Safety Performance of Part-Time Shoulder Use on Freeways, Volume 2: Conduct of Research Report (2021)

Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

5 C H A P T E R 2 - L I T E R A T U R E R E V I E W Literature Review Introduction This chapter describes the findings from a review of the domestic and international literature related to part-time shoulder use (PTSU) and bus-on-shoulder (BOS) operation on freeway facilities. The objective of the literature review was to identify freeway design elements, traffic-flow characteristics, operational strategies, and specific PTSU- or BOS-related conditions that might impact the safety performance of freeways with PTSU or BOS operation. The review also identified studies of the operational effects of PTSU. The insights obtained from this review were used to provide a better understanding of the potential tradeoffs between increased roadway capacity and safety impacts when considering PTSU operation on freeways. The information obtained from the literature review was used to guide the development of a list of research questions that are of interest to practitioners. These questions were addressed as part of this project (see Chapter 8, Supplemental Safety Findings). This chapter is organized into five sections. The first section provides background information on the PTSU and BOS types based on design, operations, and management as well as components of freeway facilities with PTSU or BOS operation. The second section describes findings from a review of the literature related to the safety and operational effects of PTSU and BOS facilities. The third section describes findings from a review of the literature related to the freeway design elements and traffic characteristics affecting freeway safety. The fourth section covers a review of the literature related to operational strategies that affect freeway safety. Identified knowledge gaps are presented at the end of this chapter. Background This section provides background information on the PTSU and BOS facilities on freeways in the United States. The objectives of this section are to provide some context for the discussion in the subsequent parts of the working paper and to establish vocabulary for this discussion. Topics addressed include PTSU design and operations options, PTSU components, and site types on a freeway facility. Part-Time Shoulder Use Options PTSU operation refers to the use of the shoulder as a travel lane during recurring congested conditions, and not a permanent conversion of shoulder to travel lane (Jenior et al. 2016). The differing physical conditions, travel patterns, and demand patterns of freeways result in different design and operational options of PTSU. These various options are provided in Table 1 (adapted from Jenior et al. 2016). Most PTSU facilities in the United States are open to all vehicles, are not tolled, and have the same speed limit as the adjacent freeway through lanes.

6 Table 1. Part-time shoulder use design and operations options. Process Category Option Design Location of shoulder used  Right shoulder  Left shoulder Operations & Management Vehicle types using shoulder  Open shoulder as a high-occupancy vehicle (HOV) lane that permits carpools and transit vehicles to use it  Open shoulder as a high-occupancy toll (HOT) lane that allows vehicles to pay a toll to use it if they do not meet HOV occupancy requirements  Open shoulder to all vehicles except trucks  Open shoulder to all vehicles Daily operating time period  Dynamically open shoulder when certain congestion thresholds are reached  Statically open shoulder during specified historical peak periods (time of day) Speed control  Same speed limit as other lanes (at posted speed limits).  Same speed as other lanes (at a reduced speed relative to normal posted speed limits)  Lower speed limit than other lanes Part-Time Shoulder Use Components PTSU facilities use many existing and established intelligent transportation technologies, which include closed-circuit television cameras, variable message signs, and lane control signs. Key features include the use of shoulder lanes as an additional temporary travel lane, lane control signs above each lane displaying the speed limits, and emergency turnouts for use in the event of crashes or vehicle breakdowns during PTSU operation. A typical freeway section with a PTSU lane was divided into different components to facilitate the identification of unique design elements that may have some influence on crash potential. The different components of a freeway facility with PTSU operation are shown in Figure 1 for a hypothetical freeway section. These components are identified in the following list:  Upstream influence area  Entry transition zone  Ramp entrance speed-change lane  Freeway segment with left-side PTSU operation  Ramp exit speed-change lane  Exit transition zone  Downstream influence area

