quality (i.e., Davis, California or Victoria, British Columbia), strategies related to education and encouragement may be effective to further increase already high bicycle mode shares. However the results of this study are likely to be informative in the majority of North American cities, where cycling modal share is low and the survey items reported to be stronger motivators and deterrents have not yet been adequately addressed.

Our findings support a shift in policy approach for the practice of bicycle infrastructure planning. In many jurisdictions, new route selection has been based largely on expediency; typically cycling routes are implemented on sections of roadway where adequate space already exists, where traffic volumes are low, or where there are developments in progress that allow for expansion of the road right of way. This paper supports a more rigorous approach to route selection that prioritizes safety, location, and environmental factors. Based on this research, key considerations for new routes should include the potential to physically separate bicycles from motor vehicles, to minimize slopes, to travel through aesthetically pleasing locations, and to serve popular destinations, including rapid transit lines. The engineering design of the facility should follow after the selection of route location, and should focus on construction of off-street paths, neighbourhood bikeways, and other designs allowing separation from traffic. Route lighting and smooth surfaces are important attendant features that should be included in both design and maintenance policies. This systematic approach

37 would use evidence of the relative preferences of all subgroups of the ―near market‖ cycling population, and should therefore serve to attract the highest number of users.

38 Table 3.1 Demographic characteristics of respondents who completed the web/mail survey (n= 1,402) according to cyclist typea

Cyclist typea Overallb potential occasional frequent regular Weighted % of total (n) 13.8 (197) 43.5 (617) 34.6 (481) 8.1 (107) 100 (1402) Gender male 45.1 49.0 57.5 58.3 52.1 A ge 19 -24 6.3 10.4 5.1 12.3 8.2 25-34 19.1 20.1 21.0 21.7 20.4 35-44 31.0 26.7 30.8 22.8 28.4 45-54 22.8 23.1 23.1 21.0 22.9 55-64 12.0 14.6 13.9 14.8 14.0 65 & older 8.8 5.0 5.9 6.5 6.0 E ducation some high school or less 2.1 0.9 0.7 1.0 1.0 graduated high school or less 13.2 11.5 9.6 12.7 11.2 some post-secondary 82.9 86.4 88.5 85.5 86.6 E mployment full time 51.9 54.9 56.2 52.1 54.7 part time 11.2 12.3 11.7 11.5 11.9 self employed 8.7 11.3 11.9 16.5 11.6 student 2.1 5.8 6.1 4.5 5.3 retired 14.4 7.6 9.5 7.3 9.2 not employed 8.8 6.4 3.3 5.3 5.6 H ousehold incomec <$30,000 10.6 7.1 8.5 14.1 8.6 $30-59,000 17.9 20.1 16.6 23.4 18.9 $60-89,000 21.6 19.5 23.8 12.3 20.7 >$90,000 30.3 32.7 30.0 33.2 31.5

M ean # of children/ household 0.9 0.9 0.7 0.5 0.8 Access to car (yes) 96.7 95.4 94.0 78.4 93.7 Mean # of motor vehicles/ household 1.8 1.9 1.8 1.4 1.8 Mean # of bicycles/ household 2.4 2.9 3.0 3.6 2.9 a potential cyclists= never in past year; yearly cyclists= 1-11 one-way trips/year; monthly cyclists=12-51 one-way trips per year; yearly cyclists=52 or more one-way trips per year b the ―near market‖ population was somewhat younger than the general population (across the region 53% of adults are between 35 and 54) and more educated (67% of the population have some post-secondary education) (Statistics Canada, 2006). c 20.3% of responses to household income were "refused/don't know"; all other variables had <2% "refused/don't know/other" responses

39 Table 3.2 Means and standard errors for each survey item, and results of factor analysis. Items grouped by factor, and ordered from highest to lowest strength of motivation or deterrence (i.e., absolute value of mean influence on likelihood of cycling). Items and factors with absolute values of mean scores ≥ 0.5 in bold.

Factor Analysis Survey Items and Influence Scores Influence on Likelihood of Cycling Loading Factors Meana for (SEb) for Meanc for on Factor Items (individual motivators and deterrents) (eigenvalue) Question Question Factor Safety 0.75 The risk from motorists who don't know how to drive safely near bicycles -0.73 (0.010) -0.58 (2.76) 0.75 The risk of injury from car-bike collisions -0.67 (0.011) 0.64 The risk of bicycle theft -0.56 (0.010) 0.68 The risk of violent crime when cycling -0.55 (0.012) 0.53 The risk from cyclists who don't know how to ride safely -0.37 (0.011) Ease of Cycling 0.46 The route is flat 0.61 (0.010) 0.56 (1.85) 0.48 Cycling to the destination takes less time than traveling by other modes 0.59 (0.011) 0.65 The distance to your destination is less than 5 km 0.53 (0.012) 0.24 I can make the trip in daylight hours 0.50 (0.013) Poor Weather 0.47 The route is icy or snowy -0.86 (0.007) -0.56 and Darkness 0.57 It is raining -0.63 (0.010) (1.41) 0.37 The route is not well lit after dark -0.59 (0.011) 0.35 The weather is hot & humid -0.16 (0.013) Pleasant Route 0.70 The route is away from traffic noise & air pollution 0.79 (0.009) 0.48 Conditions 0.77 The route has beautiful scenery 0.70 (0.010) (1.97) 0.26 The route is wide enough for cyclists to ride side-by-side 0.40 (0.011) 0.21 There are shops, banks, & grocery stores along the routed 0.34 (0.013) 0.23 I am making the trip with other people 0.18 (0.014) Interactions with 0.69 The street has a lot of car, bus, & truck traffic -0.83 (0.009) -0.48 Motor Vehicles 0.67 Vehicles drive faster than 50 km/hr -0.76 (0.010) (2.43) 0.56 The street has on-street parking -0.43 (0.012) 0.48 The street is wide enough for motorists to safely pass cyclists 0.10 (0.017)

40 Factor Analysis, cont’d Survey Items and Influence Scores Influence on Likelihood of Cycling Loading Factors Meana for (SEb) for Meanc for on Factor Items (individual motivators and deterrents) (eigenvalue) Question Question Factor Route Surfaces 0.56 The route has glass or debris -0.76 (0.008) -0.43 (2.92) 0.61 The route has surfaces that can be slick when wet or icy when cold -0.59 (0.009) 0.72 The route has potholes or uneven paving -0.55 (0.010) 0.23 There are bridges along the route where cyclists must share a narrow sidewalk -0.34 (0.012) 0.67 The route has lots of fallen leaves -0.29 (0.010) 0.57 The route surface is gravel or dirt -0.25 (0.012) 0.47 The route has speed bumps -0.25 (0.011) Integration with 0.74 You can take your bike on the SkyTrain at any time 0.50 (0.012) 0.42 Transit 0.84 The bus has racks that carry bikes 0.45 (0.011) (3.45) 0.77 There are secure bike lockers at transit stations 0.41 (0.011)

0.68 There are bike racks at transit stations 0.30 (0.011) Carrying Loads 0.79 I need to carry bulky or heavy items -0.57 (0.013) -0.40 (1.51) 0.73 I need to buy groceries -0.23 (0.014) Bike Parking 0.63 The destination has covered bike racks, to protect from rain 0.47 (0.011) 0.39 (1.19) 0.59 The destination has outdoor bike racks 0.42 (0.011) 0.31 The destination has rental bike lockers* 0.27 (0.011) End of Trip 0.48 The destination has secure indoor bike storage 0.49 (0.011) 0.37 Facilities 0.91 The destination has a place to store a change of clothing 0.38 (0.011) (3.34) 0.84 The destination has a place to dry your cycling gear 0.36 (0.011) 0.87 The destination has showers 0.34 (0.011) 0.41 The destination has bike repair facilities 0.28 (0.010)

41 Factor Analysis, cont’d Survey Items and Influence Scores Influence on Likelihood of Cycling Loading Factors Meana for (SEb) for Meanc for on Factor Items (individual motivators and deterrents) (eigenvalue) Question Question Factor Education, 0.41 Information about cycling routes to the destination is available 0.46 (0.010) 0.36 Information, 0.41 A web-based trip-planning tool is available 0.45 (0.010) Incentives 0.57 I would be eligible to receive prizes or discounts such as savings on bike gear 0.35 (0.011) (2.61) 0.80 Inexpensive or free short courses are available to help me learn how to fix my bike 0.30 (0.010) 0.77 Inexpensive or free short courses are available to help me improve my cycling skills 0.24 (0.010) Lane Marking, 0.36 The route has bicycle paths separated from traffic for the entire distance 0.69 (0.011) 0.34 Signage 0.59 A 2-way off-street bike path has a reflective centre line for night & poor weather cycling 0.49 (0.014) (6.46) 0.58 The route has bike signage, pavement markings & bike activated signals on residential streets 0.47 (0.012) 0.80 There is a consistent type of bike lane marking throughout the greater Vancouver area 0.41 (0.014) 0.54 The route has on-road bicycle lanes on major roads for the entire distance 0.36 (0.014) 0.51 Traffic calming on designated bike routes reduces the number of cars using the route 0.36 (0.013) 0.81 The bike lane has a different colour pavement than the road 0.35 (0.014) 0.77 A bicycle is stenciled every 75 meters (250 feet) along the route 0.28 (0.013) A solid white line is painted on both sides of the lane separating it from moving cars & from 0.78 0.26 (0.015) parked cars 0.73 The bike lane has one solid white line painted between moving cars & the bike lane 0.16 (0.014) 0.50 Bike lane markings end just before intersections -0.13 (0.012) Physical Challenge 0.37 The route has long steep sections -0.50 (0.011) -0.18 of Trip 0.71 The distance to your destination is 10 to 20 km -0.37 (0.014) (1.49) 0.71 The distance to your destination is 5 to 10 km 0.14 (0.015) 0.25 The route has a few small hills 0.02 (0.011)

42 Factor Analysis, cont’d Survey Items and Influence Scores Influence on Likelihood of Cycling Loading Factors Meana for (SEb) for Meanc for on Factor Items (individual motivators and deterrents) (eigenvalue) Question Question Factor Intersections, 0.49 Cyclists have to stop at many stop signs on the route -0.37 (0.010) -0.10 traffic signals 0.33 Designated bike routes on residential streets are used by cars because there are fewer stop signs -0.31 (0.013) (1.72) 0.51 The route has push-button-activated traffic signals for cyclists & pedestrians only 0.30 (0.013) 0.40 The route has rail crossings -0.13 (0.009) 0.44 Many intersections on the route have traffic circles -0.12 (0.013) 0.53 The route has regular traffic signals for all traffic (cyclists, pedestrians, cars & trucks) 0.01 (0.012) Laws Related to 0.60 Cycling on sidewalks is not allowed -0.22 (0.012) 0.00 Cycling 0.53 Cycling helmets are required 0.14 (0.013) (1.65) 0.55 Lights are required for cycling after dark 0.13 (0.013) 0.65 Cycling side-by-side on roads is not allowed -0.05 (0.010) a weighted mean score, where +1= much more likely to cycle, +0.5= more likely to cycle, 0=neutral, -0.5= less likely to cycle, and -1= much less likely to cycle b standard error of the mean c mean of weighted mean scores for each item in the factor d item could have been loaded on another factor: integration with transit

43 Figure 3.1 Bicycle route and road network of Metro Vancouver.

44 Figure 3.2 Top 10 motivating and top 10 deterring influences on cycling, of 73 items asked of survey respondents (n=1,402)a. Mean scores overall and as reported by each cyclist typeb

a less than 4% missing responses for any item b potential cyclists= never in past year; occasional cyclists= 1-11 one-way trips/year; frequent cyclists=12-51 one-way trips per year; regular cyclists=52 or more one-way trips per year

45 4. BUILT ENVIRONMENT INFLUENCES ON ROUTE SELECTION

4.1. Synopsis Current travel demand models are calibrated for motorized transportation, and perform less well for non-motorized modes. Little evidence exists on how much, and for what reasons, the routes people travel deviate from shortest-path or least-cost routes generated by transportation models. We investigated differences in total distance, road type used, and built environment features for shortest- path routes versus actual routes for utilitarian bicycle trips (n=50) and car trips (n=67) in Metro Vancouver. Bike trips were, on average, 360 m longer than the shortest possible route; car trips were 540 m longer. Regardless of mode, people do not detour far off the shortest distance route: detour ratios (actual distance/shortest distance) were similar, with ¾ of trips within 10% of the shortest path route distance, and at least 90% within 25%. Differences in the built environment measures en-route suggest why bike commuters chose to detour: the actual routes had significantly more bicycle facilities (traffic calming features, bike stencils, and signage) than did the shortest path routes. Compared to shortest-path routes, cyclists spent significantly less of their travel distance along arterial roads, and significantly more along local roads, off-street paths, and routes with bike facilities. As expected, car trips were more likely to be along highways and less likely to use local roads than predicted by the shortest distance route. Our results illustrate factors that might be included in travel models to more accurately model non-motorized transportation, and provide guidance for how dense bike facilities need to be when designing infrastructure to support cycling.

46 4.2. Introduction Progressive planning for cycling requires an understanding of where cyclists travel and what affects their decisions to cycle versus other transportation modes. One approach is to employ travel demand management models. Automobile travel-based models, such as the gravity-style EMME model (INRO, Montreal, QC), use an iterative process to predict overall travel patterns between regional sub- areas based on socioeconomic data, transportation networks, known travel patterns, and travel preferences. These models are used to evaluate the short-term impacts of infrastructure changes, to identify long-term scenarios based on land use and demographic changes, and to assess the impact of policy changes such as road pricing. They require substantial real world information for calibration, meaning that model validity is dependent upon the accuracy of input data. For most motorized vehicle models data collection is abundant and robust due to the level of investment and political pressure on motor vehicle infrastructure projects.

However, methods for modeling non-motorized travel remain underdeveloped[64]. Due to the limited availability of data for non-motorized vehicles, the errors resulting from the small numbers of non- motorized trips in travel survey data, and the lack of detailed transportation networks for non- motorized travel, it is difficult to calibrate transportation models to accurately reflect the travel patterns of cyclists. Using the idealized car route - along fast roads - as an approximation for bicycle travel is likely to produce a sub-optimal route prediction for the majority of cycle trips. Using cycling routes/paths alone to model bicycle travel is not reliable due to their relative scarcity in many urban areas. Given the slower travel speeds and greater energy requirements of cycling, an alternative option would be to assume the shortest distance path between origins and destination. However this suggests that the only factor influencing route choices is distance.

The purpose of this study was to identify factors in travel models that might need to be altered when considering cyclist travel and to make comparisons to motorist travel. We investigated differences in total distance, road type used, and built environment features along the shortest distance routes and the actual routes for car and bicycle trips in an urban area with extensive bicycle facilities, the Metro Vancouver region. The study captures travel behaviours in a population-based sample of cyclists, including potential, infrequent, and regular cyclists, a population purposefully selected based on Stages of Change Theory from health promotion to target those most willing to make behaviour changes[30, 105]. Understanding route choices for car trips and bicycle trips can provide insight into built

47 environment features and facilities that influence travel, and further improve modeling for non- motorized travel.

4.3. Methods

4.3.1. Setting The Metro Vancouver region in British Columbia, Canada, comprises 21 municipalities, covers over 2800 km2 and has a population of 2.6 million. The median work commute is 7.4 km[22], suggesting that that cycling is a viable mode for daily travel. The region has a mild climate, facilitating cycling year round. Despite this, cycling mode share for work trips is only 3.7% within the city itself, and 1.7% for the census metropolitan area[24].

The region has over 1300 km of ―designated‖ bike routes, about 170 km of which are off-road. The on-road bike routes may be along arterials, collectors or local roads, and may have a variety of enhanced facilities for cycling including: striped bike lanes; road stencils or signage to remind cars of the presence of bicycles; a variety of traffic calming features (e.g., traffic circles, median barriers to slow or block motorized traffic); and/or intersections with bike-activated crossing signals. Off-road routes may be paved or unpaved, and be bicycle only or shared with pedestrians.

4.3.2. Interview data collection As part of an ongoing research project on cycling behaviours, a survey[90] was conducted in 2006 that collected origins and destinations for over 4,000 common non-recreational trips made by 2,149 current and potential cyclists from throughout Metro Vancouver. For this study we took a random sample of 150 of the 599 participants who had indicated willingness to be contacted for future research. Those sampled were sent a letter outlining the study purpose and subsequently telephoned by the study coordinator. At least five contact attempts were made, at varying times of day, before a subject was considered a non-respondent. In the phone interview participants were reminded of their origin, destination, mode and purpose for their reported trip(s) (up to 2 per person), and asked to report a typical route for the trip(s) (Table 4.1). All study procedures were approved by the UBC Behavioural Ethics Board.

4.3.3. GIS and statistical analysis The actual travel routes reported during the interviews were digitized using ArcGIS 9.2 based on a Digital Road Atlas (DRA) centerline street network datafile enhanced with off-street cycling paths in

48 the region. Where the respondent used a commuter boat (Seabus), an alley, or an unmapped path, new line segments were created and assigned appropriate road types. Shortest routes were generated based on Dijkstra‘s algorithm using FME software (Safe Software, Surrey, Canada). The network dataset for shortest route modeling was identical to that used for actual routes and weighting based on distance, with no impedance applied for speed limits and also no consideration of restricted turn directions. This methodology generates the shortest distance paths; these do not necessarily correspond with the fastest routes, particularly for car trips. However, using a common methodology for both car and bicycle trips – comparing shortest distance routes to actual travel paths – enables direct comparison of route choice by mode. Figure 4.1 illustrates routes for a participant who reported a both a bike and a car trip for the same work commute, alongside the modeled shortest distance route.

Geographical shape files and raster files were acquired from national, institutional, and academic sources. Where possible, we aimed to build commonly used built environment measures (e.g.,[38], measures listed in Table 4.3) to facilitate comparisons across studies, while recognizing theoretical differences that may exist between cycling and other forms of physical activity, and for route choice decisions versus mode choice decisions.

To derive built environment characteristics we applied a crow-fly buffer of 250 m to routes. Gross population density was based on 2006 Census data for dissemination areas[106]. Where route buffers intersected dissemination area boundaries the population was apportioned according to the area included. The percentage of land area with green cover (defined as street trees, park park/forest trees and grasslands) was calculated from a raster file where the predominant land cover from Landsat data was assigned to each 5 x 5 m grid cell using a classification and regression tree[107]. Traffic-related pollution, based on average nitrogen dioxide concentration, was from a land use regression model[108]. Topography was measured in two ways, both relying on a Digital Elevation Model raster file. ―Hilliness‖ was captured by taking the standard deviation of the elevation for the set of 30 x 30 m grid points within the buffer zone. For ―steepness‖ we used a publicly available ―RunningSlope‖ script to split each route into 100 m segments and calculate the slope of each segment. We created a dichotomous variable with a value of 1 if the slope was greater than 10% and summarized this across each route using the Dissolve tool in ArcGIS, to give the proportion of route segments with slope > 10% for a given route. As there is not a particular slope at which a road is considered unbikeable, we investigated using cutoff values of 5% and 8% and found these were highly correlated (Pearson‘s

49 correlation coefficients > 0.70) with the 10% slope cutoff variable. For bike facilities, Hawth‘s tools ―Count Points in Polygon‖ tool was used to sum the following from detailed regional shapefiles of bicycle facilities: the number of traffic calming features (including traffic circles, median barriers); bicycle stencils; bike route signs; and traffic crossings with bike-activated signals. Street connectivity was based on the DRA centerline file enhanced with the bicycle network for off-street paths. The number of connecting roads at each intersection was calculated using an ESRI script (JPtoolsFnode_Tonode) and cleaned to address issues at intersections with boulevard roads, which are considered as multiple intersections by the script (see [109], for extensive discussion). The connectivity measure was the ratio of 4-way intersections to all intersections (4 way, 3 way, or cul-de- sac). Values closer to 1 suggest higher connectivity, likely to be better for cycling. For land use, parcel data from BC Assessment[110] was aggregated to 9 land uses (agriculture, commercial, education, entertainment, industrial, office, park, single family residence and multifamily residence). Specific uses were measured as the total parcel area with land use [X]/ total land area in the buffer. Land use mix was measured with an entropy measure (Shannon Index)[66] calculated as: Land use mix = - ∑k[(pi) ln(pi)]/ln(k); where pi is the proportion of each of 4 land uses (residential, commercial, entertainment, and office), and k is the number of those land uses present. This widely used metric captures the overall evenness of distribution of key land uses[38], but does not address the spatial grain of mix, or whether the mix is complementary from a travel perspective[111, 112].

In the road class analysis, the road type variable in the DRA was aggregated to 5 categories: highway, arterial, collector, local, or trail. Off-street paths (for bicycles, alone or shared) were added from the regional bicycle network shapefile. Road segments that were ―designated routes‖ were also identified using the regional bicycle network shapefile, and could occur along any given roadway type.

Using SAS Version 9.1 (Cary, NC) we analyzed differences between shortest routes and actual routes, separately for bike trips and for car trips. Differences in trip distance, built environment metrics and the total distance spent on different road types were analyzed using paired t-tests for normally distributed variables or the sign-rank test for highly skewed variables. To evaluate the simultaneous influence of multiple characteristics on route choice, we created a logistic model for the likelihood of a bicycle trip detouring more than 10% in distance as compared to the shortest-distance route, given the differences in built environment characteristics between the actual route and the shortest route, as well as the safety perceptions of the individual. Built environment variables offered to the model were those associated with detouring in bivariate models with significance levels of p<0.20. In two cases

50 where built environment variables were highly correlated (Pearson correlation coefficient > 0.7), the one more strongly associated with the model was retained (i.e., % green cover instead of % of land in park use; number of traffic calming features instead of number of bicycle route signs). The parsimonious model included all variables associated with detouring at p<0.10, a significance level selected based on the small sample size in this exploratory analysis. The safety variable was based on the most influential safety-related survey question: ―How does the risk of injury from car-bike collisions influence your decision to cycle?‖. The response categories were scaled from -1 (much more likely to cycle); -0.5 (more likely to cycle); 0 (no influence); 0.5 (less likely to cycle); to 1 (much less likely to cycle).

4.4. Results Of the 150 survey participants mailed letters, 41 were ineligible (returned mail or phone number out of service). During the 2-week study period in August 2008 interviews with 74 participants were completed, for a 67.9% response rate. Eleven (10.1%) individuals were confirmed to be on holiday, 1 (0.9%) refused, and no response was obtained for the remaining 23 (21.1%) individuals.

Demographic characteristics of participants are in Table 4.2. A range of different types of cyclists were included: 6 were regular cyclists, cycling at least 1/week; and 10 had not cycled in the past year. Data for a total of 132 routes were collected: 67 car trips, 50 bike trips, 1 bike/bus trip, 8 transit trips and 6 walking trips. The primary purpose was travel to work (n=41 trips, 31.0%), then shopping (n=36, 27.3%), social purposes (n=32, 24.2%), personal business (n=13, 9.9%), and school (n=10, 7.6%). All subsequent analyses were focused on car and bike trips, as the sample sizes for transit and walking trips were too small. In the final dataset, 15 individuals reported 2 bike trips each, 5 reported 1 bike trip, 21 reported 2 car trips, 10 reported 1 car trip, and 15 reported one trip each by car and by bike. Potential cyclists (n=10) contributed 13 of the trips, all car trips. The other cyclist groups made trips by both bicycle and car: those who reported cycling at least once per year contributed 37 trips to the dataset, 9 of which (24%) were by bicycle; those who cycle at least one per month contributed 56 trips, 32 (57%) by bicycle; and the regular weekly cyclists contributed 11 trips, 9 (82%) by bicycle.

4.4.1. Trip distance Trip distance statistics are summarized in Table 4.3. For both car and bike trips, the actual routes were significantly longer than the shortest routes. For car trips, the route actually taken was on average 0.54 km longer than the modeled shortest route (95% CI: 0.29-0.79). For bike trips, the mean

51 difference was 0.36 km (95% CI: 0.14-0.58). With the overlapping confidence intervals, it appears that people traveling by bike or by car detour similar distances: regardless of mode, about 39% of the actual trips were within 250 m of the shortest trip length. The ratio of the actual trip distance to shortest trip distance (the ―detour factor‖) was calculated, with histograms shown in Figure 4.2. For bike trips 75% were within 10% of the shortest distance (detour factor < 1.1), and 90% were within 25%. For car trips, 76% of trips were within 10% of the shortest distance and 90% were within 23%.

Plots of the percentage detoured versus actual distance did not suggest a relationship between detour proportion and trip length, regardless of mode (data not shown).

4.4.2. Built environment measures To understand why people choose to travel a route that was longer than the shortest path, we looked at how built environment measures differed between the shortest routes and those actually travelled. These differences may indicate more favorable conditions for a given transport mode. Results of paired t-test analyses (Table 4.4) indicate that bike trips were significantly more likely to be routed in places with enhanced bicycle facilities including traffic calming, stencils and signage.

On average, the actual routes had 2 more traffic calming features, 10 more stencils and 7 more signs than the predicted shortest routes. Other features such as hills, air pollution, greenness, street connectivity, or land use were not significantly different between the shortest and actual routes, although for several factors the differences in the mean values did fit with a priori hypothesis (actual routes had lower air pollution, a higher number of cyclist-activated traffic crossings, and less hilliness than did the shortest routes). The built environment did not have an impact on routing for car trips, as the only significant difference in the paired analysis of actual routes and shortest routes was entertainment land use, but this was not meaningfully different (only 0.14% of land area versus 0.11%).

In the logistic regression (Table 4.5) examining the likelihood of bicycle route detours greater than 10% of the shortest-distance route, both built environment variables and the individual‘s perception of safety were influential. The results indicate that cyclists are likely to detour longer in order to travel on routes with more green cover, and routes where more bicycle activated crossing signals are available. Individuals who reported that the risk of injury from car-bike crashes was a deterrent to cycling were more likely to detour. This model provides some preliminary insight on the multiple influences on route choice, but given the small sample size her replication of this work is needed in the future.

52 4.4.3. Road class analysis Figure 4.1, showing the route choices of one participant along his work commute, illustrates that different road types are used depending on the mode. The bicycle trip involved a detour to travel on a designated bike route, whereas the car trip tracked closely with the shortest distance route, mainly along arterial and collector roads. The road class analysis (Table 4.6) clearly highlights the appeal of bike facilities for cyclists. Whereas the shortest route models for bike trips predicted that 21% of the route would be along ―designated‖ bike facilities (1005 m of 4885 m mean trip distance), for the actual trips the mean distance travelled along a bike facility was 2587 m, or 49% of the total trip distance (mean 5335 m).

