Guidance for the Design and Application of Shoulder and Centerline Rumble Strips (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.

92 S E C T I O N 7 The safety evaluation of centerline rumble strips addresses two key unresolved issues: (1) the safety effectiveness of center- line rumble strips on different roadway types, and (2) the safety effectiveness of centerline rumble strips along varying road- way geometry. Also addressed in lesser detail is the safety effec- tiveness of dual applications of rumble strips (i.e., centerline and shoulder rumble strips installed along the same roadway section). Previous safety evaluations of centerline rumble strips have focused on the effectiveness of this treatment on rural two-lane roads. NCHRP Report 617 (28) indicates that center- line rumble strips reduce all crashes by 14 percent and reduce head-on and opposite-direction sideswipe crashes by 21 per- cent on rural two-lane roads. It also indicates that centerline rumble strips reduce all injury crashes by 15 percent and all injury head-on and opposite-direction sideswipe crashes by 25 percent on rural two-lane roads. NCHRP Report 617 specifically states that these expected safety estimates are only applicable to rural two-lane roads and are not applicable to other roadway types. As evident from the survey results in Section 5 of this report, in more recent years center- line rumble strips have been installed along other types of roadways, including urban multilane undivided highways (nonfreeways), urban two-lane roads, and rural multilane undivided highways (nonfreeways). Prior to the current research, it was not known whether the same safety benefits of centerline rumble strips should be expected on these other types of roadways. In all likelihood, the safety benefits of centerline rumble strips vary by roadway type because the different types of roadways are built to varying standards (i.e., lane widths, shoulder widths, etc.), accommodate vary- ing traffic volumes and distributions, and serve different driver populations. Studies concerning the safety effectiveness of centerline rumble strips have not distinguished between the safety effec- tiveness of centerline rumble strips along tangent sections as compared to horizontal curve sections. It should be recognized that the effectiveness of centerline rumble strips in reducing head-on and opposite-direction sideswipe crashes is depen- dent upon various elements, including the frequency with which vehicles cross the centerline, the angle at which vehi- cles cross the centerline, vehicle speed, lane width, and traf- fic volume (i.e., directional distribution). It should also be recognized that approximately 35 percent of all fatal acci- dents occur along horizontal curves as compared to 65 per- cent on tangent sections (7). Thus, there is reason to believe the safety benefits of centerline rumble strips differ along varying roadway geometry. Previous safety evaluations of rumble strips have focused on one type of application per study (i.e., either shoulder or centerline rumble strips). No study has investigated the com- bined safety effectiveness of installing both centerline and shoulder rumble strips along the same section of roadway. As evident from the survey results in Section 5, several state trans- portation agencies have been installing both shoulder and centerline rumble strips along the same section of roadway, but the safety effectiveness of this dual application treatment has not yet been determined. A safety evaluation was conducted to address the following key unresolved safety issues related to centerline rumble strips: • Do centerline rumble strips have the same safety effective- ness when installed along different types of roadways? • Does the safety effectiveness of centerline rumble strips dif- fer depending upon the roadway geometry (i.e., tangent vs. curve)? • What is the safety effectiveness when centerline rumble strips are installed in conjunction with shoulder rumble strips along the same roadway? This section describes the general scope of the safety eval- uation conducted to resolve these issues, the site selection process, the videolog data collection procedures, the database development, the analysis approach, and the analysis results. Safety Effectiveness of Centerline Rumble Strips

93 In general the same methodology used to perform the safety evaluation of shoulder rumble strips was used to perform the safety evaluation of centerline rumble strips. Scope of Safety Evaluation The primary objectives of the safety evaluation conducted as part of this research are to do the following: • Quantify the safety effectiveness of centerline rumble strips on specific types of roads including urban multilane un- divided highways (nonfreeways), urban two-lane roads, rural multilane undivided highways (nonfreeways), and rural two-lane roads; • Quantify the safety effectiveness of centerline rumble strips along varying roadway geometry (i.e., tangent vs. horizon- tal curve); and • Quantify the safety effectiveness of dual applications of rumble strips (i.e., centerline and shoulder rumble strips installed on the same road section). From the survey conducted in Phase I, the research team identified six states to initially contact to gather more informa- tion concerning their potential involvement in the safety eval- uation of centerline rumble strips. The intent was to involve up to three states so that data could be collected over a wide range of roadway types, ideally in different regions of the country. The primary data collection effort involved review of videologs to verify the presence or absence of centerline rumble strips (and in some cases shoulder rumble strips) and to collect (or verify) roadway characteristic data. The data were analyzed using the EB methodology for before-after analysis, similar to the EB methodology used in safety evaluation of shoulder rumble strips. Site Selection Representatives from Missouri, Oregon, Pennsylvania, Utah, Virginia, and Washington DOTs were initially contacted to inquire about potentially including sites from their respective states in the centerline rumble strip safety evaluation. A simi- lar site selection process was followed for the centerline safety evaluation as previously described for the shoulder rumble strip safety evaluation. The type of information gathered to determine whether sites from a particular state would be appropriate for this study included the following: • Has the agency installed centerline rumble strips covering the range of roadway types of interest for this study and has the agency installed any dual applications of both center- line and shoulder rumble strips at the same site (preferably installed in the same year)? • Does the agency have the ability to identify locations where centerline rumble strips and the dual applications have been installed? • Does the agency have the ability to readily provide con- struction history information, such as rumble strip instal- lation dates and information about other improvements made during the study period? • Does the agency keep a library of videologs that could be accessed by the research team? • Is the agency willing to participate in the research and work with the research team to gather the necessary data for the safety evaluation? The research team found during these initial inquiries that most state transportation agencies (a) had no readily available means to identify the locations of the installations such that they could not even provide an initial list of potential treatment sites for further investigation by the research team, (b) had only installed a limited number of miles of centerline rumble strips on roadways other than rural two-lane roads and in some cases the centerline rumble strips were installed on the other road- way types as part of a pilot project, or (c) provided locations of the potential treatment limited strictly to rural two-lane roads. Through these initial inquiries, a list of potential treatment sites with centerline rumble strips was obtained from Pennsyl- vania and Washington. The list of potential treatment sites for Pennsylvania included sites for all four roadway types of inter- est, and the list for Washington included primarily sites for rural two-lane roads, but also included a few sites on the other roadway types of interest. In a final effort to find additional sites with centerline rumble strips on roadways other than rural two-lane roads, MnDOT was contacted. MnDOT indi- cated in their survey response that they only installed center- line rumble strips on rural two-lane roads, but due to prior experience with the shoulder rumble strip safety evaluation and MnDOT’s proactive approach concerning both shoulder and centerline rumble strips, further inquiries were made. This inquiry yielded a list of potential treatment sites but only for rural two-lane roads in Minnesota. To ensure a geograph- ically representative set of sites, the decision was made to include sites from Minnesota, Pennsylvania, and Washington in the safety evaluation of centerline rumble strips. The information provided by all three state DOTs with the list of potential treatment sites included location information and installation dates of the centerline rumble strips. The road- way type was determined from the states’ roadway inventory data. For each state, it was also understood that the installations of the centerline rumble strips were the only recent improve- ments (safety or otherwise) made to the sites. The following sections summarize the tasks conducted for the respective states to select treatment sites (i.e., sites with centerline rumble strips or sites with centerline and shoulder rumble strips) for