7 Figure 1. Components of a freeway with PTSU operation. Safety and Operational Effects of PTSU and BOS This section describes facility-level safety and operational impacts associated with the change from non-PTSU to PTSU operation and non-BOS to BOS operation. Facility-level assessments focus on describing the safety performance of the overall facility. They are inclusive of all the changes to the freeway facility’s geometric design and traffic characteristics throughout its entire length. The findings from facility-level assessments are of general interest to Project 17-89 because they indicate general trends associated with design changes. However, the analysis conducted for Project 17-89 (and described in later chapters) was focused on disaggregating a facility into individual sites and quantifying the safety effect of the changes to each site. Before-After Study Results for PTSU Operation This section summarizes findings obtained from studies of facilities on which PTSU operation was implemented. Germany A study conducted in Germany reported a 20 to 25 percent increase in the capacity of a three-lane (per direction) freeway with the implementation of PTSU (Geistefeldt 2012). This increase was attributed to the addition of the PTSU lane, which effectively added a fourth lane to the freeway. The before-after evaluation for operational performance impacts was conducted using data from 1998–2000 (before period) and 2002–2006 (after period). The safety evaluation indicated that PTSU operation decreased crash frequency on the freeway. A safety study on freeway A 3 in Hessen and on A 7 in Schleswig-Holstein reported there has been a slight increase in crashes on the PTSU segment. However, this was balanced with fewer upstream crashes, specifically congestion-related (i.e., rear-end) crashes, when before and after data were compared (Geistefeldt 2012). The crashes were disaggregated into personal injury and property damage-only crashes for before-after analysis. By reducing queuing and speed reduction through a bottleneck area, PTSU operation was found to reduce upstream congestion-related crashes. Crashes decreased despite the fact that there was an increase in the vehicle-miles traveled after the implementation of PTSU operation (Jones et al. 2011). Denmark Researchers studied a 9.32-mile section of M13 from Allerod to Motorring 3. This section included a 1.24-mile segment (from Vaerlose to Bagsvaerd) where PTSU was implemented (Andersen 2016). A

8 before-after study was conducted with data from two comparable time periods – 2013 (before implementation of PTSU operation) and 2014 (after implementation). The study reported that the average travel time before implementation of PTSU operation was 22 minutes. After PTSU implementation, travel time was reduced by 1 to 3 minutes on the 9.32-mile section and 5 minutes on the segment with PTSU. The traffic volume on the freeway after PTSU implementation increased by 18 percent because much of the traffic shifted from local roads onto the freeway. The researchers did not conduct a safety evaluation. The Netherlands Before-after studies were conducted in the Netherlands for multiple freeways with dynamic PTSU operation (Veld 2009). Travel time was measured for the A20 and A27 freeways during the years 2005- 2006 (before) and 2006-2007 (after). The data showed a decrease in travel time of 24 percent on A20 and 11 percent on A27 during the morning peak hour. The study also showed a 5 percent decrease in travel time on A20 and 7 percent decrease on A27 during the off-peak hours. Similar studies were conducted on other freeways (Veld 2009). For example, during the morning peak hours, A13 showed a decrease in travel time of 12 percent and A7 showed a decrease of 60 percent. A28 showed a decrease in travel time of 21 percent during the evening peak and 14 percent during the off-peak hours. Another before-after study of multiple freeway facilities in the Netherlands was also conducted by Rijkswaterstaat (2007). The study, which used data from 2004 to 2007 with number of years changing by facility, found that PTSU operation reduced crash frequency by 25 to 28 percent due to the reduction in the upstream congestion. During low- and high-volume situations, the study from the Netherlands stated that PTSU operation on the right is more crash-prone than general-purpose lanes. However, PTSU operation has safety benefits when there is medium traffic volume on the roadway, which has been observed in other countries as well (Rijkswaterstaat 2007). The researchers did not specify any quantitative information regarding the ‘high’, ‘medium’, and ‘low’ traffic volumes. Another study of dynamic PTSU operation in the Netherlands (in which vehicles travel on the right, like the United States) found that PTSU operation on the left, during low-volume situations, is more crash prone than general-purpose lanes (Drolenga et al. 2015). This finding is likely are result of the higher speeds on the roadway during low traffic volume conditions. For medium and high traffic volume on the roadway, the PTSU operation on the left is about as safe as general-purpose lanes and safer than the PTSU operation on the right side of the roadway because there are rarely exit ramps on the left side. Colorado The Colorado Department of Transportation (CDOT) conducted an operational evaluation after the introduction of PTSU operation on I-70 in the eastbound direction. The researchers noted that there was a 14 percent increase in the throughput, a 38 percent improvement in travel time in general-purpose lanes, and an 18 percent increase in the average vehicle speeds across all lanes of eastbound I-70 during high traffic volumes on the weekends (CDOT 2017). Similarly, on US 2 in Washington, the peak period in the peak direction delays reduced from 8 to 10 minutes to 1 to 2 minutes with the implementation of PTSU operation on the roadway (Kuhn 2010). An empirical Bayes (EB) safety evaluation on US 2 showed that the safety performance degraded after implementation of PTSU operation, and there was an increase of 15 percent in expected crashes in the after period (Margiotta et al. 2014). Further evaluation showed that the site is unique, and the annual crash frequency is very low and not representative of the locations likely to have PTSU implemented in the future.

