Roadside Safety Analysis Program (RSAP) Update (2022)

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.

1 INTRODUCTION This report describes the development of an update to the Roadside Safety Analysis Program (RSAP). The report is divided into several essentially independent reports. The main body of this report describes the background, history and philosophy behind the RSAP program and provides a very brief overview of the new program, described throughout this report as RSAPv3 to distinguish it from earlier versions of RSAP. Appendix A is a USER’S MANUAL for RSAPv3 which provides detailed explanations for how to use the program. The User’s Manual is intended for any and all users of RSAPv3 from novices to experts. The User’s Manual provides several example problems to help new users in learning how to effectively manipulate and use the program and also provided input collection forms. Appendix B is the ENGINEER’S MANUAL for RSAPv3. The Engineer’s Manual provides the technical background behind the code and presents the details of the methodologies used to perform the analyses. The Engineer’s Manual is intended for those who might want to add or update information in the program like new severity models, new trajectories or new vehicle types. Appendix C is the PROGRAMMER’S MANUAL which documents how the program is structured and how it functions from a programming point of view. Most additions and improvements can be made without changing the code by following the instructions in the Engineer’s Manual but for those cases where a coding change is required the Programmer’s Manual will be essential. Lastly, the user survey conducted at the onset of this research is documented in full in Appendix D. These documents fully document the methods, data and use of RSAPv3. RSAPv3 was completely rewritten and coded as an MS Excel macro-enabled workbook. Earlier versions of RSAP used a Monte Carlo method to estimate the probability of a crash but RSAPv3 uses a deterministic method that avoids the Monte Carlo method and the convergence issues associated therewith. The program was written such that nearly all modules can be easily changed by modifying look-up tables rather than recoding. For example, if new research results in a new base encroachment model or a new severity model for a new type of guardrail, these can be incorporated simply by changing the look-up tables within Excel. This feature should make updating RSAPv3 as new research becomes available relatively easy and should ensure that the code will not need to be changed very often. BACKGROUND Motor vehicle crashes cost society more than $230 billion annually [FHWA08b]. On an average day, 117 fatal crashes occur on U.S. roadways, thirty percent of these fatalities are people under the age of twenty-five. In total, this amounts to a societal cost of $630 million lost per day. Run-off-road (ROR) traffic crashes account for almost one-third of the deaths and serious injuries each year on the Nation's highways. There were 41,059 people killed in motor vehicle crashes in 2007, of which 15,506 people (i.e., more than 37percent) were killed in single-vehicle run-off-road crashes [NHTSA07]. Collisions with fixed objects and non-collisions (e.g., rollovers which mainly occurred off the road) accounted for about 19 percent of all crashes but they were responsible for 46 percent of fatal crashes. Inattentive driving, including distracted driving, drowsy driving, or fatigued driving, has been identified as a significant causal factor in crashes of all types.[FHWA01] While inattentive driving is not always identifiable during crash investigations, such behavior is considered by

2 many to be prevalent among a large number of drivers involved in crashes. With the recent trend in “cell phone driving” and “texting while driving,” it has been suggested that inattentive driving is as serious a problem as impaired driving under the influence of alcohol and drugs. [HO08] Inattentive driving and impaired driving have been, and will continue to be, responsible for the majority of inadvertent roadside encroachments and thus run-off-the-road (ROR) crashes. The United States Government has recognized the need to reduce highway related injuries for many years. The first highway safety legislation appeared in 1966. This formal recognition has continued through today, as outlined below: • The Highway Safety Act of 1966: This legislation provided financial assistance to the states to accelerate highway safety program development and reduce highway crashes. This Act required states to develop and maintain a safety program [HSA66]. • The Highway Safety Act of 1973: This legislation established five program areas: highway-rail crossings, high hazard locations, pavement marking demonstration programs, elimination of roadside obstacles and the Federal-aid safer roads demonstration [HSA73]. • The Surface Transportation Assistance Act of 1978: This legislation consolidated the five program areas enacted in 1973 into two programs; the Highway-Rail Grade Crossings and the Hazard Elimination Program [FHWA08d]. • The Intermodal Surface Transportation Efficiency Act of 1991: This legislation was responsible for funding the two programs enacted under the 1978 legislations [FHWA08d]. • The Transportation Equity Act for the 21st Century: This