Accident Analysis and Prevention 70 (2014) 92–99
Contents lists available at ScienceDirect
Accident Analysis and Prevention journal homepage: www.elsevier.com/locate/aap
Driver behavior during bicycle passing maneuvers in response to a Share the Road sign treatment Jonathan J. Kay, Peter T. Savolainen ∗ , Timothy J. Gates 1 , Tapan K. Datta 2 Wayne State University, Department of Civil and Environmental Engineering, 5050 Anthony Wayne Drive, Detroit, MI 48202, United States
a r t i c l e
i n f o
Article history: Received 24 June 2013 Received in revised form 19 February 2014 Accepted 11 March 2014 Keywords: Bicycle safety Share the Road sign Centerline rumble strips Bicycle passing event Driver behavior
a b s t r a c t The interaction of motorists and bicyclists, particularly during passing maneuvers, is an area of concern to the bicycle safety community as there is a general perception that motor vehicle drivers may not share the road effectively with bicyclists. This is a particular concern on road sections with centerline rumble strips where motorists are prone to crowd bicyclists during passing events. One potential countermeasure to address this concern is the use of a bicycle warning sign with a “Share the Road” plaque. This paper presents the results of a controlled field evaluation of this sign treatment, which involved an examination of driver behavior while overtaking bicyclists. A series of field studies were conducted concurrently on two segments of a high-speed, rural two-lane highway. These segments were similar in terms of roadway geometry, traffic volumes, and other relevant factors, except that one of the segments included centerline rumble strips while the other did not. A before-and-after study design was utilized to examine changes in motor vehicle lateral placement and speed at the time of the passing event as they relate to the presence of centerline rumble strips and the sign treatment. Centerline rumble strips generally shifted vehicles closer to the bicyclists during passing maneuvers, though the magnitude of this effect was marginal. The sign treatment was found to shift motor vehicles away from the rightmost lane positions, though the signs did not significantly affect the mean buffer distance between the bicyclists and passing motorists or the propensity of crowding events during passing. The sign treatment also resulted in a 2.5 miles/h (4.0 km/h) reduction in vehicle speeds. Vehicle type, bicyclist position, and the presence of opposing traffic were also found to affect lateral placement and speed selection during passing maneuvers. © 2014 Elsevier Ltd. All rights reserved.
1. Introduction In 2011, a total of 677 pedalcyclists were killed and 52,000 were injured in motor vehicle crashes in the United States (NHTSA, 2013). This represents a 9 percent increase in fatalities from 2010 as bicycle safety continues to be an emerging safety issue, particularly given trends that illustrate considerable growth in cycling activity (NHTSA, 2013; Pucher et al., 2011). These changes are reflected by increases in the use of bicycles for commuter purposes (Pucher et al., 2011), as well as a 16-percent increase in the proportion of fatalities involving bicyclists between ages 25 and 64 years since 2001 (NHTSA, 2012).
∗ Corresponding author. Tel.: +1 313 577 9950; fax: +1 313 577 8126. E-mail addresses: [email protected] (J.J. Kay), [email protected] (P.T. Savolainen), [email protected] (T.J. Gates), [email protected] (T.K. Datta). 1 Tel.: +1 313 577 2086. 2 Tel.: +1 313 577 9154. http://dx.doi.org/10.1016/j.aap.2014.03.009 0001-4575/© 2014 Elsevier Ltd. All rights reserved.
