The Professional Geographer

ISSN: 0033-0124 (Print) 1467-9272 (Online) Journal homepage: https://www.tandfonline.com/loi/rtpg20

Identifying “Sensible Locations” for Separated Bike

Lanes on a Congested Urban Road Network: A
Toronto Case Study

Charles M. Burke & Darren M. Scott

To cite this article: Charles M. Burke & Darren M. Scott (2018) Identifying “Sensible Locations”
for Separated Bike Lanes on a Congested Urban Road Network: A Toronto Case Study, The
Professional Geographer, 70:4, 541-551, DOI: 10.1080/00330124.2018.1455518

To link to this article: https://doi.org/10.1080/00330124.2018.1455518

Published online: 17 May 2018.

Submit your article to this journal

Article views: 174

View Crossmark data

Full Terms & Conditions of access and use can be found at

https://www.tandfonline.com/action/journalInformation?journalCode=rtpg20
Identifying “Sensible Locations” for Separated Bike Lanes
on a Congested Urban Road Network: A Toronto Case Study
Charles M. Burke
National University of Singapore

Darren M. Scott
McMaster University

Across North America, separated bike lanes have generated political challenges from drivers concerned that such facilities might
increase their travel times. Toronto is one such city where vocal complaints have limited the development of a separated bike lane
network. To address these issues, a new approach to evaluating the travel-time impact of installing a separated bike lane is
proposed. Aided by a transportation geographic information system (GIS) and a GIS toolkit, we apply a method called the network
robustness index (NRI) to scan the Toronto road network for changes in vehicular travel times. Additionally, we use the approach
to evaluate the potential travel-time impact of adding a separated bike lane to a target corridor, one where plans have encountered
political opposition in the past. These two applications demonstrate the value of integrating the NRI approach into cycling
planning—a method that can be used to both identify potential separated bike lane locations and to measure the specific travel-time
impacts of separated bike lanes. Key Words: active travel, bike lane, cycling, network robustness index, Toronto.

在北美, 单专用车道引发了汽车驾驶担忧此般设施可能会增加其行车时间的政治挑战。多伦多正是此般城市之一, 公然的抗

议, 限制了建立单车专用道网络之发展。为了应对这些议题, 本研究提出一个评估设置单车专用道对于行车时间的冲击之崭
新方法。我们以运输地理信息系统 (GIS) 和 GIS 工具包作为辅助, 运用一个名为网络健全性指标 (NRI) 的方法, 审视多伦多
路网在汽机车行车时间的改变。此外, 我们运用该方法, 评估在过往该计画遭遇政治反对之地的目标交通廊道中增加一条单
车专用道的潜在交通影响。这两种应用显示出将 NRI 方法整合进自行车道规划的价值——一种能够同时用来指认潜在单车
专用道地点并测量单车专用道的特定交通时间冲击之方法。 关键词: 主动行车, 单车道, 骑单车, 网络健全性指标, 多伦多。

A trav
es de Norteam
erica, los carriles separados para bicicletas han generado retos políticos de parte de los conductores de carros,
preocupados porque tales facilidades p
ublicas pudiesen incrementar sus tiempos de viaje. Toronto es una de esas ciudades donde
las quejas vocales han limitado el desarrollo de una red de carriles separados para bicicletas. Para abocar estos asuntos se propone
un nuevo enfoque en el modo de evaluar el impacto que podría tener la instalaci
on de un sistema de carriles separados para
bicicletas sobre el tiempo de viaje. Con la ayuda de un sistema de informaci
on geogr
afica (SIG) para el transporte y un juego de
herramientas SIG, aplicamos un m
etodo denominado índice de robustez de red (NRI) para escanear los cambios de tiempos de
viaje vehicular en la red vial de Toronto. Adem
as, usamos el enfoque para evaluar el impacto potencial en el tiempo de viaje al
a~
nadir un carril separado para bicicletas a un corredor determinado, uno donde planes al respecto hayan encontrado en el pasado
oposici
on política. Estas dos aplicaciones demuestran el valor de integrar el enfoque NRI en la planificaci
on ciclista—un m
etodo
que puede usarse para identificar localizaciones de carriles separados para bicicleta y para medir los impactos específicos de
carriles separados para bicicleta sobre el tiempo de viaje. Palabras clave: carril de bicicleta, ciclismo, índice de robustez de
red, Toronto, viaje activo.

My number one priority is tackling traffic and transit in both directions from other road traffic (see Figure 1).
congestion. Adding separated bike lanes in sensible As a result, driver concerns over increasing urban travel
locations will give cyclists more alternatives to get times and the development of low-stress cycling networks
around the city quickly and more safely. are frequently at odds, a conflict that has fostered political
—Toronto Mayor John Tory (2014) challenges to cycling plans across North America (Fed-
eral Highway Administration [FHWA] 2015). Recogni-

I n the minds of drivers commuting across congested

urban road networks every weekday, “sensible loca-
tions” to remove capacity from their preferred routes
tion that those challenges exist has led the U.S. FHWA
to urge planners to incorporate road capacity evaluations
as part of their overall separated bike lane plans. Despite
likely do not exist. At the same time, installing a separated this recognition, though, a useful framework to evaluate
bike lane on a roadway often requires a full lane of capac- the impact of reallocating road capacity to bikes has yet
ity to implement, physically separating cyclists traveling to emerge from the planning guidelines.