Figure 4.3 shows the differences in road class choices by travel mode, across all trips. For bike trips, the most dramatic difference was in the preference for routing along designated bicycle routes, where the mean increase in distance travelled along bike routes between an actual route and the corresponding shortest route was 1583 m (95% CI: 871- 2295). Additionally, actual bike trips had significantly less distance along arterial roads than predicted by shortest route models (mean difference between pairs of actual and shortest routes = -712 m, 95% CI: -1184 to -240) and significantly more along local streets (mean difference = 570 m, 95% CI: 47 - 1092) and off-street paths (mean difference = 525 m, 95% CI: 206- 843). For car trips, the actual trips had significantly more distance along highways (mean difference = 512 m (95% CI: 111- 913), and less distance on local roads (mean difference= -564 m (95% CI: -858 to -271) than the modeled shortest routes.

Many respondents also provided explanations for their route choices. Table 4.7 provides a sample of reasons for the routes selected on bicycle trips. The list highlights the trade-offs between taking the shortest route and detouring to ride on designated bicycle infrastructure, and underscores the value of safe (away from traffic, lit) and comfortable (shade, aesthetics, road surface, topography) routes.

4.5. Discussion This study of travel behaviours in Metro Vancouver found real differences between the shortest distance routes connecting origins and destinations and the routes people actually chose to travel. While the detour ratios were similar for car and bike trips, what motivated people to take longer routes differed based on transport mode. To our knowledge, this is the first study to assess route selection using a broad set of built environment metrics, for both car and bicycle travel.

53 This is the only study of cyclist travel patterns with a representative sample (based on a random sample from a population-based survey) of both regular and non-regular cyclists who together comprise the ―near market‖ for cycling, that is, adults with access to a bicycle and who cycled in the past year or would be willing to cycle in the future. Based on the Stages of Change model from health promotion[29, 105] the near market is expected to be the easiest target for increasing cycling modal share, but they are not a homogenous group. Only 6 of the 74 people in this study could be considered regular cyclists, cycling at least once per week. Instead, the majority of trips were made by those who bicycle only once per month (n=32), or once per year (n=26). In the Cycling in Cities survey population there were real differences in route preferences and usage patterns between regular and non-regular cyclists[90], with non-regular cyclists less willing to cycle along major roads. These attitudinal and behavioural differences underscore the value in capturing the travel patterns of cyclists with different experience levels.

We found that three-quarters of trips were within 10% of the shortest route distance, regardless of whether they were made by bike or car. This corresponds with an average of 360 m detour on a bike trip, and 540 m on a car trip. A few other studies have reported on cycling detour rates, in select populations at diverse geographical locations where travel conditions, routing options, and the cyclist population (recreational versus utilitarian) may differ substantially. An intercept study of 754 bicycle commuters in Japan found that detours ranged from 6-16% depending on the city, direction, and trip purpose, with longer detours for shopping trips as compared to work or school commuters[60]. A study of 397 work/school bicycle commuters in Guelph, Ontario found a mean trip distance of 3.7 km with an average detour from shortest path of 0.4 km (11% detour)[59]. In Phoenix, 150 regular commuters (cycling > 3 days/week) reported a longer mean trip distance (6.6 km), but still a 10% increase on the mean shortest route[113]. A different approach was used in a study examining the drawing power of an off-street trail in Minnesota[114]. This intercept study found that cyclists traveled an average of 67% out of their way to use the trail. This much longer detour could result from a greater willingness to detour for recreational trips, the quality of the facility, a lack of acceptable road facilities in the area, the influence of a few trips with very long detours, or could be a result of the study sample: cyclists not willing to make a detour, either long or short, would not be found on the trail and not be included in the survey.

A major contribution of this study is the use of built environment and road network measures to examine why people take longer trips than necessary. For bike trips, people spent less time on

54 arterials, and more time on local roads and off-street paths than was predicted based on shortest routes, clearly demonstrating a preference for lower volume roads and separation from traffic. This analysis of real travel behaviour supports reported opinions on route types of Metro Vancouver cyclists, where off-street paths were preferred to residential roads, which in turn were preferred to major streets, rural roads and highways, a pattern consistent across demographics and cyclist experience level[90]. The Guelph study reported that cyclists spent relatively more of their trip distance along arterial roads, and less on off-street paths[59]. These differences likely reflect differences in the design and connectivity of the bicycle and road network (Guelph had no on-street network at the time of the study), and the availability and convenience of off-street paths in the two locales. Designated cyclist facilities are clearly the incentive for cyclists to make a trip longer than necessary. Instead of minimizing the total trip distance and being on bike routes for one-fifth of the distance on average, cyclists detoured and used bike routes for nearly half of their trip distance. The built environment analysis reflects this: actual routes had more traffic calming, bike stencils and signage than shortest paths. Similarly, a study from Portland tracked 166 regular cyclists using GPS technology and found that 49% of their total trip distance was along routes with bike facilities, most commonly secondary roads with bike lanes and bicycle/multi-use paths[57]. The Phoenix survey also evaluated road type use and found selective routing: 51% of the actual route segments were along bicycle facilities, compared with 39% of the shortest route segments[113]. These collective results, based on real travel behaviours, complement the findings of attitudinal and stated preference surveys that cyclists prefer to ride on roads with bike facilities[45, 63, 78, 115]. This body of evidence, both from stated opinions and now from actual travel choices, responds to the controversy in bicycle planning in recent decades in North America, on the need for dedicated cycling facilities (as compared to ―vehicular cycling‖) and the types that are most supportive for cycling (the traffic calmed or separated facilities prevalent in Europe, or on road facilities)[86, 116]. The literature indicates that bicycle facilities, especially those away from traffic, are perceived as safer routes by cyclists, although certainly this depends on the specific design and implementation of the facilities[117]. In the interviews for our study, participants mentioned low traffic volumes, safety, and appealing scenery as reasons for their route choices.

Other built environment features hypothesized to influence cycling, such as hills, air pollution, greenery, population density, land use mix, and street connectivity, did not differ significantly between actual and shortest routes in this analysis. Several reasons may explain this. First, operationalizing

55 built environment measures is challenging[118]. For example, very few studies have used an objective measure for hills, and no standard exists[38]. McGinn et al. created a dichotomous variable for ―steep hill‖ if any street segment had a slope greater than 8%[119] but this did not predict physical activity outcomes. Troped et al. created a variable for ―steep hill barrier‖ if any road segment along the shortest route to a trail had a slope of at least 10%[95], and the absence of a steep hill was associated with a higher likelihood of trail use (OR=1.84). Both of these studies assessed the presence of a single short steep hill, but did not differentiate between this and many steep hills, or a generally hilly topography, both of which may influence cycling. We created two measures to capture this construct: the standard deviation of elevation within 250 m of the route to capture ―hilliness‖; and the percentage of street segments along the route with slopes exceeding 10% to capture ―steepness‖. Both measures had face validity in our dataset, and notably the car trips did have higher values for both hilliness and steepness than did the bike trips. It may be that the set of route options within a reasonable detour distance from the shortest route did not differ substantially in terms of topography, greenery, air pollution, or intersection density. Variability in these characteristics may only be achieved across a wider geographical scale, so that they may be more influential on choice of travel mode than on route choice within a given cycling trip.

Conventional travel demand management characterizes regional travel patterns based on origin- destination data and uses models to assign trip routing, but such models do not reflect non-motorized travel well. We collected data about actual travel routes to understand route selection, and included both bike and car trips to contrast choices made in non-motorized and motorized travel. Applying the same analysis method for both car and bike trips illustrates not only differences in route choices by mode, but also highlights how actual travel patterns fit with travel models for each mode. For car travel our findings support the parameters applied in travel models: car trips deviate from shortest- distance routes to travel along highways, presumably reflecting faster travel paths. While this is a common assumption in transportation practice, a search the scientific literature found little evidence supporting it, and virtually none considering the influence of the built environment on route (as compared to mode) choices (TRIS online searches using keywords ―built environment‖ and ―route choice‖ resulted in 6 hits, none on car travel; using ―urban form‖ and ―route choice‖ resulted in 23 hits, with a focus on vehicle miles traveled or travel demand instead of routing). Furthermore, this study found that cyclists also detour, but in contrast, they do this directionally toward routes with bicycle facilities, along local roads, and avoid arterials. Improved travel demand models might

56 incorporate impedances such that modeled bicycle trips are routed preferentially along bicycle facilities or, where bicycle networks do not exist, off major roads. In doing this, planners must also recognize that the distance cyclists will detour is limited: we found bicycle trips were 360 m longer, on average, than the shortest route. Such distance constraints have implications on the required density of a bicycle network, if a region is to be supportive of bicycle travel.

A challenge in cycling research is the study population: existing research focuses on the regular work commuter. We sampled different types of cyclists, from daily cyclists to those who cycle only rarely, and as such our results are applicable to those who represent the latent demand for cycling. Additionally, by examining the link between non-motorized transport and the built environment, this study builds methodologically on physical activity research that has measured origin and destination characteristics[61] but not routes. It also incorporates the transportation research approach by including road class traits[57, 59]. Together these indicate that for cyclists, using a shortest route algorithm between trip origins and destinations, or the route assignment from travel demand models, will not reflect the importance of cycling facilities.

There are limitations to this research. It reflects travel patterns in a single geographic region; albeit one that includes urban and suburban areas. In addition, the sample size was small. Despite these issues, this research adds to the few prior studies on route selection, a reflection of the tremendous effort required to collect actual travel routes. The Cycling in Cities survey interviewed over 2,100 individuals on multiple aspects of their travel behaviour: travel patterns, mode share, opinions on infrastructure, motivators and deterrents, and safety perceptions. Given the broad scope of this effort, it was not feasible to survey participants about specific routes. Thus the current study collected additional data on exact routes for a subset of individuals. Several researchers have also embarked on this challenge, including a workplace-focused study in London that analyzed 46 trips[120]; a study of work/school commuters that surveyed 397 cyclists using paper-based maps[59]; and a recent study which capitalized on GPS technology to track 166 cyclists[57]. These provide some evidence on detour rates and on road class usage. However, none of these studies has examined the influence of a comprehensive set of built environment factors on route choice, nor have they compared their results to car route choices. We hope this study encourages others to gather similar built environment data to examine the influence of urban form on route choices. Future studies, in geographically disparate areas, will test the generalizability of our findings.

57 4.6. Conclusions The results of this analysis indicate that road infrastructure and bicycle-specific aspects of the built environment influence people‘s travel patterns; that car drivers detour from shortest routes to fast roads, and that cyclists deviate from shortest routes to routes with better bicycle facilities. Cyclists are heterogeneous population and not all will make the same route choices, but in this study that includes regular as well as infrequent cyclists, and both work and non-work trips, the findings clearly indicate the importance of facilities. These results highlight factors that should be included in models to more accurately reflect bicycle travel, and provide guidance about how dense a network of good bike facilities needs to be, if the goal is to attract more people to use bicycles for daily travel.

Table 4.1 Interview guide (abbreviated)

In the Cycling in Cities survey you reported (1 or 2) commonly made trips. Today we‘re interested in getting specific information on a typical route that you may have taken for those trips. In the letter we included a Metro Vancouver Cycling map. Please feel welcome to get that out if you would like to use it as a visual aid during this interview.

Example, Trip 1: The first trip you reported started at (origin location), at (postal code) and ended at (destination location). This was a trip made for (trip purpose). You reported that you typically made this trip by (trip mode).

Can you please walk me through a typical route that you may have taken between this origin and destination? Starting at the origin, can you please detail the street directions that you would have taken? If at any time you were traveling on off-street paths, or along lanes, please let me know.

Are there any reasons you selected this route?

59 Table 4.2 Demographic characteristics of the study population (n=74 regular, infrequent and potential cyclists)

n percent frequency of cycling at least 1/week 6 8.1% at least 1/month 32 43.2% at least 1/year 26 35.1% not in the past year 10 13.5% gender male 41 55.4% female 33 44.6% age 25-34 7 9.5% 35-44 25 33.8% 45-54 20 27.0% 55-64 18 24.3% >65 4 5.4% education high school or less 7 9.5% college/technical 19 25.7% some university 4 5.4% graduated university 42 56.8% refused/ don‘t know/ 2 2.7% employmentother employed 57 77.0% student 1 1.4% retired 11 14.9% unemployed 4 5.4% other 1 1.4% household income <$30,000 8 10.8% $30-59,000 13 17.6% $60-89,000 21 28.4% >$90,000 25 33.8% refused/ don‘t know 7 9.5% /other

60 Table 4.3 Distances of shortest and actual routes (km)

mean median mean std dev range difference 95% CI p-value mode: car (n=67) shortest route 4.55 6.93 6.46 0.27 - 32.7 0.54 0.29 - 0.79 <0.001 actual route 5.18 7.42 6.81 0.25 - 34.3 mode: bike (n=50) shortest route 3.38 4.89 4.32 0.29 - 17.8 0.36 0.14 - 0.58 0.017 actual route 3.65 5.25 4.52 0.34 - 18.3

61 Table 4.4 Comparison of built environment characteristicsa for each pair of actual and shortest routes

Bike Trips (n=50) Car Trips(n=67) Actual Shortest Difference Actual Shortest Difference Variable Mean Mean Mean Mean Mean Mean gross population density (per ha) 0.35 0.36 -0.01 0.23 0.23 0 % of land area with green cover 31.6 32.5 -0.92 30.4 30.5 -0.04 average air pollution (ppb NO2) 27.7 28.2 -0.56 29.4 29.1 0.35 variation in elevation 13 13.4 -0.34 18.4 17.5 0.92 % of road segments >10% slope 1.1 1.0 0 1.6 1.4 0 # traffic calming features 6 4 1.96 b 2.8 2.4 0.42 # stencils 37.2 27.6 9.66 b 21.6 24.4 -2.78 # bike route signs 25.9 18.6 7.30 b 18.1 17.8 0.28 # traffic crossings with bike activated signals 4.5 3.8 0.76 3.2 2.9 0.33 ratio of 4 way intersections: all intersections 0.42 0.41 0.01 0.35 0.35 0 % of land area with use: agriculture 4.1 4.6 -0.53 3.1 3.1 0.01 commercial 3.4 3.3 0.05 3.8 3.6 0.16 education 3.1 2.9 0.17 2.1 2.1 0.07 entertainment 0.15 0.16 -0.02 0.14 0.11 0.03 b industrial 3.6 3.3 0.32 4 3.9 0.09 office 1.5 1.3 0.17 1.4 1.4 -0.03 park 8.6 8.6 0.01 8.4 7.7 0.63 single family residence 37 36.8 0.12 37.1 38.4 -1.31 multifamily residence 2.2 2.3 -0.11 2.4 2.3 0.07 land use mix 0.27 0.26 0.01 0.29 0.28 0.01 a within a 250 m buffer of the route b p<0.05 in paired t-tests of whether the mean difference between actual route and shortest route is different than 0

62 Table 4.5 Multiple logistic regression of likelihood of detouring > 10% beyond shortest- distance route for bike trips (n=50), based on differences in the built environment features between the actual and shortest-distance routes, and reported safety perceptions

95% Odds Confidence Ratio* Interval Built Environment Compared to the shortest distance route, the actual route has….. - 2% more green cover 2.6 1.0 - 7.2 - 1 more bicycle-activated crossing signal 1.5 1.1 - 2.1 Safety Perception The risk of injury from car-bike crashes deters cycling by 0.5 (one response category) 8 1.5 - > 999 * relative odds of detouring for the specified change in the independent variable

63 Table 4.6 Comparison of distance traveled on each road class, by mode of travel

Bike Trips (n=50) Car Trips (n=67) Actual route Shortest route Actual route Shortest route Road type mean % total mean % total mean % total mean % total Total trip distance (m) 5335 100% 4885 100% 7496 100% 6931 100% highway 0 0% 38 1% 1439 19% 927 13% arterial 1628 31% 2340 48% 3532 47% 3147 45% collector 1140 21% 1111 23% 1806 24% 1595 23% local 1767 33% 1197 25% 580 8% 1144 16% off street 723 14% 198 4% 139 2% 120 2% Any designated bike routea 2587 49% 1005 21% 1096 15% 1111 16% a designated routes may be along highways, arterials, collectors, locals or off street paths.

64 Table 4.7 Examples of respondent’s reasons for route choices on bicycle trips

-Always goes along bicycle routes -Doesn't mind going extra distance to stay on bike routes, especially aesthetically pleasing ones -Selects downtown road with bike lane instead of more direct route without bike lane -Takes a longer route to avoid a dangerous on-ramp -Selects (off-street) route along dyke because there is no traffic -Selects routes through alley instead of busy arterial -Does not take shortest route, but safest; rides through regional park in daytime, but not at night -Takes route because there is not a lot of traffic, and good shade -Avoids climbing steep hills; turns to avoid hills, or narrow or rough roads -Changes route often to get favorable hills, and where there are fewer cars -Selects route to have less traffic, better scenery, and avoid hills -Takes a variety of routes to keep it interesting, along any of the residential streets -Rides different routes depending on whether trip is made fast (arterial) or safe (along local road) -Takes (unpaved) route through the park on the way home, when it is fine to get dirty

65 Figure 4.1 Bike, car and shortest distance routes for participant who commutes using both modes

66 Figure 4.2 The detour factor (Ratio of distance for actual trip: shortest trip) by modea

0.8 0.9 1

1.1 1.2 1.3

% of % trips

detour factor

a A detour factor of 0.84 (i.e., actual trip was only 84% of the distance of the computer-generated shortest distance path) occurred in the case of one bicycle trip where a commuter ferry was used instead of a bridge crossing. Other trips with detour factors < 1 typically reflect small differences in distance due to informal paths used by cyclists, or a driveway/parking lot cut-through by cars.

67 Figure 4.3 Differences in road class usage between actual routes and shortest routes, by mode of travel

2000 * 1500 bike trips car trips 1000

* *

500 *

Shortest route -

0 Actual route Actual -500

Mean difference in distance travelled (m) (m) travelled distance in difference Mean * * -1000 highway arterial collector local offstreet on any designated bike route Road type

*p<0.05 in paired t-tests of whether mean difference in distance travelled on road class X between a given actual and shortest route pair is different than 0, for each road class.

68 5. BUILT ENVIRONMENT INFLUENCES ON MODE CHOICE 5.1. Synopsis A growing body of evidence links the built environment to physical activity levels, health outcomes, and transportation behaviours. However, little of this research has focused on cycling, a sustainable transportation option with great potential for growth in North America. This study examines associations between decisions to bicycle (versus drive) and the built environment, with explicit consideration of three different spatial zones that may be relevant in travel behaviour: trip origins, trip destinations, and along the route between. We analyzed 3280 utilitarian bicycle and car trips in Metro Vancouver, Canada made by 1902 adults, including both current and potential cyclists. Objective measures were developed for built environment characteristics related to the physical environment, land use patterns, the road network, and bicycle-specific facilities. Multi-level logistic regression was used to model the likelihood that a trip was made by bicycle, adjusting for trip distance and personal demographics. Separate models were constructed for each spatial zone, and a global model examined the relative influence of the three zones. In total, 31% (1023/3280) of trips were made by bicycle. Increased odds of bicycling were associated with: less hilliness; higher intersection density; less highways and arterials; presence of bicycle signage, traffic calming and cyclist-activated traffic lights; more neighbourhood commercial, educational, and industrial land uses; greater land use mix; and higher population density. Different factors were important within each spatial zone. Overall, the characteristics of routes were more influential than origin or destination characteristics. These findings indicate that the built environment has a significant influence on healthy travel decisions, and spatial context is important. Future research should explicitly consider relevant spatial zones when investigating the relationship between physical activity and urban form.

69 5.2. Background Increasing active transportation is a promising approach to counteract issues at the forefront of both public health and transportation: the obesity and inactivity epidemics; growing congestion; and air and noise pollution[36, 39]. Many cross-sectional studies have shown that supportive built environments are associated with increased walking and overall physical activity, reduced vehicle miles traveled, and improved health outcomes[40, 121, 122]. Although few studies to date have focused on cycling, the mode warrants more attention. Since bicycle travel is 3-4 times faster than walking, it may be the better substitute for driving for short-to-medium trip distances. Making some typical utilitarian trips by bicycle, most days of the week, would fulfill the recommended physical activity guidelines for health[5].

Results from ecological (aggregate) studies[47, 48], opinion surveys[90], stated preference surveys[63, 94], and focus groups[117] all provide evidence that built environment factors influence cycling. However, few individual-level (disaggregate) studies of travel behaviour have explicitly looked at cycling outcomes, and they have found null results for built environment variables after accounting for individual characteristics[62, 123]. These inconsistencies may stem from methodological issues in travel behaviour studies that employ GIS-based measures to examine the effect of the built environment on cycling. Specifying ―place‖, or selecting the appropriate spatial zones for analysis, is a challenge in this research area. In individual-level studies a common approach is to examine how the characteristics in the area of one‘s residence correlate with activity levels. Typically the area is identified by a 1 mile or kilometer buffer around the home postal code, representing the walkable distance from home. Some cycling research has extended the 1 mile distance to 3 or even 5 miles, in recognition that the activity space should be larger to encompass bikeable distances[61, 62]. However, it is now widely recognized that these home-based areas do not accurately represent an individual‘s activity-space, or the built environment they are influenced by, since physical activity can also occur at other locations such as the workplace, parks, or a gym. Furthermore, the extent of one‘s activity-space varies by demographic characteristics[124]. Some research has tried to address this by defining physical activity outcomes by purpose (transportation, leisure) or type (walking, vigorous activity)[123, 125]. An emergent method, though technologically intensive, is to employ GPS to accurately determine where people travel, and where they engage in physical activity (i.e.,[57]).

To clarify the relationship between cycling and the built environment, methodological refinements tailored to cycling are needed. Factors such as the local availability of sidewalks or land use mix may

70 be primary motivators of walking trips, but decisions on whether to cycle may be influenced by a different suite of factors across spatial areas beyond the trip origin. For example, in a survey querying 73 factors, the top four motivators for making a trip by bicycle were related to routes: being away from traffic and noise pollution; having beautiful scenery; having separated bicycle paths for the entire distance; and having flat topography[115]. The geographic accessibility of destinations (i.e., schools, employment sites, retail) may also affect the likelihood of making trips by bicycle and since two-thirds of cycling trips are under 5 km, and 90% are less than 10 km[43], short trip distances are important.

In this study, we investigated the effects of the built environment on healthy transportation mode choices (bicycle vs. car) for trips made by 1,902 current and potential cyclists in Metro Vancouver. We addressed the issue of specificity of place by characterizing the built environment at origins, destinations, and along routes for utilitarian trips (to work or school, or for personal business or social reasons). We hypothesized that within each of the three spatial zones, different built environment features would influence decisions to travel by bicycle instead of by car.

5.3. Methods

5.3.1. Trip data Travel data came from the Cycling in Cities survey, a population-based survey of 2,149 current and potential cyclists conducted in 2006 across the Metro Vancouver region. Details of the survey are published elsewhere[90, 115]. To be eligible, respondents had to be in the ―near market‖ for cycling, defined as having access to a bicycle, and having cycled in the past year (current cyclists) or being willing to cycle in the future (potential cyclists). All study procedures were approved by the University of British Columbia Behavioural Ethics Board (Application # H05-80976).

In a telephone survey, participants were asked about destination, mode, and trip purpose for two common non-recreational trips. The two trips queried were selected by the interviewer based on reported travel patterns, using the following hierarchy: 1) the most frequent non-recreational bicycle trip (if any); 2) any other non-recreational bicycle trip (if any); 3) the most frequent non-recreational trip by any other mode; 4) the second most frequent non-recreational trip by any other mode. Data were collected for 4,260 trips. Origin and destination locations were provided by 6-digit postal code, specific address, or intersection. Geocoding (98% success rate) resulted in 3,897 trips with complete data within Metro Vancouver. As the focus was on decisions to travel by bicycle instead of car, we

71 excluded trips made by transit (n=328), walking (n=260), or other modes (n=29). The analysis dataset consisted of 3,280 trips made by 1,902 individuals.

We generated shortest distance routes connecting each origin and destination pair using FME (Safe Software, Surrey, Canada) and Dijkstra‘s algorithm with weighting based on distance only[126]. The road network dataset for creating shortest routes was the Road Atlas (DRA) centerline street network datafile[127] enhanced with off-street cycling paths in the region.

5.3.2. Demographic variables Demographic variables collected in the survey are summarized in Table 5.1. Values were imputed for variables with missing data (response = ―don‘t know/refused‖). For ordinal and nominal variables the most common response category was imputed, corresponding to: age category of 35-44 (5 missing); education level of graduated university (7 missing); and household income of 60-89K (648 missing). The mean observed value (=3) was imputed for 10 records missing the number of bicycles in the household.

Respondents were also categorized according to how often they cycled, based on their derived annual trip frequencies: regular cyclists (cycled at least weekly, i.e., ≥ 52 trips/year); frequent cyclists (cycled at least monthly, i.e., 12-51 trips/year); or rare cyclists (cycled < 12 times in the past year).

5.3.3. Spatial analysis zones For each trip we created spatial analysis zones in ArcGIS (ESRI, Redlands, CA) using buffers for routes, origins, and destinations (Figure 5.1). To create route zones, we applied a simple buffer of 250 m to each shortest route polyline. In preliminary work we evaluated the effect of different buffer sizes (100 m, 250 m, and 500 m) and found built environment measures were highly correlated across these sizes (Pearson correlation > 0.85). The final choice of 250 m was meant to maximize the variability in measures (using smaller versus larger buffers) while recognizing the imprecision of the routes (shortest routes instead of actual routes). This buffer typically included adjacent streets on each side of the shortest route and thus allowed for a set of plausible alternative routes.

To create the origin and destination buffers, we used methods developed by Oliver et al. for pedestrian travel[128] that produce irregularly-shaped buffers based on accessibility defined by the transportation network. The methodology addresses limitations of simple Euclidean buffers, which may not best represent the area experienced by and accessible to users (e.g., land area not adjacent to

72 roads or cycling paths). We used Network Analyst in ArcGIS to identify all line segments within 450 m along the street network and applied a 50 m buffer to this set. We also evaluated 250 m and 1000 m distances, and found that their built environment measures were correlated with the 500 m measures (Pearson correlation > 0.62).

5.3.4. Built environment measures Spatial data sources included the census, academic research projects, the property tax assessment authority, and the regional transit authority. Where possible we sought data from 2006 to match temporally with the trip survey; in practice the data ranged from 2003 (air pollution model) to 2006 (bicycle facility data).