94 inclusion in this safety evaluation. The Videolog Data Collec- tion part of this section provides detailed information on the actual data collection process performed for treatment site and nontreatment sites (i.e., sites without any type of centerline or shoulder rumble strips) for inclusion in this safety evaluation. Minnesota Sites MnDOT provided an initial list of treatment locations. The information provided with this list included the district, route type, route number, beginning and ending mileposts, and installation dates. The research team reviewed each of the sites using MnDOT’s videolog system, which contains videologs for calendar years 2001–2006. Initially, the 2006 videologs were reviewed to confirm the presence or absence of the centerline rumble strips at the locations. Subsequently, the 2001–2005 videologs were reviewed to confirm the installation dates of the rumble strips at each site where the centerline rumble strips were installed during calendar years 2002–2005. In the list of treatment sites, the installation dates of the centerline rumble strips ranged from 1996 to 2004. Only those installation dates between calendar years 2002 and 2005 could actually be con- firmed through this process by verifying the absence/presence of the centerline rumble strips across multiple years. This process of confirming the installation dates revealed that most of the installation dates from the initial list were cor- rect. Therefore, it appeared reasonable that the older instal- lation dates were accurate for the centerline rumble strip treatment sites. Nontreatment sites in Minnesota identified during the safety evaluation of shoulder rumble strips were also used in the safety evaluation of centerline rumble strips. The video- logs of the nontreatment sites were reviewed again as part of this safety evaluation to collect horizontal alignment data. Table 49 shows the total mileage (by roadway type) of treat- ment and nontreatment sites from Minnesota considered for inclusion in the safety evaluation. This table reflects total mileage for treatment and nontreatment sites after a series of data quality checks to make sure data were consistent and com- plete for each location. The table includes data for treatment sites with centerline rumble strips only, dual application treat- ment sites (i.e., sites with both centerline and shoulder rumble strips), and nontreatment sites (i.e., sites without either center- line or shoulder rumble strips). This table does not classify the mileage by installation dates, or by total mile-years that can be used for before- and after-period analyses in a before-after evaluation. This level of detail is provided later in the part on Descriptive Statistics in this section. Pennsylvania Sites PennDOT maintains a database of low-cost safety improve- ments made across the entire state of Pennsylvania. From this database, a list of approximately 300 safety improvement proj- ects were identified where the installation of centerline rumble strips was the only safety improvement made as part of the project. This list included the location of the safety projects on the state highway network and the installation date of the project. The installation dates ranged from 2000 to 2005. The research team reviewed the locations of these safety projects using PennDOT’s videolog, accessible via the Internet. Several dual applications sites were identified during the videolog review. To further verify the installation date of the centerline rumble strips and to determine the installation date of the shoulder rumble strips, archived videologs were reviewed at PennDOT’s Central Office. Table 50 shows the total mileage (by roadway type) of treat- ment and nontreatment sites from Pennsylvania considered for inclusion in the safety evaluation. This table is similar to Table 49 above for Minnesota sites. Washington Sites Washington State DOT (WSDOT) provided an initial list of treatment locations. The information provided with this list included the route number, beginning and ending mile- posts, installation dates, and construction information. The initial list included approximately 85 potential treatment locations with installation dates ranging from 1996 to 2007. Using WSDOT’s videolog system, the research team focused on reviewing sites where centerline rumble strips were installed between 2002 and 2005. Table 49. Total mileage of Minnesota treatment and nontreatment sites considered for inclusion in the safety evaluation of centerline rumble strips. Roadway type Treatment sites (mi) Nontreatment sites (mi) Dual application treatment sites (mi) Urban multilane undivided highways (nonfreeways) 0.00 0.00 0.00 00.000.000.0sdaorenal-owtnabrU Rural multilane undivided highways (nonfreeways) 0.00 0.00 0.00 00.089.2858.181sdaorenal-owtlaruR 00.089.2858.181sepytyawdaorllassorcaslatoT