One area of concern to the bicycle safety community is the interaction of motorists and bicyclists, particularly during passing maneuvers. Approximately 59 percent of bicycle-involved fatalities occur at non-intersection locations (NHTSA, 2013) and there is a general perception that motor vehicle drivers often do not share the road effectively with bicyclists (Chapman and Noyce, 2012). However, research in this area is limited as a 2006 study notes, “practically nothing is known about what happens when overtaking maneuvers take place” (Walker, 2007). Additionally, a report from North Carolina indicates that “virtually no research has been conducted on rural roadways” where many of these overtaking maneuvers are likely to occur (Carter and Council, 2006). Despite this gap in the knowledge base, collisions during overtaking maneuvers are cited as one of the primary causes of bicyclist fatalities (Transport for London, 2005). Several recent studies have been performed to attempt to discover more about this potentially risky interaction between motor vehicles and bicyclists. One such study employed a naturalistic experiment to gather proximity data from motorists overtaking bicycles on various highways in the United Kingdom (Walker,
J.J. Kay et al. / Accident Analysis and Prevention 70 (2014) 92–99
93
This paper examines the impacts of the combination of a bicycle warning sign (W11-1) with a “Share the Road” plaque (W16-1) on driver behavior while passing bicyclists. This study was completed in conjunction with another effort that investigated the effects of centerline rumble strips on the lateral placement of motor vehicles as they passed bicyclists on a two-lane rural highway (Savolainen et al., 2012). As rumble strips tend to shift motor vehicles away from the centerline, potentially crowding bicyclists, these types of locations represent good candidates for sign installation.
2. Methodology
Fig. 1. Bicycle Warning Sign (W11-1) with “Share the Road” Plaque (W16-1) (FHWA, 2009).
2007). This study showed that vehicles tended to crowd bicyclists in instances where the bicyclists were closer to the motor vehicle traffic stream. Interestingly, motorists tended to provide more lateral buffer space to bicyclists perceived as females than males and less space when bicyclists were wearing helmets (Walker, 2007). A study performed in Taiwan utilized an instrumented bicycle to investigate the factors that affect the initial lateral passing distance, wheel angle, and speed control behaviors (Chuang et al., 2013). Lateral distance during passing events was shown to be smaller for motorcycles than for other motor vehicle types and motor vehicles tended to give more space to female bicyclists than males (Chuang et al., 2013). One notable aspect of this study, which differentiates it from previous work, is that the research considered the periods immediately before and after the overtaking event. A similar study using an instrumented bicyclist was performed in the United Kingdom, which investigated the lateral buffer distance between motor vehicles on highways with and without bicycle lanes (Parkin and Meyers, 2010). The results showed that vehicles tended to provide less lateral space in the presence of a bicycle lane, suggesting that vehicles drive within their marked lanes with less recognition for bicycle traffic in the adjacent lane (Parkin and Meyers, 2010). Another recent study showed that centerline rumble strips reduce the overtaking distance between bicyclists and passing motor vehicles (Savolainen et al., 2012). However, the effects of the rumble strips were less pronounced than other related factors, including: the lateral placement of the bicyclists being passed; the number of bicyclists being passed; the presence of opposing traffic; and the type of motor vehicle involved in the overtaking maneuver (Savolainen et al., 2012). One safety treatment that is aimed specifically at improving interaction between motorists and bicyclists is the bicycle warning sign (MUTCD W11-1), shown in Fig. 1, which can be used in conjunction with a “Share the Road” plaque (MUTCD W161). According to the Manual on Uniform Traffic Control Devices (MUTCD), the bicycle warning sign (W11-1) is intended to warn road users of unexpected entries by bicyclists into the roadway (FHWA, 2009). The “Share the Road” plaque may be added when there is a need to warn motorists about bicyclists traveling along the roadway (FHWA, 2009). Given the relatively low costs for providing such signage, research as to its effects provides valuable evidence to guide subsequent decision-making by road agencies. To date, no research has assessed the impacts of the W11-1 sign on driver behavior, though a similar pedestrian warning sign (W11-2) was previously shown to result in marginal improvements in driver compliance at select locations (Clark et al., 1996).
