Roadway Safety Tools for Local Agencies (2003)

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.

7Ultimately, any significant reductions in roadway fatalities and crashes will require the mitigation of safety concerns along roadways under the jurisdiction of local government entities (e.g., cities, towns, counties, and villages). Effective local road safety training, technology transfer, and the resources necessary to complete these safety improvements are essential. To properly implement improvements to reduce roadway fatalities and injuries, local roadway managers and staff need to be able to • Evaluate the safety performance of their transportation network, • Identify the key locations of safety concern, • Compare the effectiveness of possible solutions, • Plan and design a chosen improvement, • Obtain appropriate funding, • Implement the safety improvement, and • Evaluate the improvement. One such mitigation approach is to evaluate crash data on the local network. This approach is reactive in that safety de- cisions emanate from a review of crash histories. To evaluate if safety improvements are needed, decisions are made based on evaluating locations where a high crash situation exists. The analysis of crash data is considered at two levels, basic and advanced. The difference between the two lies in the so- phistication of the application of statistical analysis. BASIC REACTIVE ANALYSIS TOOLS Crash analysis forms the basis of this reactive approach. Essential elements for a crash analysis program begin with decisions on the types of crashes—property damage only (PDO), personal injury, and fatal, including a means of locat- ing where crashes occur. After these decisions have been made, the locations for a more detailed analysis are identified. The initial decision is to determine the types of crashes to include in the analysis, specifically if PDO crashes should be included. Often there is a threshold value (dollars) before a PDO crash is even reported as an accident (crash) in state records. It is suggested that the policies of the state be used to help make this decision. States typically have their own reporting forms that are used when a crash occurs. The use of the term “accident” has been replaced with the term “crash,” because there are causes associated with a crash, whereas the perception of an accident is that it is a random event without a cause. These crash forms are filled out by the police responding to the scene, in addition to the forms filled out by driver(s) and others involved. The crash forms contain location information, date on the type of crash, contributing factors, and other parameters of involvement, and should provide the basis for accurately locating the crash. Accurately locating crashes is crucial to these records. When working with these data it is often necessary to ensure that the crash record applies to the spot or section being analyzed. The emerging use of geographic information system (GIS) and Global Positioning System (GPS) technology is helping to overcome this problem. If a local agency is considering such a system, a review of “GIS- Based Crash Referencing and Analysis System” could be of assistance in making the decision (39). There are a number of different techniques for locating crash sites, beginning with a simple map of the area and the place- ment of a pin where the crash occurred. The greater the num- ber of pins placed in one location, the greater the frequency and hence the greater the potential for the location to be clas- sified as a high crash location. The missing information in these data is a reflection of traffic volumes—not just total volumes, but also conflicting volumes, at locations such as intersections and driveways. Typically, one would expect lo- cations where there are higher volumes and more conflicts to have a higher potential for crashes. The consideration of the influence of volume impacts is often factored into methods of identifying high crash locations. Depending on the level of analysis sophistication, there are ways to evaluate one location relative to similar traffic volume locations to deter- mine if the location is a high crash location. Factoring in traffic data provides a greater degree of predictability. The value in a basic reactive analysis technique is to deter- mine if there are clusters of like crashes. For example, there is generally a limited ability to look at all crashes on the net- work and to conduct a detailed survey of all crash location sites. Therefore, a reactive program, based on a degree of certainty that problem locations are being identified and treated, is important. It is also important to state that just beginning a reactive crash analysis program will provide benefit. If any location is improved by evaluating crashes, then there is a good chance that crash reductions will occur. In essence, the message is to just do it. Concern about the details should not matter. CHAPTER TWO REACTIVE SAFETY TOOLS

8There are several important publications that provide more information on the reactive approach (9,43–46). Key points identified in these references include the following: • Cluster analysis requires subjective judgment. Because the analysis depends on the knowledge and judgment of the individual conducting the analysis, individuals using the same data may reach different conclusions. To over- come this difficulty, agencies often identify threshold numbers of crashes to aid in determining if the reflected location is a high crash location. • Is there a pattern of like crashes? Are these types of crashes the types expected with the type of traffic con- trol provided at the intersection? For example, one would expect a signalized intersection to be more likely to have a higher incidence of rear-end vehicle crashes and a lower number of right-angle crashes than other locations. Even with this knowledge, are there other factors that may contribute to