The selection of measures was guided by literature on the built environment and physical activity (e.g., see [38, 109]) or cycling[48, 57, 90, 115, 117]. The measures fell into four general categories: the physical environment; land use; the road network; and bicycle facilities. For each of the built environment variables described below, we generated a priori hypotheses on their direction of influence on cycling, and in which spatial zones (origin, destination, route) they might be most influential (Table 5.2).

Physical environment Greenery. The percentage of land area with green cover (defined as street trees, park/forest trees and grasslands) was calculated from a 5 x 5 m raster file where the predominant land cover was assigned based on Landsat data using a classification and regression tree[107].

Air pollution. Traffic-related pollution, measured as the average nitrogen dioxide concentration (in ppb), was based on a land use regression model at 10 m resolution[108].

Topography. Two measures were created to capture different aspects of topography, both derived from a Digital Elevation Model raster file (30 m resolution). ―Hilliness‖ was measured as the standard deviation of the elevation for the grid points within each buffer zone. ―Steepness‖ was measured for routes only as it was a polyline based method. The ArcGIS RunningSlope script was used to split each route polyline into 100 m segments and output the slope for each segment. We then calculated the percentage of route segments with slope > 5% along a given route. This cutoff slope was selected based on the Transportation Association of Canada guidelines for bicycle route design[129].

73 Land use Population density. Gross population density was measured as the total census population based on dissemination area data from 2006 Census[106], divided by the total land area in the buffer, excluding water bodies. Where buffer boundaries intersected dissemination area boundaries, the population was apportioned according to the area included.

Specific land uses. Property tax assessment data from BC Assessment[110] includes 203 land use categories. These were aggregated to 8 land use types (commercial, education, entertainment, industrial, office, park, single family residence, and multifamily residence). Land uses not hypothesized to influence cycling were excluded (e.g., agricultural, vacant, transport/utility). The land use measure used was the percentage of total land area with land use X, equal to the sum of the area of all parcels with land use X divided by the total land area in the buffer. After preliminary analyses we also reclassified commercial land use parcels according to the lot size, using a threshold of 1 hectare to differentiate between large commercial and neighbourhood commercial parcels.

Land use mix. Land use mix was calculated with an entropy measure (Shannon Index)[66]: - ∑k(pi) ln(pi)/ln(k); where pi is the proportion of each of 4 land use types (in this case: residential, commercial, entertainment, and office) and k the number of different land uses included. This widely used measure captures the overall evenness of distribution of key land uses[38] but does not address the spatial grain of land use mix, or whether the mix is complementary from a travel perspective[111, 112].

Road network Road types. Road type measures were based on the Digital Road Atlas centerline file[127]. We aggregated road class to 4 categories (highway, arterial, collector, local) and excluded non-accessible road types (private, trail, etc.). The kilometers of each road type in a buffer zone were calculated using Hawth‘s tools ―Sum line lengths in polygons‖ and divided by the total lane km of roads in the zone to give the percentage of road network of type X.

Connectivity. Intersection density was calculated as the number of intersections with 3 or more connecting road segments divided by the area of the zone in hectares. An ESRI script (JPtoolsFnode_Tonode) was used to determine the number of connecting roads, and the output cleaned to correct intersections with boulevard roads, which would otherwise be considered as multiple intersections[109]. This measure has been widely used in walking and cycling research and is correlated with other connectivity measures[130].

74 Bicycle facilities Bicycle routes. The Metro Vancouver region has about 1,400 km of designated cycling routes that are along major roads, local roads, or trails with bicycle-specific facilities; over 350 km of these are off- road paths. For the cycling route analysis, the Digital Road Atlas was merged with a digital file of designated cycling routes to produce a single transportation network file that included line segments for off-street cycling paths. The calculation of the percentage of bicycle routes within the analysis zones was analogous to the road type analysis described above, using only line segments coded as cycling routes in the numerator. To account for bi-directional route types, lane-kilometers reported for bicycle facilities. A lane-kilometer counts facilities on each side of the road. For example, one kilometer of a two-way bikeway would count as two lane-kilometers.

Bicycle facilities. Bicycle facilities were measured using Hawth‘s tools ―Count Points in Polygon‖ tool to sum the following features from detailed shapefiles of facilities: the number of traffic calming features (e.g., roundabouts, barriers that divert traffic); bicycle road markings or bicycle route signs; and traffic crossings with cyclist-activated traffic lights.

5.3.5. Statistical Analysis Descriptive summaries of built environment measures within each of the zones were calculated. Histograms were used to identify highly skewed variables or bimodal distributions. We examined correlations between built environment measures within each zone (route, origin, destination) and across zones to ensure that co-linear variables were not included in the multivariable models. The only measures with Pearson correlation exceeding 0.7 were: land use mix and % land in office use within the route zone; land use mix and % land in commercial in the origin zone; and % arterial roads and % local roads in the destination zone.

Because we had clustered data with up to 2 trips per person, we conducted multi-level logistic modeling using PROC GLIMMIX in SAS Version 9.1 (Cary, NC) to model the outcome that a trip was made by bicycle versus by car. All built environment characteristics were first tested in bivariate models for their association with cycling or driving. We then ran three separate multivariable models, one for each zone (route, origin, destination). We also created a fourth multivariable model to examine the a priori hypotheses (Table 5.2): this final model included measures from every zone. In multivariable modeling, variables that were significant at p<0.10 in bivariate regressions and in the direction of the a priori hypotheses were offered to the models. The parsimonious models included all

75 factors significant at p<0.05. Demographic variables for gender, age, education, and income were added as a block to the parsimonious models to assess if these modified the effect of the built environment on the odds of cycling. Trip distance was included in all models. For continuous variables (e.g., population density, intersection density) the odds ratios are expressed as the change in odds for a change in the interquartile range of the built environment measure. This allows comparison of effect sizes across measures with different variability. For dichotomous variables (e.g., presence of traffic calming features) the odds ratio represents the change in the odds of cycling if the feature is present within the buffer.

5.4. Results Table 5.2 summarizes the demographic characteristics of trip takers. The population was highly educated (48.6% with university degrees) and came from households with relatively high incomes (41.0% over $90,000 annually). Nearly all (96.5%) had access to a car. The population was comprised of 7.7% regular cyclists, who cycled at least once a week. The majority of participants (59.8%) were rare cyclists, cycling less than once a month.

Thirty-one percent of the trips were made by bicycle (1023 of 3280). For bicycle trips, the most common purpose was to work or school (31%), then social reasons (28%), shopping (22%) and personal business (19%). Proportionally more car trips were for work or school (43%) and shopping (23%), and proportionally less for social reasons (21%) or personal business (13%). Table 5.3 summarizes the trip characteristics by mode of travel. Bicycle trips were typically shorter than car trips (mean distances 4.7 km versus 10.2 km, respectively), though the longest trips for each mode were similar (bike = 52.7 km, car = 56.1 km). Table 5.3 also provides the dimensions of the three spatial zones. The area included within the route zones was highly variable due to the broad range of trip distances. The land area also differed within the origin and destination zones, for two reasons: we used network-based origin and destination buffers instead of airline buffers, and we excluded water areas. The differences in spatial extent between zones and within zones required that all built environment metrics be normalized (e.g., percentages, densities) as opposed raw measures (counts, kms of bike routes).

Built environment factors differed between route, origin and destination zones. Comprehensive tables are available from the authors, but we include key examples here. Destination zones had the least green cover (mean = 19.1%) as compared to origin zones (23.7%) and route zones (30.1%). It is

76 intuitive that destination zones (at workplaces, shopping places) may have less greenery than origin zones (residential areas), but these results also indicate that the travel corridors had more green cover than either. Origin zones, not surprisingly, had a higher percentage of land used for single family residences (48.6%) than route zones (34.7%) or destination zones (27.2%). Designated bicycle routes covered, on average, 10.8% of the road network in origin zones, but 12.9% in route zones, and 14.3% in destination zones. These numbers highlight that the route, origin and destination zones represented different types of built environments, and underscores the importance of considering each of these spatial zones, as opposed to only the trip origin.

5.4.1. Bivariate comparison of built environment characteristics of bicycle versus car trips Table 5.4 contrasts the built environment characteristics of the route zones for car trips and bicycle trips. Based on t-tests of the mean values, most factors differed between car and bicycle trips, and in the expected direction (e.g., bicycle trips were routed through areas that were less hilly, had more grid- based road networks, and had lower density of highways). However, two factors trended in the direction opposite to a priori hypotheses. One was green cover: route zones for bicycle trips had less green cover than those for car trips (25.0% average green cover versus 32.4%). Perhaps the unexpected result reflects areas where cyclists were limited to travel in order to make short trips or to use direct routes. The other factor was the average air pollution, which was higher in bicycle trip route zones than in car trip route zones (31.2 ppb NO2 versus 29.4 ppb), although a mean difference of 1.6 ppb is unlikely to be meaningful. Given that these two built environment factors did not conform to a priori hypotheses, and therefore may be surrogates for other built environment features, they were not considered in multivariable models.

In preliminary analyses, the association between commercial land use and the likelihood of cycling was also counter to a priori hypothesis. Bivariate models suggested people were less likely to bicycle to a destination that had a high percentage of commercial land, even though these are likely destinations for utilitarian travel. We refined the commercial land use variable, to separate large ―big box‖ or shopping mall retail from smaller, neighbourhood retail, using a threshold of 1 hectare lot size. With this differentiation, higher neighbourhood commercial land use at the destination was associated with a higher likelihood of making a trip by bicycle (unadjusted OR= 1.15, 95% CI: 1.01-1.31) and more large retail land use with a lower likelihood of cycling (unadjusted OR= 0.81, 95% CI: 0.75-0.87) (Figure 5.2).

77 5.4.2. Multivariable models for built environment characteristics of bicycle versus car trips Parsimonious models for each of the three zones, adjusted for trip distance and demographic variables, are presented in Table 5.5. Trip distance was significant in all three models, with shorter trip distances strongly associated with higher odds of cycling. Demographic factors also played an important role in mode choice. In all models, women were less likely to cycle than men, with their odds of cycling around 0.6. Younger people were more likely to cycle than older people, with those in the 19-24 age group having 5 times higher odds of making a trip by bicycle than those 65 and older. Those with higher education were more likely to cycle than those without, while those from households with lower incomes were more likely to cycle than those from higher income households. The odds ratios for the demographic variables changed little between models.

Built environment factors influenced the likelihood of cycling, even after controlling for trip distance and demographic factors. Within the route zone (Model 1), measures significantly associated with a higher likelihood of cycling were: less hilliness; higher intersection density; a lower percentage of the road network categorized as highway or arterial roads; higher population density; a lower percentage of the land used for single family residential, or large commercial use. One of these same variables was significant in the origin zone (Model 2): higher intersection density. Additionally, the presence of traffic calming features, a higher percentage of land in industrial uses, and a higher land use mix were associated with a higher likelihood of cycling. In the destination zone (Model 3), land uses were also important: a higher percentage of educational or neighbourhood commercial land uses, and less large commercial use, were associated with a higher likelihood of cycling. Other significant factors in the destination zone were higher population density, less of the road network as arterials, and the presence of road markings or signage along bicycle routes.

The cross-zonal model (Model 4) offered variables according to the a priori hypotheses (from Table 5.2) about the zone in which they might influence cycling (Table 5.6). Of four factors offered from the origin zone, none were significant, and of six offered from the destination zone, only educational land use was important. The greatest number of factors hypothesized and offered (12) was from the route zone. Of these, less hilliness, higher intersection density, and a lower density of arterial roads were associated with higher odds of cycling.

78 While avid, experienced cyclists may have no issue making long trips, trip distance was a significant influence on cycling in our models. Indeed, distances under 5 km have been found to be a strong motivator for cycling[57] and could be considered a threshold for new or casual cyclists. Model 5 (Table 5.6) used the subset of trips which were under 5 km (53% of the full dataset) where cycling may be a very reasonable substitute for driving. Results were fairly consistent with the cross-zonal full dataset model 4, with the exception that population density at the origin and the presence of cyclist- activated traffic lights along the route were both retained in the model, and educational land use at the destination was not.

City planners and health practitioners have a specific interest in understanding how to motivate reluctant cyclists to cycle more. To determine whether such cyclists might be differently influenced by the built environment, we conducted two sub-analyses of the cross-zonal model (Model 4), one restricted to the rare cyclist group, and one for frequent and regular cyclists combined. All odds ratios in these models were in the same direction as Model 4, although not all remained significant (as expected given the smaller sample sizes). In the model for rare cyclists, comprised of 1930 trips with 15% by bicycle, trip distance was a stronger deterrent (OR=0.32, 95% CI: 0.23-0.43) than in Model 4. Other significant variables highlighted connectivity as a motivator (intersection density OR=1.28, 95% CI: 1.05-1.57) and arterial roads as deterrents (OR=0.76, 95% CI: 0.62-0.94). In the model for frequent and regular cyclists (1350 trips, 54% by bicycle), hilliness, intersection density, and trip distance were significant influences, all with odds ratios similar to Model 4.

5.5. Discussion This study found that the built environment had a significant influence on the decision to use the active mode of transport, bicycling, instead of driving. For utilitarian travel, features of the physical environment, the road network, bicycle-specific facilities, and land use were all associated with the likelihood of cycling, even after accounting for personal characteristics and trip distance. The following features promoted cycling: less topographical variation; traffic calming and cyclist-activated traffic lights along bicycle routes; higher route connectivity (intersection density); local roads instead of highways and arterials; higher population density; and neighbourhood commercial, educational, and industrial land uses.

Our findings for the common constructs of connectivity, land use, and residential density were congruent with existing literature on cycling and physical activity. Higher intersection density was

79 associated with a greater likelihood of cycling. Connectivity, as measured by a variety of different but related constructs[131], has been significant in other studies of cycling using objective built environment measures[132, 133], as well as in studies that combine both walking and cycling[134]. Higher population density at the destination was associated with a higher likelihood of cycling. Population or residential density, while often considered a proxy for other unmeasured features, is pervasive in the built environment literature as the data is widely available[37, 38, 135]. No previous cycling-only studies have explicitly considered population density. A study examining transportation- related physical activity included residential density, but it was not significant[134]. Density and its variability differs from place to place (e.g., New York City versus Atlanta) and thus its impact might be expected to vary between geographical regions.

We used an entropy measure for land use mix and found that a more balanced mix of residential, commercial, entertainment, and office land uses around the origin was associated with a higher likelihood of cycling. In a San Francisco study, this land use mix measure was one component of a land use diversity index that predicted walking but not cycling[61]. The use of this entropy measure was supported by empirical research on walking and obesity[66, 136]. However, the specific mix of land uses typically included may not be the ideal for cycling, where land uses such as green space may be more highly valued[115, 123]. Limitations of this land use mix measure have been discussed elsewhere[111, 112, 118, 137]. Other measures have been applied empirically in cycling research. Moudon et al. considered an extensive array of proximity-based land use measures, but few were associated with cycling[62]. Thus to enhance the land use analysis, we also looked at intensity-based land use measures and found that zones with more neighbourhood commercial, educational, or industrial land uses were associated with a higher likelihood of cycling, and those with more single family residential or large commercial land uses were associated with a lower likelihood of cycling.

While retail destinations are important trip attractors, we found that the lot size of commercial land use affected the direction of the influence on mode choice. Where there were more large commercial lots (greater than 1 hectare), trips were less likely to be by bicycle. Where there were more neighbourhood commercial lots, trips were more likely to be by bicycle. Neighbourhood commercial lots may be characterized as human settings with easy access to multiple storefronts, in contrast to the vast parking lots that typically abut malls or big-box retail shops. While there are efforts to measure the urban design characteristics of such micro-scale environments[138] the literature linking this to physical activity is still evolving.

80 We also included built environment factors that might be specifically important to cycling. Hilly topography, as measured by variation in elevation, was an important deterrent to cycling. This concurs with results of surveys[115] and focus groups[117] where cyclists mentioned hilliness as a barrier. Studies that use perceived measures of hills have also found negative associations with cycling[139]. However, results from prior studies using objective measures have been mixed. Quartiles of average slope did not correlate with cycling in Portland[132], but a dichotomized average slope variable did in Bogota[133]. A variable for the presence of a steep hill did not predict cycling in one study[119], but predicted recent rail trail use in another[95]. Based on these discrepancies we developed new constructs for topography for the current work: the variation in elevation; and the proportion of steep road segments along the route. The steepness measure was created based on cyclist‘s input[117] but was not significant in the models, whereas the variation in elevation was. As a whole, this collection of findings relating topography to cycling is not contradictory but rather may provide detail on the specific qualities that affect cycling. For example, it may not be the presence of a single short steep hill that deters cycling, since it could be avoided with a detour or seen as a physical challenge, but instead constant up and down along a route as indicated in this analysis. Given that certain areas of Metro Vancouver are very hilly, the region provides good variability in this ―hilliness‖ measure. This new construct warrants application in other geographical locations to test its generalizability to predict cycling.

Certain aspects of bicycle facilities and the road network were significantly associated with cycling. Traffic calming around the origin, road markings or signage around the destination, or cyclist-activated traffic lights en route were associated with cycling. The current study had highly detailed spatial data on the locations of bicycle road markings, signage, traffic calming measures, and cyclist-activated traffic lights. We found no studies that had comparable detail on such amenities. As these features are relatively rare across the region, the skewed count data required that we employ dichotomous variables (present/absent). These dichotomous variables were an accurate description of exposure for smaller zones (i.e., origin or destination) but in the larger route buffers, the frequency of such facilities may have been a better measure. The measures for the three types of bicycle amenities were only moderately correlated (Pearson correlation 0.45-0.65), not enough to restrict offering all to the models, but perhaps affecting the number of factors retained in each model. Overall, it is clear that some kind of bicycle amenity is important, but our models did not indicate that one type was better than another.

81 We found that areas with a lower density of arterial roads or highways had a higher likelihood of cycling. In Bogota, Columbia, the overall street density (high versus low) predicted cycling[133]. Yet all roads are not equally valued by cyclists, indicating that road measures should be stratified by road type[90]. Other GIS-based research has found that cyclists choose local roads over major roads[57, 59, 98]. In our study, the density of designated bicycle routes or off-road paths was not significant in multivariable models. However measures of bicycle routes have been significant elsewhere. In an ecological study of 35 large US cities bike lane density was important: each additional mile of bike lane per square mile was associated with a 1% increase in the bicycle mode share[47]. In individual-level studies, living closer to a regional trail system predicted overall cycling in Portland[132] and trail use in the Twin Cities[140]. A Metro Vancouver survey of cyclist preferences found that roads and paths tailored to cycling were valued, but also that these preferences did not correspond with current use of different road and path types[90]. Instead, in areas where the ideal infrastructure (separated paths or bikeways) is not available, bicycle travel may take place along sub-optimal but best available road types (local roads). This may account for why bicycle routes and off-road paths were not significant in our models.

Trip distance is a fundamental consideration in mode choice. In our study, the median distance for bicycle trips was 2.5 km, less than half of the 6.0 km median distance for car trips. We found that distance was significant in all multivariable models and that for each additional 10 km of trip distance people were only 40-60% as likely to make the trip by bicycle. This relationship was consistent whether we included all trips or only those less than 5 km. Distance was even more important to rare cyclists, who were only 32% as likely to use a bicycle for an additional 10 km of distance. The importance of trip distance for cycling has been found by other researchers, and several studies have restricted their study populations to people who live within a bikeable distance of their workplace[133, 134, 141].

It is important to consider whether the distances travelled by bicycle in this study would contribute sufficient physical activity for health. The recommended 30 minutes per day of physical activity[5] translates to about 7.5 km total distance for average cycling speeds of 15 km/hr. Since this amount can be obtained in several bouts of exercise, a trip by bicycle to and from a destination ~4 km away would satisfy daily physical activity requirements. In our data, one third of the bicycle trips were this distance or longer. Of the car trips, about one third were at least 4 km but less than 12 km, corresponding to 15-45 minute bicycle travel times. Shifting these trips to cycling from driving can

82 therefore be expected to improve individual health. While there are increased risks of personal injury and possibly air pollution exposure associated with this shift, evidence indicates that the multiple health benefits of cycling outweigh the risks[27, 28, 142]. In addition, there is evidence that physical activity via active transportation is easier to maintain than other forms of exercise, such as going to a gym[16].

This study explicitly considered place, by measuring the built environment in different spatial zones: along the route, and around the origin and destination. We found that place was important, since in each zone different built environment factors influenced cycling. For example, hilliness was significant only in the route zone, but not in the origin or destination zones. Land use mix was significant only within the origin zone, but not along routes or at destinations. Only intersection density was significant in all three zones, but it had a stronger effect within the route and destination zones than the origin zone. In our cross-zonal model, where factors from all zones were offered, the majority of significant factors were route measures, suggesting that route characteristics were a primary consideration in making the decision to cycle instead of drive. If only origin, or origin and destination zones were considered, certain influential factors such as topography, road network composition and bicycle facilities may have been missed for planning and policy decision-making.

These findings underscore the necessity of including all three spatial zones. No prior bicycle research has analyzed this comprehensive set of zones when considering the influence of place. Studies often rely on physical activity survey data and thus have only participants‘ postal code as a spatial identifier[123, 132, 133], limiting the analysis to features around trip origins. A few studies have drawn on travel diary data, which details origins and destinations[61, 140]; only one of these explicitly considered destination characteristics in the analysis[61]. We found a single study that considered route factors, which measured the built environment along the shortest routes between home and workplace for trips made by walking or cycling[134]. A multi-zonal concept has been previously proposed in a review of neighbourhood audit tools that presented three key zones of a trip: the origin and destination, the route taken, and the area in which the trip takes place[143], but their conceptual model has yet to be applied empirically.

5.5.1. Strengths and limitations This study of the link between the built environment and physical activity is one of the few focused on cycling, and the first to include those who cycle infrequently. The travel patterns of the infrequent but

83 willing cyclist population, or the ―near market‖, are important to consider in order to have the maximum impact on behavioural change[105]. In our data, 290 (28.3%) of the bicycle trips were made by people who cycle less than once a month. Other research defines cyclists with thresholds such as those who cycle at least once per week, thus capturing only regular cyclists, and may oversample to get an adequate population size. Examples include a King County study with a high proportion of weekly cyclists (21% of 608 participants)[62], and a Portland study where 83% of the 162 participants considered themselves regular cyclists[100]. Market research indicates that certain route preferences as well as cycling motivators and deterrents can differ between regular and infrequent cyclists[90, 115]. Therefore research needs to capture the travel patterns of infrequent cyclists to identify environments that can attract new cyclists, or increase cycling rates in those who cycle rarely.

The Cycling in Cities survey collected trip origin and destination locations, but for logistical reasons the exact route traveled was not collected. This was not seen as a limitation as it is recognized that routes may vary by day or between people. A validation study conducted by the authors compared the shortest route (as used here) with the actual route for a subset of these trips[90]. It found that ¾ of the actual trips were less than 10% longer than the shortest distance path. The actual and modeled trips did not differ significantly in terms of general built environment measures (air pollution, greenness, connectivity, land use) but did in terms of cycling infrastructure and road network. Specifically, people making bicycle trips were more likely to route away from highways and arterials, and toward designated bicycle facilities, off-street paths and local roads; conversely, those making car trips detoured to highways and arterials. This validation study suggests that using shortest routes, as done here, would in fact underestimate the influence of bicycle facilities on making bicycle trips.

Data limitations meant than certain factors that could influence cycling were not included. One key factor is safety[144]; however, geocoded locations of crash-sites are not currently available for the region. End-of-trip facilities (bicycle racks, showers) are important to cyclists[115], but they are not yet identified in regional spatial datasets.

Our results provide additional evidence about built environment factors affecting cycling, where research remains rare. They describe travel behaviour decisions in one geographic region; similar analyses are needed elsewhere. Influences will vary: topography may play an influential role in hilly cities such as Vancouver or San Francisco, whereas green cover may be a serious consideration in regions with hot climates.

84 Finally, as with much of the research on the built environment, this is a cross-sectional study and the findings provide evidence of associations, not causality. Since experimental studies are impossible in this field, a combination of quasi-experimental, before-and-after, and cross-sectional studies will be needed to build the body of evidence for causality[42].

5.6. Conclusions Using novel methodology tailored to cycling, we found that the built environment influenced decisions to bicycle instead of drive after accounting for trip distance and personal demographics. This study characterized the built environment around the trip origin and destination, and along the route between the two, and found increased bicycling with less hilliness; less arterial roads and highways; higher intersection density; presence of bicycle-specific infrastructure including traffic calming, signage, road markings, and cyclist-activated traffic lights; more neighbourhood commercial, educational, and industrial land uses; less large commercial and single family housing land uses; greater land use mix; and higher population density. Different factors were important within each of the spatial zones, and when all spatial zones were considered together, more factors related to the route zone were significant influences on cycling. Future studies should explicitly consider these three spatial zones in order to fully explore the connection between urban form and travel behaviour. These findings identified features that support cycling, and can be used to develop land use and transportation planning policies that encourage active transportation for improved individual and community health.