A series of four field behavioral studies were conducted along two sections of Michigan Highway 109 (M-109), shown in Fig. 2, a two-lane rural highway in the northwestern Lower Peninsula, which serves as a popular bicyclist route. These sections included one location where centerline rumble strips had been installed and another location where rumble strips had not been installed. M109 is unique in that it includes these two consecutive roadway sections, one with and one without centerline rumble strips, separated by approximately 1.1 miles. These segments possess similar geometric and traffic characteristics, including a relatively uniform driver population, creating an optimal setting for a controlled comparison. Each study segment is 0.5 miles (0.8 km) in length and includes 11 ft (3.4 m) wide lanes, a 4 ft (1.2 m) paved shoulder (neither site included shoulder rumble strips), and minimal horizontal or vertical curvature. Separate studies were conducted before and after the installation of the W11-1/W16-1 combination sign at each location. During both the pre- and post-installation studies, data were collected concurrently at the two study locations over a period of approximately 5 h. The first series of studies (prior to sign installation) were conducted on a Saturday in July of 2011. Several weeks later, the signs were installed on each study segment and postinstallation data were collected along both segments on a Saturday in August of 2011. Two signs were installed at each study location (four signs total) as shown in Fig. 2. One sign was placed immediately upstream of each 0.5 miles (0.8 km) study segment in each direction. During the study periods, the weather was clear with temperatures of 81 F (27 ◦ C) and 78 F (26 ◦ C), respectively. The principal objective of this study was to determine the impacts of the “Share the Road” sign combination on the lateral placement and speed of vehicles as they passed bicyclists. Conceptually, the study design is complicated by the fact that several external factors may also affect vehicle lateral placement and speed in such a setting besides the presence (or absence) of the sign treatment. Such factors include the following: • Lateral placement of nearest bicyclist to travel lane – Bicyclists traveling nearer to, or within, the travel lane are expected to result in a greater lateral shifts and lower travel speeds by vehicles in comparison to when bicyclists are positioned farther outside the travel lane. • Number of bicyclists encountered – Bicyclists riding in a group may be more conspicuous or elicit a different response from motorists than a bicyclist riding alone. • Type of motor vehicle – Larger vehicles (e.g., trucks, buses, and recreational vehicles) require more space and, as such, may tend to shift over a lesser distance when encountering a bicyclist. The speed profiles for such vehicles are also likely to differ in comparison to passenger cars. • Presence of opposing traffic – If traffic is present in the opposing lane, vehicles are inhibited from moving laterally and may be forced to slow down while passing or to crowd an adjacent bicyclist.
94
J.J. Kay et al. / Accident Analysis and Prevention 70 (2014) 92–99
Fig. 2. Map of study segment locations.
• Presence of rumble strips – If centerline rumble strips are present, drivers may tend to alter their travel speeds and shift away from the centerline and toward the adjacent bicyclist.
As these factors are a function of the bicyclist and driver population, it would be difficult to evaluate their impacts based solely upon observations under a natural setting. To address this issue, data were collected by observing driver response to bicyclists under a natural setting, as well as in response to bicyclists from the research team under a quasi-experimental setting. The research team participants were instructed to ride in one of three specific lateral positions for a predetermined amount of time through an assigned study segment. In order to assess the lateral placement of each vehicle, a series of four pole-mounted, high-definition video cameras were set up along each of the 0.5 mile (0.8 km) segments. These cameras were mounted on top of 20 ft (6.1 m) telescoping poles that were secured to roadside signposts. Further details of this elevated camera installation, which has been used previously in a series of field studies of road user behavior, have been reported elsewhere (Savolainen et al., 