95 Table 51 shows the total mileage (by roadway type) of treat- ment and nontreatment sites from Washington considered for inclusion in the safety evaluation. This table is similar to Tables 49 and 50 above for Minnesota and Pennsylvania sites. Summary of Sites Across All States Table 52 shows the total mileage (by roadway type) of treat- ment and nontreatment sites summed across all three states (Minnesota, Pennsylvania, and Washington) for use in the safety evaluation of centerline rumble strips. Based on the total available mileage of both treatment and nontreatment sites across all three states, Table 52 suggests that analyses of the data for rural and urban two-lane roads have the greatest potential to provide reliable results for investigating the safety effective- ness of centerline rumble strips by themselves. Formal analy- ses of dual application sites were not performed due to limited mileage of dual applications sites. Crash statistics of the dual application sites are presented later in the section along with general observations of the data. Videolog Data Collection Data were collected in a similar manner as the videolog data collection effort conducted to assess the safety effective- ness of shoulder rumble strips (see Section 6), with some vari- ations. The primary purposes of the videolog data collection effort for the safety evaluation of centerline rumble strips were to (a) confirm the absence/presence of the centerline rumble strips, (b) confirm the absence/presence of both center- line and shoulder rumble strips, (c) record the beginning and ending locations of the centerline rumble strips, (d) col- Table 50. Total mileage of Pennsylvania treatment and nontreatment sites considered for inclusion in the safety evaluation of centerline rumble strips. Roadway type Treatment sites (mi) Nontreatment sites (mi) Dual application treatment sites (mi) Urban multilane undivided highways (nonfreeways) 3.00 3.73 0.00 00.007.0405.52sdaorenal-owtnabrU Rural multilane undivided highways (nonfreeways) 4.74 1.84 0.00 08.304.63287.081sdaorenal-owtlaruR 08.376.28220.412sepytyawdaorllassorcaslatoT Roadway type Treatment sites (mi) Nontreatment sites (mi) Dual application treatment sites (mi) Urban multilane undivided highways (nonfreeways) 1.08 0.00 0.00 00.093.441.4sdaorenal-owtnabrU Rural multilane undivided highways (nonfreeways) 0.00 0.12 0.00 05.295.8644.35sdaorenal-owtlaruR 05.201.3766.85sepytyawdaorllassorcaslatoT Table 51. Total mileage of Washington treatment and nontreatment sites considered for inclusion in the safety evaluation of centerline rumble strips. Roadway type Treatment sites (mi) Nontreatment sites (mi) Dual application treatment sites (mi) Urban multilane undivided highways (nonfreeways) 4.08 3.73 0.00 00.090.5446.92sdaorenal-owtnabrU Rural multilane undivided highways (nonfreeways) 4.74 1.96 0.00 03.679.78370.614sdaorenal-owtlaruR 03.657.83435.454sepytyawdaorllassorcaslatoT Table 52. Total mileage of treatment and nontreatment sites considered for inclusion in the safety evaluation of centerline rumble strips (includes data from Minnesota, Pennsylvania, and Washington).

96 lect horizontal alignment data, (d) collect roadside hazard ratings for each site, (e) verify certain roadway characteristic data, and (f) gather information concerning the construction history of sites and installation dates. The main difference between the videolog data collection effort for the safety evaluation of centerline rumble strips com- pared to the safety evaluation of shoulder rumble strips is the collection of horizontal alignment data. Horizontal alignment data were specifically collected to assess the safety effectiveness of centerline rumble strips along varying roadway geometry. From the videologs, the beginning and ending locations of tan- gents and horizontal curves were recorded through manual observations. The sharpness of a curve was noted based upon the presence or absence of a curve warning sign, and the direc- tion of the curve (i.e., left vs. right) in the direction of increas- ing milepost/offset was recorded as well. In summary, the following horizontal alignment data were gathered during the videolog data collection effort for each state: • Beginning and ending mileposts/offsets of tangents, • Beginning and ending mileposts/offsets of horizontal curves, • Sharpness of curve (i.e., presence/absence of curve warn- ing signs), and • Direction of curve (left vs. right). Some final points regarding the data collection effort, rel- evant to the analysis approach and analysis results, are the following: • Even when centerline rumbles strips are installed continu- ously along a segment, there are many breaks in the rumble strips for various reasons such as bridges, intersections, driveways, etc. Depending upon the roadway type and the policy of the individual states, the frequencies of these breaks vary considerably. For those treatment sites where a signifi- cant length of contiguous mileage of centerline rumble strip installation existed, long breaks in the rumble strips may have been recorded during the data collection process, but the boundaries of the treatment site were not modified to reflect the breaks in the rumble strips. Thus, there are loca- tions along the roadways of treatments sites where center- line rumble strips are not present. Only for those treatment sites where centerline rumble strips were not installed over a very long stretch of highway were the boundaries of the treatment sites modified to reflect numerous or significant breaks in the centerline rumble strips. • The ideal type of treatment site to include in a before-after analysis is one in which the only type of treatment made during the analysis period is the safety improvement (i.e., installation of centerline rumble strips). To the best of our knowledge, the only improvements made to the treatment sites during the analysis period were the installation of the centerline rumble strips (and shoulder rumble strips in the case of dual application treatment sites). • In Pennsylvania, the initial list of treatment sites was gener- ated from the low-cost safety improvement database devel- oped and maintained by PennDOT. Thus, it is likely that many of the treatment sites in Pennsylvania were initially identified as being high-crash locations compared to the rest of the highway network. For Minnesota and Washington, the policy for determining the need and location for instal- lation of centerline rumble strips is not known. • All centerline rumble strips are treated as being equivalent in their alerting properties. No effort was made to obtain either rumble strip dimensions for each treatment site or information concerning the placement of the centerline rumble strips relative to the centerline pavement markings. • No information was recorded concerning the presence of passing zones for either treatment or nontreatment sites. Based upon the written or unwritten centerline rumble strip policies in the three states, several of the treatment sites do include passing zones, but as indicated no information was recorded concerning the presence of passing zones, so neither the percentage of sites with passing zones nor the total mileage of sites with or without passing zones can be determined. The decision to not collect passing zone infor- mation was based upon previous research (38) indicating that centerline rumble strips have minimal impact on driver behavior in passing zones. Database Development The final database(s) utilized for analysis consisted of the roadway characteristic data (including traffic volume), the videolog data, and crash data. In summary, the database(s) for each state included the following roadway inventory and videolog data for a given site: • Location reference information (i.e., beginning and ending mileposts/logpoints, or route, county, segment, and offsets), • Presence/absence of milled centerline rumble strips, • Presence/absence of dual application (i.e., both centerline and shoulder rumble strips), • Area type (rural vs. urban), • Roadway type (i.e., multilane undivided or two-lane), • Number of lanes, • Lane widths, • Shoulder widths, • RHR, • Horizontal alignment (i.e., tangent or curve), • Sharpness of curve (i.e., presence/absence of curve warning signs), • Direction of curve (i.e., left or right),