The Professional Geographer, 70(4) 2018, pages 541–551 © 2018 by American Association of Geographers.
Initial submission, March 2017; revised submissions, July and October 2017; final acceptance, October 2017.
Published by Taylor & Francis Group, LLC.
542 Volume 70, Number 4, November 2018

Figure 1 A separated bike lane on Cannon Street, Hamilton, Ontario, Canada. Reproduced with permission from Burke
and Scott (2016). (Color figure available online.)

When the current mayor of Toronto, John Tory, capacity and then recalculates a new travel time after
precedes his commitment to separated bike lanes by some or all of the capacity from a link has been
stating that his first priority is tackling traffic conges- removed. The process is then repeated by the Caliper
tion, we know that in Toronto, like many North Amer- Script toolkit, again and again, until a capacity loss,
ican municipalities, driver concerns will play a role in travel-time impact estimate is produced for every link
his support of those facilities. For politicians, drivers in the road network (the method and the toolkit are
make up the majority of their constituents and there- described in detail in the “Data and Methods” sec-
fore driver concerns often can define what is considered tion, under the heading “NRI Methodology”).
“sensible” when public resources are allocated to bikes. In the past, the NRI has been used in planning to
Geographic information systems (GIS) have been prioritize highway construction projects in Vermont
used by urban planners for many years, stemming (Novak, Sullivan, and Scott 2012). This approach has
from their ability to maintain inventories of the many not yet been directed toward cycling plans as part of
components comprising urban environments. More a real-world case study, however. Fortunately though,
recently, GIS has been used within planning depart- because installing a separated bike lane is theoretically
ments to facilitate the planning of cycling infrastruc- and operationally similar to removing a lane of capac-
ture. As of October 2017, a Google Scholar search of ity for motor vehicle traffic, the NRI approach can
the keywords “GIS” AND “Cycling” AND “Plan- be used to measure the potential travel-time impact
ning” yielded nearly 30,000 results, of which about of a separated bike lane. Once that impact is under-
12,000 articles have been published in the past five stood, planners can then evaluate which locations are
years alone. The approach proposed in this article to “sensible” should that concept’s meaning be subject
evaluate cycling networks can assist a professional to the politics of driver concern.
geographer working within a planning department It should be pointed out before continuing with
under the political constraints of prioritizing motor this approach that the entire process of cycling plan-
vehicles. This approach, borrowed from the literature ning is more complex than just accounting for a
concerning critical link analysis, is known as the facility’s impact on vehicular travel time. One must
network robustness index (NRI), a method first devel- recognize, though, that those impacts often do play a
oped by Scott et al. (2006). The NRI approach meas- role in whether or not a facility is built. Therefore,
ures the impact on vehicular travel time attributed to by integrating the NRI approach into a larger cycling
reduced road capacity. It is implemented as a Caliper planning framework, a planner is provided more
ScriptÒ toolkit within TransCADÒ , a powerful trans- information to improve decision making and address
portation GIS. Using traffic simulations, the tool first driver concerns moving forward. That integration can
calculates travel time for a network at full road be achieved in at least two ways.
 “Sensible Locations” for Separated Bike Lanes 543