2012). This data collection method provides several advantages as it is completely unobtrusive, involves no interaction with road users, and allows for covert data collection without influencing driver or bicyclist behavior. Another benefit of this study design is that it allows for an assessment of driver behavior upstream and downstream of the passing event, in addition to capturing the effects of various confounding factors, such as the presence of opposing traffic. However, one drawback of this design is that the precision of the video review process is limited to 0.5 ft (0.15 m). Consequently, some imprecision is introduced with respect to the buffer distance estimates, which may mask small differences in comparison to alternate designs (Walker, 2007; Parkin and Meyers, 2010; Chuang et al., 2013). However, some of the other metrics (e.g., proportion of vehicles in right-most lane position) are largely unaffected. During the study periods, pairs of bicyclists from the research team rode continuous loops around each of the study segments. These bicyclists were staggered such that a bicyclist was on each side of the roadway at all times. The ends of each loop were clearly marked on the shoulder in order to provide visual cues for the
bicyclists’ reference during data collection. All bicyclists rode in the prescribed lateral position for approximately 1 h before taking a break and continuing in a different lateral position during the subsequent one-hour period. The bicyclist loops were evenly distributed among three predetermined lateral positions, which included: (1) within the center of the shoulder; (2) on the left edge of the shoulder; and (3) on the right edge of the travel lane. A schematic of the data collection plan is shown in Fig. 3. After completion of the field data collection, project staff reviewed the video data. Each video reviewer completed a classroom-based training program, which provided targeted instructions on data collection and coding procedures. Each data collector independently reviewed pilot videos to ensure internal consistency among the review team. During the subsequent fullscale review, data were randomly checked to ensure consistency among observers. Fig. 4 shows an example screenshot from one video review. A total of 1200 passing events (an event defined as any instance where a motor vehicle passed a bicyclist) were observed during the pre-installation studies and 1225 events were observed postinstallation. During the review of the field videos, the following data were collected to control for some of the potentially confounding factors described previously: • • • •
Number of bicyclists encountered during the passing event; position of the leftmost bicyclist when each event occurred; type of motor vehicle performing the passing maneuver; whether opposing traffic was present at the time of the passing maneuver; • whether the study segment included centerline rumble strips.
Summary statistics for these passing events are provided in Table 1, aggregated by time period (i.e., pre-installation or postinstallation). The aggregate statistics show that most of these factors were relatively consistent between the two study periods. The data distributions for bicyclist position, motor vehicle type, and presence of opposing traffic were not significantly different between the two periods as determined by Kolmogorov–Smirnov tests (P-values of 0.82, 0.17, and 0.28, respectively). Furthermore, the numbers of observations were also approximately equal
J.J. Kay et al. / Accident Analysis and Prevention 70 (2014) 92–99
95
Table 1 Summary statistics for lateral positioning study, aggregated by presence/absence of “Share the Road” sign treatment. Factor
3. Results and discussion This research aimed to assess the impacts of the bicycle warning sign (W11-1) and “Share the Road” plaque (W16-1) on driver behavior while passing bicyclists. In order to examine this issue, four specific measures of effectiveness were examined: • Buffer distance – The buffer distance represents the lateral distance between the motor vehicle and bicyclist at the time a
Total
No.
No.
%
%
%
439 427 169 83 60 22
36.6 35.6 14.1 6.9 5.0 1.8
461 431 166 98 42 27
37.6 35.2 13.6 8.0 3.4 2.2
900 858 335 181 102 49
37.1 35.4 13.8 7.5 4.2 2.0
Bicyclist position Outside edge of travel lane Inside edge of shoulder Center of shoulder
288 397 515
24.0 33.1 42.9
319 361 545
26.0 29.5 44.5
607 758 1060
25.0 31.3 43.7
1137 63