85 Table 5.1 Descriptive characteristics of trip takers (n=1902) (before imputation for missing data)

N % Cycling frequency rarea(< 12 trips /year) 1138 59.8 frequent (at least 1 trip/month) 618 32.5 regular (at least 1 trip/week) 146 7.7 G ender male 925 48.6 female 977 51.4 A ge (4 missing) 19-24 119 6.3 25-34 296 15.6 35-44 613 32.3 45-54 495 26.1 55-64 262 13.8 65&older 113 6 E ducation (22 missing) some high school or less 34 1.8 graduated high school 240 12.8 vocational/college/technical 532 28.3 some university 160 8.5 graduated university 914 48.6 H ousehold income (383 missing) < $30,000 146 9.6 $30,000-$59,999 350 23 $60,000-$89,999 400 26.3 > $90,000 623 41 E mployment status (7 missing) working 1504 79.4 student 66 3.5 retired 170 9 unemployed/homemaker/other 155 8.2 M ean number of people in household 3.08 Has children in household 961 50.5 Access to car 1831 96.3 Mean number of bikes in household 2.98 Home residence City of Vancouver 494 26 rest of Metro Vancouver region 1408 74 a this group comprises both the occasional and potential cyclists from Chapters 2 and 3

86 Table 5.2 A priori hypotheses on measureable built environment measures that affect the likelihood of making a trip by bicycle, with hypothesized direction of influence, and relevant spatial zone

Primary spatial zone of influence Domain Metric route origin destination

air pollution average NO2 - Physical greenery % of land area with green cover + environment topography variation in elevation (hilliness); - % road segments with slope > 5% (steep hills). - - - street connectivity ratio of 4 way intersections to all intersections + Road % of road network that is: highway; - network arterial road; - local road. + % of road network that is: off-street path; + designated bike route. + Bicycle presence of: traffic calming features; + facilities road markings or signage; + crossings with cyclist-activated traffic lights. + density population per hectare + land use mix + + % of land area that is: single family residential; - multi-family residential; + commercial; + Land use educational; + entertainment; + industrial; + office; + park. + + : indicates that higher values are expected to be associated an increased likelihood of making a trip by bicycle versus car; - indicates that higher values are expected to be associated with a decreased likelihood of making a trip by bicycle

87 Table 5.3 Trip distance and spatial analysis zone dimensions, by mode

Car trips (n=2257) Bike trips (n=1023) mean median std dev min max mean median std dev min max Trip distance (km) 10.21 6.02 10.5 0.04 56.1 4.65 2.47 5.96 0.04 52.72 Land area (ha) route zones 518.8 312.5 518.9 21.6 2797.8 245.1 136 293.8 21.4 2648.3 origin zones 26.2 24.8 10.7 0.9 51 30.9 31.2 11.8 0.9 50.3 destination zones 34.6 34.1 9.8 3.7 54.4 38.1 38.9 9.7 5.6 56

88 Table 5.4 Descriptive statistics for built environment measures for route zones, by trip mode

Car trips (n=2257) Bike trips (n=1023) interquartile interquartile mean std dev range mean std dev range t-test Physical environment % of land area with green cover 32.4 15.4 19.8 25 15.7 21.6 * average air pollution (ppm NO2) 29.4 8.4 11.2 31.2 9.9 12.7 * variation in elevation (std dev of elevation points) 23.9 21.7 24.5 14.4 13.6 15.1 * % of road segments >5% slope 10.9 13.5 15.3 9.6 13.2 14.5 * R oad network intersection density (# intersections/ha) 0.4 0.2 0.2 0.5 0.2 0.2 * % road network: highway 4.1 7.9 4.5 1.2 4.4 0 * % road network: arterial 22.5 11.3 15.6 19.1 12.3 17.2 * % road network: local road 57.1 13.2 18.1 63.3 13.5 18.8 * B icycle facilities presence of traffic calming features 0.5 0.5 1 0.5 0.5 1 presence of road markings or signage 0.8 0.4 0 0.8 0.4 0 presence of cyclist-activated traffic lights 0.4 0.5 1 0.5 0.5 1 % road network: designated bike route 12.9 8.8 11.1 12.8 9.5 13.7 % road network: off street 2.4 3.1 3.8 2.2 3.6 3.3 L and use population density (population/ha) 0.2 0.5 0.2 0.5 0.8 0.5 * % of land area with the following use a: single family residence 34.2 12.8 17.7 35.7 14.7 22.2 * multifamily residence 2.8 3.7 2.9 3.4 4.2 4.4 * neighbourhood commerical 2.1 2.1 2.3 2.9 2.5 3.5 * large commercial 2.2 2.7 2.7 1.6 2.8 2.1 * education 2.2 2.8 1.8 2.4 3.7 2.6 * entertainment 0.2 0.2 0.2 0.2 0.2 0.2 industrial 4 4.8 5.7 3 4.8 3.7 * office 1.4 2.1 1.5 1.6 2.3 2.1 * park 9.2 6.4 6.9 8 7.2 7.3 * land use mix 0.3 0.2 0.3 0.3 0.2 0.3 a generalized land use categories that were not hypothesized to influence cycling were excluded from analysis (e.g., agricultural, vacant, transportation/utility) * significant result (p<0.05) in t-test for difference in mean value between car trips and bike trips

Table 5.5 Results of multi-level logistic models for effect of built environment measures on the likelihood that a trip is made by bicycle instead of car

ROUTE ORIGIN DESTINATION (Model 1) (Model 2) (Model 3) ORa 95% CI ORa 95% CI ORa 95% CI Physical environment variation in elevation (std dev of elevation points) 0.59 0.49 0.72 R oad network intersection density (# intersections/ha) 1.54 1.32 1.8 1.17 1.05 1.31 1.38 1.21 1.57 % of road network that is highway 0.93 0.88 0.98 % of road network that is arterial 0.75 0.64 0.89 0.81 0.68 0.97 B icycle facilities presence of traffic calming featuresb 1.95 1.49 2.54 presence of road markings or signageb 1.25 1.02 1.54 L and use population density 1.09 1.01 1.19 % land area that is single family residential 0.74 0.63 0.87 % land area that is neighbourhood commercial land use 1.22 1.04 1.43 % land area that is large commercial land use 0.75 0.63 0.9 0.81 0.74 0.89 % land area that is educational land use 1.04 1.01 1.07 % land area that is industrial land use 1.01 1 1.02 land use mix (entropy measure) 1.14 1.02 1.27 T rip distance (km)d 0.61 0.5 0.74 0.39 0.34 0.46 0.38 0.32 0.44

D emographics gender (ref: male) female 0.57 0.47 0.7 0.57 0.47 0.7 0.57 0.47 0.7 age (ref: >65) 19-24 6.07 3.28 11.22 5.41 2.95 9.95 6.41 3.47 11.87 25-34 2.89 1.71 4.89 2.65 1.57 4.47 3.46 2.05 5.82 35-44 2.77 1.69 4.53 2.51 1.54 4.09 3.08 1.88 5.03 45-54 2.83 1.72 4.67 2.62 1.59 4.3 3.07 1.87 5.06 55-64 1.72 1.01 2.94 1.61 0.94 2.73 1.81 1.06 3.08

cont’d ROUTE ORIGIN DESTINATION (Model 1) (Model 2) (Model 3) ORa 95% CI ORa 95% CI ORa 95% CI education (ref: vocational)c some high school or less 1.05 0.73 1.49 1.03 0.72 1.47 1.01 0.7 1.44 graduated high school 1.38 1.08 1.75 1.41 1.11 1.8 1.29 1.01 1.64 some university 1.88 0.9 3.95 1.91 0.91 4.02 1.73 0.82 3.62 graduated university 1.22 0.82 1.84 1.2 0.8 1.79 1.19 0.79 1.78 household income (ref: > $90,000) <$30,000 2 1.35 2.97 1.87 1.26 2.77 2.1 1.42 3.11 $30,000-59,000 1.59 1.2 2.09 1.42 1.08 1.87 1.59 1.2 2.09 $60,000-89,000 1.58 1.22 2.04 1.5 1.16 1.93 1.62 1.25 2.1 a OR: odds ratios are the change in odds for a change of the interquartile range of dependant variable: 95% CI: 95% confidence interval b OR for cycling facilities is the odds associated with it being present in buffer c education is only p=0.059 in route model; p=0.18 in destination model d OR is for a change in 10 km, approximately the interquartile range (9.5 km)

Table 5.6 Cross-zonal multi-level logistic modelsa for effect of built environment on the likelihood that a trip is made by bicycle instead of car, for all trips, and for short trips only

ALL TRIPS SHORT TRIPS: < 5km (Model 4) (Model 5) (n=3280) (n=1737) OR for change in IQR OR for change in IQR Z one built environment feature ORb 95% CI ORb 95% CI Origin factors population density (population/ha) 1.09 1 1.19 D estination factors % of land area that is educational land use 1.05 1.02 1.08 R oute factors variation in elevation 0.61 0.5 0.73 0.75 0.64 0.87 intersection density (intersections/ha) 1.74 1.51 2.01 1.43 1.19 1.72 % of the road network that is arterial 0.83 0.73 0.97 0.79 0.66 0.95 presence of cyclist-activated traffic lights 1.7 1.27 2.26 T rip distance (km) 0.59 0.49 0.71 0.64 0.52 0.79 aadjusted for gender, age, and household income (results not presented here, because odd ratios were very stable across models, as shown in Table 5.5) b for built environment measures, OR is the change in odds of bike associated with a change in the interquartile range of the variable, with these exceptions: for cyclist-activated traffic lights it is the odds associated with it being present in zone; for trip distance in model 4 it is the odds associated with a change in 10 km (~ interquartile range = 9.5 km); for trip distance in model 5 it is the odds associated with a change in 2 km (~ interquartile range = 2.1 km).

Figure 5.1 Potential zones influencing decisions to cycle: route, origin, and destination zones

Figure 5.2 Differential effect of large commercial and small commercial land use on likelihood that a trip is made by bicycle, instead of car

6. DEFINING AND MAPPING BIKEABILITY

6.1. Synopsis

Evidence indicates that aspects of the built environment have a significant influence on decisions to cycle instead of drive. Geospatial data describing these environmental features are increasingly available. The promotion of walking has capitalized on this to create walkability indices and mapping tools such as ―WalkScore‖. To date, however, there has been little effort to use existing data and knowledge to define and map ―bikeability‖ as an approach to promote cycling as a form of active transportation. Given that determinants of walking and cycling are different, and that cycling potentially reaches a different target market, new metrics are needed. Our goal was to build a tool to identify areas that are more and less conducive to cycling. We used empirical research to develop a bikeability index and used GIS to map the index across the Metro Vancouver region. Results of an opinion survey, travel behaviour studies, and focus groups were used to identify the components of the index and their relative importance. Pertinent geospatial data layers were scored and combined using a flexible weighting scheme to create a composite map highlighting both high and low bikeability areas. The bikeability index was comprised of five factors shown to consistently influence cycling: bicycle facility availability; bicycle facility quality; street connectivity; topography; and land use. For mapping purposes, we created corresponding metrics: density of bicycle facilities; separation from motor vehicle traffic; connectivity of bicycle friendly- roads (local streets, bicycle routes and off-street paths); slope; and density of destination locations. Using empirical evidence to combine data layers for these metrics we generated a high-resolution (10 m) bikeability surface for the region, depicting bike-friendly areas and areas where cycling conditions need to be improved. Built environment interventions for specific locations are informed by evaluating scores for the five individual component layers. Mapping bikeability provides a powerful visual aid to identify zones that need improvement to support healthy travel choices. This evidence- based tool presents data in a user-friendly way for planners and policy makers. The overall bikeability score and its five component scores can guide local action to stimulate changes in cycling rates. It uses widely available data types, thus facilitating easy application in other cities.

Furthermore, the flexible parameters and weighting scheme enable users elsewhere to tailor it to evidence about local preferences and conditions.

6.2. Introduction Faced with global challenges of congestion, air pollution, climate change, energy scarcity, and physical inactivity, the fields of public health, urban planning, and transportation are collectively focused on strategies to reduce auto travel and promote active transportation[39]. A growing body of research has explored how the built environment influences physical activity, with findings that people who live in more walkable neighbourhoods walk more, have lower rates of obesity and chronic disease, and travel less by car, even after accounting for demographics and personal preferences[36, 145]. Features of these walkable neighbourhoods include high population density, mixed land use, grid-based street networks, and human scale design[37, 66]. The existing focus on walking is justifiable given that it is the most common form of leisure-time physical activity, with few barriers and no cost. However bicycle travel, being faster and more efficient while nearly as accessible and economical, is a more reasonable substitute for auto travel when trip distances exceed a kilometer[15]. The utility of cycling for transportation has been recognized in model cities such as Copenhagen and Amsterdam, where cycling mode shares are as high as 30%, with little differentiation by age and gender[23]. The potential of cycling for transportation in North America has yet to be realized, although promising initiatives are taking place in select cities (e.g., New York, Portland, and Montreal).

The literature provides some evidence regarding built environment correlates of cycling behaviours, based on both opinion and behaviour-based studies. Surveys have documented cyclists‘ opinions on factors that motivate and deter cycling, highlighting issues of safety, infrastructure, and the physical environment[63, 115, 146]. Stated preference studies have quantified trade-offs between such factors, given hypothetical travel situations[45, 58, 93]. The initial behaviour- based research drew on aggregated data, and found that cities with more bicycle lanes have higher cycling rates[47, 48]. Other endeavours have used disaggregate origin-destination data from travel surveys[61] or tracked cyclists‘ route selection[57, 59, 98, 99] and found that features of the built environment are associated with both mode choice and route choice.

While the research indicates that environmental factors are associated with cycling, the data has not been used in a systematic way to stimulate changes in cycling rates. Over the past decade,

government departments and agencies have built spatial databases of environmental features, and the data and the technology to use them have become more accessible. Research on the promotion of walking has drawn on these, using geospatial data to create walkability maps and popular online tools such as WalkScore (http://www.walkscore.com/).

Much less has been done to use existing knowledge to define and map ―bikeability‖. Existing indices for walkability, sprawl, or urban and suburban built environments[121, 138, 143, 147, 148] are comprised of components such as density (population, employment, park), land use, transportation network (road connectivity, bus networks) and some specific pedestrian features (e.g., sidewalk coverage). Some indices include select factors related to cycling (e.g., the presence of bicycle lanes) but few have applied evidence from the cycling literature. Given that walking and cycling are recognized as functionally different[42] and the dimensions of the built environment associated with each mode may differ[62, 149], there are certain to be differences between a measure for walkability and one for bikeability.

Our goal was to define bikeability and build a planning tool that identified areas more and less conducive to cycling. We used empirical evidence to derive a bikeability index and applied it to Metro Vancouver as a case study. Capitalizing on geospatial technology, we mapped bikeability as a continuous surface, displaying a continuum of conditions from high to low. Our ultimate goal was to build a flexible tool relying on commonly available data to facilitate widespread adoption, and enabling it to be tailored to local conditions.

6.3. Components of bikeability Three types of evidence were used to identify components of the bikeability index and their relative importance: an opinion survey; travel behaviour studies; and focus groups (Figure 5.1).

6.3.1. Identifying built environment factors Opinion Survey

We used the findings from a population-based survey to identify built environment factors that influence cycling. The Cycling in Cities Survey included 1,402 current and potential cyclists from Metro Vancouver. The survey queried the relative importance of 73 potential motivators and deterrents, about a third of which were related to the built environment. Survey participants reported that strong influences were bicycle facilities, aesthetics, topography, traffic and trip distance

(or speed of cycling relative to other modes)[115]. The survey also asked about usage and preferences for 16 types of cycling infrastructure. Results highlighted the desirability, to all types of cyclists, of off-street or physically separated routes, followed by local street bike routes[90].

Travel Behaviour Studies

We identified objective measures of the built environment that are associated with cycling based on two studies of actual travel. The trip data also came from the Cycling in Cities Survey, which collected information about 2 common trips from each participant. The first study used the 3,280 trips made by car and bicycle to explore which built environment measures were associated with a higher likelihood of cycling versus driving[98]. The second study used a subset of the data to compare the actual route used with the shortest distance route between the origin and destination, to understand how built environment influences route selection[99]. The results suggest the importance of the following general domains:

1. Bike facilities  Cyclists detour en route to use cycling facilities;  More bicycle-friendly facilities (traffic calming, road markings or bicycle signage, or bicycle-activated crossing signals) were positively associated with the odds of cycling. 2. Connectivity  Intersection density was positively associated with the odds of cycling;  Arterials and highways were inversely associated with odds of cycling. 3. Topography  Hilliness was inversely associated with odds of cycling. 4. Land Use  Certain land use types were associated with higher odds of cycling: neighbourhood commercial, education, entertainment, and office. Focus groups

We conducted a series of focus groups to provide context on how the built environment contributes to making a neighbourhood bikeable[117]. The emphasis was on physical environment factors that are modifiable through planning and zoning, as opposed to those beyond the scope of municipal or regional decision making (e.g., climate, helmet legislation). Based on the results of the Cycling in

Cities Survey and existing literature on the built environment[37, 38], eight broad factors were selected for exploration: topography, distance traveled, environment (i.e., air and noise pollution), traffic, street network, bicycle facilities, land use, and population density.

We recruited four groups of 8-10 participants: one group of cycling advocates and three groups of Cycling in Cities Survey respondents. Participants in the latter groups were stratified according to cycling frequency: regular cyclists (who cycled at least once a week); occasional cyclists (who cycled less than once a week); and potential cyclists (who had not cycled in the last year, but were willing to consider cycling in the future).

Participants started by completing questionnaires about how the factors might increase or decrease their likelihood of cycling for a utilitarian trip, and what other factors were important. A moderator then facilitated an hour long discussion exploring the context in which each factor influenced cycling. Finally, each participant was asked to prioritize their top three factors. The prioritization was not completed by the cycling advocacy group.

The focus groups provided information on the relative importance of built environment factors as indicated in the ranking in Table 6.1. Bicycle facilities were clearly the most important component, scoring about twice as high as traffic or the street network. The participant quotes further highlighted the importance of separated facilities. Conversely, land use and population density were not as highly ranked by participants.

The focus group discussions also gave insight as to how concepts could be operationalized. For example, participants felt that highly connected grid-based road networks allowed for more route choice and efficient travel, but that busy streets with bus, truck, or high car volumes deterred cycling. This suggested that a conventional connectivity measure (intersection density) should be modified for cycling to include only bicycle-friendly roads (e.g., local roads and bicycle paths). Another theme was that bike routes needed to connect with other bike routes, suggesting that a density measure be used instead of a present/absent dichotomous measure.

6.3.2. Developing the bikeability index The empirical evidence from the opinion survey, travel behaviour studies, and focus groups is summarized in Table 6.2. A comparison of results across studies, looking for consistency and readily mapped features, produced four main domains: bicycle facilities; street connectivity;

topography; and neighbourhood land use. These met the criteria of data availability: topography and road networks exist in national datasets, and bicycle route networks and land use data (e.g., parcel data from tax assessments) are held by most municipalities.

Some synthesis of the results was conducted, based on the focus group discussions. We integrated the connectivity and road type measures as ―connectivity of bicycle-friendly streets‖, and the neighbourhood land use and distance concepts as a ―destination density‖ measure. Additionally, both the focus groups and opinion survey specified that not all bicycle routes are equally desirable, but that facilities separated from traffic were preferred over others. There are often detailed facility types in bicycle network datasets. As such, where data on facility separation is available we propose its inclusion as a fifth factor.

The final index is a combination of these factors: Bikeability =B1* Bicycle route density + B2* Bicycle route separation + B3* Connectivity of bicycle-friendly streets + B4* Topography + B5* Destination density where B1-B5 are the weights that may be applied to each layer. The scoring from focus groups (Table 6.2) proposes a weighting that is higher for bicycle facilities. Where bicycle route density is the only facility measure available, the scoring indicates it should be weighted twice the other factors (i.e., B1=2, B3=B4=B5=1); where both bicycle route density and separation are available, the weighting could be balanced (i.e., B1 through B5=1).

6.4. GIS procedures This section outlines manipulations of the geographical data to create and combine data layers to generate the bikeability map. In brief, we generated metrics to correspond to the five components and scored them (Table 6.3). All manipulations were done in ArcGIS 9.3 (ESRA, Redlands, CA). We worked with 10 m grid cell raster files in order to generate a high resolution surface, with the exception of topography where input files were 30 m grids. Where these processes required a search radius we used a 400 m buffer (~1/4 mile). This distance was identified in our previous research as the average distance that cyclists are willing to detour[99].

100

6.4.1. Generating component raster files 1. Bike route density. We converted a shapefile of all the designated bicycle routes (on-road and off-road) to a density surface by applying the ―Line Density‖ tool. For ease of interpretation we used Math Tools to convert this raster to have units of ‗m of bicycle route‘ in a 400 m radius circular buffer around a given cell. 2. Bike route separation. To generate an indicator file for high quality routes we selected all route sections that were physically separated from motor vehicle traffic (i.e., off-street paths and cycle tracks). We applied a buffer of 200 m to each side of these before converting it to a raster file. 3. Connectivity of bicycle-friendly streets (i.e., a local road, an off-street path, or a designated route[90]). The Digital Road Atlas[127] was merged with a shapefile of designated cycling routes. An ESRI script (JPtools Fnode Tnode) was used to generate a point file with the number of connecting roads at each intersection. The ―Select by Location‖ function was used to select all intersections with 3 or more legs (i.e., at least a T- junction) where at least one road was favourable for cycling. We applied the ―Point Density‖ tool to create a connectivity raster and used Math Tools to convert to units of ‗number of intersections‘ in buffer. 4. Topography. A slope raster file was created from the 30 m grid Digital Elevation Model raster file using the ―Slope‖ Spatial Analyst tool with percent rise as the output. This generates the value of the maximum slope between a cell and its neighbouring cells. 5. Destination density. Our land use data was property tax assessment data from BC Assessment[110]. We selected those parcels that were potential destinations for cycling based on land uses that were positively associated with cycling in the travel behaviour study (neighbourhood commercial (n=6986 parcels), education (n=910), entertainment (n=352) and office (n=1754)). The polygon file was converted to a point file, and the ―Point Density‖ tool was used to create a density raster. We used Math Tools to produce units of ‗number of destination parcels‘ in the buffer.

6.4.2. Scoring and combining component files

1. Each raster file was reclassified to a scale of 1-10, where 1 was the least bikeable environment and 10 the most bikeable (Table 6.3). For continuous values (topography layer

101

and density layers) cutoffs were approximated to deciles. The bicycle route separation layer was classified as 10 (present) or 1 (absent). 2. Reclassified raster files were combined using the ―Weighted Overlay‖ tool. This tool combines layers with a common measurement scale and allows the user to assign relative weighting to the layers. As empirical evidence suggested, the five component layers were assigned equal weighting (20% each, to sum to 100%).

6.5. Results Figure 6.2 shows the bikeability surface for Metro Vancouver where bike-friendly areas are in green, and areas where cycling conditions need to be improved are in yellow to red. The map provides a regional overview and illustrates the great diversity of cycling conditions.

This composite map is most powerful when used in combination with the individual component maps also included in Figure 6.2. Built environment interventions for specific locations can be informed by evaluating scores for the component layers. For example, certain areas had high scores for topography (no hills) and a reasonably high density of shops and other destinations, but scored low in terms of the density of bicycle facilities. Such areas could be prioritized for new bicycle routes.

Figure 6.3 shows the mean values for the bikeability index and the component layers for each municipality. It illustrates marked differences in local planning policy: the cities of Vancouver, Burnaby and New Westminster had denser bicycle infrastructure compared to other municipalities. The City of Richmond had excellent topography for cycling, but scored lower for the more modifiable factors that support cycling. Rural areas such as Delta also had good topography but low scores for destination density, reflecting land use practices that do not encourage utilitarian trips by bicycle.

There was not just between-city variability, but also within-city variability. Figure 6.4 shows mean values for neighbourhoods within Vancouver, demonstrating the scalable nature of this tool and its flexibility in delineating boundaries. While Vancouver scored well for bikeability in comparison with other municipalities (Figure 6.3), Figure 6.4 shows there is variability between neighbourhoods within the city. Downtown scored well for separated bicycle facilities, and Downtown and the surrounding neighbourhoods also scored well for bicycle route density. However, areas to the southeast (Kensington-Cedar Cottage, Renfrew-Collingwood) had good connectivity and destination

102

density, but lower scores for bicycle facilities. This illustrates that there is room to improve cycling conditions within Vancouver, and these maps suggest area-specific strategies to consider.

6.6. Discussion We developed a ―bikeability‖ index to characterize and map a region‘s suitability for cycling. Evidence from a series of studies suggested that the index be a composite of five factors: bicycle route density, bicycle route separation, connectivity of bicycle-friendly roads, topography, and density of destinations. We demonstrated how the index could be used as a planning tool, using Metro Vancouver as a case study. The overall bikeability score and its five component scores can guide local action to improve cycling environments and stimulate changes in cycling rates.

Our focus was to develop a new measure specific to cycling conditions, based on empirical evidence. This bikeability index differs from existing walkability or sprawl indices[143, 148]. For example, the walkability index that has been applied to Vancouver, Atlanta, and Seattle, has four components: residential density, land use mix (an entropy measure), street connectivity, and retail floor area ratio, with connectivity weighted double the other factors[148, 150]. In comparison, this bikeability index also includes a connectivity component (although modified to focus on bike friendly streets) and a measure of land use (accessibility instead of entropy). The topography and bike facility-related components are unique. This work supports the hypothesis that bikeability differs from walkability, and warrants separate consideration.

Bikeability is clearly an area of growing interest, as confirmed by efforts undertaken by other health researchers and planners in the past year. These projects (summarized in Table 6.4) each focused on a premier North American cycling city (Montreal, New York, Portland) and combined a number of bike-related measures into a composite score[151-153]. A common theme across the indices is the importance of bicycle facilities. However each of these projects was initiated with a different objective and, as is often the case with concurrent efforts, used disparate approaches. The criteria used in selecting components in these projects, where described, are based mainly on expert opinion or intuition. We compare and contrast our results with theirs below.

We designed a system that relies on widely available spatial data to increase utility and facilitate widespread implementation. Our focus was on existing physical conditions, and used data for bicycle network, road network, topography, land use. Larsen at al. incorporated travel behaviours and cyclists‘ opinions about each area[151]. This adds a dimension of demographic data and local

103

input, but requires data that is not necessarily attainable in many locations. Both Larsen and El- Geneidy [151] and Richards et al. [153] included crash and injury data. Unfortunately this type of data (where available) is subject to reporting bias and typically lacks adjustment for exposure (e.g., cyclist counts)[154-156]. Voros et al.‘s[152] Cycle Zone Analysis had a similar focus to ours, i.e. on existing conditions, however, their Bikeway Quality Index component requires very rich facility- specific data (e.g., pavement quality, lane width, stop signs). While we used regionally-generated data in this case study, future applications of the tool could be built using open data access sources, such as those that exist for street networks (www.openstreetmap.org), topography (maps.nrcan.gc.ca/topo_e.php), and location density (from Google Local Search API, as is used in the WalkScore, www.walkscore.com) to enhance accessibility. Bicycle network data would be the only data requirement for participating municipalities.

We have presented a straightforward methodology with flexible parameters to enable others to build this tool and tailor it to evidence about local preferences and conditions. The stepwise data manipulations detailed herein rely on tools available in ArcGIS. This differs from the scoring procedures in Voros et al.‘s Cycle Zone Analysis[152] and allows for the tool to be routinely updated by planners rather than developed de novo from time to time at considerable expense. We used a flexible methodology to score and combine factors, and have provided cut-off values that can be replicated or used as comparators in other locations. For example, a given municipality could alter the weighting if topography was not an issue, or could leave out the bike route separation component if data were not available on separation of facilities. Alternatively, a municipality with particular interest in attracting the next wave of cyclists could put more emphasis on factors especially important to potential cyclists, perhaps using sub-analyses from the opinion survey[90] or the focus groups[117] to guide weighting. A plan for future implementation is to design an interface with interactive slider bars to change the relative weights of layers.