First, a planner can begin by using the NRI toolkit methodology, and the different road configurations
to perform a full network scan, measuring the poten- possible following a one-lane loss on an arterial road
tial travel-time impact that a one-lane loss has on link. The section that follows is separated into two
each road link to identify all of the locations where subsections, each demonstrating a way to incorporate
travel time changes are considered acceptable. In the NRI as an evaluation method into the cycling
theory, because bicycle travel is only prohibited on planning process. Finally, we offer a brief summary
highways, every road in the network could possibly and suggestions for future research.
act as the location of a cycling facility. This option
presents one of many ways to narrow the entire net-
work to fewer possibilities, especially if driver con- Motivation
cerns have limited development in the past. After the
full scan, further analysis of the remaining locations The Deputy Mayor’s Roundtable on Traffic Conges-
can take into account many other criteria to priori- tion estimates that the average commute for Toronto
tize projects for installation. office workers is forty-two minutes and based on that
Another potential fit is to begin a separated bike average Toronto commuters as a whole spend one
lane plan with an in-depth analysis of a target location additional day each week stuck in traffic (City of Tor-
and then afterward measure the potential travel-time onto 2014). Congested travel speeds in Toronto’s
impact of installing a separated bike lane along that urban core and the nearby surrounding suburbs can
corridor. Using the NRI as a final step between plan- fall to as low as 40 km/h on average during peak peri-
ning and implementation allows for the initial selec- ods (Sweet, Harrison, and Kanaroglou 2015). Overall,
tion process to be made based on cyclists’ needs and the traffic congestion problem in the region is esti-
other site-specific attributes in the absence of driver mated to cost the regional economy at least $6 billion
concerns but still provides a planner with the ability to (Canadian) each year (C. D. Howe Institute 2013).
address those possible challenges should they arise. Frustration over traffic congestion has slowed the
The aim of this study is to demonstrate both poten- growth of Toronto’s urban cycling network, where
tial uses of the NRI as a technique to evaluate the opposition to separated bike lanes has been particularly
travel-time impact of installing separated bike lanes. limiting. The City’s current network consists of just
First, the toolkit is used to conduct a full scan of the 18.7 km of separated bike lanes (City of Toronto 2017).
Toronto road network, identifying each link where a New York City, on the other hand, has installed 157 km
one-lane removal is free of negative travel-time conse- of separated bike lanes—40.6 km of which were added
quences. Conversely, this analysis also identifies loca- in 2017 (New York City Department of Transportation
tions where a one-lane loss does affect driver travel 2017). At times, Toronto City Councilors have called
times. Both of these results are mapped for display. Sec- separated bike lane plans “controversial” and character-
ond, the NRI approach is used to measure the potential ized the City’s past experience with implementing them
travel-time impact of adding a separated bike lane a “disaster” (Grant 2010). Perhaps the most clear politi-
across a target corridor, the Bloor–Danforth. Recently, cal opposition to urban bike lanes came from former
part of this corridor was chosen by the City of Toronto Toronto Mayor Rob Ford, who once stated, “Roads are
for a separated bike lane pilot project in its 2016 built for buses, cars and trucks. Not for people on bikes”
Cycling Network Plan (City of Toronto 2017). To pro- (Mahoney 2010). In 2011, Ford successfully spearheaded
duce an impact estimate of the entire corridor, one lane the removal of a separated bike lane from downtown, cit-
of capacity is removed from each road link along it at ing driver complaints over perceived travel-time
once, rather than use the toolkit to iteratively remove increases along the corridor (Flack 2011).
capacity from each individual link. This aggregate cor- Although former Mayor Ford embodies Toron-
ridor approach measures the impact of adding a sepa- to’s political opposition to on-road bicycle facilities,
rated bike lane to the whole stretch of road at once. one particular arterial corridor stands as the culmi-
The remainder of this article is structured as fol- nation of the political stagnation that opponents
lows. The next section provides a brief background have created in the City over time. The Bloor–
concerning the political challenges that separated bike Danforth corridor has been discussed as a possible
lane installations have created in Toronto, experiences target for a separated bike lane in Toronto City
that, in part, motivate our selection of the city as a Hall for an incredible forty years (see Figure 2).
case study. We then describe some of the most recent The first studies began in 1976, then again in
cycling planning considerations with a special focus 1992, 2010, and 2013. Yet by the end of 2015, no
on a common rule presented in planning guides to definitive action had been taken (Davis 2015). In
select bicycle facilities based on a road’s traffic speed 2016, however, a small section of the Bloor–Dan-
and volume. In this section, a critical review of this forth corridor was selected as a separated bike lane
rule illustrates the potential for a new approach to pilot project (CBC News 2016). Should the results
balance bicycle facility and traffic planning needs with of this pilot prove “sensible” to drivers and the cur-
the NRI. We next provide a synopsis of the data and rent mayor, John Tory, the addition represents a
methods used in this study. This includes a descrip- tremendous opportunity to develop a low-stress
tion of the Toronto road network trip data, the NRI urban cycling network in Toronto. Given that
544 Volume 70, Number 4, November 2018

Figure 2 The Bloor–Danforth corridor and its location spanning Toronto’s downtown core. (Color figure available online.)

separated bike lanes have not survived political multicriteria analysis to account for a facility’s impact
challenges in the past, though, every possible analy- on cyclist’s safety, demand, and accessibility (Larsen,
sis of the potential impact a separated bike lane Patterson, and El-Geneidy 2013; Lovelace et al. 2017).
might have on corridor traffic is important to the Geographers have contributed to multicriteria planning
project’s future. The potential travel-time impact methods by adding location-specific variables like ride
of adding a separated bike lane across the Bloor– topology, junction density, and network centrality to
Danforth corridor is explored in the “Results and the decision-making process (Rybarczyk and Wu 2010;
Discussion” section, under the heading “NRI Eval- Milakis and Athanasopoulos 2014). Despite the fact
uation of an East–West Separated Bike Lane that these recent techniques attempt to account for
Corridor.” every factor to generate a comprehensive cycling plan-
ning framework, none of them consider the role driver
impacts can play as a barrier to implementation.
Cycling and Traffic In its recent guide, the FHWA recognizes that
evaluating road capacity for all users should be a “top
Most cycling planning frameworks are rightly built with priority” for planners, one driven by the political
the needs of cyclists at the center. The latest techniques challenges separated bike lanes have created in many
to aid a planner developing a cycling network use cities in North America (FHWA 2015, 47). The only