97 • Analysis period(s) (including year(s) without rumble strips, installation year(s), and years with rumble strips), and • ADT for each year in the analysis period(s). The original roadway inventory files obtained from the states did not contain ADTs for all sites for each year in the analysis period(s). Therefore, rules were established for inter- polating and extrapolating the ADT data so that the final data- base included ADTs for each site and year in the analysis periods. The analysis periods were determined based upon the construction history and installation data gathered and the years of available crash data for each state. Crash data were obtained for the following calendar years (inclusive) for each state: • Minnesota (1997 through 2006), • Pennsylvania (1997 through 2006), and • Washington (2001 through 2006). The crash data available for use in the analyses consisted of the following: • Crash report number or crash ID number, • Date of crash, • Location information (county, route, direction, segment and offset or logpoint), • Number of vehicles involved, • Crash severity, and • Accident type or manner of collision. The final database only included crashes assigned to road- way segments. Rules were established to eliminate (i.e., screen out) intersection-related crashes. As a final note concerning the crash data, logic was developed to identify SVROR-left and SVROR-right crashes from the electronic crash data for Penn- sylvania. The accuracy of the results was assessed by reviewing about 100 sample crash reports. Based upon the results of the sampling, it was determined that SVROR-left and SVROR- right crashes could not be accurately determined from the elec- tronic files for Pennsylvania. Therefore, SVROR-left crashes are not included in analyses of target crashes for centerline rumble strips. At least one study (75) indicated that SVROR- left crashes should potentially be considered as target crashes of centerline rumble strips. Several other rules were established for developing the final database(s). Most of these rules pertained to establishing a rationale for combining adjacent sites to create longer homo- geneous sites. Several of these rules pertained to the following: • Selected roadway characteristics (e.g., lane widths, shoul- der widths, number of lanes) and • Desirable minimum lengths (e.g., 0.1 mi [0.16 km]). The following section provides descriptive statistics of the information contained in databases developed for the safety evaluation of centerline rumble strips. Descriptive Statistics To address the three objectives of this safety evaluation of centerline rumble strips, a database has been assembled such that data from a given site could be used to address one or more of the objectives. This section provides descriptive sta- tistics in either tabular and/or graphical form for the inde- pendent variables (i.e., ADT, site geometrics) and dependent variables (i.e., crash data) of interest in the safety evaluation of centerline rumble strips. The data are presented in several lay- outs, designed to provide basic summaries of the available information. The data for each site were collected over time periods of varying lengths. For comparison, the site length and the num- ber of years were combined into a single variable, mile-years, for each site. The following is nomenclature used in the safety evaluation of centerline rumble strips to describe the types of sites included in the evaluation: • BA-No RS: Nontreatment site of the matched before-after site pair in the before period; • BA-RS: Treatment site of the matched before-after site pair in the after period; and • NT-No RS: Nontreatment reference site. Table 53 summarizes the basic layout of the available data in the three states for evaluation of the safety effectiveness of cen- terline rumble strips on specific types of roads. The data pro- vided separately for each state, roadway type, and type of site are the number of sites, total site length, and mile-years. Due to an insufficient number of treatment sites and mile-years for a number of roadway types and states, it was decided to focus the safety evaluation on the following roadway types: • Urban two-lane roads and • Rural two-lane roads. For urban two-lane roads a decision was made to only analyze data from Pennsylvania and not include data from Washington because the number of mile-years of data from Pennsylvania was far greater than for Washington. The Penn- sylvania data would likely dominate the analysis such that including data from Washington would not contribute to the results. Thus, the analysis for urban two-lane roads is based on data from Pennsylvania only. ADT volume. For each site, ADTs were averaged across years within an analysis period. This allowed for a fair

Minnesota Pennsylvania Washington Roadway type Site type Treatment status Number of sites Length (mi) Mile- years Number of sites Length (mi) Mile- years Number of sites Length (mi) Mile- years SRoNBA RS 0 0 0 Urban multilane undivided highways (nonfreeways) NT No RS 0 8 3.73 37.28 0 SRoN 138.78 12.90 BA RS 0 74 25.50 90.64 6 4.14 7.40 Urban two-lane roads NT No RS 0 85 40.70 407.01 22 4.39 21.95 SRoNBA RS 0 0 0 Rural multilane undivided highways (nonfreeways) NT No RS 0 4 1.84 18.42 0 No RS 1,033.15 893.14 190.88 BA RS 301 181.85 603.48 526 180.78 722.28 135 53.44 53.97 Rural two-lane roads NT No RS 243 82.98 747.82 518 236.40 2,364.01 206 68.59 342.95 a Shaded cells are focus of statistical analysis. Table 53. Summary study layout—total number of sites, site length, and mile-years by state, roadway type, and site typea.