Also included in Table 6.4 is an earlier project on ―bicycle level of service‖ [157]. This road- segment audit tool is a measure of perceived safety and comfort of a hypothetical cyclist with respect to motor vehicle traffic, and relies exclusively on motor vehicle-related measures: traffic volumes and mix; speed limits; and lane widths. There is little overlap between our approach and bicycle level of service, although there are some complementary outcomes: our connectivity measure excludes heavy traffic roads that would have a low level of service, and facilities with no traffic also received good scores in the bicycle facility separation component.

104

Finally, our methodology produces a bikeability surface that reduces the biases generated by data aggregation. The modifiable areal unit problem (MAUP) is a persistent geographical analysis issue whereby the boundaries (i.e., municipalities or cycle zones) selected to aggregate spatial data can affect findings[158-160]. This can result not only from the size of spatial units but also the shape of zones. For example, destination density values will be different depending on whether a zonal boundary falls along a major road (i.e., the center of a main shopping district) or between major roads. While practical applications of our tool may require some aggregation, as demonstrated in Figures 6.3 and 6.4, it can be done post-hoc and can be tailored to the question at hand, whether it is a particular transportation corridor or neighbourhood. The other projects used varying units of analysis, from a reasonably small 300 m grid, to neighbourhood-based or larger cycle zones.

6.6.1. Limitations This paper presents a novel planning tool for Metro Vancouver which can be adapted for other locations. As we used evidence from local research to derive the bikeability index, some might argue that these results are location-specific and not generalizable to other places. However, while Vancouver is reasonably dense, has relatively high cycling rates (journey-to-work mode share of 3.8% in the city and over 10% in some neighbourhoods[24]) and has made substantial investments in cycling, the Metro region also includes suburban and rural areas with much different cycling conditions (as shown in Figures 6.2 and 6.3), and the car remains the dominant transport mode across the region. In previous articles we have situated our findings in the broader context of cycling literature[90, 115] and found consistencies across geographical locations in what constitutes bike-friendly design. Moreover, the flexible nature of the tool allows users to supplement it with locally available data. For example, the discussions from our focus groups could be augmented with local insight, to apply weights to any additional layers[117].

Our focus was on existing conditions. While the maps represent conditions at a snapshot in time, the straightforward methodology utilized means that users can easily incorporate new data as it becomes available. The tool can test the impact of proposed changes on bikeability scores (e.g., adding a new connection on the bicycle network) but it is not designed to predict resultant changes in cycling rates. Determining the latent demand for cycling, or the direct effect of interventions on cycling rates, remains a sizeable challenge[161].

105

There were several assumptions made in the development of this model. First, the GIS procedures employed required decisions on categories for each component and on a method for combining the components. We selected decile cutoff values as a reasonable starting point for a regional model and provided these cutoffs (Table 6.3) for clarity. While this decision was admittedly somewhat arbitrary, it was vetted by local planners. For example, an earlier categorization for topography (e.g., only round numbers) resulted in maps that did not resonate with local municipal staff. Note this categorization step is flexible and can be modified if evidence exists to guide selection of specific cut-offs, or if location-specific conditions suggest non-linear relationships between density and cycling conditions (e.g., extremely high density commercial areas may be poor cycling environments). We also made the decision to build the index as an additive model, similar to existing walkability and sprawl indices. While other methods to combine component scores exist (i.e., multiplicative models), the additive approach is a reasonable one since this tool allows users to examine the results of the index and each of its components separately. Additionally, we used 400 m search buffers to calculate density measures, based on our previous research. There is little information to guide selection of buffer sizes[38] but a sensitivity analysis could be done with a range of values. Finally, our land use data is based on tax parcels and thus indicates the number of lots, but not the number of individual offices or stores in a given parcel. This is a common limitation of land use data.

6.7. Conclusions Mapping bikeability provides a powerful visual aid and a quantitative metric to identify zones that can be improved to support healthy travel. This is an evidence-based tool that presents data in a user-friendly way for planners and policy makers. This tool has a number of potential applications. It can be used to identify and prioritize locations for new infrastructure for cycling. It can also be used for research, to select areas of high and low bikeability for studies on health disparities related to physical activity and the built environment. It can be used to engage the public in planning processes for the promotion of cycling. Its regional focus is suited to population-level planning, and it complements the growing number of travel planning tools targeted to individuals (e..g, Vancouver‘s www.cyclingvancouver.ubc.ca, Portland or Milwaukee‘s http://byCycle.org, and Los Angeles‘: http://opt.berkeley.edu/). In future studies we hope to evaluate the bikeability index with actual cycling behaviours, and integrate walkability and bikeability maps, to capture overall conditions for active transportation.

106

Table 6.1 Ranking of built environment factors in focus groups with current and potential cyclists (n=23)

Rank Factor Sample participant quote Score* 1 Bicycle Facilities ―The more you feel separated from traffic, the safer you feel‖ 50 ―The encroaching volume of traffic is rendering the bike lanes 2 Traffic 25 inadequate‖ ―If I‘m commuting I find [grid network] much easier, I feel I can 3 Street network progress through much faster‖ 17 ―Hills are a big problem for me because I don‘t have the strength or 4 Topography 16 endurance‖ ―I just hate it—all types of pollution—I would go a long way out of my 5 Environment 12 way to avoid it.‖ ―If I could ride to work from home in a half hour, I wouldn‘t think twice 6 Distance about it.‖ 9 ―I don‘t find the suburbs set up for bicycles. It‘s not easy to do your 7 Neighbourhood Land Use everyday chores because of the distances‖ 4 8 Population Density [no comments] 2 * factors were prioritized in order of importance to participant. Priority 1 was assigned 3 points, priority 2, 2 points, and priority 3, 1 point. Total points = 135, as 23 individuals completed the ranking exercise, and one person assigned only 2 factors.

107

Table 6.2 Developing an evidence-based bikeability index

108

Table 6.3 Scoring the components of bikeability to create maps, using empirical data from Metro Vancouver

Connectivity of Bike route Bike route Destination Score bicycle friendly Topography density separation Density streets (1= low # intersections on # of bike- bikeability, bike routes, local m of bike route a, b separation % rise a friendly 10= high roads, off street destinations a,b bikeability) paths a,b 1 0 no 0 >20 0 2 >0-250 - 1 10-20 0 3 >250-450 - 2-3 7-10 1-2 4 >450-600 - 4-6 5-7 3 5 >600-750 - 7-10 3-5 4-5 6 >750-850 - 11-15 2-3 6-8 7 >850-1100 - 16-20 1-2 9-10 8 >1100-1400 - 21-25 0.5-1 11-20 9 >1400-1800 - 26-30 0-0.5 21-40 10 >1800-6000 yes 31-60 0 40-300 a deciles based on empirical data from the Metro Vancouver area b in 400 m (1/4 mile) buffer around grid cell

109

Table 6.4 Concurrent projects using spatial data to capture “bikeability”

Reference Objective Factors Unit of Analysis (Location) Current paper To build a planning tool that Bike route density; Bike route separation; Continuous (Vancouver) identifies areas more and less Connectivity of bike friendly roads; surface: 10 m grid conducive for cycling, based on Topography; and Destination density cell the on-the-ground conditions, using widely available data

Larsen et al., To determine optimal locations Current cyclists' trips; short car trips; 300 m grid 2010[151] for new facilities suggested routes from local survey; (Montreal) reported bicycle-car crashes

Richards et al., To determine if neighbourhood % bike commuters; % of streets with bike Neighbourhood 2010[153] * bikeability predicts BMI lanes; bike lane density; bike injuries (New York)

Voros et al., To help planners assess existing Cycle Zone Analysis: Bikeway quality*; Homogeneous 2010[152] conditions and cycling potential, Road network density; Bike network cycle zones, (Portland) and maximize return on density; Permeability; Connectivity defined based on investment (connected node ratio); Average slope; expert assessment Distance to commercial establishment. (36 across *Bikeway quality: Auto speed; Auto Portland) Volume; Number of lanes; Bike lane drop; Difficult transition; Bike lane width; Jogs; Pavement quality; Intersection crossing quality; Stops

Landis et al., To quantify the ―bike- Bicycle level of service: Motor vehicle Road segment 1997[157] friendliness‖ of a roadway, related volume; Number of lanes; Effective speed (Various US to bicyclist comfort level for limit; % of heavy vehicles; Surface cities) specific roadway geometries and conditions; Width of outside lane. traffic conditions

110

Figure 6.1 Data sources and methodology for derivation of bikeability index

111

Figure 6.2 Bikeability and component maps for Metro Vancouver

Bikeability = Bicycle route density + Bicycle route separation + Connectivity + Topography + Destination density

112

Figure 6.3 Mean bikeability and component values for Metro Vancouver municipalities

113

Figure 6.4 Mean bikeability and component values for Vancouver neighbourhoods

114

7. SYNTHESIS: CONTRIBUTIONS, IMPACTS, FUTURE DIRECTIONS

The research chapters of this thesis were designed to be used for publication elsewhere. As such, specific conclusions and implications are included in each. This final chapter serves to synthesize the overall findings, bringing to light the unique contributions of this body of work to both research and policy, and discussing challenges and limitations that point to future research.

7.1. Summary

7.1.1. Objectives The objectives of this dissertation were: (1) to characterize the near market for cycling, according to demographic characteristics, stated cycling patterns and route preferences, and opinions on factors that influence their cycling; (2) to map built environment characteristics that may impede or facilitate bicycle travel; (3) to link built environment characteristics to actual travel data to determine which characteristics influence (a) decisions on route selection and (b) decisions on mode choice; (4) to develop a measure for ―bikeability‖ based on evidence; and (5) to create outputs that promote the uptake and application of findings by governments and other agencies that plan urban areas, fund and build cycling infrastructure, or promote cycling.

7.1.2. Findings The first research chapters (Chapters 2 and 3) reported the opinions of the near market cyclist population, tackling the first objective of this dissertation. Chapter 2 focused on preferences for different types of infrastructure. There was a clear inclination for off-street paths and routes separated from motor vehicle traffic. Women, people with children, and potential cyclists were especially drawn to these types of routes over on-road facilities. Moreover, this chapter highlighted a great disparity between route types that currently exist and those that cyclists would prefer to use. Chapter 3 assessed the relative importance of 73 potential motivators and deterrents of cycling, grouped according to the 5E framework (engineering, environment, education, encouragement, and enforcement). In the factor analysis, the domains with the strongest reported influence on cycling were safety, ease of cycling, route conditions, weather, and interactions with motor vehicles.

The next two research chapters (Chapters 4 and 5) comprise the behaviour-based analyses, addressing objectives 2 and 3. These chapters relied on GIS to map trips, generate travel routes, and

115

develop measures of the built environment. Chapter 4 responded to the scarcity of data on cyclists‘ route choices. It compared characteristics and built environment features of shortest distance routes and actual travel routes, for both bicycle and car trips. We found that detour rates were similar by mode: three-quarters of trips were less than 10% longer than the shortest distance route. But the motivations for detouring differed: bicycle trips were more likely to be along routes with better bicycle facilities and away from major roads; car trips, conversely, were more likely to be along highways and arterials. Chapter 5 identified features of the built environment associated with a higher likelihood of cycling, versus driving, using a cyclist-specific lens. It applied a novel spatial approach, measuring features of the trip origin, destination, and the route between them, and employed multi-level logistic modeling to account for demographic characteristics and trip distance. The findings indicate that the built environment is associated with mode choice and specifically that urban form is important not only at the trip origin, but also along the route and around the destination.

The final research chapter (Chapter 6) integrated the findings from earlier chapters and from focus groups associated with this work (Appendix 1), to fulfill objectives 4 and 5. The result is an evidence-based measure for bikeability and a planning tool based on the bikeability index, modeled for Metro Vancouver as a case study.

7.2. Unique contributions The novel aspects of this research have been detailed in each of the study chapters (2-6). However, taken collectively, the research papers yield several unique contributions to the literature worthy of mention.

7.2.1. Interdisciplinarity This dissertation transcends the traditional boundaries of epidemiological study. Research linking health, the built environment, and transportation spans the disciplines of physical activity, health promotion, urban planning, and transportation engineering. The origins of the current project were grounded in the public health-based objective to increase population-level rates of physical activity. However, in order to generate appropriate and policy-relevant conclusions from this research, I had to complete independent study and literature reviews in urban planning, transportation, and geographic information sciences. This interdisciplinarity is reflected in the journals where the individual chapters have been published: the American Journal of Health Promotion, the Journal of

116

Urban Health, Transportation, and Transportation Research Record. Correspondingly, the results have been shared with broad audiences: presentations at academic conferences for health, planning, geography, and beyond; at public forums and in popular media; in webinars and at meetings of public health practitioners; and in presentations to city councils, municipal staff, and regional transportation bodies. The wide interest in this work is a hallmark of its necessity and value.

7.2.2. A new population A distinct contribution to the cycling literature is the population studied. The Cycling in Cities‘ characterization of the ―near market‖ population complements the traditional focus on regular cyclists. Chapter 2 emphasizes the need to capture a broader cycling population: where regular cyclists were willing to cycle on most types of route infrastructure (13 of the 16 queried), potential cyclists were much more selective in routes they were willing to use (only 6 of the 16). The focus groups, split by cycling frequency, also showed variation between groups with respect to which environmental factors were important: for example, air pollution was a strong consideration for cycling advocates and regular cyclists, but less so for occasional or potential cyclists.

The choice of this particular population was made by the research team that preceded me; they recognized from health promotion experience the benefit of reaching the ―near market‖[30]. Again and again as I presented these findings, it was this particular aspect that resonated with the audience. In highlighting the preferences of potential cyclists, so many planners, policy makers, and avid commuters recognized that it is not them but their children, their parents, or their more hesitant colleagues and friends who form the next wave of cyclists. Anecdotally, I have been told that these results shaped their understanding that a shift in city design is required to induce population-level changes in travel behaviour.

7.2.3. Integration of multiple methodologies The truly novel aspect of this thesis is the blend of methodologies, both quantitative and qualitative, and both opinion- and behaviour-based. This diversity in approach strengthens conclusions where there is consistency in the findings across methods. In this work, the importance of bicycle facilities as a cycling motivator was such a finding. The opinion-based results from Chapter 2 and 3 highlighted this. There was a clear preference for bicycle-specific facilities, for separation from motor-vehicles and on-street parking; and for route ease and enjoyment. While opinions are a marker for behavioural intentions, they are subject to social desirability bias and may differ from true

117

behaviours. Within this dissertation, the opinion-based results could be cross-validated with the behaviour-based analyses in Chapters 4 and 5. In general, the findings of these mapped analyses corroborated what was learned in earlier chapters: that cyclists went out of their way to access high quality bicycle facilities, and that features of the route affected decisions to cycle. Chapter 6 introduced the qualitative portion of this research, focus groups conducted to contextualize what is meant for a neighbourhood to be ―bikeable‖. In these sessions, bicycle facilities were ranked as the most important built environment factor. Hence, the consistent theme across chapters emphasizes that bicycle facilities, of high quality and protected from motor vehicles, are crucial for urban environments to support cycling.

The two research components that provided opportunities to interact directly with study participants provided a deeper understanding of cyclists‘ experience of the city. The focus group discussions and the personal interviews (when collecting travel routes) provided context that enriched the survey‘s opinion-based and behaviour-based findings. These interactions highlighted differences in opinions between types of cyclists, and also the breadth of experience between individuals within each group. Their quotes and experiences often provided the backdrop to interpret the results of the objective analyses. A simple example is related to air pollution and topography. The factor ―the route is away from traffic noise and air pollution‖ was the top motivator in decisions to cycle (Chapter 3). In retrospect, this factor really contained two components related to the built environment – traffic and pollution. The focus group sessions allowed these two issues to be unpacked, and clarified that traffic was a major influence across all types of cyclists. Air pollution was ranked lower overall, although its impact varied with cycling experience. The discussions uncovered that air pollution in Metro Vancouver was considered to be not bad in comparison with other cities.

7.3. Policy messages and impacts Beyond the purely academic contributions of this dissertation, there was an explicit intention to create outputs that would promote uptake and application of the findings by government, decision- makers, and advocates (objective 5). This dissertation has substantive findings with policy relevance. Each of the research chapters (Chapter 2-6) provides recommendations for policies specific to the findings of that study. Here we summarize the policy implications and describe knowledge exchange activities that have taken this research into action.

118

7.3.1. Policy implications Chapter 2, ―Route preferences among adults in the near market for bicycling‖, provided key direction on how to design the transportation infrastructure to increase cycling modal share. Key messages were:  Preferred routes were those separated from motor vehicle traffic, or, where routes are along the road network, those with enhanced bicycle amenities such as bike lanes and traffic calming.  The presence of on-street vehicle parking made a route type less desirable.  These preferences were especially important to women, adults with children and others who cycle less frequently than the young male population.

Chapter 3, ―Motivators and deterrents of bicycling‖, provided focused direction for policy and planning related to cycling promotion initiatives. Key messages were:  The strongest motivators and deterrents were related to engineering and the environment.  Design features that encourage cycling were: o Cycling routes near beautiful scenery, away from noise and air pollution, separated from heavy and high speed traffic; o Minimum slopes and distances, options to take bikes on transit; o Smooth, non-slip surfaces, free of debris; o Good lighting, lanes marked with reflective paint; and o Safe indoor bike storage.  The main deterrents of cycling were related to safety, either from motorists, or climate- related conditions (rain, ice, darkness).

Chapter 4, ―Built environment influences on route selection for bicycle and car travel‖, characterized the types of facilities that people detour to use, and provided insight for how dense a network of bike facilities need to be. Key messages were:  Cyclists deviated from shortest routes to take routes with better bicycle facilities, including traffic calming features, bike stencils and signage.  While cyclists detoured, they did not go far – only about 2 city blocks on average. In urban areas where there is a desire to increase the modal share of cycling, the bicycle route network should have high quality facilities spaced every 500 m.

119

Chapter 5, ―Built environment influences on healthy transportation choices‖, highlighted the importance of characteristics of the travel route in decisions to cycle instead of drive. Key messages were:  Bicycle travel was more likely where there were: o Bicycle-specific infrastructure including traffic calming, signage, road markings, and cyclist-activated signals; o Fewer hills; o Fewer arterial roads and highways, and more grid-based street networks (instead of cul-de-sacs); and o Greater land use mix and population density, more land use as neighbourhood commercial, educational and industrial parcels, and less land use as single family residential or large big-box commercial parcels.

Chapter 6, ―Building a bikeability index‖, synthesized this body of research to create a bikeability index that could serve as a planning tool. The output itself is designed to have policy impact by suggesting area-specific strategies to improve conditions for cycling.

7.3.2. Knowledge translation and exchange Both knowledge translation and knowledge exchange models were used to promote the uptake and application of these policy-relevant findings. Knowledge translation can be defined as strategically planned efforts to promote or accelerate the natural diffusion of innovation or knowledge to affect practice, programs or policy. In this regard, countless presentations were made to a host of audiences, from public forums, to academic networks, to city councils and beyond. A website (http://www.cher.ubc.ca/cyclingincities/) created for the full Cycling in Cities program of research (of which mine was one part) was updated regularly with conference presentations, reports, and new manuscripts. Many researchers, practitioners, and advocates discovered our research program through this web presence.

The dissertation also embraced the knowledge exchange model, where the focus is on a two-way dialogue between researchers and users to incorporate the user's needs and thus increase the impact of the findings. Here we highlight several examples of knowledge exchange. First, two brochures were generated to summarize key findings from Chapters 2, 3 and 4 in lay terminology, highlighting key policy messages with pictures, figures, and summary points. The brochures were produced with

120

assistance from Christie Hurrell, the Director of the Centre for Health and Environment Research, and early drafts were evaluated by local advocates and planners for utility and clarity. Knowledge users have since distributed these brochures far beyond the audience that could be reached by our team alone. Second, partners from the regional transportation authority contributed as co-authors on the manuscripts published from Chapters 2, 3 and 4. A main part of their role was the interpretation of results from a practical policy standpoint. This involvement strengthened messaging and uptake. A third example was the bikeability planning tool from Chapter 6. Local planners and transportation engineers were consulted early in the process to understand how such a tool might be used, and later to comment on the index components and scoring. Their insights underscored the need to include the bicycle facility separation component and challenged previous scoring algorithms for topography. The knowledge exchange model was proven in this last example: we have now submitted a grant proposal to develop the tool for 10 cities across the country and received enthusiastic letters of support from municipal staff from almost every city we approached. Finally, the findings of this research were direct inputs into the development of an online bicycle trip planner for the Metro Vancouver Region (www.cyclevancouver.ubc.ca)[104]. Early versions of the tool were pilot tested with local planners and the public and, at the request of the City of Vancouver and TransLink, a modified version of the tool was launched to encourage non-motorized travel during the 2010 Olympic and Paralympic Games.

7.3.3. Impacts to date Certain pieces of evidence from this dissertation have already been used to develop policy and tools that support cycling. Vancouver municipal staff used survey results (Chapters 2 and 3) to gain council approval for a $25-million plan for cycling[162] and to argue specifically for the need for separated bicycle facilities downtown[102, 103]. The UBC TREK program applied the survey evidence and focus group report in development of the Campus Cycling Plan and as guidance for bicycle facility design on campus (Adam Cooper, personal communication, December 14, 2010). Both the survey results and the metrics from Chapters 4 and 5 were informative in shaping methodology for a TransLink-funded Cycling Zone Analysis project to measure cycling environments across the Metro region[163]. The recommended density for bicycle networks (Chapter 4) provided a framework for urban designer Erick Villagomez‘s panel presentation in the public forum hosted by David Byrne, ―Cities, Bicycles and the Future of Getting Around‖. Finally,

121

the results were used in materials for a new Bicycle Facility Design course to train planners and engineers, funded by the BC Recreation and Parks Association.

7.4. Limitations and future endeavors While the section above speaks to the strengths and contributions of this dissertation, a discussion of limitations is needed to ensure that the results are meaningfully interpreted. Specific limitations are mentioned in each of the chapters; overarching concepts are presented here. The challenges faced in the current work suggest pathways for future research.

With 2,149 participants, the Cycling in Cities survey was large relative to other cycling research efforts. However, each chapter used only a subsample for analysis. Chapters 2 and 3 were limited to the 1,402 participants who completed the follow-up study (65% of total). Of the cyclist segments, the regular cyclists were the fewest (n=107), although they were not the main population of interest. The next smallest group were the potential cyclists (n=197). Chapter 5 was based on 3,280 car and bicycle trips made by 1,902 individuals. Chapter 4, initially conceived as a validation study which required de novo data collection, had by far the most limited sample. Results are based on the travel routes of only 74 participants (177 trips). This last sample warrants some discussion. Ultimately, sample size is a function of the data collection method and the funds available. The Cycling in Cities survey collected origin and destination data (similar to other travel diary surveys) but it was not feasible to capture travel routes in the phone interview. The few prior studies that have gathered route data had small samples[59, 113, 120]. Advancing technology is now changing this and offers promise for future projects. Recently, GPS has been used to track individual cyclists[57]. Another project has compiled data from computers mounted on bicycles in the bikeshare program in Lyon, France, providing origin, destination, and travel time for over 11 million trips[164]. A forthcoming technology, ―The Copenhagen Wheel‖, collects data not only on the exerted effort while cycling, but also simultaneously on road conditions, air pollution, noise levels, temperature (http://senseable.mit.edu/copenhagenwheel/). This technology is in prototype phase but to our knowledge has not yet been applied or validated for research purposes.

The results in this dissertation reflect the opinions and travel behaviours of near market cyclists in Metro Vancouver, Canada. It is difficult to assess the generalizability of the findings to populations in other cities. The City of Vancouver proper is considered a cycling-friendly city by North American standards, and has one of the highest cycling rates in Canada[24, 85]. However, to note,

122

the Metro region has a wide variety of urban form and cycling conditions (as illustrated in Chapter 6) and regionally the cycling mode share is on par with many Canadian cities. This suggests that the physical environment in many North American cities may be analogous to some portion of the Metro Vancouver region. The Cycling in Cities Survey had geographically stratified sampling with 31% of the respondents from Vancouver proper (weighted proportion, 664/2149) and 69% from the surrounding municipalities. Compared with respondents from other areas, Vancouver residents were more likely to be regular cyclists (10.9% versus 6.2% of the survey population) and less likely to be potential cyclists (14.8% versus 21.6%), but sub-analyses showed few differences in terms of route type preferences, or factors that motivate and deter cycling. Another opportunity to examine generalizability is via uptake of our study tools elsewhere. Through the course of this research we have had many requests to share our survey tools; while to date this has been a passive effort on our part, it is hoped that our methodologies are adopted elsewhere to enable comparisons across populations.

This research was cross-sectional in nature, and, as always, this brings caveats in terms of causal relationships. The links between the built environment and travel reported in each of these chapters are statistical associations and do not alone provide evidence of causality. Nearly all existing studies of the built environment are cross-sectional[37]. Furthering causal conclusions requires consistency of results across a range of study designs and populations. Considering the evidence in Chapters 2-6 collectively, the fact that findings were consistent across the range of study methodologies employed serves to increase the strength of evidence. Reviews of the built environment literature consistently call for prospective studies to complement the existing body of evidence. Ideally, such study designs would be quasi-experimental or natural experiments that evaluate physical activity levels before-and- after an environmental intervention, or in populations that relocate between areas with differing built environments[6, 36, 37, 165]. These study designs bring additional research challenges[42], but examples exist[165-169]. One particular cycling-related natural experiment to follow is England‘s ―Cycling Demonstration Towns‖, where over £7 million was provided to six towns (populations ~100,000) in 2005 to invest in cycling over a 3 year period[170]. Initial markers of success – a 27% increase in cycling across the towns and a benefit-cost ratio estimated at 3:1 – supported further specific and targeted investment in the 2008 Cycling City and Towns Programme in 2008. The outcomes of these initiatives will provide direction on successful strategies to increase cycling that can be adopted elsewhere.

