Improving Pedestrian and Motorist Safety Along Light Rail Alignments (2009)

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.

52 This chapter provides information about how data and observation can help to indicate why collisions happen and how that information can be used to determine treatments. The discussion starts with a general description of the con- cepts of safety analysis, including root causes and contribut- ing factors. After this initial overview, the chapter explores the major categories of root causes and contributing factors for collisions on LRT alignments, as suggested by the literature, the survey, the data analysis, the transit agency consultations, and the site visits. Such information can help practitioners understand their safety issues and identify how their concerns align with those of other systems. The final section outlines four treatment strategies that address the causes and contributing factors for LRT collisions. These strategies are based on the cumulative judgment of the project team after compiling anecdotal and statistical informa- tion with safety engineering experience and in-situ observa- tions during the site visits. Root Causes and Contributing Factors Safety initiatives are often intended to respond to one or a series of crashes. To reduce the number of future collisions, it is necessary to determine the root causes. In most cases, there are many factors that influence an incident. These factors come together in a specific way to influence the likelihood of colli- sion, to result in the collision, and to affect the resulting sever- ity of the collision. Root cause analysis is a structured procedure for examining the reasons an undesirable event occurred, and for identify- ing ways to prevent the event from happening again. RCA is employed in the industrial and medical communities for safety and quality management, and it is particularly well doc- umented by NASA. (See NASA Procedural Requirement NPR 8621.1.) A presentation explaining the relationship between root causes and contributing factors and how to conduct a RCA can be found on the NASA website at www.hq.nasa.gov/ office/codeq/rca/rootcauseppt.pdf. The procedure recognizes the distinction between root causes and contributing factors. Root causes are one or more fundamental flaws or problems in a system that lead to the undesirable outcome, and without the root cause(s), the undesirable outcome (in our case a colli- sion) could not have occurred. It is noted that root causes are not the proximate (or direct) causes of a collision, but lead to those causes that in turn lead to the collision. This is an important distinction because direct causes often vary from incident to incident, and are beyond the reasonable control of the system. For example, a collision may be directly caused by a pedestrian walking in front of an LRV. A root cause could be a lack of sufficient warning to the pedestrian, but it may not be reasonable to expect that the system could have stopped that particular pedestrian and avoided that particular collision. It is, however, reasonable to expect the system to address the root cause that people in general do not have sufficient warning in the location or circumstances involved. Contributing factors influence the occurrence or severity of a collision but are not actually root causes and if eliminated would not have prevented the collision. Contributing fac- tors provide a context for the collision. As an illustration, con- sider the case of a fatality on an interstate highway facility. The collision occurred on a straight portion of roadway in good driving conditions. Police determine that the driver was fatigued, fell asleep, and left the right-of-way. Driver fatigue was a causal factor. Police also determine that the vehicle was trav- eling at 70 mph when it left the right-of-way. The speed was a contributing factor. It did not on its own cause the collision, but would have reduced the time available for the driver to recover control upon leaving the pavement (e.g., if awoken by a rumble strip), and it affected the severity once the incident occurred. In the same case, the driver would not have left the roadway had there been barriers, but the lack of barriers did not cause the crash. The absence of barriers is a contributing factor. C H A P T E R 4 Safety Issues and Their Treatment

Understanding the roles of root cause(s) and contributing factors is important. If the fundamental weaknesses of the sys- tem are not identified and improved, the likelihood of inci- dents recurring is increased. Ideally, safety treatments should address the root causes to have the best effect on conditions leading to a collision. If pos- sible, a holistic approach should be used to determine root causes and contributing factors, and safety treatments that will provide the best overall safety effects for the system. By consid- ering the entire system, practitioners can avoid taking a narrow view that may modify one aspect of the system without taking into account other issues and repercussions. It some cases where the root cause is difficult to control, it may be possible to mitigate the number and/or severity of crashes by treating a contributing factor in an effective matter. Determining LRT Safety