94.8 5.3
1208 17
98.6 1.4
2345 80
96.7 3.3
Centerline rumble strips Not present Present
629 571
52.4 47.6
535 690
43.7 56.3
1164 1261
48.0 52.0
Opposing traffic Not present Present
872 328
72.7 27.3
862 363
70.4 29.6
1734 691
71.5 28.5
Buffer distance 5 ft (1.5 m) or less 5–8 ft (1.5–2.4 m) More than 8 ft (2.4 m)
122 803 275
10.2 66.9 22.9
151 842 232
12.3 68.7 18.9
273 1645 507
11.3 67.8 20.9
Motor vehicle lateral position Center of lane Left edge of lane Touched centerline One wheel over centerline Halfway over centerline
41 254 309 399 197
3.4 21.2 25.8 33.3 16.4
19 298 354 359 195
1.6 24.3 28.9 29.3 15.9
60 552 663 758 392
2.5 22.8 27.3 31.3 16.2
1200
100.0
1225
100.0
2425
100.0
Total
between the two study locations (with and without centerline rumble strips). The post-installation study included a lower percentage of bicyclists riding in groups (P-value
Contents lists available at ScienceDirect
Accident Analysis and Prevention journal homepage: www.elsevier.com/locate/aap
Driver behavior during bicycle passing maneuvers in response to a Share the Road sign treatment Jonathan J. Kay, Peter T. Savolainen ∗ , Timothy J. Gates 1 , Tapan K. Datta 2 Wayne State University, Department of Civil and Environmental Engineering, 5050 Anthony Wayne Drive, Detroit, MI 48202, United States
a r t i c l e
i n f o
Article history: Received 24 June 2013 Received in revised form 19 February 2014 Accepted 11 March 2014 Keywords: Bicycle safety Share the Road sign Centerline rumble strips Bicycle passing event Driver behavior
a b s t r a c t The interaction of motorists and bicyclists, particularly during passing maneuvers, is an area of concern to the bicycle safety community as there is a general perception that motor vehicle drivers may not share the road effectively with bicyclists. This is a particular concern on road sections with centerline rumble strips where motorists are prone to crowd bicyclists during passing events. One potential countermeasure to address this concern is the use of a bicycle warning sign with a “Share the Road” plaque. This paper presents the results of a controlled field evaluation of this sign treatment, which involved an examination of driver behavior while overtaking bicyclists. A series of field studies were conducted concurrently on two segments of a high-speed, rural two-lane highway. These segments were similar in terms of roadway geometry, traffic volumes, and other relevant factors, except that one of the segments included centerline rumble strips while the other did not. A before-and-after study design was utilized to examine changes in motor vehicle lateral placement and speed at the time of the passing event as they relate to the presence of centerline rumble strips and the sign treatment. Centerline rumble strips generally shifted vehicles closer to the bicyclists during passing maneuvers, though the magnitude of this effect was marginal. The sign treatment was found to shift motor vehicles away from the rightmost lane positions, though the signs did not significantly affect the mean buffer distance between the bicyclists and passing motorists or the propensity of crowding events during passing. The sign treatment also resulted in a 2.5 miles/h (4.0 km/h) reduction in vehicle speeds. Vehicle type, bicyclist position, and the presence of opposing traffic were also found to affect lateral placement and speed selection during passing maneuvers. © 2014 Elsevier Ltd. All rights reserved.
1. Introduction In 2011, a total of 677 pedalcyclists were killed and 52,000 were injured in motor vehicle crashes in the United States (NHTSA, 2013). This represents a 9 percent increase in fatalities from 2010 as bicycle safety continues to be an emerging safety issue, particularly given trends that illustrate considerable growth in cycling activity (NHTSA, 2013; Pucher et al., 2011). These changes are reflected by increases in the use of bicycles for commuter purposes (Pucher et al., 2011), as well as a 16-percent increase in the proportion of fatalities involving bicyclists between ages 25 and 64 years since 2001 (NHTSA, 2012).
∗ Corresponding author. Tel.: +1 313 577 9950; fax: +1 313 577 8126. E-mail addresses: [email protected] (J.J. Kay), [email protected] (P.T. Savolainen), [email protected] (T.J. Gates), [email protected] (T.K. Datta). 1 Tel.: +1 313 577 2086. 2 Tel.: +1 313 577 9154. http://dx.doi.org/10.1016/j.aap.2014.03.009 0001-4575/© 2014 Elsevier Ltd. All rights reserved.