Figure 3 Bike facility preselection nomograph. A similar chart is found in Ontario Traffic Manual Book 18: Cycling Facili-
ties (Ministry of Transportation Ontario 2014, 30). Reproduced with permission from Burke and Scott (2016).
 “Sensible Locations” for Separated Bike Lanes 545

specific direction that many authorities’ guidelines safety especially align. As a result, they might provide the
offer a planner relative to vehicle traffic, however, is a best opportunity for easy implementation as the seeds of
simple “rule of thumb” chart used to select the type a planner’s larger cycling network.
of bicycle facility to employ. This chart, shown in
Figure 3, appears in transport authority manuals in
Data and Methods
Canada, Denmark, Australia, and New Zealand, as
well as guides from agencies like Sustrans in the
United Kingdom and CROW in the Netherlands Toronto Road Network and Trip Data
(Ministry of Transportation Ontario 2014, 30). The
American Association of State Highway and Trans- To conduct an NRI evaluation of a road network, an
portation Officials (AASHTO) uses similar rules for analyst must have access to a transportation GIS with
facility guidance, but presents them in tabular rather the ability to perform traffic simulations. This study uses
than graphic format (AASHTO 2012, Tables 2–3). TransCAD software to conduct congested traffic simula-
The chart uses a road’s 85th percentile speed and daily tions using a Wardrop (1952) user equilibrium traffic
traffic volume to guide a planner to choose between vary- assignment algorithm. Performing these simulations in
ing degrees of bicycle facility width. Higher speeds and the GIS requires two key inputs: a network file and an
higher volumes dictate the provision of more space origin–destination (OD) matrix of flows between the
between cyclists and traffic, whereas lower order roads traffic sources and sinks within the study area.
demand less separation. When cyclists’ safety is the main A TransCAD road network for the Greater Toronto
objective, this rule makes perfect sense. As traffic speed and Hamilton Area (GTHA) was developed from data
and volume rise, more separation between traffic and obtained from Desktop Mapping Technologies, Inc.
cyclists should reduce the frequency and severity of acci- (DMTI). In addition to editing some links in the road
dents. Cyclists’ safety should be paramount, but this objec- network to ensure correct network connectivity for traf-
tive is often tempered by the goal of eliminating fic simulations, we added an attribute for design capacity
congestion and maintaining adequate traffic flow. Faced based on the guidance of the Highway Capacity Manual
with these seemingly conflicting aims, a planner might 2000 (Transportation Research Board 2000). Centroids,
purposely avoid reallocating significant capacity away corresponding to the center points of the 5,253 GTHA
from traffic on congested high-volume roads regardless of traffic analysis zones used by the OD matrix as units of
the rule for fear of the potential travel-time consequences. geography, were also added to the network.
What the original NRI study conducted by Scott The OD matrix itself was created from data obtained
et al. (2006) found, however, was that road capacity through the 2006 Transportation Tomorrow Survey
loss does not necessarily result in equal consequences. (TTS), a travel diary survey of household travel patterns in
Viewing a corridor in isolation, a significant removal the GTHA. In the survey, 5 percent of the population
of road capacity from a high-volume link might appear within the GTHA is asked to provide information on each
counterintuitive. This perspective, though, does not trip made by every person eleven years or older in the
account for the potential of the road network to act as household for the previous day. Because a quarter of daily
a system. Traffic is fluid and can reroute onto addi- trips occur during the peak morning commute from 6
tional roads when faced with a change in network a.m. to 9 a.m. (Data Management Group 2014), this study
capacity. This flexibility can increase network travel uses the respondents’ trips over that peak period and uses
time, should that rerouting lead to further congestion; a weighted average to create the one-hour morning peak
create no impact, should capacity on parallel routes OD matrix used in the traffic simulations. Finally, to
absorb additional volume; or, in the case of a Braess account for commercial vehicles that are not included in
paradox, improve network travel time overall, should the survey, a ratio of one truck to every twenty-seven cars
rerouting alleviate bottlenecks elsewhere (Braess, was added to the peak period OD matrix. This corre-
Nagurney, and Wakolbinger 2005). As a result of these sponds to the ratio of cars to trucks on the road estimated
potential outcomes, the NRI approach can be used to from provincial registration data by Statistics Canada
identify “sensible locations” for separated bike lanes, (2009). It should be noted that the estimate of trucks con-
even on roads where traffic volume is high. tributing to traffic congestion on Toronto’s arterials and
Scott et al.’s (2006) NRI study indexed results by list- local roads is a ballpark figure, based on available informa-
ing them from greatest impact to least impact. The tion (detailed information on truck volumes or movements
authors were primarily concerned with a road’s criticality within the city is not readily available).
in case of its failure due to extreme weather or a deliber-
ate attack. The novelty of applying this approach to
cycling planning stems from the value created by knowl- NRI Methodology
edge of the least impactful locations. Roads with enough The NRI was developed by Scott et al. (2006) to mea-
connected capacity to allow for a low-consequence or sure how critical a link is to overall traffic flow through
even consequence-free capacity reduction can be used in a road network. Although they defined the NRI of a
this context for some purposeful advantage—installing link as the change in total travel time attributed to the