Issues and Identifying Treatments Chapter 2 identified a list of the safety issues noted by agen- cies and SSOs. Chapter 3 outlined the data quality and quan- tity problems facing LRT safety analysis. Because of these data problems, it is difficult to determine the root causes of LRT collisions in a statistically significant analysis. It is also difficult to undertake a statistical analysis of the effectiveness of a given safety treatment. This section outlines methods that can be used to determine LRT safety issues and identify treatments. Anecdotal informa- tion available from agency and SSO staff who work with LRT alignments, formal LRT safety analysis (where available), the safety literature, consultation with professionals, and the site visits to LRT agencies were used to identify some of the most common or severe LRT collision types. The section then iden- tifies some surrogate (proxy) measures that can help agencies identify potential safety hazards. Studying LRT Safety Issues and Treatments Most previous studies of safety along LRT alignments have examined treatments using simple before-and-after compar- isons of collisions, anecdotal evidence, collision surrogate measures such as violations, or a combination of two or three of these approaches. Standard “t” tests require sufficiently long periods to elapse both before and after studies. Where the number of years is limited and/or the numbers involved are small, an unusually “high” year and a “low” year after could reflect regression to the mean rather than an effect of the treatment. Although the effectiveness of treatments (such as LRV- activated signs) in reducing the number of incidents of risky behavior on the part of motorists has been well demonstrated (1, 2), LRT safety studies to date have not been based on sta- tistically significant analysis of the reduction in the number of collisions following the implementation of a given treatment. The state of the art in safety study methodology generally requires the use of empirical Bayes analysis, but contemporary statistical approaches such as empirical Bayes analysis are still relatively new to transportation studies, and have not been applied to the field of LRT safety. No references to LRT and empirical Bayes analysis were found during the literature review, survey, or other inquiries. The literature review found no statistical analyses of LRT collision data that were statisti- cally defensible in terms of contemporary statistical analysis methods. The lack of studies using statistically sophisticated and defen- sible methods can be attributed to several issues. Firstly, as sug- gested above, transportation and LRT safety practitioners are not usually familiar with contemporary statistical methods. In addition, data issues include the lack of essential and sufficient collision data; the lack of vehicular, pedestrian, and LRV vol- ume data; and the lack of rail and highway inventory informa- tion, including the dates on which safety treatments were implemented. In the survey of local LRT agencies, for example, 10 of the 17 agencies that answered the question on four- quadrant gates did not record the installation date. The survey found that non-recording of date information ranged from 27% to 88%, depending on the treatment. To determine the feasibility of adopting an empirical Bayes analysis to examine the safety impacts of selected treatments along LRT alignments, it is essential to first determine the data that are needed to carry out the analysis. (Data availability and data quality are discussed in Chapter 3.) The studies available are generally limited in scope and do not examine the holistic safety impacts of the various treat- ments being studied. For example, devices such as pre-signals and advance signals have been widely implemented through- out North America. The focus of these studies of pre-signals and advance signals, however, is on signal violations or the impact on LRV–motor vehicle crashes. No studies have exam- ined the system-wide impacts of such treatments—for exam- ple, the possibility that the implementation of a new traffic signal at a location could result in an increase in crashes, such as rear end collisions, that involve only motor vehicles. Determining the Highest Risk LRT Safety Issues Unlike motor vehicles, LRVs cannot swerve, and even emer- gency stop conditions do not enable the LRV to avoid pedes- trians who are errant or walking counter to traffic control devices. Nevertheless, Korve et al. in TCRP 17 (2) found that collisions between pedestrians and LRVs are the least common type of LRT-related collision. Collisions between pedestrians and LRVs in the systems reviewed represented only about 10% 53