9 Virginia An operational evaluation was conducted on the I-66 PTSU facility in Virginia. The study, which was conducted in 2007, revealed that the PTSU operates at near capacity with volume-to-capacity ratios of 0.90 to 1.00 in the eastbound direction and 0.83 to 1.00 in the westbound direction. These numbers indicate that the PTSU operation was able to help significantly augment throughput during the peak periods (Jenior et al. 2016). A safety evaluation was also conducted on the I-66 PTSU facility in Virginia in 2007. The segment has a left-side HOV lane and a right-side, dynamic, part-time shoulder use lane. The study only included “after” data with the HOV lane and PTSU operation in place, while focusing on the crash frequency differences between the hours when the PTSU operation was open and closed to traffic. Results showed that the crash frequency reduced by 8 percent. The rear-end crashes reduced by 13 percent, and the injury crashes reduced by 6 percent. However, there was no significant change in the crash frequency along the entire corridor of I-66. The crash modification factors (CMFs) for all types/severities of crashes on the PTSU facility are 0.75, 0.71, and 0.69 for total, multiple-vehicle, and rear-end crashes, respectively. Similarly, the CMFs for the PTSU facility fatal and injury crashes are 0.69, 0.59, and 0.61 for total, multiple-vehicle, and rear-end crashes, respectively (Suliman 2017). Other safety evaluations of I-66 in Virginia concluded that there are negative effects in the safety performance between the hours when the part-time shoulder use was open and closed to traffic. Researchers for a 2015 study on PTSU operation stated that for the right shoulder-specific crashes, motorists’ behaviors at the merge and diverge areas during adverse light conditions are significant, and there was an increase of about 38 percent in all crashes (Lee et al. 2007; Kuhn et al. 2013). Minnesota Researchers for a Minnesota Department of Transportation project observed that following the implementation of PTSU operation on I-35W, the rear-end crash frequency increased in certain roadway sections in the vicinity of PTSU operation (Davis et al. 2017). The EB method showed an increase of 28 percent in the expected number of crashes in the after period (Margiotta et al. 2014). Additional analysis determined that the observed change/increase in the crash frequency was attributed to the change in traffic volume and traffic patterns. The analysis also indicated no direct effect on the likelihood of rear- end crashes due to the operation of the PTSU lane (Davis et al. 2017). Before-After Study Results for BOS Operation In TCRP Report 151, Martin and Levinson (2012) summarized the available information on safety outcomes of BOS implementations in the United States as of January 2011. They found few before-and- after safety studies that had been performed for BOS implementations and none with statistically significant results. At the same time, they found that no BOS implementation had been discontinued due to safety or other concerns, and that several BOS implementations had been, or were planned to be, expanded because of successful pilot projects. In the time since TCRP Report 151 was published, several more BOS implementations have been documented. Table 2 summarizes the safety experience of BOS presented in the literature.