99 comparison of the distribution of ADTs across site types, analysis periods, and states since the sample size is reduced to the number of sites within each category and thus not unduly influenced by the length of the varying analysis periods. Figures 11 and 12 show the ADT distributions in the form of side-by-side boxplots for both urban and rural two-lane roads. Within each figure, the data are organized by the states included in the analysis; within each state, the data are ordered by site type—before-after sites then nontreatment sites; the mean ADTs of nontreatment sites are colored white; those of the treatment sites are black. Each figure also contains a table of basic descriptive ADT statistics for num- ber of sites, mean, standard deviation, minimum, median, and maximum. Lane width. Lane widths ranged from 10 to 14+ ft (3.0 to 4.3+ m) across all sites and states, with the majority of lanes being 11 or 12 ft (3.3 or 3.6 m) wide. The distribution of lane width is summarized in Table 54 by state and site type. Shoulder width. Outside shoulder widths ranged from 0 to 10+ ft (0.0 to 3.0+ m) across sites and states. The distribu- tion of shoulder width is summarized in Table 55 by state and site type. RHR. Roadside hazard ratings were recorded as integers ranging from 1 (low RHR) to 7 (high RHR); it is treated as a continuous variable in the statistical model development. Table 56 presents basic descriptive RHR statistics (i.e., num- ber of sites and minimum, maximum, mean, and standard deviation) by state and site type. Non-integer values for min- imums and maximums in Table 56 are the result of combin- ing adjacent segments into homogeneous sites for analysis purposes. When adjacent segments with different RHRs were combined into a single site for analysis purposes, a weighted average, based on segment length, of the RHR was calculated for the site. Horizontal alignment. Table 57 provides the mile- years and number of horizontal curve sites and tangent sites, respectively, used to assess the safety effectiveness of centerline rumble strips along varying roadway geometry. The analysis of the safety effectiveness of centerline rumble strips along varying roadway geometry is based upon data strictly from rural two-lane roads. Sites in Minnesota and Washington are defined slightly different than sites in Penn- sylvania. The SPFs for Pennsylvania nontreatment sites were also developed in a slightly different manner than the Figure 11. ADT distribution by site type for urban two-lane roads in Pennsylvania.

100 SPFs for Minnesota and Washington. The differences are as follows: • For Minnesota and Washington treatment and nontreat- ment sites, curve sites begin at the beginning of a curve and end at the end of a curve. Similarly, tangent sites begin at the beginning of a tangent and end at the end of a tangent. • For Pennsylvania sites, treatment sites were classified as a horizontal curve if 75 percent or more of the length was on a horizontal curve, and treatment sites were classified as a tangent if 100 percent of the length was on a straight portion of roadway (i.e., without any horizontal curves). Treatment sites with less than 75 percent of the length on a horizontal curve were not included in either classifica- Figure 12. ADT distribution by site type for rural two-lane roads in Minnesota, Pennsylvania, and Washington. Table 54. Distribution of lane width by state and site typea. Treatment status 0 5,000 10,000 15,000 20,000 25,000 AD T (av g v e hs /d a y) 0 boxes clipped MN PA WA BA NT BA NT BA NT No. of sites Mean 6,407 Std dev 2,591 Min 1,495 Median 6,307 Max 13,240 301 6,773 2,700 1,336 7,145 12,478 243 3,265 3,121 176 1,247 10,508 526 6,510 3,315 574 5,821 17,205 526 6,359 3,220 596 5,760 17,591 518 6,204 3,794 777 4,992 22,794 135 7,602 3,644 3,222 6,495 20,448 135 7,746 3,902 3,167 6,394 20,784 206 7,851 3,723 3,097 7,300 16,749 301 No RS RS Minnesota Pennsylvania Washington Lane width (ft) Before–after sites Nontreatment sites Before–after sites Nontreatment sites Before–after sites Nontreatment sites 9 – – – – – – 10 – – 48 68 – 33 11 36 20 313 304 61 96 12 251 218 208 155 74 71 13 7 5 7 17 – 2 14+ 7 – 24 59 – 4 a Includes data for all roadway types for which analyses are conducted.

101 tion. During the SPF development for nontreatment sites, all nontreatment sites were included in the model and a curve/tangent independent variable was considered. The curve/tangent variable proved to be not significant; therefore, all Pennsylvania nontreatment sites, whether they included a horizontal curve or tangent sections were considered appropriate for analysis as nontreatment sites. Thus, the number of sites, length, and mile-years for nontreatment sites for Pennsylvania rural two-lane roads in Tables 53 and 57 are equivalent. More details on the analysis approach to quantifying the safety effectiveness of centerline rumble strips along vary- ing roadway geometry are provided on the next page in Analysis Approach. Dual application sites. Table 58 provides the number of sites, length, and mile-years for the dual application sites found in Pennsylvania and Washington. For these sites, the centerline and shoulder rumble strips were installed during the same calendar year. No dual application sites were found Minnesota Pennsylvania Washington Average shoulder width (rounded) (ft) Before–after sites Nontreatment Before–after sites Nontreatment Before–after sites Nontreatment 0 – 2 4 53 – 3 1 – 60 – 9 – 2 2 9 - 11 41 7 26 3 8 27 66 50 63 17 4 37 48 155 139 27 50 5 – - 70 93 11 10 6 31 10 123 74 22 41 7 22 33 49 20 – 4 8 100 59 91 91 – 39 9 31 3 8 3 1 6 10+ 63 1 23 30 4 8 a Includes data for all roadway types for which analyses are conducted. RHR State Site type Number of sites Minimum Maximum Mean Standard deviation BA 301 2 5.5 3.55 1.14 MN Nontreatment 243 2 6 3.50 1.23 BA 600 2 5 3.57 0.57 PA Nontreatment 603 2 6 3.67 0.83 BA 135 2 5 3.72 0.49 WA Nontreatment 206 3 5 4.21 0.43 a Includes data for all roadway types for which analyses are conducted. Minnesota Pennsylvania Washington Roadway type Site type Treatment status Number of sites Length (mi) Mile- years Number of sites Length (mi) Mile- years Number of sites Length (mi) Mile- years Horizontal Curve Sites 80.1401.43160.261SRoNBA RS 135 28.41 93.07 144 29.32 125.84 62 10.42 11.02 Rural two-lane roads NT No RS 105 14.89 134.04 518 236.40 2,364.01 104 17.07 85.35 Tangent Sites 08.94176.01190.178SRoNBA RS 166 153.44 509.87 73 23.25 98.24 73 38.55 42.95 Rural two-lane roads NT No RS 138 68.09 612.79 518 236.40 2,364.01 102 51.52 257.60 Table 55. Distribution of shoulder width by state and site typea. Table 56. RHR statistics by roadway type, state, and site typea. Table 57. Summary study layout—total number of horizontal curve and tangent sites, site length, and mile-years by state and site type.