mation on the effectiveness of corrective actions. This study provides a starting point by gathering the LRT safety informa- tion available into one, easily accessible catalog of information about specific LRT safety treatments (Appendix A). Even with excellent data collection processes in place, analy- sis of LRT collision data will still be limited by the number of collisions. LRT collisions are even more statistically rare than vehicle collisions. LRT data and the information obtained from the site visits suggest that a high accident location along an LRT alignment has about one incident per year. This collision fre- quency is too low for conducting before-and-after studies, or even for identifying problem locations. The site visit agencies all reported using various other measures to determine which locations may require safety treatment. These “surrogate” measures could be used in formal safety studies (before-and- after, empirical Bayes, etc.) to determine the effects of specific safety treatments without needing to wait for a significant number of collisions to occur. Surrogate measures described by the LRT agencies con- sulted by the project team include: • Operator reports of near misses or other significant safety- related events. • Risky behavior: – Metro Transit in Minneapolis conducted a series of pedestrian violation counts at a downtown station. The counts showed a decrease in the number of pedestrians crossing the tracks illegally mid-station as the level of advertising and signage increased. The data and follow- up analysis allowed Metro to show that there was still significant risk of collision due to pedestrians crossing mid-station, despite the impact of advertising and sign- age. This led to the municipal authorities approving inter-track fencing at this location. • Emergency braking: – Automated emergency braking records allowed UTA to identify a location where a pedestrian signal was badly timed. After the signal timing was changed, the num- ber of emergency braking incidents was substantially reduced. • Insurance and non-recoverable cost records: – UTA and Metro Transit both have detailed recordkeep- ing for insurance purposes. These records could be used to identify problem locations. UTA has a large number of locations with crossing gates and can identify poten- tial problem locations by the frequency of crossing gate replacements. • Customer complaints: – SF Muni keeps a detailed database of customer com- plaints. This could be used to identify problem locations, or operators with an unusual history that might indicate a need for additional training. 54 of the total, but these collisions are the most severe and account for at least 50% of all fatalities resulting from LRT collisions (3). These general findings are supported by the data reviewed in Table 11 of Chapter 3, which showed that pedestrian colli- sions accounted for 168 (22%) of the 773 collisions (SEPTA excluded) in the NTD, but 80% of all fatalities (47 of the 59 fatalities in the years from 2002 to 2007). Although the comparable numbers for cyclists was very small, with only 24 total collisions in the 2002–2007 NTD data, cyclists also appear to be overrepresented in terms of severity: they were involved in only 3% of collisions with LRVs, but accounted for 10% of all fatalities. Together, cyclists and pedes- trians accounted for nearly 90% of all LRV collision fatalities. Improving pedestrian and bicycle safety is essential. Conversely, road vehicles accounted for the majority of col- lisions with LRVs, but only a small proportion of the fatalities. On the other hand, motor vehicle occupants were involved in 65% (261) of the 404 injuries (Table 11, Chapter 3) in all injury- causing collisions, and almost 48% of all motor vehicle–LRV collisions resulted in injuries (261 of 545 motor vehicle colli- sions, as shown in Table 11 of Chapter 3). It is clear that each group of road users has significant risks associated with LRV collisions, and that the risks vary. It is, however, difficult to hypothesize the root causes for these events without detailed exposure information and more infor- mation on the circumstances of each collision reported. More data, and more historical data than are currently available (many LRT systems are relatively new), are needed to deter- mine the most appropriate and effective combination of safety treatments. An additional concern in the evaluation of LRT collision data is that collisions between non-LRV vehicles and other vehicles or pedestrians that may have been related to the LRT facility or operations but did not physically involve an LRV are not recorded in the LRT collision data. For example, an inci- dent in which a pedestrian is hit by a car while walking to or from a center-of-street transit stop is arguably related to the LRT, but the incident will not appear in LRT collision data- bases. A comprehensive analysis of LRT safety will therefore need to include information from local transportation agencies and police to determine whether any safety implications of LRT operations extend beyond the LRT right-of-way itself. Building a Safety Analysis Toolkit It is important that all agencies (transit agencies, SSOs, city transportation departments, police, and research organiza- tions) take an active role in adding to the available knowledge. Good data collection guidelines are needed to increase the quality and quantity of LRT safety data assembled for analysis. As the database improves, it will become possible to perform more rigorous statistical analyses and to develop better infor-