123

Data inadequacies limited certain lines of inquiry in the current work. Perhaps the most crucial example was related to safety. Safety-related issues were highlighted as major deterrents of cycling in Chapter 3, and came up repeatedly in the focus group discussions. However data limitations meant that safety factors could not be included in the built environment analyses in Chapter 4 and 5. The literature calls for better measurement of safety[37]. Indeed, given appropriate effort and political will, geographical cycling safety data could be improved for Metro Vancouver. Three organizations hold unique data sets related to bicycle safety, but each of the sources has deficiencies from a research perspective (emphasized in italics below). The Insurance Corporation of BC (ICBC) has the locations of all reported cyclist-motor vehicle crashes. Police Traffic Accident Reports capture any crashes that police attended, with location, weather, time-of-day, and road geometry data. The BC Injury Prevention Research Unit has hospitalization data for serious cycling injuries. The geographic locator is the home postal code (3-digit), not the crash location. Privacy concerns restrict the linkage of these datasets. The ICBC data or Police Reports could be mapped to identify hotspot locations for crashes, as has been done for pedestrian safety[41]. To date there has been no effort to do this. Still, such an output would represent numerator information only (i.e., crash events) and not a measure of risk, as it includes no denominator data (the number of cyclists that are exposed to that site). Moreover, the existing data sources do not cover the full realm of cycling safety; even beyond the widespread underreporting of bicycle crashes[154], the local data do not capture adults who visit the emergency room for injuries; less severe crashes or injuries that do not involve hospital, police, or ICBC; or any counts of ―near misses‖ or avoided crashes. Given these formidable limitations with existing administrative data, another alternative is research studies designed to characterize the risk of specific infrastructure. We recently conducted a review of existing studies of this issue and found that we were able to make conclusions only across broad infrastructure categories (e.g., sidewalks more dangerous than on road, bicycle-specific facilities safer than none)[171]. This evidence gap is being addressed in a concurrent study on which I am a co-investigator, Bicyclists‘ Injuries and the Cycling Environment. It is examining the types of cycling routes that are associated with higher or lower rates of injury, using a case-crossover study design to address issues of measuring exposure and confounding by personal and environmental factors[172].

Finally, this dissertation took a multi-faceted approach, considering environmental and demographic factors associated with cycling, but there are complementary questions of interest that need to be addressed by other research projects. This work considered the decision to cycle as the primary

124

outcome (i.e., healthy travel choice), instead of possible outcomes of minutes of cycling/physical activity, sufficient activity to meet recommended guidelines, or downstream health outcomes such as obesity or chronic disease. Second, the mapped analyses in Chapters 4 and 5 used objective (GIS- based) measures only. The correlation and explanatory power of objective versus perceived measures of the built environment is the subject of research elsewhere[173, 174]. Third, the study population included only adults, suitable for the focus on utilitarian travel and mode substitution of bicycle for car. Concurrent research is tackling the escalating physical inactivity and obesity rates in children, where a popular topic is the complex issue of the link between urban form, parents‘ perceptions, and active travel to school behaviours[175, 176]. Fourth, this was a study of actual behaviour, but did not address the task of estimating latent demand for cycling or predicting changes in cycling rates with changes to the built environment, an area of formidable challenge[161]. Finally, this was not a risk-benefit or cost-benefit study, although recent research elsewhere indicates that cycling has a net positive health impact[26] and that investment in cycling has good returns[177].

7.5. Conclusion Prominent cycling researcher John Pucher is leading a study investigating whether a ―cycling renaissance‖ is underway in North American cities. I would argue that the findings of this thesis can direct planners on how to create urban settings that are ripe to capitalize on just such a renaissance. Using this evidence to ensure that the healthy transportation choice is the easy transportation choice can lead to increased physical activity and associated health improvements for the population.

125

REFERENCES

[1] WHO. Transport, Environment and Health; 2000.

[2] Hoek G, Brunekreef B, Goldbohm S, Fischer P, van den Brandt PA. Association between mortality and indicators of traffic-related air pollution in the Netherlands: a cohort study. Lancet. 2002 Oct;360(9341):1203-9.

[3] Brook RD, Franklin B, Cascio W, Hong YL, Howard G, Lipsett M, et al. Air pollution and cardiovascular disease - A statement for healthcare professionals from the expert panel on population and prevention science of the American Heart Association. Circulation. 2004 Jun;109(21):2655-71.

[4] Babisch WF, Beule B, Schust M, Kersten N, Ising H. Traffic noise and risk of myocardial infarction. Epidemiology. 2005 Jan;16(1):33-40.

[5] Center for Disease Control. Physical activity and health: A report of the Surgeon General. Atlanta, GA: CDC; 1996.

[6] Badland H, Schofield G. The built environment and transport-related physical activity: What we do and do not know. J Phys Act Health. 2005;2(4):433-42.

[7] Sallis JF, Cervero RB, Ascher W, Henderson KA, Kraft MK, Kerr J. An ecological approach to creating active living communities. Ann Rev Pub Health. 2006;27:297-322.

[8] Pratt MC, Macera CA, Blanton C. Levels of physical activity and inactivity in children and adults in the United States: current evidence and research issues. Med Sci Sports Exerc. 1999;31(11):S526-S33.

[9] Katzmarzyk PT, Gledhill N, Shephard RJ. The economic burden of physical activity in Canada. CMAJ. 2000;163(11):1435-40.

[10] Raine KD. Overweight and obesity in Canada: A population health perspective. Ottawa, Ontario: Canadian Institute for Health Information; 2004.

[11] Colditz GA. Economic costs of obesity and inactivity. Med Sci Sports Exerc. 1999;31(11):S663-S7.

[12] WHO. Physical activity: Facts and Figures. 2011 http://www.euro.who.int/en/what-we- do/health-topics/disease-prevention/physical-activity/facts-and-figures/10-key-facts-on-physical- activity-in-the-who-european-region. Accessed November 15, 2010.

[13] Warburton DE, Nicol CW, Bredin SS. Health benefits of physical activity: the evidence. CMAJ. 2006 Mar 14;174(6):801-9.

[14] Statistics Canada. Where Canadians work and how they get there. Ottawa, Canada; 2003.

[15] European Commission. Cycling: the way ahead for towns and cities Brussels, Belguim; 1999.

126

[16] Dunn AL, Marcus BH, Kamper JB, Garcia ME, Kohl HW, Blair SN. Comparison of lifestyle and structure interventions to increase physical activity and cardiorespiratory fitness: A randomized control trial. JAMA. 1999;281(4):327-34.

[17] Oja P, Vuori I, Paronen O. Daily walking and cycling to work: their utility as health- enhancing physical activity. Pat Edu Counsel 1998 Apr;33(1 Suppl):S87-94.

[18] Gordon-Larsen P, Nelson MC, Beam K. Associations among active transportation, physical activity, and weight status in young adults. Obesity Res. 2005 May;13(5):868-75.

[19] Andersen LB, Schnohr P, Schroll M, Hein HO. All-cause mortality associated with physical activity during leisure time, work, sports, and cycling to work. Arch Int Med 2000 Jun 12;160(11):1621-8.

[20] Hamer M, Chida Y. Active commuting and cardiovascular risk: A meta-analytic review. Prev Med 2008:9-13.

[21] Hendriksen IJ, Simons M, Garre FG, Hildebrandt VH. The association between commuter cycling and sickness absence. Prev Med. 2010 Aug;51(2):132-5.

[22] Statistics Canada. 2006 Census: Employed labour force by commuting distance (www.statcan.gc.ca) Accessed March 2, 2009.

[23] Pucher J, Dijkstra P. Promoting safe walking and cycling to improve public health: Lessons from the Netherlands and Germany. Am J Pub Health. 2003;93(9):1509-16.

[24] Statistics Canada. 2006 Census: Mode of Transportation (www.statcan.gc.ca) Accessed November 1, 2009.

[25] Jacobsen PL. Safety in numbers: more walkers and bicyclists, safer walking and bicycling. Inj Prev. 2003 September 1, 2003;9(3):205-9.

[26] de Hartog JJ, Boogaard H, Nijland H, Hoek G. Do the Health Benefits of Cycling Outweigh the Risks? Environ Health Perspect. 2010 Aug;118(8):1109-16.

[27] Hillman M. Cycling and the promotion of health. Policy Studies. 1993;14:49-58 [28] Woodcock J, Edwards P, Tonne C, Armstrong BG, Ashiru O, Banister D, et al. Public health benefits of strategies to reduce greenhouse-gas emissions: urban land transport. Lancet. 2009 Dec 5;374(9705):1930-43.

[29] Prochaska JO, Velicer WF. The transtheoretical model of health behavior change. Am J Health Promot. 1997 Sep-Oct;12(1):38-48.

[30] Prochaska JO, Velicer WF, Rossi JS, Goldstein MG, Marcus BH, Rakowski W, et al. Stages of change and decisional balance for 12 problem behaviors. Health Psychol. 1994 Jan;13(1):39-46.

[31] Sallis JF, Owen N, Fisher E. Ecological Models of Health Behavior. In: Glanz K, Rimber B, Viswanath K, eds. Health Behavior and Health Education: Theory, Research, and Practice 4th ed: Jossey-Bass 2008:465-82.

127

[32] Frank LD, Kerr J, Sallis JF, Miles R, Chapman J. A hierarchy of sociodemographic and environmental correlates of walking and obesity. Prev Med. 2008 Aug;47(2):172-8.

[33] Pucher J, Dill J, Handy S. Infrastructure, programs, and policies to increase bicycling: An international review. Prev Med. 2010 Jan;50:S106-S25.

[34] Heath GW, Brownson RC, Kruger J, Miles R, Powell KE, Ramsey LT, et al. The effectiveness of urban design and land use and transport policies and practices to increase physical activity: A systematic review. J Phys Act Health. 2006;3 (Suppl 1):S55-S76.

[35] Handy SL, Boarnet MG, Ewing R, Killingsworth RE. How the built environment affects physical activity - Views from urban planning. American journal of preventive medicine. 2002 Aug;23(2):64-73.

[36] Transportation Research Board & Institute of Medicine of the National Academies. Does the built environment influence physical activity?: Examining the evidence. Washington, DC: National Academies of Sciences 2005.

[37] Saelens BE, Handy SL. Built environment correlates of walking: a review. Med Sci Sports Exerc. 2008 Jul;40(7 Suppl):S550-66.

[38] Brownson RC, Hoehner CM, Day K, Forsyth A, Sallis JF. Measuring the built environment for physical activity: state of the science. Am J Prev Med. 2009 Apr;36(4 Suppl):S99-123 e12.

[39] Frumkin H, Frank LD, Jackson R. Urban sprawl and public health: Designing, plannning and building for health communities. Washington, D.C.: Island Press 2004.

[40] Frank LD, Sallis JF, Conway TL, Chapman JE, Saelens BE, Bachman W. Many pathways from land use to health - Associations between neighborhood walkability and active transportation, body mass index, and air quality. J Am Plan Assoc. 2006 Win;72(1):75-87.

[41] Schuurman N, Cinnamon J, Crooks VA, Hameed SM. Pedestrian injury and the built environment: an environmental scan of hotspots. BMC Public Health. 2009;9:233.

[42] Krizek KJ, Handy SL, Forsyth A. Explaining changes in walking and bicycling behavior: challenges for transportation research. Environ Plann B. 2009 Jul;36(4):725-40.

[43] Translink. Greater Vancouver Trip Diary Survey 2004. Burnaby, British Columbia: Greater Vancouver Transportation Authority; 2005.

[44] Decima Research. City of Toronto 1999 Cycling Study: Final Report on Quantitative Research Results. Toronto: Decima Research Inc.; 2000.

[45] Stinson MA, Bhat CR. Commuter bicyclist route choice - Analysis using a stated preference survey. Transp Res Rec. 2003(1828):107-15.

[46] Titze S, Stronegger WJ, Janschitz S, Oja P. Environmental, social, and personal correlates of cycling for transportation in a student population. J Phys Act Health. 2007 Jan;4(1):66-79.

128

[47] Dill J, Carr T. Bicycle Commuting and Facilities in Major U.S. Cities: If You Build Them, Commuters Will Use Them- Another Look. Transp Res Rec. 2003;1828:116-23.

[48] Nelson AC, Allen D. If you build them, commuters will use them: Association between bicycle facilities and bicycle commuting. Transp Res Rec. 1997;1578:79-83.

[49] Rietveld P, Daniel V. Determinants of bicycle use: do municipal policies matter? Transport Res A-Pol. 2004;38:531-50.

[50] Antonakos CL. Environmental and travel preferences of cyclists. Transp Res Rec. 1994 1994;1438:25-33.

[51] Baltes MR. Factors influencing nondiscretionary work trips by bicycle determined from 1990 U.S. census metropolitan statistical area data. Transp Res Rec. 1996;1538:96-101.

[52] Moritz WE. Survey of North American Bicycle Commuters: Design and Aggregate Results. Transp Res Rec. 1997;1578:91-101.

[53] City of Vancouver Engineering Services. 1999 Bicycle Plan; Reviewing the Past, Planning the Future. Vancouver: City of Vancouver; 1999.

[54] Nankervis M. The effect of weather and climate on bicycle commuting. Transport Res A- Pol. 1999;33:417-31.

[55] Pucher J, Renne JL. Socioeconomics of Urban Travel: Evidence from the 2001 NHTS. Transport Q. 2003;57(3):47-77.

[56] Winters M, Friesen MC, Koehoorn M, Teschke K. Utilitarian bicycling: a multilevel analysis of climate and personal influences. American journal of preventive medicine. 2007 Jan;32(1):52-8.

[57] Dill J. Bicycling for transportation and health: the role of infrastructure. J Public Health Pol. 2009;30 Suppl 1:S95-110.

[58] Krizek KJ. Two Approaches to Valuing Some of Bicycle Facilities‘ Presumed Benefits. J Am Plan Assoc. 2006;72(3):pp 309-20.

[59] Aultman-Hall L, Hall FL, Baetz BB. Analysis of bicycle commuter routes using Geographic Information Systems: Implications for bicycle planning. Transp Res Rec. 1997;1578:102-10.

[60] Hyodo T, Suzuki N, Takahashi K. Modeling of bicycle route and destination choice behavior for bicycle road network plan. Transp Res Rec. 2000;1705:70-6.

[61] Cervero R, Duncan M. Walking, bicycling, and urban landscapes: evidence from the San Francisco Bay Area. Am J Pub Health. 2003 Sep;93(9):1478-83.

[62] Moudon AV, Lee C, Cheadle AD, Collier CW, Johnson D, Schmid TL, et al. Cycling and the built environment, a US perspective. Transport Res D-Tr E. 2005;10:245-61.

[63] Hunt JD, Abraham JE. Influences on Bicycle Use. Transp Policy. 2007;34(4):453-70.

129

[64] Meyer M, Miller E. Urban Transportation Planning. 2nd ed: McGraw-Hill Science 2000.

[65] Alliance for Biking and Walking. Bicycling and Walking in the United States: 2010 Benchmarking Report. Washington, DC; 2010.

[66] Saelens BE, Sallis JF, Frank LD. Environmental correlates of walking and cycling: findings from the transportation, urban design and planning literatures. Ann Behav Med. 2003;25(2):80-91.

[67] Naess P. Urban structure matters: Residential location, car dependence and travel behaviour. New York: Routledge 2006.

[68] Pucher J, Buehler R. Cycling for Everyone Lessons from Europe. Transp Res Rec. 2008(2074):58-65.

[69] Decima Research. Survey for Region of Waterloo Regional Cycling Plan. Toronto: Decima Research Inc.; 2002.

[70] Bureau of Transportation Statistics, US Department of Transportation Statistics. How bike paths and lanes make a difference. BTS Issue Brief No11. 2004.

[71] Garrard J, Rose G, Lo SK. Promoting transportation cycling for women: The role of bicycle infrastructure. Prev Med. 2008 Jul 17;46(1):55-9.

[72] Gatersleben B, Appleton KM. Contemplating cycling to work: Attitudes and perceptions in different stages of change. Transport Res A-Pol. 2007 May;41(4):302-12.

[73] Aultman-Hall L, Hall FL. Research design insights from a survey of urban bicycle commuters. Transp Res Rec. 1998(1636):21-8.

[74] Shannon T, Giles-Corti B, Pikora T, Bulsara M, Shilton T, Bull F. Active commuting in a university setting: Assessing commuting habits and potential for modal change. Transp Policy. 2006 May;13(3):240-53.

[75] Metro Vancouver. 2006 Census Bulletin 3- Population by Age. 2006 http://www.gvrd.bc.ca/growth/pdfs/Census2006_PopbyAge_Bulletin_3.pdf

[76] Translink. Cycling Performance Scorecard. Vancouver: Greater Vancouver Transportation Authority; 2004 June.

[77] Cavill N, Davis A. Cycling and health: What's the evidence? London, England: Cycling England; 2007 August.

[78] Tilahun NY, Levinson DM, Krizek KJ. Trails, Lanes, or Traffic: Valuing Bicycle Facilities with an Adaptive Stated Preference Survey. Transport Res A-Pol. 2007;41(4):287-301.

[79] Katzmarzyk PT. The Canadian obesity epidemic: an historical perspective. Obesity Res. 2002 Jul;10(7):666-74.

130

[80] Colman R, Walker S. The Cost of Physical Inactivity in British Columbia: GPI Atlantic, for B.C. Ministry of Health Planning; 2004.

[81] Dora C. A different route to health: implications of transport policies. BMJ. 1999 Jun 19;318(7199):1686-9.

[82] Sallis JF, Frank LD, Saelens BE, Kraft MK. Active transportation and physical activity: opportunities for collaboration on transportation and public health research. Transport Res A-Pol. 2004 May;38(4):249-68.

[83] Go For Green. National active transportation survey 2004.

[84] Pucher J, Buehler R. Cycling trends and policies in Canadian cities. World Transport Policy & Practice. 2005;11(1):43-61.

[85] Pucher J, Buehler R. Why Canadians cycle more than Americans: A comparative analysis of bicycling trends and policies. Transp Policy. 2006;13:265-79.

[86] Pucher J, Buehler R. Making cycling irresistible: Lessons from the Netherlands, Denmark and Germany. Transport Rev. 2008;28(4):495-528.

[87] Steer Davies Gleave. Cycling in London. London, England: Prepared for: Transport for London; 2008.

[88] Accent Marketing and Research. The near market for cycling in London: Prepared for: Transport for London Surface Transport; 2004.

[89] Environment Canada. Climate Normals [http://www.climate.weatheroffice.ec.gc.ca/climate_normals/index_e.html]. Accessed March 2, 2009.

[90] Winters M, Teschke K. Route preferences among adults in the near market for cycling: Findings of the Cycling in Cities Study. Am J Health Promot. 2010;25(1):40.

[91] FHWA: US Department of Transportation Federal Highway Administration. FHWA Course on Bicycle and Pedestrian Transportation. http://safety.fhwa.dot.gov/PED_BIKE/univcourse/swtoc.htm

[92] Giles-Corti B, Donovan RJ. The relative influence of individual, social and physical environment determinants of physical activity. Soc Sci Med. 2002 Jun;54(12):1793-812.

[93] Sener IN, Eluru N, Bhat CR. An analysis of bicycle route choice preferences in Texas, US. Transportation. 2009 Sep;36(5):511-39.

[94] Stinson MA, Bhat CR. Frequency of bicycle commuting Internet-based survey analysis. Transp Res Rec. 2004(1878):122-30.

131

[95] Troped PJ, Saunders RP, Pate RR, Reininger B, Ureda JR, Thompson SJ. Associations between self-reported and objective physical environmental factors and use of a community rail-trail. Prev Med. 2001 Feb;32(2):191-200.

[96] Schneider RJ, Dunbar LC, Toole JL, Fink C. Avoiding biased interpretation of bicycle surveys: Comparing results from four distribution methods in Winston-Salem, North Carolina. Transp Res Rec. 2006;1982:174-86.

[97] Demographia. World urban areas and pouplation projections, 5th Comprehensive edition. Belleville, Illinois; 2009.

[98] Winters M, Brauer M, Setton E, Teschke K. Built environment influences on healthy transportation choices: bicycling versus driving J Urban Health. In Press.

[99] Winters M, Teschke K, Grant M, Setton E, Brauer M. How far out of the way will we travel? Built environment influences on route selection for bicycle and car travel. . Transp Res Rec. In press.

[100] Dill J, Gliebe J. Understanding and Measuring Bicycle Behavior: A Focus on Travel Time and Route Choice, OCREC-RR-08-03, Center for Urban Studies/Center for Transportation Studies. Portland, Oregon: Oregon Transportation Research and Education Consortium,; 2008.

[101] Walker L, Tresidder M, Birk M, Weigand L, Dill J. Fundamentals of Bicycle Boulevard Planning & Design. Portland, Oregon: Initiative for Bicycle and Pedestrian Innovation, Center for Transportation Studies, Centre for Urban Studies, Portland State University; 2009.

[102] City of Vancouver Engineering Services. Downtown Separated Bike Lane Connection to Burrard Bridge and Seawall, RTS No.:08741; 2010 September 27, 2010.

[103] City of Vancouver Engineering Services. Separated Bike Lanes in Downtown, RTS No.:08181; 2010 January 21, 2010.

[104] Su JG, Winters M, Nunes M, Brauer M. Designing a route planner to facilitate and promote cycling in Metro Vancouver, Canada. Transport Res A-Pol. 2010 Aug;44(7):495-505.

[105] Prochaska JO, Redding CA, Evers KE. The transtheoretical model and stages of change. In: Glanz K, Rimer BK, Lewis FM, eds. Health behavior and health educaiton: theory, reserach, and practice. San Francisco, CA: Josseey-Bass 2002:99-120.

[106] Statistics Canada. 2006. Complete cumulative profile, including Income and earnings, and Shelter costs, Canada, Provinces, Territories, Census Divisions, Census Subdivisions and Dissemination Areas: Census of Canada, 2006 Data. [http://data.library.ubc.ca/java/jsp/database/production/detail.jsp?id=1057].

[107] Su JG, Brauer M, Buzzelli M. Estimating urban morphometry at the neighborhood scale for improvement in modeling long-term average air pollution concentrations. Atmos Environ. 2008 Nov;42(34):7884-93.

132

[108] Henderson SB, Beckerman B, Jerrett M, Brauer M. Application of Land Use Regression to Estimate Long-Term Concentrations of Traffic-Related Nitrogen Oxides and Fine Particulate Matter. Environ Sci Technol. 2007;41(7):2422-8.

[109] Forsyth A, D'Sousa E, Koepp J, Oakes JM, Schmitz KH, Riper DV, et al. Environment and Physical Activity: GIS Protocols Version 4.1. University of Minnesota and Cornell 2007.

[110] BC Assessment. 2006. http://www.bcassessment.bc.ca/products/index.asp. Accessed May 2, 2008.

[111] Hess PM, Moudon AV, Logsdon MG. Measuring land use patterns for transportation research. Land Development and Public Involvement in Transportation 2001:17-24.

[112] Song Y, Rodriguez DA. The measurement of the level of mixed land uses: A synthetic approach: Chapel Hill, NC; 2005.

[113] Howard C, Burns EK. Cycling to work in Phoenix: Route choice, travel behavior, and commuter characteristics. Transp Res Rec. 2001;1773:39-46.

[114] Krizek KJ, El-Geneidy A, Thompson K. A detailed analysis of how an urban trail system affects cyclists' travel. Transportation. 2007 Sep;34(5):611-24.

[115] Winters M, Davidson G, Kao D, Teschke K. Motivators and deterrents of bicycling: comparing influences on decisions to ride Transportation. 2010;DOI: 10.1007/s11116-010-9284-y.

[116] Forester J. The bicycle transportation controvery. Transport Q. 2001;55(2):7-17.

[117] Winters M, Cooper A. What Makes a Neighbourhood Bikeable? Reporting on the Results of Focus Group Sessions. Vancouver, BC: University of British Columbia for Translink; 2008.

[118] Krizek KJ. Operationalizing neighborhood accessibility for land use-travel behavior research and regional modeling. J Plan Educ Res. 2003;22:270-87.

[119] McGinn AP, Evenson KR, Herring AH, Huston SL. The relationship between leisure, walking, and transportation activity with the natural environment. Health Place. 2007;13(3):588-602.

[120] Raford N, Chiaradia A, Gil J. Space syntax: The role of urban form in cyclist route choice in central London. UC Berkeley Traffic Safety Center (Paper UCB-TSC-TR-2007-5). 2007:527-40.

[121] Ewing R, Schmid T, Killingsworth R, Zlot A, Raudenbush S. Relationship between urban sprawl and physical activity, obesity, and morbidity. Am J Health Promot. 2003 Sep-Oct;18(1):47-57.

[122] Leyden KM. Social capital and the built environment: the importance of walkable neighborhoods. Am J Pub Health 2003 Sep;93(9):1546-51.

[123] Wendel-Vos GCW, Schuit AJ, De Niet R, Boshuizen HC, Saris WHM, Kromhout D. Factors of the physical environment associated with walking and bicycling. Med Sci Sports Exerc. 2004 Apr;36(4):725-30.

133

[124] Kwan MP. Gender, the home-work link, and space-time patterns of nonemployment activities. Econ Geogr. 1999 Oct;75(4):370-94.

[125] Lovasi G, Moudon A, Pearson A, Hurvitz P, Larson E, Siscovick D, et al. Using built environment characteristics to predict walking for exercise. Int J Health Geogr. 2008;7(1):10.

[126] Cormen TH, Lesiserson CE, Rivest RL, Stein C. Introduction to Algorithms, 2nd ed.: MIT Press and McGraw-Hill 2001.

[127] BC Government. Digital Road Atlas. December 2006 version. http://ilmbwww.gov.bc.ca/crgb/products/mapdata/digital_road_atlas_products.htm. Accessed May 20, 2008.

[128] Oliver LN, Schuurman N, Hall AW. Comparing circular and network buffers to examine the influence of land use on walking for leisure and errands. Int J Health Geogr. 2007;6:41.

[129] Transportation Association of Canada. Geometric Design Guide for Canadian Roads, Part 2. Ottawa, ON. Transportation Association of Canada 1999.

[130] Dill J. Measuring Network Connectivity for Bicycle and Walking. Transportation Research Board Annual Meeting, Compendium of Papers. Washington, DC. 2004.

[131] Tresidder M. Using GIS to measure connectivity: An exploration of issues: Portland State University; 2005.

[132] Dill J, Voros K. Factors affecting bicycling demand - Initial survey findings from the Portland, Oregon, Region. Transp Res Rec. 2007(2031):9-17.

[133] Cervero R, Sarmiento OL, Jacoby E, Gomez LF, Neiman A. Influences of Built Environments on Walking and Cycling: Lessons from Bogota. Int J Sustain Transp. 2009;3(4):203- 26.

[134] Badland HM, Schofield GM, Garrett N. Travel behavior and objectively measured urban design variables: Associations for adults traveling to work. Health Place. 2008 Mar;14(1):85-95.