this statistic? For the previous example, is the clearance interval of the signal cycle too short? • Implied in the first two points is that the crash spot (inter- section) has been correctly identified. Checking the actual crash records is generally necessary when con- ducting an in-depth analysis of a high crash location. Statements made by the police and others involved become helpful in understanding factors that contrib- uted to the crashes and what types of safety actions may reduce the incidence of these types of crashes. • A threshold number of crashes may also be used to establish a decision starting point. This number helps in deciding if a more detailed analysis is justified. These threshold numbers generally depend on the crashes asso- ciated with the different types of areas and also the different types of facilities. The city or size of the local area plays a major role in reaching a decision. • In many situations, crashes are not reported unless there is an injury, fatality, or major property damage. Local road users can often provide good information on unreported crashes, other factors, and special conditions that may have existed at the time of the crash. These data should not be overlooked in the analysis. • There are similar approaches employed for linear or road- way segments, although crash clusters most often occur at intersection locations. If road segments are used in the analysis, it is generally a good idea to normalize the effect of length and also traffic volume. • A cluster of crashes involving like occurrences may or may not be abnormal depending on factors such as inter- section geometry, environmental conditions, and special events. Learning the details associated with the crashes and crash patterns helps to isolate these factors. • The experience of the individual conducting the analysis, the consistency of decisions, and continued evaluation of the effects of the improvements makes reactive analysis a potentially valuable tool. The tables contained in Appendix H have been provided to aid in decisions regarding the potential countermeasures and the issues associated with the clusters of crashes that may occur. • In the investigation of intersection crashes, approach speeds, vehicle types (trucks, passenger cars, etc.), and prevailing sight distances at the corners of the inter- sections are important factors to be considered. At rural intersections, a lack of recognition that the vehicle is approaching an intersection may be a factor. This may be particularly true during certain times of the year when crops, trees, or other vegetation obscure the definition of a crossroad or other junction, such as a railroad cross- ing. Well-maintained advanced warning signs are essential to prevent rural crashes in these locations. Morning or evening sun glare also may be a significant contributing factor. • Maintenance factors such as adverse pavement condi- tions may also be identified through crash analysis. These situations may result in clusters of incidents during certain times of the year. • Many of these safety issues will be identified in the specific analysis of a location, whether or not it is a high crash location. Even if a lower incident location were selected, the decisions and implementation of the improvements may reduce crash potential. These factors have not taken into consideration the num- ber of years of crash data that should be used in the analysis. That is because an in-depth analysis takes time and resources, and that concentrating on the worst locations will potentially provide the greatest safety benefit. How many years of data are needed to normalize trends? Is 1 year’s worth of crash data adequate? What is the ideal number of years of data? To answer these questions, the location must also be carefully evaluated for changing conditions that may affect the magni- tude of the crash picture. In most cases, a 1- to 3-year period is considered adequate. If there is a major change in the net- work or the traffic volume, or the facility in question has been under construction, then a briefer period may be appropriate. It is generally accepted that a 3-year period of crash data is ideal for a before evaluation of crash clusters. Once a change or improvement has been implemented, then the question to ask is, “How effective is the change?” Evalu- ating the effectiveness of a change is based on a before-and- after statistical methodology. If the change produces the desired result, then implementing these changes at similar locations would be warranted. The after evaluations and before-and-after analysis decisions need to reflect the following: • How much time is needed after the change has been made before beginning the after period of time? In essence, an initial benefit may appear to occur, but the effect diminishes with time as the traffic picture changes. A

9brief period of time is often used (1 to 3 months) and these immediate effects are not included in the analysis. • What changes have occurred in the area and/or adjacent to the facility that may have affected the results? • What may have introduced other bias into the analysis results? • An analysis of the crash history at similar locations is often used to address such questions. This analysis involves locations where no change was made. Adding these sites provides a greater degree of certainty when evaluating the effectiveness of the improvement. Given the discussion of reactive analysis issues and the potential analysis pitfalls in drawing the wrong conclusions, is this an effective safety tool for local agencies? Is there a good way for local agencies with limited expertise and prac- tice in reactive analysis to employ the crash analysis reactive safety tool? How can local agencies achieve the greatest ben- efit from reactive analysis? The following guidance is pro- vided for the use of reactive tools. SUGGESTIONS FOR LOCAL AGENCIES TO CONSIDER IN APPLYING BASIC REACTIVE SAFETY TOOLS The suggestions that follow are provided to aid local jurisdic- tions in maximizing the benefits of including reactive analy- sis in an affordable approach to improving their safety pro- grams (9,43,45,47). 