55 These surrogate measures can provide agencies the infor- mation required to identify potential hazards. They may also be useful as before-and-after studies and other safety research. General Treatment Strategies The analysis and agency input obtained during site visits enabled the project team to identify four general treatment strategies that would help to mitigate the safety issues and risks reported by local LRT agencies. The strategies listed below are a summary of commonly encountered observations and recommendations from the staff of the LRT agencies visited. Information about specific safety treatments is provided in Appendix A, and documentation of the site visits and discus- sions with the local staff are provided in the site visit memos in Appendix D. The four strategies are: 1. Give responsibility to the operators: Representatives from NJT were especially clear on the need to give responsibil- ity to the operators. An LRV is not a bus. LRV operators should be given special, intensive, and location-specific training. NJT LRT operators have four weeks of training, followed by one week of hands-on experience with a sea- soned operator. NJT staff noted that operator training is essential where an agency cannot install the optimum treatment because of a physical limitation. NJT recom- mended giving operators increased responsibility, and sug- gested that automation should be limited. Other transit agencies visited also reported that their operators were trained to drive defensively. One agency noted the fre- quency of reviewing their operators was based on the oper- ators’ incident and complaint records. 2. Increase motorist, pedestrian, and cyclist awareness by providing active, appropriate information: All the transit agencies visited reported that motorists, cyclists, and pedes- trians respond better to active signage than to passive sign- age. Agencies have observed that motorists, cyclists, and pedestrians who cross the LRT alignment on a regular basis can become desensitized to warnings. This problem is more pronounced where the warnings provide general informa- tion instead of specific information (e.g., train sign instead of a no turn sign), or where the duration of the warning is longer than necessary. The project team visited one NJT crossing location where pedestrians clearly disregarded the audible warning device. The agency guide noted that the warning sound was a standard length and much longer than required for a pedestrian to clear the tracks. Local pedestri- ans learned to disregard the warning because it did not pro- vide relevant information about the approach of a train. In fact, because of the length of the signal, its warning seemed to sound “all the time.” The same principle can be applied to second train (some- times termed “approaching train”) warning signs. None of the second train warning signs viewed during the site visits to transit agencies indicated the direction of the approach- ing train. The impacts of providing this additional infor- mation should be tested using before-and-after surrogate measures such as risky actions by pedestrians or emergency brake records at a crossing where a directional second train warning sign is installed. 3. Education: Agencies commented that the public does not seem to have the same respect for LRT as for freight and commuter rail trains. Further, in cities where LRT is new, motorists, cyclists, and pedestrians may not under- stand how to behave in the alignment. Safety campaigns improve public understanding of how to act in the light rail alignment. Arizona, for example, has included a section concerning driving around LRT in their driver training handbook. Other agencies broadcast safety reminders over their platform and LRV audio systems and/or run public awareness campaigns. Operation Life- saver has an LRT branch that works with agencies to increase awareness. All of these strategies aim to teach citizens how to drive, walk, or cycle around LRT, but they also increase public awareness of the serious nature of LRV collisions. 4. Separation of LRT space from the space occupied by other modes: While active information (as described in strat- egy 2, “Increase motorist, pedestrian, and cyclist aware- ness by providing active, appropriate information”) provides useful and direct information to motorists, cyclists, and pedestrians, the separation of LRT space can provide environmental cues for safety. The separation can be a clear physical barrier, such as landscaping or channel- ization, or a more subtle measure such as a change in pavement type. Pavement type has an effect: in the tran- sit agency site visits, all the locations identified as problem areas because pedestrians crossed mid-station or mid-block had surface treatments that were conducive to walking between the tracks. The locations that had gravel between the tracks were not identified as problem areas for pedes- trian incidents. The separation can also be complete by separating the grades (conversion to a type a.1 alignment) which precludes errant movements. This offers the additional benefit of reducing delay for road vehicles. There are, however, major drawbacks in terms of capital cost, land requirement, envi- ronmental impacts, disruption of existing operations, and so on. Complete separation of the ROW has not been con- sidered as a practical safety treatment in this study (in terms of something that can be applied to a safety problem), but it would clearly improve safety if it can be designed-in before construction starts.