10 Table 2. Safety experience with BOS projects. Region Facility Length (miles) Safety Experience Source(s) Minneapolis, Minnesota (MN) Various facilities, serving 14 routes and 400 buses daily as of 2016 300 as of 2016  1991–2001: 20 total accidents, all property damage only  2001–2007: 1 fatal crash during this period, transit not at fault  System continuously expanded since 1991 Martin and Levinson 2012; Douma 2007; FDOT 2017 Minneapolis, Minnesota Trunk Route 77, right shoulder with Driver Assist System (DAS), 10 buses equipped with DAS Total length not provided, test segment evaluating DAS was 1.7  No accidents in 6 months prior to start of DAS use (BOS only)  No accidents in 17 months after start of DAS use in April 2012  DAS reduced side-to-side movement within lane by 4.7 inchesa. Pessaro 2013 San Diego, California State Route (SR) 52 & I-805, right shoulder, 1 route, 11 buses daily 8  No accidents in first 6 months of operation (2005–2006) Martin and Levinson 2012 Old Bridge, New Jersey US 9, right shoulder, 440 buses daily 4  No accidents since startup (2006– 2010) Martin and Levinson 2012 Miami, Florida State Route (SR) 874 & SR 878, right shoulder, 3 routes, 116 buses daily (initially), 200 buses daily (2016) 9 (total of both directions), only 1 route uses the full facility length  No accidents during 3-year pilot (2006–2009)  As of 2015, BOS planned to be expanded to other toll facilities in the region Parsons 2009; FDOT 2017 Atlanta, Georgia SR 400, right shoulder, 2 routes initially, 3 routes and 119 daily buses as of 2016, operated as BOS only 2005–2011, converted to PTSU operation in 2012 6 as BOS only  No bus-involved accidents reported during either BOS-only or PTSU operation  Facility length extended in 2015 FDOT 2017 Chicago, Illinois I-55, left shoulder, 4 routes, 60 daily buses 15 (total of both directions)  No safety issues reported during pilot (2011–2013)  Facility made permanent in 2014 and is planned to expand to I-90 & I- 94 FDOT 2017 Columbus, Ohio I-70, right shoulder, 4 routes, 20 daily buses initially (2006) 10  Expanded to I-670 in 2015, with further expansion planned Martin and Levinson 2012; FDOT 2017 Columbus, Ohio I-670, left shoulder, 4 routes 6  Opened in 2015, no information reported FDOT 2017 Kansas City, Kansas I-35, right shoulder, 6 routes, 43 daily buses 8 (2012–2014), 11.5 (2015–)  No accidents in first year of operation (2012)  Facility expanded in 2015 FDOT 2017 Raleigh, North Carolina I-40, right shoulder, 9 routes, 161 daily buses 60 (total of both directions)  No accidents in first year of operation (2012) FDOT 2017 Haifa–Tel Aviv, Israel Hwy 2, right shoulder 7  No “hazardous events” observed in reviewing 60 hours of video Gitelman et al. 2016 a Statistically significant at 95 percent level. Despite the general lack of long-term safety data, it is notable that none of the documented BOS implementations experienced safety issues in the first 6 to 12 months after opening. In contrast, safety

11 issues with other types of exclusive transit facilities (e.g., motorists not observing new or changed traffic signal operations associated with new light rail and busway facilities) tend to appear immediately after the facility opens. The lack of documented safety issues with BOS implementations suggests that the signing, striping, bus operating procedures, and public education efforts used with these implementations have been sufficient to avoid safety problems. Bus drivers also have the discretion not to use the shoulder if they feel it would be unsafe. Design Elements and Traffic Characteristics Affecting Freeway Safety This section summarizes the findings on the topics related to design elements and traffic-flow characteristics that affect freeway safety. Number of Lanes The relationship between number of lanes and safety is complex; therefore, generalization is difficult. Crash data for urban freeways were examined in three states in the United States in an effort to understand the relationship between number of lanes and traffic safety (Kononov and Allery 2008). The study found that number of lanes had a unique influence on the relationship between traffic demand and crash frequency. At lower traffic demand, a four-lane cross section is shown to have more crashes than the six-lane cross section. However, as the annual average daily traffic volume reaches above 65,000 vehicles per day, the four-lane cross section was found to have fewer crashes than the six-lane cross section. Horizontal Curve Two studies document the development of a horizontal curve CMF for highways based on curve radius and speed limit (Bonneson et al. 2005; Bonneson and Pratt 2008). These CMFs were developed by using curve radius as an independent variable and assuming zero degrees as the base condition. Both CMF functions indicate that crash frequency is larger on freeway curves with a smaller radius. Lane Width The NCHRP Project 17-45 identified an association between lane width and freeway safety (Bonneson et al. 2012a). The lane width CMF developed as part of this research is incorporated in Chapter 18 of the Highway Safety Manual Supplement (HSM Supplement) (AASHTO 2014). The CMF function indicates that crash frequency is larger on freeways with a smaller average lane width. Shoulder Width Similar to lane width, the research by Bonneson et al. (2012a) as part of the NCHRP Project 17-45 has identified an association between inside and outside shoulder width and freeway safety. The CMFs were developed as part of this research and are incorporated in Chapter 18 of the HSM Supplement (AASHTO 2014). The CMF function indicates that crash frequency is larger on freeways with a smaller average inside or outside shoulder width.