separated bike lanes.1 These low-impact locations are rerouting of traffic through a network given the com-
ones where the dual priorities of traffic flow and cyclist plete disruption or removal of that link—a 100 percent
546 Volume 70, Number 4, November 2018
capacity
loss—the
NRI
of
a
link
can
be
calculated
for
and can also be normalized to better communicate the
any
level
of
reduced
capacity,
as
demonstrated
in
later
estimate to the general public. Normalization in this study
work
by
Sullivan
et
al.
(2010).
In
this
article,
we
for- is achieved by dividing the NRI value by the number of
mally
generalize
the
NRI
to
denote
this
flexibility: trips on the link where the capacity removal occurs (i.e.,
the road where the separated bike lane would be imple-
NRIar D cra ¡ c; (1) mented). Whereas the total change in travel time mea-
sured by the NRI is a network measure, this normalized
value represents the potential impact on drivers at the epi-
where NRIar is the value of the index (change in sys- center of impact. Therefore, this normalized value likely
tem-wide travel time) for link a when its capacity is represents the maximum impact of a capacity removal on
reduced by r, which is a value greater than 0 percent, driver travel time, should delay be assumed to be felt
but less than or equal to 100 percent; c is the system- greatest at the location of the loss.
wide travel time when all links in the network are Mathematically, this expression of the normalized
operating at full capacity (i.e., base case scenario); and NRI is defined as
cra is the system-wide travel time attributed to the
reduced capacity on link a after traffic has reached a
new equilibrium (i.e., capacity reduction scenario). cra ¡ c
nNRIar D ; (4)
X da
c D ti xi ; (2)
i2I
where nNRIar is the normalized value of the NRI (change
where ti and xi are, respectively, the travel time and in system-wide travel time per link trip) for link a when
traffic flow across link i at equilibrium. I is the set of its capacity is reduced by r, and da is the total travel
all links comprising the road network. demand on link a. In this study, the NRI is measured in
seconds per trip, as we believe that it is easier to commu-
X nicate the impact of a separated bike lane to drivers and
cra D tia;r xa;r
i ; (3) passengers in terms of how it might affect them person-
i2I/a ally. One should note that the NRI does not estimate the
potential mode shift from car to bike that might follow
where tia;r and xa;r
i are, respectively, the travel time and the installation of a bike lane. Therefore, the NRI results
traffic flow across link i when link a’s capacity has should be considered higher end estimates of any poten-
been reduced by r and all traffic has been rerouted tial travel-time impacts that might follow the loss of road
through the network achieving a new equilibrium. capacity due to a bike lane installation.
The key to deriving the NRI of a link is computing
realistic link-level travel times and traffic flows as input to
Equations 2 and 3. In practice, this is accomplished using Installing a Separated Bike Lane
a traffic assignment model, such as Wardrop’s (1952) user The level of reduction set for removal in the NRI Calcu-
equilibrium, which is used in this study. The inputs for lator is based on the percentage of road capacity that
such models are a topologically correct road network and equals a loss of one lane or otherwise the amount needed
an OD matrix of vehicular trips for a given time interval, to install a separated bike lane. Arterial roads within the
such as the morning peak period or a day. Toronto road network with posted speeds of 60 km/h or
To automate computation of the NRI for links in a road less range between two and seven lanes in number. Real-
network, a toolkit called the NRI Calculator was devel- locating a lane to a separated bike lane on a two-lane arte-
oped in TransCAD, a powerful GIS for transportation rial (one lane in each direction) would either convert that
applications, using its native programming language, Cali- road to unidirectional or affect parking. These types of
per Script. This software tool is designed for maximum changes are not modeled in this study. Arterials with
flexibility. It first prompts the user for a traffic assignment more than two lanes, however, can still maintain reduced
model that is available within TransCAD. Further, it flow in both directions. Assuming that each lane is set at a
allows the user to specify a capacity reduction value greater standard North American width of twelve feet or 3.7 m
than 0 percent but less than or equal to 100 percent. Using (AASHTO 2011), links that consist of seven lanes of traf-
the input, the tool calculates iteratively the NRI for all links fic would require a 14 percent reduction to dedicate one
in a road network or a subset of links specified by the user of those seven lanes completely to bikes. To approximate
through the tool. In total, the NRI Calculator runs the the impact of installing a separated bike lane, six-lane arte-
chosen traffic assignment model n C 1 times—once for rials are reduced by 17 percent, five-lane arterials by 20
the base case scenario (Equation 2) and once for each link percent, four-lane arterials by 25 percent, and three-lane
identified by the user as warranting investigation under a arterials by 33 percent in the NRI Calculator. When the
given capacity reduction scenario (Equation 3). number of lanes in each direction becomes uneven, the
The NRI Calculator outputs values in hours as an total impact of the capacity loss will be felt by drivers trav-
aggregate travel time for all trips in the network. The eling in the direction of that loss. Figure 4 shows these
NRI can be measured in other units of time, however, changes to arterial roads with differing numbers of lanes.
 “Sensible Locations” for Separated Bike Lanes 547

Figure 4 Arterial road conversions following a separated bike lane installation and the amount of capacity necessary to
facilitate that transition.