One area of concern to the bicycle safety community is the interaction of motorists and bicyclists, particularly during passing maneuvers. Approximately 59 percent of bicycle-involved fatalities occur at non-intersection locations (NHTSA, 2013) and there is a general perception that motor vehicle drivers often do not share the road effectively with bicyclists (Chapman and Noyce, 2012). However, research in this area is limited as a 2006 study notes, “practically nothing is known about what happens when overtaking maneuvers take place” (Walker, 2007). Additionally, a report from North Carolina indicates that “virtually no research has been conducted on rural roadways” where many of these overtaking maneuvers are likely to occur (Carter and Council, 2006). Despite this gap in the knowledge base, collisions during overtaking maneuvers are cited as one of the primary causes of bicyclist fatalities (Transport for London, 2005). Several recent studies have been performed to attempt to discover more about this potentially risky interaction between motor vehicles and bicyclists. One such study employed a naturalistic experiment to gather proximity data from motorists overtaking bicycles on various highways in the United Kingdom (Walker,
J.J. Kay et al. / Accident Analysis and Prevention 70 (2014) 92–99
93
This paper examines the impacts of the combination of a bicycle warning sign (W11-1) with a “Share the Road” plaque (W16-1) on driver behavior while passing bicyclists. This study was completed in conjunction with another effort that investigated the effects of centerline rumble strips on the lateral placement of motor vehicles as they passed bicyclists on a two-lane rural highway (Savolainen et al., 2012). As rumble strips tend to shift motor vehicles away from the centerline, potentially crowding bicyclists, these types of locations represent good candidates for sign installation.
2. Methodology
Fig. 1. Bicycle Warning Sign (W11-1) with “Share the Road” Plaque (W16-1) (FHWA, 2009).
2007). This study showed that vehicles tended to crowd bicyclists in instances where the bicyclists were closer to the motor vehicle traffic stream. Interestingly, motorists tended to provide more lateral buffer space to bicyclists perceived as females than males and less space when bicyclists were wearing helmets (Walker, 2007). A study performed in Taiwan utilized an instrumented bicycle to investigate the factors that affect the initial lateral passing distance, wheel angle, and speed control behaviors (Chuang et al., 2013). Lateral distance during passing events was shown to be smaller for motorcycles than for other motor vehicle types and motor vehicles tended to give more space to female bicyclists than males (Chuang et al., 2013). One notable aspect of this study, which differentiates it from previous work, is that the research considered the periods immediately before and after the overtaking event. A similar study using an instrumented bicyclist was performed in the United Kingdom, which investigated the lateral buffer distance between motor vehicles on highways with and without bicycle lanes (Parkin and Meyers, 2010). The results showed that vehicles tended to provide less lateral space in the presence of a bicycle lane, suggesting that vehicles drive within their marked lanes with less recognition for bicycle traffic in the adjacent lane (Parkin and Meyers, 2010). Another recent study showed that centerline rumble strips reduce the overtaking distance between bicyclists and passing motor vehicles (Savolainen et al., 2012). However, the effects of the rumble strips were less pronounced than other related factors, including: the lateral placement of the bicyclists being passed; the number of bicyclists being passed; the presence of opposing traffic; and the type of motor vehicle involved in the overtaking maneuver (Savolainen et al., 2012). One safety treatment that is aimed specifically at improving interaction between motorists and bicyclists is the bicycle warning sign (MUTCD W11-1), shown in Fig. 1, which can be used in conjunction with a “Share the Road” plaque (MUTCD W161). According to the Manual on Uniform Traffic Control Devices (MUTCD), the bicycle warning sign (W11-1) is intended to warn road users of unexpected entries by bicyclists into the roadway (FHWA, 2009). The “Share the Road” plaque may be added when there is a need to warn motorists about bicyclists traveling along the roadway (FHWA, 2009). Given the relatively low costs for providing such signage, research as to its effects provides valuable evidence to guide subsequent decision-making by road agencies. To date, no research has assessed the impacts of the W11-1 sign on driver behavior, though a similar pedestrian warning sign (W11-2) was previously shown to result in marginal improvements in driver compliance at select locations (Clark et al., 1996).