102 in Minnesota. Due to limited mileage and mile-years of treat- ment sites, a detailed safety evaluation of dual applications was not performed, but crash statistics are presented below followed by several general observations of the crash statistics in the section on Analysis Results. Crash data. Four crash types are analyzed as part of the safety evaluation of centerline rumble strips: • TOT crashes, • FI crashes, • Total head-on and sideswipe opposite-direction (TOT tar- get) crashes, and • Fatal and injury head-on and sideswipe opposite-direction (FI target) crashes. Analyses of TOT crashes are performed primarily because several previous safety evaluations of centerline rumble strips analyzed this crash type. However, analyses of TOT crashes include many other crash types besides head-on and opposite- direction sideswipe crashes (i.e., the target crash types). No strong argument can be made to support why centerline rumble strips should affect crashes other than the primary target crashes (i.e., head-on and opposite-direction sideswipe crashes), with the possible exception of SVROR-left crashes. Analyses of FI crashes were also conducted because there is great interest in reducing crashes that result in fatalities and injuries, but again, analyses of FI crashes include many other crash types besides the target crashes. Analyses of head-on and opposite-direction sideswipe crashes are expected to pro- duce more reliable results than analyses of TOT and FI crashes because the analyses include only those crashes expected to be most directly impacted by centerline rumble strips. Finally, analyses based on head-on and opposite-direction sideswipe FI crashes are of interest because these analyses address the more severe target crashes. The crash data across all years are summarized in Table 59 and are shown as both total number of crashes and crash fre- quency (crashes/mi/yr) and separately for each type of site of a given roadway type within a given state. The two statistics are presented separately for TOT crashes, head-on crashes, and sideswipe opposite-direction crashes. Table 59 also provides the number of sites and their total length and mile-years to facilitate comparison between groups of data. For before-after site pairs (i.e., same site paired in time), the number of sites and length are shown only once since the sites are the same before and after treatment; however, since the study periods changed from site to site, mile-years vary between nontreat- ment and treatment before-after site pairs. The crash data are summarized by roadway type, state, site type, and treatment status in Table 60 for all FI crashes and the target crashes. Crash counts and average frequencies per mile per year are presented. Table 61 summarizes crash data used to evaluate the safety effectiveness of centerline rumble strips along varying roadway geometry. TOT crashes and the target crashes are presented for horizontal curve sites and tangent sites for rural two-lane roads by state, site type, and treatment sta- tus. Crash counts and average frequencies per mile per year are presented. Table 62 presents the corresponding data for FI crashes. Table 63 summarizes crash data for dual application sites in Pennsylvania and Washington. Data are presented for TOT crashes, FI crashes, and TOT target crashes for before-after sites only, for rural two-lane roads. For dual application sites, the target crashes include head-on, sideswipe opposite- direction, and SVROR crashes. Crash counts and average fre- quencies per mile per year are presented. Analysis Approach The EB methodology, as described in Section 6, Analysis Approach, was used to evaluate the safety effectiveness of centerline rumble strips on different roadway types and along varying roadway geometry. The evaluations included analyses of TOT, FI, and TOT target crashes. Due to small sample sizes, analyses of FI target crashes were performed in the evaluation of centerline rumble strips on different road- way types, based upon proportions. Rather than developing SPFs for FI target crashes, the SPF for TOT target crashes was used along with the percentage of FI target crashes to TOT target crashes. The same approach for analyzing FI tar- get crashes for centerline rumble strips was used as in the supplemental analyses of shoulder rumble strips (see Analy- sis Approach, Section 6). No analyses of FI target crashes were performed in the evaluation of centerline rumble strips along varying roadway geometry due to extremely small sample sizes. Table 58. Summary study layout—total number of dual application sites, site length, and mile-years by state and site type. Minnesota Pennsylvania Washington Roadway type Site type Treatment status Number of sites Length (mi) Mile- years Number of sites Length (mi) Mile- years Number of sites Length (mi) Mile- years 0.58.22SRoNRural two- lane roads BA RS 0 11 3.80 11.4 5 2.50 7.5