Figure 5 Arterial road links where the NRI estimates no travel-time cost for a separated bike lane. NRI D network robust-
ness index. (Color figure available online.)
548 Volume 70, Number 4, November 2018

Figure 6 Arterial road links where the NRI estimates a travel-time cost for a separated bike lane. NRI D network robust-
ness index. (Color figure available online.)

Results and Discussion planner knew a particular impact threshold, much

greater connectivity could be generated. For the pur-
pose of this demonstration, a strict separation of no
Full Network Scan of Separated Bike Lane impact and very impactful links serves to highlight the
Travel-Time Impacts information produced by the full scan.
The NRI Calculator is used to conduct a full network In addition to the mapped results, Table 1 presents
scan of the Toronto network. As mentioned, Toronto’s the top ten locations where separated bike lane
arterial network is made up of roads that range between impacts are found to be the lowest. In Table 2, we
two and seven lanes. Two-lane roads are eliminated present the top ten locations where impacts are the
from the analysis. To measure the potential impact of highest. In all, a total of 1,504 arterial road links were
removing one lane from each arterial type, several scans evaluated by the NRI Calculator scans.
of the NRI Calculator at different levels of capacity loss
(14 percent, 17 percent, 20 percent, 25 percent, and 33
NRI Evaluation of an East–West Separated Bike
percent) were performed. The results of those scans
were matched to arterials based on their number of Lane Corridor
lanes, creating an index of estimated changes in travel In addition to the full network scan, the NRI approach
time following a separated bike lane addition. is used to measure the travel-time impact of adding a
The links where a separated bike lane has no travel-
Table 1 Locations of the top ten least costly NRI esti-
time impact or even improves network travel time—a mates following a one-lane reduction
Braess paradox—are mapped in Figure 5. These are
the ideal locations for a planner to uncover as they Location Total vehicles Normalized NRI
could be considered by both cyclists and drivers to be (street–cross street) affected (sec/trip)
sensible locations for a separated bike lane. Figure 6
Lakeshore Blvd. 1,568 ¡108
maps the road links where a one-lane loss of capacity E.–Woodbine Ave.
does affect travel time. Even the most affected links on Sheppard Ave. 1,980 ¡108
the network carry a travel delay of less than one minute E.–Warden Ave.
Danforth Ave.–Broadview 2,958 ¡76
on average. Drivers and cyclists, too, might consider Ave.
some of these areas sensible, but at this time, we do not Bayview Ave.–York Mills Rd. 2,314 ¡90
know the impact that Toronto drivers are willing to Ellesmere Rd.–Birchmount 2,252 ¡78
Rd.
accept. Most of these impacts are small and might very Steeles Ave. W.–Jane St. 1,303 ¡76
well be palatable. By separating the Toronto road net- Kennedy Rd.–Finch Ave. E. 1,571 ¡76
work into an impact–no impact dichotomy, the maps York Mills Rd.–Bayview Ave. 2,449 ¡76
show the potential to create a cycling network that in Albion Rd.–Islington Ave. 2,422 ¡73
Finch Ave. W.–Yonge St. 3,323 ¡71
many places offers little connectivity. Indeed, many no-
impact links are adjacent to significant-impact links. If a Note: NRI D network robustness index.
 “Sensible Locations” for Separated Bike Lanes 549

Table 2 Locations of the top ten most costly NRI esti- commutes. Should that impact be acceptable, the consul-
mates following a one-lane reduction tation might disarm concerns that later could lead to the
installation’s removal.
Location Total vehicles Normalized NRI
(street–cross street) affected (sec/trip)