12 Median Width A CMF for freeway median width is provided in Chapter 18 of the HSM Supplement (AASHTO 2014). The median width used in the CMF function excludes the width of the inside shoulder. The associated CMF function indicates that a narrower median is associated with more frequent crashes. Barrier Presence The effect of median barrier presence on crash frequency and severity was examined by Tarko et al. (2008). The researchers found that the conversion of a depressed median to a flush median with rigid barrier increased single-vehicle crashes by 120 percent. This conversion also reduced the same-direction crashes by 20 percent. A CMF for freeway barrier presence is provided in Chapter 18 of the HSM Supplement (AASHTO 2014). This document also includes a severity distribution function (SDF) that includes a variable for barrier presence. The combined SDF and CMF function indicate that freeways with barrier on the roadside or in the median are associated with more frequent, but less severe, crashes. Narrow Lanes and Shoulders Curren (1995) studied freeway facilities where lane widths were narrowed to 11 feet. The research findings show slightly lower speeds for a given volume range. The findings also show a slightly greater tendency to fall into level-of-service F conditions for the altered sites. Field visits and observations indicated that that some operational impacts were particularly notable in the transition area of the altered sites. The safety evaluation of the conversion showed that the truck crash rates were higher on altered sites (i.e., sites with 11-ft lanes) when compared with unaltered sites (i.e., site with 12-ft lanes). The crash rates were higher in three of the five facilities that used narrow lanes and shoulders on a continuous basis for an extended length (more than a mile). The safety effect of narrow lanes and shoulder-use lanes was investigated by Bauer et al. (2004). They examined before-after data for 490 sites in California where the freeway was converted from four to five lanes, and five to six lanes. In most instances, the shoulders and lanes were reduced in width and the added lane was used as an HOV lane. The before-after study was conducted using data from 1991–1992 (before) and 1994–2000 (after) period. The safety evaluation found that the projects that converted four lanes to five lanes resulted in an increase of 10 to 11 percent in crash frequency. The researchers speculated that this finding might result from the speed differential between the general-purpose lanes and the added HOV lanes, in addition to the effects of narrower lanes and shoulders. The researchers also noted that the conversion projects might decrease crashes upstream of the project and increase crashes within and downstream of the project due to the relocation of the traffic operational bottleneck on the freeway. The safety effect of various roadway geometric characteristics (including shoulder width, number of lanes, and speed limit) was investigated by Islam et al. (2014). They used data from freeways in Connecticut for this investigation. For both total and fatal-and-injury single-vehicle crashes, the safety benefit was found to increase as shoulder width increased (greater than 10 feet). A 2015 study examined the operational and safety implications of reduced lane and shoulder widths to accommodate an additional travel lane using speed, crash, and geometric data for freeways in Texas (Dixon et al. 2016). The operational analysis identified an increase of about 2.2 miles per hour in travel speed for a 12-ft lane as compared to an 11-ft lane. The effect of shoulder width was found to be more significant with narrower lanes. This finding suggests that shoulder width is more important with a reduced-lane width. The researchers also found that crash frequency decreases with each additional travel lane, with increased left shoulder widths for narrow lanes, and with increased right shoulder widths.