[135] Forsyth A, Oakes JM, Schmitz KH, Hearst M. Does Residential Density Increase Walking and Other Physical Activity? Urban Stud. 2007 April 1, 2007;44(4):679-97.

[136] Frank L, Schmid T, Sallis J, Chapman J, Saelens B. Linking objectively measured physical activity with objectively measured urban form: findings from SMARTRAQ. Am J Prev Med. 2005;28:117-25.

[137] Brown BB, Yamada I, Smith KR, Zick CD, Kowaleski-Jones L, Fan JX. Mixed land use and walkability: Variations in land use measures and relationships with BMI, overweight, and obesity. Health Place. 2009 Dec;15(4):1130-41.

[138] Ewing R, Handy S, Brownson RC, Clemente O, Wintston E. Identifying and measuring urban design qualities related to walkability. J Phys Act Health. 2006;3(Suppl 1):S223-S40.

134

[139] Titze S, Stronegger WJ, Janschitz S, Oja P. Association of built-environment, social- environment and personal factors with bicycling as a mode of transportation among Austrian city dwellers. Prev Med. 2008 Sep;47(3):252-9.

[140] Krizek KJ, Johnson PJ. Proximity to trails and retail: Effects on urban cycling and walking. J Am Plan Assoc. 2006 Win;72(1):33-42.

[141] de Geus B, De Bourdeaudhuij I, Jannes C, Meeusen R. Psychosocial and environmental factors associated with cycling for transport among a working population. Health Educ Res 2008 Aug;23(4):697-708.

[142] Int Panis L, de Geus B, Vandenbulcke G, Willems H, Degraeuwe B, Bleux N, et al. Exposure to particulate matter in traffic: A comparison of cyclists and car passengers. Atmos Environ. 2010 Jun;44(19):2263-70.

[143] Moudon AV, Lee C. Walking and bicycling: An evaluation of environmental audit instruments. Am J Health Promot. 2003 Sep-Oct;18(1):21-37.

[144] Pikora T, Giles-Corti B, Bull F, Jamrozik K, Donovan R. Developing a framework for assessment of the environmental determinants of walking and cycling. Soc Sci Med. 2003 Apr;56(8):1693-703.

[145] Handy S, Xinyu C, Mokhtarian PL. Self-Selection in the Relationship between the Built Environment and Walking. J Am Plan Assoc. 2006 Winter;72(1):55-74.

[146] Sener IN, Eluru N, Bhat CR. Who Are Bicyclists? Why and How Much Are They Bicycling? Transp Res Rec. 2009;2134:63-72.

[147] Schneider RJ, Rodriguez DA, Young HM. Easy-to-Compute Index for Identifying Built Environments That Support Walking, Paper #06-1489. Transportation Research Board Annual Meeting, Compendium of Papers. Washington, DC. 2006.

[148] Frank LD, Sallis JF, Saelens BE, Leary L, Cain K, Conway TL, et al. The Development of a Walkability Index: Application To the Neighborhood Quality of Life Study. Brit J Sport Med. 2009. doi:10.1136/bjsm.2009.058701.

[149] Lee C, Moudon AV. The 3Ds+R: Quantifying land use and urban form correlates of walking. Transport Res D-Tr E. 2006 May;11(3):204-15.

[150] Marshall JD, Brauer M, Frank LD. Healthy neighborhoods: walkability and air pollution. Environ Health Perspect. 2009 Nov;117(11):1752-9.

[151] Larsen J, El-Geneidy A. Build it but where? A GIS methodology for guiding the planning of new cycling facilities, Paper #10-0514. Transportation Research Board Annual Meeting, Compendium of Papers. Washington, DC 2010.

[152] Voros K, Birk M, Rose M, Geller R, Igarta D, Patterson B. Cycle Zone Analysis: An Innovative Approach to Bicycle Planning, Paper #10-2680. Transportation Research Board Annual Meeting, Compendium of Papers. Washington, DC 2010.

135

[153] Richards C, Lovasi GS, Bader MDM, Quinn J, Weiss C, Purciel M, et al. Neighborhood Bikeability and BMI in NYC. Active Living Research Annual Conference 2010. San Diego, CA 2010.

[154] Aertsens J, de Geus B, Vandenbulcke G, Degraeuwe B, Broekx S, De Nocker L, et al. Commuting by bike in Belgium, the costs of minor accidents. Accid Anal Prev. 2010;42(6):2149-57.

[155] Elvik R, Mysen AB. Incomplete accident reporting: meta-analysis of studies made in 13 countries. Transp Res Rec. 1999;1665:133-40.

[156] Aultman-Hall L, Hall FL. Ottawa-Carleton commuter cyclist on- and off-road incident rates. Accid Anal Prev. 1998 Jan;30(1):29-43.

[157] Landis BW, Vattikuti VR, Brannick MT. Real-Time Human Perceptions: Toward a Bicycle Level of Service Transp Res Rec. 1997;1578:119-26.

[158] Flowerdew R, Manley DJ, Sabel CE. Neighbourhood effects on health: Does it matter where you draw the boundaries? Soc Sci Med. 2008;66(6):1241-55.

[159] Jelinski DE, Wu JG. The modifiable areal unit problem and implications for landscape ecology. Landscape Ecol. 1996 Jun;11(3):129-40.

[160] Horner MW, Murray AT. Excess commuting and the modifiable areal unit problem. Urban Stud. 2002 Jan;39(1):131-9.

[161] Barnes G, Krizek K. Estimating bicycling demand. Transp Res Rec. 2005;1939:45-51.

[162] City of Vancouver Engineering Services. Cycling in Vancouver: Looking Forward to 2010/2011, RTS No.: 08127; 2010 April 30, 2010.

[163] Urban Systems and Alta Planning + Design. Regional Cycling Network Background Study Burnaby, BC.: For Translink; 2009.

[164] Jensen P, Rouquier J-B, Ovtracht N, Robardet C. Characterizing the speed and paths of shared bicycle use in Lyon. Transport Res D-Tr E. 2010;15(8):522-4.

[165] Colabianchi N. Does the built environment matter for physical activity? Current Cardiovascular Risk Reports. 2009;3(4):302-7.

[166] Gibson DM. The Neighborhood Food Environment and Adult Weight Status: Estimates From Longitudinal Data. Am J Pub Health. 2011 January 1, 2011;101(1):71-8.

[167] Evenson KR, Herring AH, Huston SL. Evaluating change in physical activity with the building of a multi-use trail. Am J Prev Med. 2005;28(2):177-85.

[168] Nancy MW, Yizhao Y. Neighborhood Design and Walking: A Quasi-Experimental Longitudinal Study. Am J Prev Med. 2008;34(4):313-9.

136

[169] Hou N, Popkin BM, Jacobs Jr DR, Song Y, Guilkey D, Lewis CE, et al. Longitudinal associations between neighborhood-level street network with walking, bicycling, and jogging: The CARDIA study. Health Place. 2010;16(6):1206-15.

[170] Cycling England. Cycling City and Towns Programme Overview; 2010.

[171] Reynolds CC, Harris MA, Teschke K, Cripton PA, Winters M. The impact of transportation infrastructure on bicycling injuries and crashes: a review of the literature. Environ Health. 2009;8:47.

[172] Harris MA, Reynolds CC, Winters M, Chipman M, Cripton PA, Cusimano MD, et al. Tackling methodological issues of studying bicycling safety: an illustration using the Bicyclists‘ Injuries and the Cycling Environment study. Transportation Research Board Annual Meeting, Compendium of Papers 2011.

[173] McGinn A, Evenson K, Herring A, Huston S, Rodriguez D. Exploring Associations between Physical Activity and Perceived and Objective Measures of the Built Environment. J Urban Health. 2007;84(2):162-84.

[174] McCormack G, Giles-Corti B, Lange A, Smith T, Martin K, Pikora TJ. An update of recent evidence of the relationship between objective and self-report measures of the physical environment and physical activity behaviours. J Sci Med Sport. 2004;7(Supplement 1):81-92.

[175] Ahlport KN, Linnan L, Vaughn A, Evenson KR, Ward DS. Barriers to and Facilitators of Walking and Bicycling to School: Formative Results From the Non-Motorized Travel Study. Health Educ Behav. 2008 April 1, 2008;35(2):221-44.

[176] McMillan TE. Urban Form and a Child's Trip to School: The Current Literature and a Framework for Future Research. J Plan Lit. 2005 May 1, 2005;19(4):440-56.

[177] Cavill N, Kahlmeier S, Rutter H, Racioppi F, Oja P. Economic analyses of transport infrastructure and policies including health effects related to cycling and walking: A systematic review. Transp Policy. 2008;15(5):291-304.

137

APPENDIX 1: FOCUS GROUP REPORT

138

TransLink and the University of British Columbia What Makes a Neighbourhood Bikeable Reporting on the Results of Focus Group Sessions

Prepared by: Meghan Winters , PhD Candidate, School of Population and Public Health, UBC Adam Cooper, M.A. Candidate, School of Community and Regional Planning, UBC November, 2008.

139

Table of Contents Notes to the Reader ...... 140 Executive Summary ...... 141 Background and Purpose ...... 146 Research Objectives ...... 146 Method ...... 147 Regular Cyclists ...... 148 Group Summary ...... 148 Unlikely to Cycle Factors ...... 148 Likely to Cycle Factors ...... 149 Focus Group Discussion on the Factors Affecting the Decision to Ride ...... 149 Other Factors ...... 151 Occasional Cyclists ...... 151 Group Summary ...... 152 Unlikely to Cycle Factors ...... 152 Likely to Cycle Factors ...... 153 Focus Group Discussion on the Factors Affecting the Decision to Ride ...... 153 Other Factors ...... 155 Potential Cyclists ...... 155 Group Summary ...... 156 Unlikely to Cycle Factors ...... 156 Likely to Cycle Factors ...... 156 Focus Group Discussion on the Factors Affecting the Decision to Ride ...... 157 Other Factors ...... 159 Vancouver Area Cycling Coalition ...... 159 Group Summary ...... 159 Unlikely to Cycle Factors ...... 160 Likely to Cycle Factors ...... 160 Focus Group Discussion on the Factors Affecting the Decision to Ride ...... 160 Other Factors ...... 162

Notes to the Reader The reader is reminded that while qualitative research provides a rich source of information in clarifying existing theories, creating hypotheses, and giving direction to future research, the intention in qualitative research is to uncover and explore people’s motivations, perceptions, attitudes, beliefs and feelings, not to count the number of people who demonstrate a particular attribute. Although the participants in this study were drawn from the group(s) in the population from whom we seek answers, they were not chosen on any statistical basis. The findings presented in this report must therefore be considered directional in nature. No statistical inferences should be drawn from the research results.

Participant responses included in this report are quoted verbatim. Because they are verbatim quotes, they have not been edited for content or grammar. Respondents’ names have not been included in this report, in order to protect their anonymity and the confidentiality of their contributions.

140

Executive Summary This report details the results of four focus group sessions conducted by TransLink in partnership with the Cycling in Cities study from the University of British Columbia. The focus groups were conducted in order to gain a deeper understanding of what makes a neighbourhood bikeable. Participants for one session were recruited from the Vancouver Area Cycling Coalition (VACC), while the other three groups were recruited from a pool of survey respondents from a previous Cycling in Cities survey who agreed to be contacted again. Participants in the latter three groups were stratified based on cycling ability; regular cyclists (who cycle at least once a week for transport purposes), occasional cyclists (who cycle less than once a week) and potential cyclists (who have not cycled for transportation purposes in the last year, but were willing to consider cycling in the futre). The results of these sessions will be used to inform other components of the Cycling in Cities Study, a bikeability index, as well as well as all long range cycle planning in Metro Vancouver.

During the four sessions, participants were asked to fill out a questionnaire that asked how eight different factors would increase or decrease their likelihood of cycling for a utilitarian trip. Following this, the moderator discussed these eight factors with them for approximately 45 minutes. Finally, the moderator asked the participants to prioritize the eight factors, by choosing the top three that would influence their decision to cycle. The results of the prioritization exercise were calculated by awarding 3 points to priority # 1, 2 points to priority # 2 and 1 point to priority # 3, giving a total of pool of 135 point to be distributed to the 8 factors.

The results of the prioritization exercise (Table 1) matched well with what was heard in the talk aloud sessions, with a few minor discrepancies. The first two factors - bike routes and traffic - account for 75 points or 55% of the total points that could be awarded during the exercise, suggesting that these two factors have a major influence on how bikeable a neighbourhood is perceived to be. The VACC group did not complete the prioritization exercise, as one member suggested these factors could not be disaggregated. However, the discussion and results of the questionnaire help to illustrate what about the 8 factors are important to cyclists.

The following sections examine the prioritized factors in greater detail, outlining the percentage of participants that indicated on the questionnaire a given aspect of a particular factor would make them more or less likely to cycle for a utilitarian trip, with notes on key points from all discussions. Please note that these percentages are not statistically significant for Metro Vancouver and are meant to demonstrate preferences among the 31 participants. For a detailed examination of each group please see the following sections which outline each group’s questionnaire results, prioritization results and discussion.

141

Table 5: Prioritized Factors

Prioritization 1 2 3 4 5 6 7 8

route -

Factor

Traffic

The Distance

The The Topography The Street Network The Street

routes bike There are

The Population Density The Population Neighbourhood Land Use Land Neighbourhood En The Environment

Points (135 Total)

Priority 1 = 3 points 50 25 17 16 12 9 4 2 Priority 2= 2 points Priority 3 = 3 points

Factor # 1: There are Bike Routes … Characteristics of bike routes that make people more likely to cycle

 Lead to my destination: 94%  Connect with each other: 94%  Have a barrier separating bikes from traffic: 90%  Have bike symbols marked on the pavement: 71%  Along major roads: 52%

Bike routes were of critical importance to all groups, although VACC members were more likely to cycle in places where bike routes were not present. Comments made during the sessions suggest that people feel much safer when cycling on a designated route. This is not true for parts of the region where a route is designated on a map, but no facilities have actually been installed. Participants defined a safe route as a place that; has a physical barrier separating them from traffic, is away from parked cars, is on a street with low traffic volume and away from large rucks and buses. Additionally many participants placed a high value on the aesthetics of the route.

Factor # 2: Traffic … Characteristics of traffic that make people less likely to cycle

 Has many trucks: 87%  Is fast moving: 77%

142

 Has many buses: 74%

Traffic was a major deterrent for cyclists of all skill levels and served as a barrier to entry for potential cyclists who typically had limited experience riding on the road. Many cyclists commented on their fear of having a door open onto them while riding next to parked vehicles and expressed a dislike for routes that sandwiched them between moving and stationary vehicles. The effect of traffic was magnified when riding on hills, as this made cyclist speed even lower compared to vehicles. Only the VACC group had participants that indicated they would be very likely to cycle in heavy traffic conditions.

Factor # 3: The Street Network … Characteristics of the street network that make people more likely to cycle

 Is on a grid mainly with long blocks (well connected): 67%  Is on a grid mainly with short blocks (well connected): 61%  Characteristics of the street network that make people less likely to cycle  Has tunnels: 65%  Has highways running through it: 61%

The street network was important to cyclists as it dictated their ability to maintain momentum during their trip. A common theme across all of the groups was that start and stop situations should be avoided or that cyclist activated intersections should be in higher use. Maintaining momentum was especially important in a commuting or utilitarian trip. Surprisingly, many participants felt that cul-de-sacs offered a safe and direct route, if they offered pedestrian and cyclist linkages. The presence of highways, bridges and tunnels was also important in the decision to cycle. Most cyclists felt they should be avoided if possible. If not possible, cyclists would like to have a route that is wide, physically separated and with a low grade, as this would help them navigate these obstacles with confidence. This is especially true for cyclists with less experience, although as experience is gained these factors become more neutral in the decision to cycle.

Factor # 4: The Topography … Characteristics of topography that make people more likely to cycle

 Is totally flat: 81%  Is moderately hilly: 52%  Characteristics of topography that make people less likely to cycle  Is very hilly: 58%

Topography was most important to cyclists taking trip for commuting purposes, not because of the extra effort required (some even viewed climbs as a personal challenge) but because it meant their trip took longer to complete. Cyclists riding for recreational purposes also suggested that topography provided a physical challenge. Several individuals mentioned that if they could not manage the hill on bicycle, they would walk and push their bicycle up the hill, or that they could use transit. Topography was especially important when put in the context of other factors. Topography combined with traffic, pollution or

143

noise, made it less bearable to the cyclists. Additionally, very steep terrain was a deterrent for cyclists with physical limitations and those who did not have end of trip facilities at their destination.

Factor # 5: The Environment En-route … Characteristics of the environment en-route that make people less likely to cycle

 Has heavy air pollution: 77%  Is very noisy: 58% Although all of the groups had one or two participants who said they would ride in noisy or polluted situations, it was not the preference of any participant. Being exposed to air pollution was a very serious deterrent to cycling. Some indicated that they would go substantially out of their way to avoid heavily polluted roadways, especially frequent commuters and those with health impairments. Noise was less critical as a stand alone factor. However, the combination of noise, air pollution and heavy traffic would create an environment where some cyclists would be overloaded by the number of stimuli they had to pay attention to, making them more likely to make mistakes or have an accident.

Factor # 6: The Distance Characteristics of the distance that make people more likely to cycle

 Is between 3 - 5 km: 74%  Is between 1 - 3 km: 71%  Is under 1 km: 58%  Is over 5 km: 48%

The discussion regarding distances that people were willing to cycle yielded some very interesting and surprising information. In almost every case it was not distance that was important to cyclists, but the time it took to complete a trip. The time quoted most often as being appropriate was approximately 30 minutes of cycling.

Factor # 7: Neighbourhood Land Use … Characteristics of neighbourhood land use that make people more likely to cycle

 Is mainly parks: 84%  Is mainly residential: 77%  Is a mix of the above uses (parks, stores, industrial, residential): 55%  Characteristics of neighbourhood land use that make people less likely to cycle  Is mainly industrial: 45%

The discussion around neighbourhood land use captured the participants desire to cycle in an area that was safe, calm and aesthetically pleasing. For experienced riders land use was not important; getting to their destination safe and quickly was the priority. For inexperienced cyclists, residential neighbourhoods offered an increased perception of safety as they believed traffic to move slower in these areas.

144

Factor # 8: The Population Density … Characteristics of the population density that make people more likely to cycle

 In the area is low (i.e., single family dwellings, large lots): 58%  Characteristics of population density that are neutral in the decision to cycle  In the area is moderate (i.e., 1 - 3 storey apartments and shared homes): 68%  In the area is very high (i.e., mainly high-rise apartments): 61%

The discussions around population density failed to generate much interest within the groups. It was the factors that people associate with increased population density, such as increased traffic (pedestrian, cyclist and vehicle) that people were more concerned about. In general the built form of the area was not important to any of the participants.

Other Important Factors

Safe Storage and End of Trip Facilities: Many of the comments received on the questionnaire were focused on providing safe/secure places to store bicycles at destinations. Additionally, end of trip facilities at work places seem to be critically important in attracting more people to cycle for commuting purposes.

Aesthetic Appeal: Having an attractive route was a major incentive to cycling more. Riders liked park like settings, places with good views or places that made them feel connected to nature in some way.

Road Condition: Multiple participants commented that the shoulders of roads need to be better maintained to encourage greater cycling uptake. The participants felt that currently bicycle lanes have too many potholes, as well as too much garbage and gravel littering the route.

Driver and Cyclist Education: Many participants commented that drivers and cyclists each did not have a clear understanding of the rules of the road. A greater understanding by each group of the actions of the other would make cyclists feel safer on the road.

Wayfinding: Improved signage would ensure cyclists do not get lost when attempting to navigate the bicycle routes. A common complaint was that cyclists wanted to ride and attempted to do so, but ended up lost.

Conclusions There are many actions that TransLink can take in conjunction with the regional municipalities to encourage higher levels of cycling across Metro Vancouver. Some of these actions are only feasible in the long run and will require significant planning and investment. Improving the quality of designated cycling routes and building new routes should be the highest priority for improving cycling mode share across the region. However, there are other actions that can be taken in the near term to move towards a more cycling friendly region. Cyclist and motorist education, improved wayfinding and the provisions of secure storage areas are actions that can be taken now to help encourage the population to cycle more.

145

Background and Purpose TransLink and the University of British Columbia require research on cyclists’ perceptions and preferences to better understand what objective factors (i.e., observable information, such as the prevalence of hills, nature of street network , population density, land-use mix, prevalence of bike routes, air and noise pollution) make neighbourhoods more cycling friendly.

This research will complement a number of past and ongoing studies on the topic, and will serve to fill in a number of informational gaps, as well as to “validate” earlier findings by examining these issues from a qualitative perspective.

In the 2006 Cycling in Cities survey, 73 factors and 16 route types were rated on the extent to which they increased or decreased respondents likelihood of cycling. Factors that more strongly encouraged cycling included: cycling routes away from traffic noise and air pollution; cycling routes separated from traffic; cycling routes on flat terrain; and shorter distances to key destinations. Factors that strongly discouraged cycling included: bridges along the route where cyclists must share a narrow sidewalk; routes with on street parking; and routes with long steep sections. Although the 2006 Cycling in Cities survey included a fairly comprehensive list of factors and routes, several key objective factors were not included such as density and land use, as they are more complex topics difficult to address in a self administered survey.

The focus groups will serve to more qualitatively determine if all key aspects of the built environment that influence cycling choices have been captured, and to investigate in greater depth what aspects of these factors influence the bikeability of Metro Vancouver.

An important outcome of this research will be the development of a bikeability index , which will evaluate the capacity of neighbourhoods to achieve increased cycling mode shares. The bikeability index will allow for the funding of cycling improvements to b allocated across the region in a more systematic manner.

Feedback was sought from three types of cyclists (regular; who cycle at least once per week for transportation, occasional; who cycle at less than once a week and potential; who have not cycled in the last year for transportation purposes) that were identified in the 2006 Cycling in Cities survey, as well as from selected members of the Vancouver Area Cycling Coalition, a local cycling advocacy group.

Research Objectives The objectives of the research are as follows;

To better understand what objective factors make neighbourhoods more conducive to cycling;

To use this information to validate results obtained from objective sources and aid in the development of a bikeability index;

To determine If any key objective measures have been missed between the cycling segments.

146

The emphasis of the research will be on cycling for utilitarian or non-recreation purposes, but may touch upon aspects of recreation cycling, where useful and relevant. During the groups, participants will be asked to complete a brief questionnaire rating the influence of various objective factors on their decision to cycle, and then discuss their ratings in a group discussion led by a professional moderator. For more information please see the moderators guide and bikeability index questionnaire.

Method The focus group sessions were designed to encourage dialogue and interaction, which is useful for uncovering perceptions and issues on which cyclists may hold a range of opinions. Four group sessions were conducted, with ten individuals recruited to each, in order to achieve groups composed of 6-8 individuals with similar cycling habits.

Group participants were recruited by UBC, and provided with a 24 hour recall. Non-VACC participants were recruited from a database of approximately 600 respondents from the 2006 Cycling in Cities survey who indicated they were willing to participate in future research. All participants signed consent forms and were verbally informed that they would be sound and video recorded for review purposes after the completion of the session. UBC and TransLink provide all participants with a $75 incentive upon completion of the session.

The four 1½-hour focus groups were moderated by Adam Di Paula of NRG Research in Vancouver and were held on August 19th and 20th 2008. All sessions were held in NRG Research’s Focus Group Facility, where they were video recorded for future academic purposes.

Each focus group session followed a similar format; an introduction session lasting ~15min, discussion of the various factors which affect the decision to cycle (topography, distance, the environment en route, traffic, the street network, bike routes, land uses and population density) for ~ 45 minutes, followed by a discussion of any missing factors ~15 minutes and ending with a prioritization of the factors ~15 minutes.

147

Regular Cyclists

Sex: 5 male, 3 female

Place of Residence: Burnaby, Langley, Richmond, Tsawwassen North Vancouver, and Vancouver

Age Range and Number of Participants: 35-44 (2) 45-54 (1) 55-64 (3) 65 or older (2)

The Regular Cyclist group represents cyclists in Metro Vancouver who have a high level of confidence cycling in this region. They are commuters, utilitarian cyclists, bicycle racers and recreational cyclists as well. They take at least one trip per week, typically for commuting purposes, although some ride much more frequently than this. Group Summary

Using a scoring system from 1 to 3 - with 1 being the highest - participants were asked to rank the top 3 factors that would influence their decision to cycle trip for a non-recreational purpose. In order to calculate the most important factors across the group, scores of 1 were given 3 points, scores of 2 were given 2 points and scores of 3 were given 1 point. The results of the prioritization exercise are shown below. For a better understanding of the sub-components of the prioritization chart read the sections; likely to cycle factors, unlikely to cycle factors and factors affecting the decision to ride.

Factor Prioritization

Group 1 2 3

The environment en- There are bike routes Traffic Regular Cyclists route (20 points) (10 points) (9 points)

The number next to the factor indicates the number of participants who indicated that the factor would make them unlikely to bicycle. Only those factors that 50% of the group agreed on have been included.

Unlikely to Cycle Factors Traffic has many trucks (8/8)

Traffic has many buses (8/8)

Traffic is fast moving (7/8)

The street network has highways running through it (7/8)

The street network has tunnels (7/8)

The environment en-route has heavy air pollution (6/8)

148

The environment en-route is very noisy (6/8)

The street network has many cul-de-sacs (disconnected) (5/8)

The street network has bridges (5/8)

The topography is very hilly (5/8

Likely to Cycle Factors There are bike routes that lead to my destination (8/8)

There are bike routes that connect with each other (8/8)

Neighbourhood land use is mainly residential (8/8)

Neighbourhood land use is mainly parks (7/8)

There are bike routes that have bike symbols marked on the pavement (7/8)

There are bike routes that have a barrier separating bikes from traffic (7/8)

The topography is totally flat (6/8)

The distance is between 1 - 3 km (6/8)

The distance is between 3 - 5 km (6/8)

The street network is on a grid with mainly long blocks (well connected) (6/8)

Neighbourhood land use is a mix of uses (6/8)

The population density in the area is low (6/8)

The distance is under 1 km (5/8)

The street network is on a grid with mainly short blocks (very connected) (5/8)

There are bike routes along major roads (5/8)

The topography is moderately hilly (4/8)

The distance is over 5 km (4/8)

Neighbourhood land use is mainly stores (4/8)

Focus Group Discussion on the Factors Affecting the Decision to Ride Topography: “If it’s a really hilly one, you might think twice.” Topography does affect decisions to cycle and routing for regular riders. On the questionnaire, five participants indicated that very hilly topography would make them less likely to cycle. In discussion, they said they plan their routes (using

149

bike maps) to avoid the steepest areas of Metro Vancouver. However, if presented with hills once en- route, regular cyclists would do their best to tackle them. “If its too steep then just push it,” commented one participant. Alternatively, one regular cyclist in this group, commented, “I like the hills…. they are not a consideration.”