Bayview Ave.–Lawrence 797 67

Ave. W. Conclusion
Ellesmere Rd.–Neilson 729 57
Rd. Building separated bike lanes in Toronto has been a
Sheppard Ave. W.–Yonge 4,364 53
St. challenge for decades, one shared by many municipali-
Don Mills Rd.–York Mills 1,741 50 ties across North America. To overcome these chal-
Rd. lenges, driver concerns must be addressed with a
Keele St.–Eglinton Ave. 1,079 47
W.
communicable data-driven travel-time metric. A means
Eglinton Ave. W.–Mount 3,309 39 to achieve this is to use the NRI approach either as a
Pleasant Rd. full scan or as a targeted corridor impact evaluation.
The East Mall–Evans Ave. 961 37
Kingston Rd.–Eglinton 3,524 36
The study reported here demonstrates each
Ave. E. approach to integrating the NRI into cycling plan-
Dixon Rd.–Martin Grove 1,711 35 ning, illustrating the valuable information the results
Rd. might potentially provide. A planner can use the NRI
Kingston Rd.–Lawrence 3,543 34
Ave. E. to identify every potential location for a separated bike
lane and address potential concerns of travel-time
Note: NRI D network robustness index. increases on a targeted corridor. In many cases, the
travel-time impacts following an installation could be
relatively small. Once known, these impact estimates
Table 3 NRI approach cost estimate of installing a sepa-
rated bike lane on the Bloor–Danforth can be used to address and eliminate concern.
The NRI approach to estimating the travel-time
Bloor–Danforth cost evaluation Total Unit impact of a separated bike lane still needs to be vali-
dated. If proven reasonable, these simulated estimates
Total Toronto network peak travel time 119,240 Hours could supersede the need for a pilot project. In many
Bloor–Danforth NRI estimated cost 56 Hours
Normalized NRI 3–5 Sec/trip ways, the approach fulfills a similar purpose of a trial
Bloor–Danforth peak volume 114,560 Vehicles lane, to evaluate impacts, except in this case through
modeling rather than construction. This virtual
Note: NRI D network robustness index.
approach provides several advantages over a real-world
pilot, chief among them avoiding the initial dollar cost
separated bike lane along the span of the entire Bloor– and potential travel-time impacts of constructing a
Danforth corridor (shown in Figure 2). For this evalu- potentially temporary facility and the possible conflicts
ation, the NRI Calculator is not used, as the results of incurred should the pilot project be removed later.
the scan are estimates of individual link impacts and These factors considered, ex-post observation would
therefore do not possess additive properties. provide the most accurate evaluation of a facility, and
Most of the Bloor–Danforth corridor consists of although the NRI approach should provide a reason-
four lanes of traffic, two lanes in each direction. To able impact assessment, it still requires validation. &
measure the potential travel-time impact of adding a
separated bike lane to the entire corridor we reduce
capacity across the Bloor–Danforth on each four- Note
lane stretch by 25 percent of the total road capacity.
1
The estimated impact of that reallocation is pre- We use the NRI approach to identify roads where capacity
sented in Table 3. can be reduced with little impact to aid planners in separated
The total estimated travel-time impact of installing a bike lane installations. This method can be used, though, to
separated bike lane is small, just three to five seconds per help evaluate the impact of any reallocation of road capacity
trip. The range in travel-time impact depends on which away from motor vehicles, including light rail transit, bus
rapid transit, or street pedestrianization.
direction of traffic flow the separated bike lane is added
to along the corridor. Depending on direction, though,
the impact of the installation affects between 38,000 and
76,000 Toronto drivers during the peak period. As a Acknowledgments
result, the nonnormalized NRI value is quite substan-
tial—fifty-six hours. This information can be used to We thank the editor, Dr. Barney Warf, and two anon-
begin a dialogue with those drivers, conveying to them a ymous reviewers for providing insightful comments to
data-driven estimate of the potential impact on their improve our article.
550 Volume 70, Number 4, November 2018
Funding https://www.theglobeandmail.com/news/toronto/bloor-
danforth-bikeway-plan-rolls-on/article4310305/.
This work was supported by the Social Science and Larsen, J., Z. Patterson, and A. El-Geneidy. 2013. Build it.
Humanities Research Council of Canada (752-2015- But where? The use of geographic information systems
1584) and the Natural Science and Engineering in identifying locations for new cycling infrastructure.
Research Council of Canada (RGPIN-2016-06153). International Journal of Sustainable Transportation 7:299–
317.
Lovelace, R., A. Goodman, R. Aldred, N. Berkoff, A.
Literature Cited Abbas, and J. Woodcock. 2017. The Propensity to
Cycle tool: An open source online system for
American Association of State Highway and Transportation sustainable transport planning. Journal of Transport and
Officials (AASHTO). 2011. A policy on geometric design of Land Use 10:505–28.
highways and streets. 6th ed. Washington, DC: American Mahoney, J. 2010. In quotes: Rob Ford and a decade of