legislation added a provision that states must consider bicycle safety [FHWA08d]. Much of this legislation, including the most recent, also was a means to provide funding for the operation of the U. S. National Highway System (NHS). Many different factors contribute to the cost of operating a transportation system such as the network of roads that comprise the NHS. Costs are realized from the planning of the network through the design, construction and maintenance of the network. Many of these costs are obvious: the design cost, the construction costs and the costs to maintain the infrastructure. Some costs, however, are less obvious and are a result of decisions made during the early stages of designing a new roadway or upgrading an existing roadway. For example, a decision to route an existing stream through a culvert and provide a headwall protected by a guardrail may appear to be the most cost effective decision during a Value Engineering (VE) review. When the potential for vehicles striking the guardrail is considered, however, the societal cost of crashes and increased guardrail repair costs throughout the design life of that section of road may result in costs not considered in the traditional design or VE process. If the safety costs of decisions are included when considering alternatives, the choice of a guardrail and headwall may not be as economically attractive as, for example, moving the culvert intake farther from the road such that a guardrail is no longer necessary. While the construction cost for moving the culvert out of the clear zone would be greater, it may be the best alternative if it results in decreasing the number of crashes and the resulting societal costs. The National Highway System (NHS) Designation Act, passed by Congress in 1995, included a provision that required states to conduct a VE analysis for all NHS projects with an estimated construction cost of $25 million or more [NHS95]. In 1997, VE regulations were

3 published by the Federal Highway Administration (FHWA) establishing the program [VE08] The FHWA’s website devoted to VE analysis suggests that “Federal, State and local highway agencies are responsible for getting the best overall project value for the taxpayer. …VE is an organized application of common sense and technical knowledge directed at finding and eliminating unnecessary costs in a project.” [VE08] The FHWA suggests that design, right-of- way and construction costs shall be included in a VE analysis. Unfortunately, this 1995 NHS Designation Act requiring VE analysis and the Transportation Equity Act for the 21st Century with a focus on highway safety do not share the same goal of increasing highway safety throughout the United States, as the foregoing definition of VE does not specifically address safety. A VE analysis of the design, right-of-way and construction costs is limited in scope and has the potential to overlook or eliminate design features which have the potential to improve safety therefore increasing the potential for crashes and the resulting societal costs. Safety or the potential for crashes is not often directly considered during the planning and design phases of a project, but highway designers often rely on established design standards as a means for producing a “safe” design. These established design standards, including the American Association of State Highway and Transportation Officials’ (AASHTO) A Policy on Geometric Design of Highways and Streets (i.e., the Green Book) and the Roadside Design Guide (RDG), also published by AASHTO, provide the designer with criteria for designing a roadway [AASHTO01; AASHTO06]. While these guidelines are simple to follow, they may not result in the safest designs. For example, the Green Book suggests in Exhibit 3-26 that for a design speed of thirty miles per hour (30 mph or 50 km/h) a minimum horizontal curve radius of 3,350 feet (1,110 meters) should be used when there is no super-elevation [AASHTO01]. This criterion does not consider the interaction of the horizontal curvature with other highway characteristics like the vertical alignment, the clear zone, the number of lanes, the side slopes or the expected traffic. The horizontal alignment is considered independent of the other characteristics of the road even though common sense would suggest that all the highway alignment characteristics are interrelated. Mathematical models have been developed to predict where crashes might occur along the road and the roadside as a function of traffic and geometric characteristics. These complex statistical models often consider the vertical and or horizontal alignment of the highway, the placement of roadside objects, the speed of the traffic and other factors in relation to each other. The use of these models during planning and design in conjunction with established design standards will bring the issue of maximizing highway safety to the forefront of the highway design process. An informed discussion of the cost of changes to an alignment can be assessed over the design life of a highway with the economic impacts of safety also included as a factor in the analysis. These models, however, are complex and scattered throughout various literatures and are not easily accessible to the engineer or planner who works on highway designs every day. Additionally, they are not easy to integrate into the typical highway design process which uses computer software to generate detailed designs. Highway engineers are constantly designing and building, and redesigning and rebuilding roadways to meet higher standards to provide ever safer highways with increased mobility. This includes designing and building roadways that are more forgiving when a driver inadvertently encroaches onto the roadside. There are, however, impediments that keep highway engineers from achieving the desired design and operational goals of safety and mobility including the need to operate, maintain, and improve a vast highway system with limited resources. Consequently, highway engineers are often required to make incremental improvements over