Crash type TOT Head-on Sideswipe opposite-direction Roadway type State Site type Treatment status Number of sites Length (mi) Mile- years Total number of crashes Crash frequency (crashes/ mi/yr) Total number of crashes Crash frequency (crashes/ mi/yr) Total number of crashes Crash frequency (crashes/ mi/yr) 41.09122.00358.259387.831SRoNBA RS 74 25.50 90.64 222 2.44 7 0.08 6 0.07 PA NT No RS 85 40.70 407.01 1,168 2.87 82 0.20 17 0.04 13.0461.0209.69809.21SRoNBA RS 6 4.14 7.40 31 4.19 0 0.00 1 0.08 Urban two-lane roads WA NT No RS 22 4.39 21.95 43 1.96 0 0.00 1 0.23 30.05360.04679.0000,151.330,1SRoNBA RS 301 181.85 603.48 523 0.87 39 0.06 16 0.03 MN NT No RS 243 82.98 747.82 445 0.60 39 0.05 16 0.02 80.07632.090271.2639,141.398SRoNBA RS 526 180.78 722.28 1,256 1.74 54 0.08 40 0.06 PA NT No RS 518 236.40 2,364.01 4,140 1.75 301 0.13 117 0.05 80.05170.03152.292488.091SRoNBA RS 135 53.44 53.97 133 2.46 2 0.04 3 0.06 Rural two-lane roads WA NT No RS 215 69.85 349.25 967 2.77 19 0.05 28 0.08 Crash type All FI Head-on Sideswipe opposite-direction Roadway type State Site type Treatment status Number of sites Length (mi) Mile- years Total number of crashes Crash frequency (crashes/ mi/yr) Total number of crashes Crash frequency (crashes/ mi/yr) Total number of crashes Crash frequency (crashes/ mi/yr) 90.02102.07227.193287.831SRoNBA RS 74 25.50 90.64 116 1.28 2 0.02 4 0.04 PA NT No RS 85 40.70 407.01 600 1.47 40 0.10 2 0.00 61.0280.0133.33409.21SRoNBA RS 6 4.14 7.40 11 1.49 0 0.00 0 0.00 Urban two-lane roads WA NT No RS 22 4.39 21.95 20 0.91 0 0.00 0 0.00 20.09140.07483.079351.330,1SRoNBA RS 301 181.85 603.48 193 0.32 15 0.02 9 0.01 MN NT No RS 243 82.98 747.82 179 0.24 23 0.03 7 0.01 40.08391.086152.1411141.398SRoNBA RS 526 180.78 722.28 664 0.92 44 0.06 24 0.03 PA NT No RS 518 236.40 2,364.01 2,237 0.95 83 0.04 13 0.00 80.05170.03161.122288.091SRoNBA RS 135 53.44 53.97 63 1.17 2 0.04 2 0.04 Rural two-lane roads WA NT No RS 215 69.85 349.25 424 1.21 18 0.05 19 0.05 Table 59. Crash statistics by roadway type, state, site type, treatment status, and severity (TOT crashes). Table 60. Crash statistics by roadway type, state, site type, treatment status, and severity (FI crashes).

Crash type TOT Head-on Sideswipe opposite-direction Roadway type State Site type Treatment status Number of sites Length (mi) Mile- years Total number of crashes Crash frequency (crashes/ mi/yr) Total number of crashes Crash frequency (crashes/ mi/yr) Total number of crashes Crash frequency (crashes/ mi/yr) Horizontal Curve Sites 70.01190.05132.100260.261SRoNBA RS 135 28.41 93.07 98 1.05 9 0.10 2 0.02 MN NT No RS 105 14.89 134.04 113 0.84 7 0.05 7 0.05 60.0813.01458.184201.431SRoNBA RS 144 29.32 125.84 223 1.77 7 0.06 8 0.06 PA NT No RS 518 236.40 2,364.01 4,140 1.75 301 0.13 117 0.05 70.0350.0243.26980.14SRoNBA RS 62 10.42 11.02 27 2.45 1 0.91 1 0.09 Rural two-lane roads WA NT No RS 104 17.07 85.35 288 3.37 4 0.05 9 0.10 Tangent Sites 30.04260.09429.000890.178SRoNBA RS 166 153.44 509.87 425 0.83 30 0.06 14 0.03 MN NT No RS 138 68.09 612.79 332 0.54 32 0.05 9 0.02 40.0481.00221.253276.011SRoNBA RS 73 23.25 98.24 160 1.63 6 0.06 4 0.04 PA NT No RS 518 236.40 2,364.01 4,140 1.75 301 0.13 117 0.05 80.02170.01142.233308.941SRoNBA RS 73 38.55 42.95 106 2.47 1 0.02 2 0.05 Rural two-lane roads WA NT No RS 102 51.52 257.60 636 2.47 15 0.06 18 0.07 Table 61. Crash statistics for curve/tangent sites by roadway type, state, site type, treatment status, and severity (TOT crashes).

Crash type All FI Head-on Sideswipe opposite-direction Roadway type State Site type Treatment status Number of sites Length (mi) Mile- years Total number of crashes Crash frequency (crashes/ mi/yr) Total number of crashes Crash frequency (crashes/ mi/yr) Total number of crashes Crash frequency (crashes/ mi/yr) Horizontal Curve Sites 40.0670.02185.04960.261SRoNBA RS 135 28.41 93.07 34 0.36 2 0.02 1 0.01 MN NT No RS 105 14.89 134.04 42 0.31 4 0.03 2 0.02 20.0342.02381.185101.431SRoNBA RS 144 29.32 125.84 120 0.95 6 0.05 6 0.05 PA NT No RS 518 236.40 2,364.01 2,237 0.95 83 0.04 13 0.00 70.0320.0163.16580.14SRoNBA RS 62 10.42 11.02 10 0.91 1 0.09 3 0.27 Rural two-lane roads WA NT No RS 104 17.07 85.35 130 1.52 10 0.12 7 0.08 Tangent Sites 20.03140.05353.030390.178SRoNBA RS 166 153.44 509.87 159 0.31 13 0.02 8 0.02 MN NT No RS 138 68.09 612.79 137 0.22 19 0.03 5 0.01 30.0371.09142.173176.011SRoNBA RS 73 23.25 98.24 79 0.80 6 0.06 2 0.02 PA NT No RS 518 236.40 2,364.01 2,237 0.95 83 0.04 13 0.00 40.0650.0811.166108.941SRoNBA RS 73 38.55 42.95 53 1.23 2 0.05 1 0.02 Rural two-lane roads WA NT No RS 102 51.52 257.60 274 1.06 14 0.05 13 0.05 Crash type TOT FI TOT Target (head-on, sideswipe opposite-direction, and SVROR) Roadway type State Site type Treatment status Number of sites Length (mi) Mile- years Total number of crashes Crash frequency (crashes/ mi/yr) Total number of crashes Crash frequency (crashes/ mi/yr) Total number of crashes Crash frequency (crashes/ mi/yr) 01.15279.15446.3388.22SRoNPA BA RS 11 3.80 11.4 30 2.63 18 1.58 2 0.18 00.1504.1704.3710.5SRoN Rural two-lane roads WA BA RS 5 2.50 7.5 30 4.00 16 2.13 12 1.60 Table 62. Crash statistics for curve/tangent sites by roadway type, state, site type, treatment status, and severity (FI crashes). Table 63. Crash statistics for dual application sites by roadway type, state, site type, treatment status, and severity.