13 A 2016 study examined the overall operational and safety effects of narrow lanes and shoulders on freeways throughout the United States (Neudorff et al. 2016). The evaluations from the literature show that adding a lane by narrowing the existing lanes and shoulders generally improves operations and level of service but decreases the average speed on the roadway. Exhibit 11-8 of the Highway Capacity Manual 2010 (HCM 2010) shows the reduction in free-flow speed values as a function of lane width. This exhibit is reproduced as Table 3. Table 3. Adjustment to free-flow speed for lane width. Average Lane Width (feet) Reduction In Free-Flow Speed (mph) 12 0.0 11-12 1.9 10-11 6.6 Source: Highway Capacity Manual 2010 (HCM 2010). Freeway Speed-Change Lane A study conducted in 2009 evaluated and compared the impacts of various types of ramp exit speed- change lanes on the safety performance of freeway diverge areas (Chen et al. 2009). Four different types of speed-change lanes were considered in the study. Cross-sectional data were used to compare crash rates, crash frequency, and crash severity among different speed-change lane types. The study found that the following characteristics had a statistically significant association with diverge area safety performance: speed-change lane type, ramp and freeway annual average daily traffic volume, posted speed limit, deceleration lane length, right shoulder width, and type of exit ramp (i.e., single lane versus two-lane exit; lane balanced versus not balanced). Research conducted for NCHRP Project 17-45 led to the development of a predictive model for speed- change lane sites. A speed-change lane site was defined by the researchers as a one-directional length of freeway that includes the speed-change lane and the adjacent through lanes on the freeway. The model includes CMFs for the following freeway elements: speed-change lane type (i.e., ramp entrance or ramp exit), speed-change lane length, lane width, shoulder width, and freeway median width. This model is described in Chapter 18 of the HSM Supplement (AASHTO 2014). Operational Strategies Affecting Freeway Safety This section summarizes the findings related to the operational characteristics that affect freeway safety. Traffic Density The German Highway Capacity Manual includes the design capacities for freeways with dynamic part- time shoulder lane presence, in and outside of the urban areas (FGSV 2015). The design capacities for basic freeway segments with PTSU operation, with a gradient of less than or equal to 2 percent, range from 4,700 to 7,000 vehicles per hour. This is for heavy vehicle percentages ranging from less than 5 percent to 30 percent for two-lane and three-lane roadway sections, with an additional PTSU lane (Geistefeldt 2016). A 2012 study examined the relationship of safety to traffic-flow parameters such as traffic volume, density, and speed (Kononov et al. 2012). These relationships for selected freeways in Colorado suggest that as flow increases, crash rate initially remains constant until a certain critical threshold combination of

14 speed and density is reached. Once this threshold is reached, the crash rate rapidly increases. The researchers concluded that the crash rates decline with the implementation of PTSU operation because of the lower traffic volume per density per lane, and the safety benefits outweigh the adverse effects of not providing a shoulder on travel lanes at all. A 2012 study investigated the relationships between traffic safety and congestion for urban freeways, using data from Seattle, Washington, and Minneapolis/St. Paul, Minnesota, freeways during the years 2005 to 2007 (Harwood et al. 2013). The traffic volume and speed information were obtained for 564 urban freeway segments during the study periods from detectors in individual lanes. The results show U- shaped relationships between crash rates per million vehicle-miles of travel and traffic density—i.e., higher crash rates correspond with the lower and higher traffic densities, with lower crash rates associated with the middle range of traffic densities. A study by Aron et al. (2013) examined the safety impacts of implementing PTSU operation on the A4- A86 freeway (weaving section) in Paris, France. They used the before-after study method outlined in Hauer (1997) to quantify the change in crash frequency. The researchers found that the main effect of opening an additional lane—i.e., implementing PTSU operation during the peak period—is a reduction in the traffic density and, therefore, a reduction in crash frequency on the freeway. They rationalized that this outcome is because crash risk is lower during fluid traffic flow than during dense traffic flow on the freeway. The researchers also noted an increase in crash risk downstream of PTSU operation because of the relocation of the traffic operational bottleneck on the freeway. Speed Limits The safety effect of implementing variable speed limits (VSL) on a section of freeway in Oregon was investigated using the EB method (Chambers et al. 2016). The freeway segments used in the safety evaluation begin and end at every ramp gore point, i.e., they include the speed-change lanes and the basic freeway segments in the analysis. The data used for analysis include the July 2011–July 2014 before data and the July 2014–April 2016 after data. The crash frequency and severity analysis show that fatal and injury crashes decreased in the after period, with slight increases in property damage-only crashes. The frequency of rear-end crashes also increased. The researchers rationalized that these changes in crash frequency may be due to the lower compliance of VSL by the drivers, as found previously by Downey and Bertini (2016). A similar study on the safety effects of implementing VSL on freeway mainlines was evaluated on the VSL system on I-5 in Washington (Pu et al. 2016). The system was implemented in 2010, and the crash data used for analysis ranged from 2007 to 2012. The before-after study method was used for the safety evaluation. The researchers found that the total crash frequency was reduced by 29 percent after the VSL system was applied on the freeway facility. A 2016 study investigated the safety effects of an advisory VSL system deployed along OR 217 freeway in Oregon (Siddiqui and Al-Kaisy 2016). Due to the limited crash data available after deployment of the VSL system, this study employed surrogate measures in addition to the crash data to assess the safety impacts of the VSL system. The surrogate measures include speed and speed variability. They were measured at point locations across two directional lanes of travel, at multiple locations along the corridor. The results of the safety evaluation show that rear-end, injury, and property damage-only crashes decreased 10 percent, 9 percent, and 18 percent, respectively, after VSL implementation. VSL system was found to generally reduce mean speed and speed variability within the same lane and between the median and shoulder lanes.