A series of four field behavioral studies were conducted along two sections of Michigan Highway 109 (M-109), shown in Fig. 2, a two-lane rural highway in the northwestern Lower Peninsula, which serves as a popular bicyclist route. These sections included one location where centerline rumble strips had been installed and another location where rumble strips had not been installed. M109 is unique in that it includes these two consecutive roadway sections, one with and one without centerline rumble strips, separated by approximately 1.1 miles. These segments possess similar geometric and traffic characteristics, including a relatively uniform driver population, creating an optimal setting for a controlled comparison. Each study segment is 0.5 miles (0.8 km) in length and includes 11 ft (3.4 m) wide lanes, a 4 ft (1.2 m) paved shoulder (neither site included shoulder rumble strips), and minimal horizontal or vertical curvature. Separate studies were conducted before and after the installation of the W11-1/W16-1 combination sign at each location. During both the pre- and post-installation studies, data were collected concurrently at the two study locations over a period of approximately 5 h. The first series of studies (prior to sign installation) were conducted on a Saturday in July of 2011. Several weeks later, the signs were installed on each study segment and postinstallation data were collected along both segments on a Saturday in August of 2011. Two signs were installed at each study location (four signs total) as shown in Fig. 2. One sign was placed immediately upstream of each 0.5 miles (0.8 km) study segment in each direction. During the study periods, the weather was clear with temperatures of 81 F (27 ◦ C) and 78 F (26 ◦ C), respectively. The principal objective of this study was to determine the impacts of the “Share the Road” sign combination on the lateral placement and speed of vehicles as they passed bicyclists. Conceptually, the study design is complicated by the fact that several external factors may also affect vehicle lateral placement and speed in such a setting besides the presence (or absence) of the sign treatment. Such factors include the following: • Lateral placement of nearest bicyclist to travel lane – Bicyclists traveling nearer to, or within, the travel lane are expected to result in a greater lateral shifts and lower travel speeds by vehicles in comparison to when bicyclists are positioned farther outside the travel lane. • Number of bicyclists encountered – Bicyclists riding in a group may be more conspicuous or elicit a different response from motorists than a bicyclist riding alone. • Type of motor vehicle – Larger vehicles (e.g., trucks, buses, and recreational vehicles) require more space and, as such, may tend to shift over a lesser distance when encountering a bicyclist. The speed profiles for such vehicles are also likely to differ in comparison to passenger cars. • Presence of opposing traffic – If traffic is present in the opposing lane, vehicles are inhibited from moving laterally and may be forced to slow down while passing or to crowd an adjacent bicyclist.
94
J.J. Kay et al. / Accident Analysis and Prevention 70 (2014) 92–99
Fig. 2. Map of study segment locations.
• Presence of rumble strips – If centerline rumble strips are present, drivers may tend to alter their travel speeds and shift away from the centerline and toward the adjacent bicyclist.
As these factors are a function of the bicyclist and driver population, it would be difficult to evaluate their impacts based solely upon observations under a natural setting. To address this issue, data were collected by observing driver response to bicyclists under a natural setting, as well as in response to bicyclists from the research team under a quasi-experimental setting. The research team participants were instructed to ride in one of three specific lateral positions for a predetermined amount of time through an assigned study segment. In order to assess the lateral placement of each vehicle, a series of four pole-mounted, high-definition video cameras were set up along each of the 0.5 mile (0.8 km) segments. These cameras were mounted on top of 20 ft (6.1 m) telescoping poles that were secured to roadside signposts. Further details of this elevated camera installation, which has been used previously in a series of field studies of road user behavior, have been reported elsewhere (Savolainen et al., 2012). This data collection method provides several advantages as it is completely unobtrusive, involves no interaction with road users, and allows for covert data collection without influencing driver or bicyclist behavior. Another benefit of this study design is that it allows for an assessment of driver behavior upstream and downstream of the passing event, in addition to capturing the effects of various confounding factors, such