1. Knowing and identifying local crashes is important to improving local road safety. Care needs to be taken to ensure that crash reports reflect the location being studied. Consider using the tools of GIS and GPS to locate crashes and to ensure that these data are accurate (39). 2. Contact your state DOT to see if your local jurisdiction can receive a yearly summary of crashes. If the answer is yes, then request a map display of the crashes by loca- tion and type of crash and proceed to Step 7. Again, the use of GPS and GIS technology is valuable to location accuracy. Post a map of the network of roads and streets in your local jurisdiction. 3. Contact all law enforcement jurisdictions that may investigate crashes on this local network and request that crash reports be sent to you. 4. On a map, plot each crash by type of crash by using different colored pins to reflect the different types of crashes. 5. Follow your state’s reporting requirements for PDO crashes. [See the primary annotated references (Appen- dix C) for different ways to plot crashes and examples of different types of these pin maps.] 6. Identify the locations with the highest number of crashes. (See Appendix H for examples of crash maps, as well as collision and condition diagrams.) 7. Determine if there is a pattern of crashes. Appendix H contains an example of a collision diagram and a condi- tion diagram that are used to evaluate the patterns of crashes and also link the patterns to the conditions at the site. These diagrams are easily developed by using aerial photographs or plan maps or by visiting the site being evaluated. Check to see if your state DOT will provide computer-generated spot maps and collision diagrams and their review of high crash locations. This service was indicated as being available in some states, although often only for cities with larger populations. 8. Identify possible issues associated with the crash loca- tion. Determine if there are site-related factors that cause crashes to occur there. This step may require an on-site visit or the evaluation of aerial photographs of the location. 9. Consult referenced listings of crash types and pos- sible countermeasures, which are also contained in Appendix H. 10. Implement the selected decision. Even though it may not be the ideal implementation decision, identify what was done, when it was done, and what other improve- ments were considered. 11. Document the agency’s use of a reactive safety program tool, the immediate implementation decisions, and the actions that are desired when additional resources become available. The listings of alternative treatments contained in Appendix H are based on analysis of before-and-after crash data conducted in a number of different locations. These nationally recognized successful alternatives provide a simple decision process. Using crash data may also be undertaken with a more rigorous analysis. ADVANCED REACTIVE ANALYSIS TOOLS This section presents statistical analysis options for reactive analysis beyond that of relying on nationally established improvement alternatives. The issue for local agencies to con- sider is the added benefit gained from methodologies that focus on a more in-depth analysis of local data. Many larger jurisdictions with adequate staffs and larger volumes of crashes may benefit from an approach based on using their local crash data. Many of the issues presented in the basic section for reactive tools also apply to advanced methods of analysis. It should be noted that good statistical analyses require high-quality input data. Again, state DOTs are the best initial source for data. Several states provide yearly sum- mary reports and special location analysis upon request from a local agency. Those documents enable the development of high-quality input for analysis. They also provide excellent starting points for reviewing crash histories within the com- munity. The analyses are also made with or without control sites. It has been generally accepted that the inclusion of

10 control sites with similar characteristics benefits the reliability of the analysis by normalizing changing conditions such as traffic growth and weather effects. It is important to recognize that significant expertise is required to conduct sound before-and-after safety analyses. As indicated by Mike Griffith of the FHWA, nationally, analyses of crash data provide local agencies with excellent links to crash causation and probable crash reduction tech- niques for patterns of crashes (9). In 1999, the Institute of Transportation Engineers (ITE) produced a report, Statisti- cal Evaluation in Traffic Safety Studies (48), which also pro- vides an excellent background for local agencies in making the decision on conducting their own local advanced reactive analysis. In assessing the benefits of conducting advanced analyses, it is recommended that local jurisdictions look closely at the established relationships between crash data and possible treatments based on rigorous analyses that have been conducted nationwide (use the experience- and knowledge- based solutions that are presented in Appendix H). Another excellent source for more advanced reactive safety analysis is NCHRP Report 440: Accident Mitigation Guide for Congested Rural Two-Lane Highways (43). Although this report focuses on rural two-lane highways, the recommended approach is applicable for any road or street network and is also complete in describing a crash mitigation process. The following six specific steps are identified in the report: 1. Identify sites with potential safety problems, 2. Characterize crash experience, 3. Characterize field conditions, 4. Identify contributing factors and appropriate