4 time and make difficult trade-offs between cost, safety and mobility. Modeling tools that enable highway engineers to explicitly evaluate the costs and benefits of highway design decisions are essential in making these trade-offs. Historically, models of the relationships between ROR crashes and roadside features, such as utility poles, traffic sign posts, trees, guardrail, median barriers and side slopes, have been categorized as either crash-based or encroachment-based approaches. The following sections describe, compare and contrast these basic analysis approaches. ENCROACHMENT APPROACH One alternative available for making decisions about roadside design is the encroachment approach. There is a long history of the use of the encroachment approach for cost-benefit and probabilistic methods in roadside safety dating back to at least 1974. Not surprisingly, the literature devoted to cost-benefit methods in roadside safety, crash prediction modeling, encroachment modeling, and severity modeling is extensive going back as far as Hutchinson in 1962 and proceeding right up to present-day ongoing NCHRP projects like Projects17-11 and 17-22.[Hutchinson62, Bligh08] In his PhD dissertation, Ray provided an extensive review of the research related to probabilistic modeling of roadside crashes as of 1993.[Ray93] Turner wrote an NCHRP Synthesis report in 1994 collecting the latest research on severity indices of roadside features. [Turner94] In addition there are several ongoing NCHRP projects that either are using RSAP or are directly working on improved methods that could be integrated into a new version of RSAP in the future. Probably the first researcher to suggest the use of risk-based cost-effectiveness analysis for roadside safety was Glennon in 1974.[Glennon74] Glennon noted that the warranting approach, the approach where a set of guidelines about the selection, location and placement of roadside hardware, failed to provide an assessment of priorities or effectiveness. If there were unlimited funding, then all roadside safety problems could be treated using the warranting approach. Of course, there is never sufficient funding so an approach was needed to prioritize possible alternatives in order to select the roadside design alternative that would make the greatest improvement in safety at the least cost. This would allow designers and highway agencies to maximize the benefit of scarce construction and maintenance funding. Ross included Glennon’s basic risk-based cost-benefit procedure for roadside safety in Chapter VII and Appendix E of the 1977 Barrier Guide, the first document promoted by the American Association of State Highway and Transportation Officials (AASHTO) that provides guidance on designing the roadside. [AASHTO77] The method as presented in the 1977 Barrier Guide was not particularly practical for roadside design practitioners since there was a lot of tedious hand calculation required. In 1988, AASHTO revised the 1977 Barrier Guide transforming it into the Roadside Design Guide. [AASHTO88] Appendix A of the Roadside Design Guide included a revision of the cost- effectiveness procedures and provided a computer program called ROADSIDE to ease the calculation burden on designers and policy makers. The ROADSIDE program was essentially a direct computer implementation of the procedure in the 1977 Barrier Guide as proposed by Glennon. A similar cost-effectiveness procedure called BCAP was included in the 1989 AASHTO Bridge Specification for designing bridge railings.[AASHTO89] In their day, ROADSIDE and BCAP were innovative implementations of encroachment-based and risk-based models for cost-benefit design of roadsides. Of course, as computer applications became more sophisticated and additional research was performed to refine and improve encroachment models, severity indices and other aspects of the procedures, it became apparent that a new