as the presence of opposing traffic. However, one drawback of this design is that the precision of the video review process is limited to 0.5 ft (0.15 m). Consequently, some imprecision is introduced with respect to the buffer distance estimates, which may mask small differences in comparison to alternate designs (Walker, 2007; Parkin and Meyers, 2010; Chuang et al., 2013). However, some of the other metrics (e.g., proportion of vehicles in right-most lane position) are largely unaffected. During the study periods, pairs of bicyclists from the research team rode continuous loops around each of the study segments. These bicyclists were staggered such that a bicyclist was on each side of the roadway at all times. The ends of each loop were clearly marked on the shoulder in order to provide visual cues for the
bicyclists’ reference during data collection. All bicyclists rode in the prescribed lateral position for approximately 1 h before taking a break and continuing in a different lateral position during the subsequent one-hour period. The bicyclist loops were evenly distributed among three predetermined lateral positions, which included: (1) within the center of the shoulder; (2) on the left edge of the shoulder; and (3) on the right edge of the travel lane. A schematic of the data collection plan is shown in Fig. 3. After completion of the field data collection, project staff reviewed the video data. Each video reviewer completed a classroom-based training program, which provided targeted instructions on data collection and coding procedures. Each data collector independently reviewed pilot videos to ensure internal consistency among the review team. During the subsequent fullscale review, data were randomly checked to ensure consistency among observers. Fig. 4 shows an example screenshot from one video review. A total of 1200 passing events (an event defined as any instance where a motor vehicle passed a bicyclist) were observed during the pre-installation studies and 1225 events were observed postinstallation. During the review of the field videos, the following data were collected to control for some of the potentially confounding factors described previously: • • • •
Number of bicyclists encountered during the passing event; position of the leftmost bicyclist when each event occurred; type of motor vehicle performing the passing maneuver; whether opposing traffic was present at the time of the passing maneuver; • whether the study segment included centerline rumble strips.
Summary statistics for these passing events are provided in Table 1, aggregated by time period (i.e., pre-installation or postinstallation). The aggregate statistics show that most of these factors were relatively consistent between the two study periods. The data distributions for bicyclist position, motor vehicle type, and presence of opposing traffic were not significantly different between the two periods as determined by Kolmogorov–Smirnov tests (P-values of 0.82, 0.17, and 0.28, respectively). Furthermore, the numbers of observations were also approximately equal
J.J. Kay et al. / Accident Analysis and Prevention 70 (2014) 92–99
95
Table 1 Summary statistics for lateral positioning study, aggregated by presence/absence of “Share the Road” sign treatment. Factor
3. Results and discussion This research aimed to assess the impacts of the bicycle warning sign (W11-1) and “Share the Road” plaque (W16-1) on driver behavior while passing bicyclists. In order to examine this issue, four specific measures of effectiveness were examined: • Buffer distance – The buffer distance represents the lateral distance between the motor vehicle and bicyclist at the time a
Total
No.
No.
%
%
%
439 427 169 83 60 22
36.6 35.6 14.1 6.9 5.0 1.8
461 431 166 98 42 27
37.6 35.2 13.6 8.0 3.4 2.2
900 858 335 181 102 49
37.1 35.4 13.8 7.5 4.2 2.0
Bicyclist position Outside edge of travel lane Inside edge of shoulder Center of shoulder
288 397 515
24.0 33.1 42.9
319 361 545
26.0 29.5 44.5
607 758 1060
25.0 31.3 43.7
1137 63
94.8 5.3
1208 17
98.6 1.4
2345 80
96.7 3.3
Centerline rumble strips Not present Present
629 571
52.4 47.6
535 690
43.7 56.3
1164 1261
48.0 52.0
Opposing traffic Not present Present
872 328
72.7 27.3
862 363
70.4 29.6
1734 691
71.5 28.5
Buffer distance 5 ft (1.5 m) or less 5–8 ft (1.5–2.4 m) More than 8 ft (2.4 m)
122 803 275
10.2 66.9 22.9
151 842 232
12.3 68.7 18.9
273 1645 507
11.3 67.8 20.9
Motor vehicle lateral position Center of lane Left edge of lane Touched centerline One wheel over centerline Halfway over centerline
41 254 309 399 197
3.4 21.2 25.8 33.3 16.4
19 298 354 359 195
1.6 24.3 28.9 29.3 15.9
60 552 663 758 392
2.5 22.8 27.3 31.3 16.2
1200
100.0
1225
100.0
2425
100.0
Total
between the two study locations (with and without centerline rumble strips). The post-installation study included a lower percentage of bicyclists riding in groups (P-value