counter- measures and select the most appropriate, 5. Implement the most appropriate countermeasure, and 6. Evaluate the effectiveness. The higher level of statistical reactive analysis compares and evaluates the effectiveness of different countermeasures. Documentation is an essential component of this process. Measuring whether or not the countermeasure was effective is also essential to the process. Although, as mentioned, the focus is on two-lane rural roads, the examples provided are often urban situations, consistent with the concept that most crashes are related to intersection conflicts. The following two lists derived from Tables 2 and 3 of NCHRP Report 440 were modified to reflect crash terminol- ogy and are provided for consideration of the concepts asso- ciated with the different ways in which a high crash location may be determined. The focus becomes identifying sites which are likely candidates for crash countermeasures as opposed to identifying high-crash locations. In some respects the concept reflects a shift to using crash data in a more proactive manner. The conventional methods require a large set of data, expensive to perform every year, sub- ject to regression to the mean biases, and likely to identify sites with no obvious cost-effective remedy (43). Crash Evaluation Methods Number of crashes—the number of crashes at a location. Locations with more than a predetermined number of crashes are classified as high crash locations. Crash density—the number of crashes per unit length for a section of highway. Sections with more than a predeter- mined number of crashes are classified as high crash loca- tions. Crash rate—crash numbers divided by vehicle exposure to provide rates such as crashes per million entering vehicles per spot location and crashes per million vehicle-miles for sections of highways. Locations with higher than a predeter- mined rate are classified as high crash locations. Number rate—a combination of number of crashes and crash rate. Locations with more than the prescribed mini- mum number of crashes and higher than the minimum crash rate are classified as high crash locations. Number quality control—locations with the number of crashes that is greater or significantly greater than the aver- age number of crashes for the state or a similar region. Rate quality control—locations with a crash rate that is greater or significantly greater than the average crash rate for the state or a similar region. Crash severity—the number of fatal and/or injury crashes at a location or per unit length for a section of highway. Severity index—the number of fatal and/or injury crashes at a location or section of highway being given a greater weight than PDO crashes. Crash index—combining some of the aforementioned methods to determine a single index value for a group of sites. Crash Rate Method This method is practiced through these steps: 1. Locate all crashes in accordance with accepted coding practices. 2. Identify the number of crashes in each established section and at individual intersections and spots. 3. Calculate the actual crash rate for each established section during the study period. Rate/MVM = (no. of crashes on section) (106) (ADT) (no. of days) (section length) where MVM = million vehicle-miles, and ADT = average daily traffic.

11 4. Calculate the actual crash rate for each intersection or spot during the study period. Rate/MV = (no. of crashes at intersection or spot) (106) (ADT at location) (no. of days) where MV = million vehicles, and ADT = average daily traffic. (ADT at location represents the sum of all vehicles enter- ing the intersection.) 5. For the same period, calculate the systemwide average crash rates for sections, intersections, and spots, using the formulas in steps 3 and 4 and the summation of total crashes, total vehicle-miles, and total vehicles, respec- tively, for each category of location. 6. Select crash rate cutoff values as criteria for identifying high crash locations. A value approximately twice the systemwide rate usually is realistic and practical. 7. If actual rates exceed the minimum established criteria, the location is identified as a high crash location and placed on the list for investigation and analysis. Selection of the cutoff value (Step 6) is not as critical as it might appear. The principal purpose is to control the size of the list of locations to be investigated—a shorter list with high values or a longer list with low values. Experience will indi- cate the appropriate level for a particular agency. Another excellent informational table in NCHRP Report 440 is Table 10, which identifies basic field observations used to study problem locations. These include exercising good judgment while simply driving through the location. Also, considering the issues associated with the site, such as the conditions at nighttime and at nighttime with opposing traffic, is an important step toward ensuring a thorough evaluation. An on-site observation report form is also included in this study. [Table 10 and the on-site report (Figure 3) from NCHRP Report 440 are contained in Appendix H of this synthesis.] Table 11 in NCHRP Report 440 points out other supple- mental studies that may provide more detailed information to assess the relationship of the crash data and the conditions that may indicate the need for a particular type of study. The final series of informational tables in the report are the listing of contributing factors for crashes and the potential counter- measures. (These are contained in Tables 14 and 15 and also provided in Appendix H of this synthesis.) Reactive crash analysis provides an excellent basis for a safety program, whether conducted at the simplest or most complex stages. The improvement alternatives that have been developed when applying the reactive safety tool also result in excellent implementation decision aids for the proactive tools discussed in the next chapter.