Distance: “Time is more important.” Echoing the views of the VACC group, the regular cyclists thought about this issue in terms of time, rather than distance. Several participants identified 30 minutes of riding as an appropriate amount of time. One participant indicated that the decision to purchase a new home was based on a 30 minute commute time to work by bicycle.

The Environment: “When I can smell that diesel or gas fumes, I avoid it” In the discussion around air and noise pollution the regular cyclists were more concerned about the conditions responsible for generating high level of pollution (large trucks, high volumes of traffic and the presence of diesel buses) rather than the pollution itself. Some commented they would choose a route that was more aesthetically pleasing, or a route with less traffic, and that they would consider the traffic at the time of day when picking a route. Questionnaire results indicated that the noise and air pollution were extremely important to this group, 6 participants indicated that both of these would affect their route choice.

Traffic: “In Langley you get on the sidewalk, or you get run over.” Traffic was a very important issue for regular cyclists. The presence of trucks and buses on the route was a major deterrent to riding in traffic. Some riders would avoid traffic by taking residential streets to their destination, even if this meant taking a route that required more travel time. This is an important trade off to note, as travel time is more important to riders than distance. Others participants would avoid high traffic situations, unless they were provided with a designated bike lane on the street.

Network: “With curvilinear you’re almost always going to be headed in a direction you don’t want to go.” The consensus reached by the regular cyclists was that a grid network is typically easier to ride on than a curvilinear pattern, unless the latter is marked as a designated bike route. One participant noted that in certain areas of Richmond the city has made an effort to connect cul-de-sacs, which can provide a direct route to destinations away from heavy traffic flow. The preference of this group was for long blocks, allowing the rider to maintain speed. Highway crossings on the route were a deterrent, as participants did not feel confident judging the speeds of cars travelling at 90 - 100 km/hr.

Bridges & Tunnels: “If there is an appropriate lane, I don’t mind crossing bridges.” The regular cyclists had mixed views about travelling on bridges in Metro Vancouver. They agreed that some bridges were well designed for cyclists, while others such as the Burrard Bridge and the Second Narrows Bridge were not comfortable. For this group of cyclists to feel safe on a bridge, the grade must not be too steep and there must be adequate space to make them feel comfortable. When asked what about the Burrard Bridge made them feel uncomfortable, one participant commented that “it is theoretically divided between pedestrians and bikes, but it is too narrow for either of them.”

Bike Routes: “One thing that I find awkward is where I have a designated area on a road, where I’ve got parked cars and a lane of traffic and I’m sandwiched between the two, I find that very unsafe.”

150

Bike facilities were a major issue for this group. While some said they would not ride on anything other designated as a bike route, a few mentioned they felt safer riding as another vehicle in traffic. Several individuals were very concerned with speed- they wanted a very fast route where they did not have to lose momentum down at lights, traffic diverters or paths. All were concerned about the safety of bike routes- mentioning downtown routes- the preference was not to be riding next to parked cars, for fear of drivers pulling out, or opening doors. This group also wanted bicycle specific roads, where only local traffic would be allowed to travel. This group was suggested better facilities and signage was needed on bike routes in the suburbs.

Land Use: “I don’t find the suburbs set up for bicycles. Its not easy to do your everyday chores because of the distances.” The regular cyclists were mixed on their opinions about land use. They prefer to ride in areas with trees to screen them from nearby smog or near parks. However, they also wish to minimize the amount of time it takes them to reach their destination. The more confident riders (a former road racer and a tri-athlete) has less concerns about land use. “To me neighbourhood land use is not important.” If the traffic is not bad, and the pavement is good they will ride through. Other riders wanted to ride through calmer, residential areas, although they still had concerns about visibility through intersections in these areas.

Residential Density: The density of a neighbourhood does not appear to be a factor for regular cyclists. Only one participant indicated that a high density residential area might make them less likely to cycle there. One participant recognized that high density living is often associated with less parking and suggested that “If I couldn’t park my car, I’d be more likely to on a bicycle.”

Other Factors Safety: The main theme that came out of this session was that regular cyclists will ride in many conditions regardless of how safe it is. This does not mean they do not want improved facilities. Safety was a critical factor to this group and was defined in a number of ways. Safety meant improved facilities to store their bicycles, dedicated bicycle lanes, better lighting at night and improved road surfaces.

Bikes on Transit: The commuter cyclists within the group suggested that they would like to have more access to transit with their bicycles, specifically the Sky Train. Several participants wanted to see a dedicated car on the train for cyclists.

Road Condition: Riders who spent time commuting along major streets were upset at the amount of garbage and gravel present at the edge of the bike lane. They felt this made them less safe riding in these spaces, as a portion of their lane was now full of garbage, pushing them closer to cars. Additionally, the regulars were concerned about pot holes, drains and other disturbances to the bike lane that were not designed with cyclists in mind.

Occasional Cyclists

Sex: 4 male, 3 female

151

Places of Residence: Burnaby, North Vancouver, Port Moody, Coquitlam, Vancouver

Age Range and Number of Participants: 25-34 (1) 35-44 (1) 45-54 (3) 65 or older (2)

The occasional cyclists were Metro Vancouver residents who are comfortable riding their bicycles, although they tend to do so in fair weather conditions and for recreational rather than utilitarian purposes (6 of 7 self identified as recreational). Collectively they have many years of cycling experience for recreational, utilitarian and sport purposes. Riders in this group expressed a desire to cycle more frequently, but were held back by barriers to be discussed in more detail. Group Summary

Factor Prioritization

Group 1 2 3

There are bike routes Traffic The street network Occasional Cyclists (14 points) (7 points) (4 points)

Unlikely to Cycle Factors Traffic has many trucks (6/7)

Traffic is fast moving (5/7)

Traffic has many buses (5/7)

The street network has highways running through it (5/7)

The street network has tunnels (5/7)

The environment en-route has heavy air pollution (5/7)

The environment en-route is very noisy (4/7)

Neighbourhood land use is mainly industrial (4/7)

152

Likely to Cycle Factors There are bike routes that have a barrier separating bikes from traffic (7/7)

The distance is between 3 - 5 km (7/7)

Neighbourhood land use is mainly parks (7/7)

Neighbourhood land use is a mix of uses (7/7)

The distance is between 1 - 3 km (6/7)

There are bike routes that lead to my destination (6/7)

There are bike routes that connect with each other (6/7)

There are bike routes that have symbols marked on the pavement (6/7)

The topography is totally flat (5/8)

The distance is under 1 km (5/8)

The street network is on a grid mainly with short blocks (very connected) (5/8)

The street network is on a grid mainly with long blocks (well connected) (5/8)

Neighbourhood land use is mainly residential (5/8)

The distance is over 5 km (4/8)

The street network has bridges (4/8)

There are bike routes along major roads (4/8)

Focus Group Discussion on the Factors Affecting the Decision to Ride Topography: “If I were to ride to work I would prefer a sort of flat or mildly hilly terrain. I don’t want to kill myself before I get to work.” The varied level of cycling experience was evident in this group during the discussion on topography. Some riders found the hills challenging and fun, especially when riding for recreational purposes. When they considered topography from a utilitarian standpoint, there was consensus that having detours around hills or minimizing the impact of topography would be critical for attracting new riders into commuting. They also mentioned the importance of this factor for people with health problems or low energy levels after a day at work.

Distance: “Its time more than distance.” “If I could ride to work from home in a half hour, I wouldn’t think twice about it.” The occasional cyclists commented that a half an hour trip time on a bicycle was a reasonable limit for utilitarian purposes. In the discussion time and distance did not seem to be as important to this group as to the VACC and Regular cyclist groups. Interestingly, the questionnaire revealed that distance was more important in the decision to cycle than the previous two groups.

153

Particularly, distances over 5km, which three of seven participants indicated would negatively influence their decision to ride.

The Environment: “I remember going for bike rides in Tokyo and commuting around there and I would come home and there was black gunk in my nose because I had been riding, and we don’t have that, and that’s a wonderful thing to not have that.” Pollution as a factor in the decision to cycle for the occasional cyclists, did not ignite as much conversation as it did with the VACC and Regular cyclist groups. Comments were made that when riding next to traffic it was bothersome, but they didn’t find pollution in Vancouver very irritating. Some mentioned this is because much of their time cycling was for recreation.

Traffic: “Rush hour is good because traffic is slower, the energy is lower” On the issue of traffic, this group had less to say than the VACC or Regulars, possibly due to their typically recreational cycling allowing less exposure to traffic. The group felt it was a combination of elements rather than just traffic volume that was important to them, including motor vehicle traffic, signage, pedestrians and limited road space. With too many factors to worry about they become anxious and this is when they make mistakes or have accidents. Interestingly, the questionnaire reveals that fast moving traffic or traffic with many trucks and buses is a major disincentive in the decision to cycle

Network: “If I’m commuting I find this much easier (grid network), I feel I can progress through much faster” “If you are able to reduce your downtime by having long blocks, then it is attractive to me.” The occasional cyclists were very interested in the discussion of street networks, and had varied perceptions. Several participants liked he curvilinear pattern, as it reminded them of riding through a park area. Others believed that a curvilinear street pattern could be dangerous, as there is less ‘eyes on the street’, opportunities to get lost in cul-de-sacs and more traffic on these roads (as they are the only option for motorists). Additionally, some felt the grid pattern offered increased safety, as there would be a greater number of people on the street. For commuting purposes it was agreed that long blocks would be advantageous.

Bridges & Tunnels: “Bridges are a nightmare.”“If there’s room it can be a comfortable experience.” “I like the Burrard Bridge simply because it is very clear where bikes must be. Having said that, my wife dislikes it intensely because you are right next to traffic.” Except for one cyclist, the occasional cyclists had a limited amount of experience riding on bridges. The experience was highly variable to them and was very dependent on the presence of a pathway or barrier which made them feel comfortable and protected from traffic. Interestingly, only two participants indicated that bridges would make them less likely to cycle, while five participants felt tunnels would deter them.

Bike Routes: “I’ve seen some bike lanes to Coquitlam Centre and I notice them there and I thought oh that’s great. So I tried them out, but unfortunately the bike lane ended and then I found myself in traffic and I was uncomfortable.” There was strong consensus among the group that bike lanes were the preferred route choice when cycling, however they could use major improvements. Continuity, signage and surface quality were the most important factors. Outside of Vancouver proper, many of the participants felt they might get lost on the bike routes and suggested that signage between jurisdictions

154

and at jog points be improved. “I don’t go enough to know the routes … I like to go on the ones that are well marked, well signed… safer, suited to riding and an attractive choices”

Land Use: “I don’t like riding near stores because people are parking, the doors are opening, its scary!” “I would never consider riding downtown, just driving here was scary enough.” Land use was important to the occasional cyclists. They had a preference for places that were calm, aesthetically pleasing and more residential in character. The occasional cyclists also indicated on the questionnaire that industrial and commercial land uses would negatively affect their decision to cycle for a utilitarian trip, something not mentioned by the VACC and Regular cyclist groups.

Residential Density: “I find it mostly neutral.” In the discussion there was group consensus that residential density was not a factor, nor the height of buildings as long as the route was safe at the ground level. Interestingly, three participants from the occasional cyclist group indicated on the questionnaire that a very high residential density would make them less likely to cycle in an area. This is in contrast to the two previous groups who would cycle in essentially any condition.

Other Factors Safety: “I think if were really going to try to see greater ridership we have to ensure that there is better accommodation so that its safe and people want to use them (bike facilities).” The perception of safety was the primary barrier to cycling more of for the occasional cyclists. If this group of riders could be made to feel more safe cycling, they would likely increase the amount they ride. Safety for this group seemed to mean safe routes, separated from traffic in safe areas of the city, with secure storage spaces at their destinations.

Safe Storage: In addition to having a safe route, this group was very concerned about safe storage spaces. The group perception was that theft of bicycles was high in Vancouver. Five of seven participants made written comments on their questionnaires related to theft and the provision of safe storage spaces, especially at transit hubs and workplaces.

Potential Cyclists

Sex: 3 male, 5 female

Places of Residence: North Vancouver, Coquitlam, Vancouver, University Endowment Lands

Age Range and Number of Participants: 25-34 (1) 35-44 (2) 45-54 (4) 55-64 (1)

Potential cyclists represent Metro Vancouver riders who cycle typically less than once a month to not at all in the last year. They may cycle more for recreation, and many mentioned they cycled for many years in the past. Members of this group would consider cycling more if the right conditions were in place. Health, family and job commitments are barriers to increased levels of cycling.

155

Group Summary

Factor Prioritization

Group 1 2 3 4

There are Bike The Street Network The Topography Traffic Potential Cyclists Routes (13 points) (9 points) (8 points) (16 points)

Unlikely to Cycle Factors Traffic has many trucks (7/8)

Traffic has many buses (7/8)

The environment en-route has heavy air pollution (7/8)

Neighbourhood land use is mainly industrial (7/8)

Traffic is fast moving (6/8)

The distance is over 5 km (6/8)

The topography is very hilly (5/8)

The street network has highways running through it (5/8)

The street network has bridges (5/8)

The street network has tunnels (5/8)

The environment en-route is very noisy (4/8)

Likely to Cycle Factors There are bike routes that lead to my destination (8/8)

156

There are bike routes that connect with each other (8/8)

Neighbourhood land use is mainly parks (8/8)

The topography is totally flat (7/8)

There are bike routes that have a barrier separating bikes from traffic (7/8)

Neighbourhood land use is mainly residential (7/8)

There are bike routes that have bike symbols marked on the pavement (6/8)

The distance is under 1 km (5/8)

The distance is between 1 - 3 km (5/8)

The street network is on a mainly with short blocks (well connected) (5/8)

The topography is moderately hilly (4/8)

The distance is between 3 - 5 km (4/8)

The street network is on a grid mainly with long blocks (4/8)

There are bike routes along major roads (4/8)

The population density in the area is low (4/8)

Focus Group Discussion on the Factors Affecting the Decision to Ride Topography: “I will avoid going up a hill at any cost.” “Hills are a big problem for me because I don’t have the strength or endurance.” Participants in the potential group were polarized in their views towards topography. Those with health impairments were very unlikely to cycle in hilly environment. Alternatively, participants who did not specifically state a health impairment and were in good physical shape commented that hills were not a factor, unless they commuting to work and needed to look professional. Five of the eight participants indicated that very hilly topography would make them unlikely to cycle.

Distance: “It’s how long it takes you to get there, more than distance.” The potential cyclists achieved group consensus that travel time is more important than travel distance. Six of eight participants indicated on their questionnaire that they would be less likely to cycle for trips over 5km, very different than the results of the other three groups. The data all suggests that their trips are short and mainly for recreational purposes. Throughout the session these cyclists had a hard time imagining cycling for any purpose other than recreation.

The Environment: “I wouldn’t want to cycle if it was dirty and smoggy.” “I do check the pollution index. If you’re asthmatic or if you have any lung conditions, biking is good for you, but not in certain times.” Cyclists in this group did not have the same level of experience riding in pollution or noise as the other

157

three groups. This group tends to ride for pleasure, suggesting that they may ride in areas away from environmental nuisances. The results of the prioritization exercise indicate that the environment is of low importance relative to the other seven factors. However, seven of the eight participants indicated that heavy air pollution would deter them from riding and four of the eight indicated that very noisy routes would make them less likely to cycle.

Traffic: “It’s kind of a fact of life, but in an ideal world it would be nice to get rid of it all.” In the decision to cycle, traffic is a major deterrent for potential cyclists. This includes not only moving traffic but parked cars as well. The type of traffic was very important to this group. Roads with heavier bus or truck traffic would make them less likely to cycle. Results from the questionnaire indicated that traffic was the fourth most important factor overall. All of the sub-components of traffic were very important to the group. Seven of the eight participants indicated that traffic with heavy trucks and many buses would make them less likely to cycle, while six of the eight participants indicated that fast moving traffic would make them less likely to cycle.

Network: “Absolutely the (grid) street pattern would be my preference if I was going to work.” The potential cyclists raised some interesting concerns about street network that had not been mentioned by previous groups. Specifically, one participant drew attention to the fact that some utilitarian trips are made by families, so the network should be designed for the safety of families who may be travelling to school or extra curricular activities (i.e., cut-throughs, routes without alleys). Additionally the perception of safety on grid and curvilinear streets was extremely varied. Some felt safer on curvilinear because of the reduced traffic flow, while other disliked it because you could not see the traffic coming. Overall, the network was the second most important factor.

Bridges & Tunnels: “I avoid the Burrard Street Bridge. I’ll take it, but I try to avoid.” In the discussion of bridges and tunnels many of the group members agreed that bridges were a major deterrent to cycling. This group felt exposed to traffic on bridges and worried about conflicts with pedestrians; some would take the aqua bus or sea bus with their bikes in order to avoid having to cycle on a bridge. One bridge that potential cyclists liked was the Number 2 Road bridge in Richmond, which felt safe in this space thanks to the physical barrier. Questionnaire results showed that five of the eight participants would be less likely to if their route had bridges and tunnels.

Bike Routes: “Love them.” “The more you feel separated from traffic, the safer you feel.” The presence of safe bike routes, especially those that allow for continuous travel, is the most important factor for potential cyclists. Some cyclists within the potential group felt terrified at times and were not convinced of the safety of the existing bike lanes in Vancouver, especially when travelling with children. While riding potential cyclists are worried about conflicts with pedestrians and other cyclists. Some group members suggested that if their confidence level could be increased they would feel comfortable riding on the existing routes.

Land Use: I grew up cycling all around my neighbourhood – I can’t imagine letting my kids bike in our neighbourhood now – to nurture generations of cyclists we must focus on kids and safety. Due to time constraints with the potential cyclists group, the moderator did not have an opportunity to discuss the

158

issue of land use in any detail. However, questionnaire results indicated that industrial land uses would negatively influence six of eight potential cyclists likelihood of taking a trip by bicycle. Alternatively, land use that is residential or mainly parks would positively influence the decision to cycle for seven and eight of eight participants, respectively.

Residential Density: Due to time constraints residential density was not discussed by this group. However, questionnaire results are consistent with the other groups and indicate that potential cyclists regard residential density as a neutral factor. The group was split between those who would be likely to cycle under very high, moderate and low residential densities and those who identified it as not a factor in their decision.

Other Factors Safety: For potential cyclists safety seemed to be perceived as having a wide separated lane, preferably with a physical barrier from traffic. This was mentioned over and over by participants as being extremely important to them. This was especially true for one participant who wanted to ride an adult tricycle which is wider than a typical bicycle.

Secure Storage: “I’d leave a bike in the city all the time, so I could ride in the city.” Secure storage again arose as a major factor. Potential cyclists were very worried about the safety of their bicycles, as their bikes represent a major investment to them. All eight participants mentioned the need for secure long term storage on the comments section of their questionnaire.

End of Trip Facilities: “Shower facilities at work” Potential cyclists made it clear that their current employers were not doing enough to promote cycling through the provision of appropriate shower and change facilities in the work place.

Vancouver Area Cycling Coalition

Sex: 6 male, 2 female

Places of Residence: Surrey, Burnaby, Port Moody, New Westminster, Richmond and Vancouver

Age Range and Number of Participants: 35-44(1) 45-54 (4) 55-64(1) 65 or older (2)

The VACC group represents cyclists in Metro Vancouver with a very high level of confidence cycling in this region. They are cycling advocates who participate in local cycling events (such as bike to work month and commuter cycling skills workshops). The participants in this group cycle daily for utilitarian purposes in most weather conditions. Group Summary

The VACC group failed to prioritize the 8 factors on the questionnaire during the session.

159

Unlikely to Cycle Factors The topography is very hilly (6/8),

The Environment en route has heavy air pollution (6/8)

Traffic is fast moving (6/8),

Traffic has many trucks (6/8),

The environment en route is very noisy (4/8),

Likely to Cycle Factors To topography is totally flat (7/8)

There are bike routes that lead to my destination (7/8)

There are bike routes that connect with each other (7/8)

There are bike routes that have a barrier separating bikes from traffic (7/8)

The distance is between 3 - 5 km (6/8)

The distance is over 5 km (6/8)

The street network is on a grid - mainly with long blocks (well connected) (6/8)

The topography is moderately hilly (5/8)

The street network is on a grid - mainly with short blocks (very connected) (4/8)

Neighbourhood land use is mainly residential (4/8)

Neighbourhood land use is mainly parks (4/8)

Neighbourhood land use is a mix of uses (4/8)

The population density in the area is low (4/8)

Focus Group Discussion on the Factors Affecting the Decision to Ride Topography: “I don’t know anybody that likes to ride up hills, I try to keep my altitude level.” Although the group agreed topography was a consideration, there was consensus that other factors (pollution level, traffic level and facility type) play into the route choice. In discussion, it was clear that most people took longer routes to avoid the steepest hills. Concerns regarding steep topography included breathing extra exhaust fumes while climbing, the additional time required to climb hills and the

160

decreased safety associated with travelling next to moving traffic. Topography was very clearly an important factor with several participants commenting they would travel out of their way to avoid the steepest terrain in Vancouver.

Distance: “Time is more of a constraint than distance. Its not that its too far, its just going to take too long.” Distance between origin and destination did not seem to be of particular importance to VACC participants, although they did recognize that for new riders, the distances could be much more important. The focus on this factor shifted to the amount of time it took to get to a destination. Particularly related to distance was the amount of time it took to get prepared for a bike ride. “It takes too long to get the biking gear ready, if its under 1 km I’m not going to bike, I will walk. Its not worth turning into a Martian to bike 1 km. You just go out with an umbrella.”

The Environment: “I’ll grit my teeth and put up with it (pollution). But I hate it and I find it an insult that there is such a disregard for our health generally.” Pollution was a major factor in the decision to ride for VACC members. Participants used strong words, such as hate, to describe how they felt about areas with high noise or air pollution. There was a consensus that if a less polluted route was within close proximity, they would detour to avoid being exposed to pollution. Noisy environments also raised strong negative feelings. “I just hate it- all types of air and pollution- I would go a lot out of my way to avoid it”

Traffic: “The encroaching volume of traffic is rendering the bike lanes inadequate” Traffic is an important issue. Pparticipants would definitely travel on alternative routes to avoid heavy traffic. The impact of traffic was dependant on the quality of bike facility- most VACC riders felt comfortable riding next to traffic, but did have concerns regarding the speed of the traffic relative to their own speed, the volume of traffic and the presence of parked cars along the side of the road were important considerations. Participants would definitely travel on alternative routes to avoid heavy traffic, even if that meant taking a steeper or more circuitous route.

Network: “All gets back to this issue of encountering vehicles” VACC participants realized that the design of the network impacts travel speed, and suggested a preference for the network type that best minimizes down time at intersections and bike-car interactions. Participants felt frustrated by constant starting/stopping at intersections, as this forced them to loose momentum and be exposed to air pollution from idling cars. One participant suggested that streets with cul-de-sacs could be faster than travelling on a grid pattern, especially in Surrey, where these dead end streets are connected for pedestrian and cyclist use only. “The purpose of the trip dictates the importance of the network.”

Bridges & Tunnels: “The bike lane is next to a drop (Burrard Bridge) and if your front wheel goes off that drop, you’re toast!” The issues regarding bridges were mainly about connectivity to bicycle facilities at either end and about perceived rider safety, both from cars and with pedestrians on shared facilities. There was consensus that the Burrard Bridge as it currently exists is not comfortable for cyclists and that it could be a deterrent to getting more people cycling in the region. Safety was the main concern around bridges, especially the rider perception of safety. There was consensus that the Burrard

161

Bridge as it currently exists is not comfortable for cyclists and that it could be a deterrent to getting more people cycling in the region.

Bike Routes: “The bike routes are really key for me. Without those bike routes I would not have been happy to start cycling” (for commuting purposes). VACC members rely on bike routes for daily routing. In terms of improvements, standardization across the region was a main theme. It was noted that the bike route design south of the Fraser River differs from north of the Fraser. Another issue was signage, as it is possible to miss the connections between the existing bike routes, especially outside Vancouver. Traffic calming on bike routes was also of key, adding both traffic circles and diverters, and one participant suggested that bike routes should be fully closed to all non-local traffic.

Land Use: “it doesn’t matter (land use), safety is the paramount factor” Land use (commercial, industrial or residential) was not a key factor for the VACC group when deciding whether to cycle or not and that speed and safety were more important than the type of area they were riding through. However, other said “land use certainly enhances my enjoyment” with many comments that routing is affected by aesthetics, given adequate time.

Residential Density: Due to the amount of time spent on other factors, residential density was not discussed by the VACC group. Examination of the questionnaire’s completed by this group suggests residential density is not a deterrent in their decision to cycle, as all participants rated the various residential densities as “neutral (not a factor)” or “very likely to cycle”.

Other Factors In addition to the 8 factors on the questionnaire VACC members provided the following as additional factors in their decision to cycle. These comments were provided verbally or on in the space for additional comments in the questionnaire.

Safety: “are there eyes on the street?” Is it safe for a middle school child to ride? Can any age or skill level safely use this bike route, without being bothered by traffic?

Weather: “There are a lot of things that could be done to minimize weather factors. Curbs could be painted yellow so that when there is fog out there, you can see the curb coming up.”

Bike Helmets: “A disincentive for a large number of people for riding a bike is the compulsory helmet law, and I’m very much against it. Its amazing how many people will not ride to work because of that.” “If the route is safe, we don’t need one” (a helmet).

Aesthetics: “the beauty of the street makes a difference, even if I’m commuting somewhere, unless I really have to get there fast, I will choose a route that is more attractive. Being in Europe has reminded (me) that our roads are so ugly”

Bike Parking: Many participants commented on their questionnaire that additional secure bike parking facilities are needed. A downfall of the current system is that they are not available for short term use.

162

Factor Prioritization: Members of this group failed to prioritize the various factors presented to them in the questionnaire. One participant suggested that it would be impossible to prioritize these factors as they are too inter-related. This comment caused the rest of the participants to decide that they too would not be able to prioritize the factors.

One participant commented that “I think this unlikely to cycle and very likely to cycle might not affect us seasoned cyclists. I think it has a significant impact on the other 80% of the people in the world who are potential bike riders. I know my kids and my wife, there’s no chance they’re going to ride half the places I ride. These (factors) are important, just maybe not for us.”

163