5 computer program was needed. The resulting program, RSAP, was completed in 2003 and documented in NCHRP Report 492 by Mak and Sicking. [Mak03] The basic RSAP procedure was included in the 2002 revision of the AASHTO Roadside Design Guide and has remained a feature of subsequent editions of the Roadside Design Guide ever since. [AASHTO02] Rationale The encroachment-based approach divides the encroachment event into a series of sub- events; from vehicle departure to collision, to the severity of the crash and the determination of expected benefit-cost ratios for various alternatives. The approach uses a conditional probability model consisting of a series of conditionally independent probabilities which quantify possible consequences of inadvertent encroachments onto specific roadside design conditions at various stages of the encroachments. In terms of evaluating the benefit-cost of alternative roadside designs, there are basically two. arguments favoring the use of encroachment-based approach over the crash-based approach. The first argument is that not all roadside encroachments result in crashes; those that do result in crashes do not necessarily always get reported; and those that do get reported do not always make it into the traffic crash database. Thus, examining the reported traffic crash records alone may not allow the full benefits of good roadside designs/features to be realized. This argument can be better understood when encroachments are classified by possible outcomes and crash reporting requirements as follows. • No Crash: The encroaching vehicle safely returns to the traveled way without serious vehicle damage and without occupant and non-occupant (i.e., pedestrians and cyclists) injury or death and without striking anything; or, putting it differently, the driver is able to either successfully stop the vehicle or regain control of the vehicle without striking a roadside object, incurring serious vehicle damage or causing injury or death to vehicle occupants and non-occupants after encroaching into the roadside. Often, the driver simply drives the vehicle away and no one knows that a potential crash has occurred. • Unreportable Crashes: The encroaching vehicle experiences only minor damage and the collision does not result in an injury or death and the damages do not exceed the lawful reporting threshold in the jurisdiction (e.g., drivers must report crashes if $1,000 or more damage results from the crash in states like Michigan and Wyoming whereas only crashes involving injuries or towing need be reported in Texas). An example of this type of crash is a vehicle that leaves the roadway, strikes a cable median barrier but is so lightly damaged that it is able to drive away. • Unreported Crashes: Some encroachments result in what should be a reportable crash (i.e., the damage exceeds the legal threshold) but the crash is not reported for a variety of reasons (e.g., vehicle leaves the scene or is towed without notifying the police). Previous studies have found that many reportable crashes involve hitting safety features, especially more forgiving breakaway devices such as sign posts. For example, a vehicle may leave the roadway and strike a luminaire support. The luminaire may activate and cause damage to the vehicle, but the driver may be able to have the vehicle towed from the scene and may elect not to report the crash to the police even though the damage to the vehicle and luminaire exceed the reporting threshold. This is particularly common for single-vehicle run-off-road crashes since there is only one vehicle involved and the driver may see no need to notify the police.

6 • Reported Crashes Not Recorded into Database: Depending on the particular state, some reported crashes, especially minor single-vehicle crashes, may not be entered into the state computerized traffic crash database. Generally states do this to minimize the cost of maintaining the database. As an example, a single vehicle may run off the road and strike a guardrail. Even if the police are notified, if the damage to the guardrail is minor, the police report may not be entered into the State-wide database so a record of the crash is lost. • Reported Crashes Recorded into Database: When the encroaching vehicle leaves the roadway and strikes an object such that the vehicle is damaged and/or vehicle occupants or non-occupants are injured, the crash is generally reported to the police. Oftentimes, this is the only data that is available to the researchers in studying traffic safety. Setting aside how states maintain their crash databases, it should be clear that the first argument centers on the potential benefits of good roadside designs. Good roadside design provides a better chance of resulting in the first three types of encroachments: no crash, an unreportable crash, or an unreported crash. It is a fairly reasonable assumption that crashes resulting in injuries will be reported. Relying only on reported crashes can distort the effectiveness of roadside hardware since injury and larger amounts of property damage will be over-represented in the data. For example, say that before a crash cushion was used in front of a particular bridge pier, there were ten reportable crashes in a five year period, five of which caused injuries. After a crash cushion is installed, only two crashes are reported, one of which involves injury but eight additional crashes occurred but were never reported to the police because the crash cushion was effective in preventing occupant injuries or serious vehicle damage. Based on the reported data, the crash severity for the treated and untreated site are the same (i.e., 50 percent injury rate) when, in fact, the crash cushion reduced the severity of eight of the crashes to a level where they were no longer reported at all. The crash cushion really had a 10 percent injury rate, five times less than the untreated site. Crash-based data analysis would by definition miss the unreported low severity crashes that represent