15 Queue Length A 2016 study focused on evaluating the safety effects of an end-of-queue warning system at freeway work zones (Ullman et al. 2016). The researchers conducted a study of a warning system deployed on I- 35 in Texas. The end-of-queue warning system alerted drivers when they entered the lane-closure work zone through the presence of warning signs, law enforcement officers, and portable rumble strips. This warning system was deployed upstream of nighttime lane closures, where queues were expected to develop. The safety effects of this warning system were analyzed using before-after data for the years 2003 to 2009. The results show a 44- percent reduction in crash frequency when the warning system was present, relative to when the system was not in place. The percent of severe crashes decreased from 58 percent to 41 percent, and rear-end crashes decreased from 58 percent to 36 percent. Transit Facilities BOS implementations in the United States typically provide the following features that promote safe operations:  Buses only use the shoulder when general traffic slows (e.g., speeds drop below 35 mph).  Buses maintain a relatively low speed differential between themselves and other vehicles while operating on the shoulder (e.g., 10–15 mph).  Buses use four-way flashers when operating on the shoulder.  Signage is provided along the freeway mainline (e.g., SHOULDER — AUTHORIZED BUSES ONLY) and on freeway ramps (e.g., WATCH FOR BUSES ON SHOULDER).  BOS operation is allowed only when adequate shoulder width is available (typically 10 feet without a barrier and 11.5 to 12 feet with a barrier).  Bus drivers receive training prior to being allowed to drive on the shoulder, and only use the shoulder when they feel it is safe to do so. Some transit and roadway agencies have also conducted public awareness campaigns prior to the start of pilot BOS implementations. Knowledge Gaps This section summarizes the knowledge gaps identified in the previous sections and provides an indication of the relative importance of each gap as it relates to achieving project objectives. In general, the literature review indicated that there is relatively little quantitative information available that describes the influence of various PTSU or BOS design and operational elements on crash potential. The number of facilities with PTSU and BOS operation is limited in the United States, and many have opened in the years just prior to the start of NCHRP Project 17-89. Nevertheless, research on freeway safety and international research on PTSU safety does provide general safety trends and an indication of which elements may have a greater influence on freeway crash potential. Literature generally discusses safety performance in terms of the entire cross-section of one or both directions of the freeway and does not distinguish lane or shoulder on which crashes occurred. The research team believes this may be due to the lack of reliable lane location data in agency crash databases. The knowledge gaps identified for the PTSU- and BOS-related conditions influencing safety are listed in Table 4. These conditions are used to guide the development of topics that will help achieve the project objectives. The list is not intended to be comprehensive. Rather, it is intended to reflect conditions that likely influence the safety of a freeway with PTSU or BOS operation. The last column of the table indicates the relative importance of each topic. A topic’s importance was based on consideration of both its association with crash potential and practitioner interest in information about this association.

16 Table 4. Knowledge gaps for PTSU- and BOS-related conditions potentially affecting safety. Category Topic Importance Basic Descriptors Number of Lanes High Area type (urban or rural) High Type of operation (BOS or PTSU) High Conventional Freeway Attributes Elements captured in CMFs of HSM Supplement freeway models Medium PTSU-Related Conditions Effect of traffic density/congestion Medium Effect of 2nd edge line markings Medium Effect of static versus dynamic signs High Effect of static versus dynamic PTSU operation High Effect of speed limits Medium Effect of left versus right shoulder use High Effect of shoulder width (of shoulder used for PTSU or BOS) Medium Effect of PTSU component (see Figure 1) Medium Differences when shoulder is open or closed High Effect of taper versus parallel ramp types Low Effect interchange spacing Low