Association of State Highway and Transportation Officials. controversy. The Globe and Mail, August 19. Accessed
———. 2012. Guide for the development of bicycle facilities. 4th December 22, 2017. http://www.theglobeandmail.com/
ed. Washington, DC: American Association of State news/toronto/rob-ford-and-a-decade-of-controversy/arti
Highway and Transportation Officials. cle4330595/.
Braess, D., A. Nagurney, and T. Wakolbinger. 2005. On the Milakis, D., and K. Athanasopoulos. 2014. What about
paradox of traffic planning. Transportation Science 39:446–50. people in cycle network planning? Applying
Burke, C. M., and D. M. Scott. 2016. The space race: A participative multicriteria GIS analysis in the case of the
framework to evaluate the potential travel-time impacts of Athens metropolitan cycle network. Journal of Transport
reallocating road space to bicycle facilities. Journal of Geography 35:120–29.
Transport Geography 56:110–19. Ministry of Transportation Ontario. 2014. Ontario traffic
CBC News. 2016. Bloor bike lane pilot one step closer to manual book 18: Cycling facilities. St. Catherines, ON,
reality. CBC News, March 8. Accessed December 22, 2017. Canada: Ministry of Transportation Ontario.
http://www.cbc.ca/news/canada/toronto/bike-lanes-bloor- New York City Department of Transportation. 2017.
street-1.3480907. Protected bicycle lanes. Accessed December 22, 2017.
C. D. Howe Institute. 2013. Cars, congestion and costs: A http://www.nyc.gov/html/dot/downloads/pdf/nyc-pro
new approach to evaluating government infrastructure tected-bike-lanes.pdf.
investment. Accessed December 22, 2017. https://www. Novak, D. C., J. L. Sullivan, and D. M. Scott. 2012. A
cdhowe.org/sites/default/files/attachments/research_pa network-based approach for evaluating and ranking
pers/mixed/Commentary_385_0.pdf. transportation roadway projects. Applied Geography
City of Toronto. 2014. Deputy mayor’s roundtable on traffic 34:498–506.
congestion. Accessed December 22, 2017. http://www.tor Rybarczyk, G., and C. Wu. 2010. Bicycle facility planning
onto.ca/legdocs/mmis/2014/ex/bgrd/backgroundfile- using GIS and multi-criteria decision analysis. Applied
72899.pdf. Geography 30:282–93.
———. 2017. Cycling network status. Accessed December Scott, D. M., D. C. Novak, L. Aultman-Hall, and F. Guo.
22, 2017. http://www1.toronto.ca/wps/portal/contenton 2006. Network Robustness Index: A new method for
ly?vgnextoidDd7c3970aa08c1410VgnVCM10000071d identifying critical links and evaluating the performance
60f89RCRD. of transportation networks. Journal of Transport
Data Management Group. 2014. 2011, 2006, 1996 & 1986 Geography 14:215–27.
travel survey summaries for the Greater Toronto and Statistics Canada. 2009. Too many trucks on the road?
Hamilton Area. Accessed December 22, 2017. http://dmg. Accessed December 22, 2017. http://www.statcan.gc.ca/
utoronto.ca/transportation-tomorrow-survey/tts-demo pub/11-621-m/11-621-m2005028-eng.htm.
graphic-and-travel-summaries/2011travel-summaries-gtha. Sullivan, J. L., D. C. Novak, L. Aultman-Hall, and D. M.
Davis, S. 2015. Bike lanes on Bloor ever elusive for Toronto Scott. 2010. Identifying critical road segments and
planners and cyclists. Toronto Star, October 7. Accessed measuring system-wide robustness in transportation
December 22, 2017. https://www.thestar.com/news/gta/ networks with isolating links: A link-based capacity-
2015/10/07/bike-lanes-on-bloor-ever-elusive-for-toronto- reduction approach. Transportation Research Part A: Policy
planners-and-cyclists.html. and Practice 44:323–36.
Federal Highway Administration (FHWA). 2015. Separated Sweet, M. N., C. J. Harrison, and P. S. Kanaroglou.
bike lane planning and design guide. Washington, DC: 2015. Gridlock in the Greater Toronto Area: Its
FHWA, U.S. Department of Transportation. Accessed geography and intensity during key periods. Applied
December 22, 2017. https://www.fhwa.dot.gov/environ Geography 58:167–78.
ment/bicycle_pedestrian/publications/separated_bikela Tory, J. 2014. Tory to expand Toronto cycling network.
ne_pdg/separatedbikelane_pdg.pdf. Accessed December 22, 2017. http://www.johntory.ca/
Flack, D. 2011. Rob Ford on removing the Jarvis Street bike tory-to-expand-torontos-cycling-network/.
lanes. Accessed December 22, 2017. http://www.blogto. TransCAD (version 6.0). Newton, MA: Caliper Corporation.
com/city/2011/06/rob_ford_on_removing_the_jarvis_ Transportation Research Board. 2000. Highway capacity manual
street_bike_lanes/. 2000. Washington, DC: Transportation Research Board.
Grant, K. 2010. Bloor-Danforth bikeway plan rolls on. The Wardrop, J. G. 1952. Some theoretical aspects of road traffic
Globe and Mail, March 16. Accessed December 22, 2017. research. ICE Proceedings: Engineering Divisions 1:325–62.
 “Sensible Locations” for Separated Bike Lanes 551

CHARLES M. BURKE is a Lecturer in the University DARREN
M.
SCOTT
is
a
Professor
in
the
School
of
Geogra-

Scholars Programme at the National University of Singa- phy
and
Earth
Sciences
at
McMaster
University,
Hamilton,

pore, Singapore 138593. E-mail: [email protected]. His ON
L8S
4K1,
Canada.
E-mail:
[email protected].
His

research interests include transportation geography and pub- research
interests
include
active
transportation,
big
data
(GPS),

lic policy. GIScience,
sustainable
transportation,
and
travel
behavior.