successful roadside hardware design. The second argument embodied in the encroachment-based approach is that the crash- based approach has not been able to provide the kind of detailed relationships required to conduct the necessary benefit-cost evaluation. Roadside conditions can be complex and variable along the roadway and thus their safety effects can be difficult to quantify. In evaluating roadside safety improvements, many roadside safety features and their natures and configurations need to be considered for different classes of roadways which have different traffic volumes and posted speed limits. Typical roadside features include foreslope, backslope, parallel ditches, intersecting slopes, fixed objects, breakaway sign posts, culvert ends, longitudinal barriers, terminals and crash cushions, shoulder rumble strips and pavement markers, to name just a few. The encroachment-based approach provides a probabilistic framework that allows for more detailed modeling of vehicle roadside interactions (e.g., various driver input, including braking and steering, behaviors), manner of collision with roadside objects (e.g., a combination of impact positions, speeds, and angles), and consequence of collisions (e.g., damage/injury severity distributions for striking different types of objects). This probabilistic modeling framework is more sensitive to the safety design parameters associated with complex roadside features and configurations and the needs of roadside safety evaluations. The price that must be paid for this increased analysis capability is that the detail required by the

7 encroachment-based approach demands more detailed data which may be difficult to obtain, or, require more assumptions about the way crashes occur which may be difficult to validate. CRASH-DATA APPROACH Crash-data based approaches generally take advantage of police-level crash reports that are collected by all the states. The main strength of using the crash-based approach is the sheer size of the available crash and road inventory data which are collected and maintained routinely by state Departments of Transportation (DOTs), Departments of Public Safety, and the National Highway Traffic and Safety Administration (NHTSA). The limitation of using this real-world crash data is its lack of detail regarding the roadside, vehicle, and collision conditions. In addition, minor crashes tend to be underreported and/or under-coded in the state traffic crash database so lower severity crashes are usually not included. Since the 1980s crash-based roadside studies, including many conducted as part of the Federal Highway Administration’s (FHWA) Interactive Highway Safety Design Model (IHSDM) program discussed in a later section, have largely been limited to a small number of roadside features or some sort of roadside hazard indexes on rural two-lane roads and intersections. The crash-based approach relies mainly on statistical regression models to develop macro-level relationships between reported ROR crashes and associated traffic, highway geometric and roadside characteristics. Typically the regression models are formulated to predict reported ROR crash frequencies by severity for a selected set of sites (e.g., typically intersections or road segments). The crash severity predictions are the dependent or outcome variables and traffic, highway geometrics, and roadside characteristics associated with these sites are the independent or explanatory variables. The predictive values from these regression models, together with estimated uncertainties, are then used in risk and benefit-cost analysis. In the last four decades there have been dramatic developments in the statistical and biomedical sciences on specialized statistical regression methods for handling discrete/categorical types of response data, such as event frequency or event count data like highway crashes. These developments “reflect the increasing methodological sophistication of scientists and applied statisticians, most of who now realize that it is unnecessary and often inappropriate to use methods for continuous data with categorical responses.” [Agresti96] Many textbooks have been published to introduce these new techniques under such titles as categorical data analysis, generalized linear models, models for discrete data, models for limited-dependent variables, and, recently, generalized linear mixed models. In addition, statistical procedures for carrying out these regressions are now available or programmable in many popular statistical software packages like SAS and SPSS. The advent of statistical techniques for treating discrete response data has triggered an explosive number of crash-based studies with more appropriate statistical regression models and associated goodness of fit measures in the last 20 years [Miaou92; Miaou93; Miaou94]. In these newer crash-based studies, the focus has largely been on traffic volume and highway design characteristics, such as AADT, lane width, horizontal curvature, grade, and different lane and intersection configurations. The promising results from many of these studies over the past several years have led to four NCHRP projects under an extensive initiative by the Transportation Research Board (TRB), FHWA, and AASHTO aimed at developing the first edition of the Highway Safety Manual (HSM).