106 Due to limited mileage and mile-years of dual application sites, a detailed safety evaluation of dual applications was not performed. Crash statistics are presented in Descriptive Statistics above; general observations of these data are pre- sented in the next section. Cross-sectional analyses using GLM were not used to assess the safety effectiveness of centerline rumble strips since all treatment sites had before-period information. Regarding the dual application sites, several sites were found during data col- lection that initially had shoulder rumble strips installed for several years, followed by the installation of centerline rumble strips. Thus, rather than having a before-treatment period without any centerline and shoulder rumble strips, the before- treatment period had shoulder rumble strips present for a sig- nificant number of years. A decision was made to report only those sites where the before-treatment period had no shoulder rumble strips present and the after period had both centerline and shoulder rumble strips present. The SPFs developed for use in the EB methodology to eval- uate the safety effectiveness of centerline rumble strips were based on negative binomial regression. Model estimation used PROC GENMOD in SAS. Variable selection was done for each state and roadway type based on trials with various combina- tions of logical potential dependent variables. Each potential model was assessed based on the statistical significance of the variable coefficients and overall model fit. Several final points follow that concern the analysis approach to the safety evaluation of centerline rumble strips on different roadway types and along varying roadway geometry: • It has been noted that the research by Persaud et al. (4) is the most comprehensive and reliable evaluation on the safety effectiveness of centerline rumble strips on rural two-lane roads. The research team had access to the raw data collected by Persaud et al. (4) for the Insurance Institute for Highway Safety (IIHS) and also the EB calculations. This enabled the data collected for rural two-lane roads during this research and the IIHS to be combined to provide reliable and com- prehensive estimates on the safety effectiveness of centerline rumble strips on rural two-lane roads. • In the evaluation of the safety effectiveness of centerline rumble strips along varying roadway geometry, horizontal curve and tangent sites were defined in a slightly different manner for Minnesota and Washington than for Pennsylva- nia. The differences in the way horizontal curve and tangent sites were defined were due to differences in the location ref- erencing systems for Minnesota and Washington compared to Pennsylvania. The information on rumble strip place- ment in Minnesota and Washington was used to define curve and tangent segments when overlapped with the road- log location referencing system. For Pennsylvania, a similar exercise would have resulted in many very short roadway segments, which would not be reliable for analysis. For Pennsylvania, a horizontal curve treatment site was classi- fied as being a horizontal curve site if 75 percent or more of the segment length was curvilinear. Only treatment sites with at least 75 percent or more of the segment length being curvilinear were included in the analysis of the safety effectiveness of centerline rumble strips along horizontal curves. A tangent treatment site was classified as a tangent site if 100 percent of the segment length was on a straight portion of roadway without any horizontal curves. During the SPF development of nontreatment sites in all states, mod- els were developed in which all sites (i.e., horizontal curves and tangents) were included and a curve/tangent indicator variable was included as a predictor (i.e., independent) vari- able in the model. The curve/tangent variable was not sig- nificant; therefore, all nontreatment sites, whether they included horizontal curves, tangents, or both were con- sidered appropriate for analysis as nontreatment sites. Although the data for the nontreatment sites did not sup- port different models for curve and tangent sites, the empir- ical Bayes procedure does reflect differences between sites when combining SPF predictions and observed crash counts and therefore would reflect actual differences between the two groups should they exist. Even though horizontal curve and tangent sites were defined slightly differently for Min- nesota and Washington than for Pennsylvania, the results for all three states were combined in the separate analyses for horizontal curves and tangents, and the combined results from all three states are considered the most reliable and comprehensive compared to the individual results by state. Analysis Results Analysis results are first presented for estimating the safety effectiveness of centerline rumble strips on different roadway types and then followed by the analysis results for estimating the safety effectiveness of centerline rumble strips along vary- ing roadway geometry. Finally, several general observations on the safety effectiveness of dual application sites based on the basic crash statistics are presented. Estimating the Safety Effectiveness of Centerline Rumble Strips on Different Roadway Types The EB analysis consisted of two steps: 1. Develop SPF models based on all nontreatment sites, and 2. Using the SPFs, evaluate the safety effectiveness of center- line rumble strips using crash data from the before-after sites only.

107 SPF results. Negative binomial regression models were developed for total crashes and the proportion of FI and tar- get crashes were applied to the models to predict for these crash types, thus combining data for horizontal curve sites and tangent sites. For Minnesota and Washington, the curve/ tangent characteristic was not a statistically significant variable for rural two-lane roads, while for Pennsylvania, curvature was not known for all of the reference sites used. For Wash- ington, ADT was the only variable used in the model, while Minnesota also included roadside hazard rating and Pennsyl- vania included roadside hazard rating, posted speed, and road width as predictor variables. Only nontreatment sites were used for SPF development. Tables H-1 through H-3 in Appen- dix H summarize the crash frequency models for TOT, FI, and TOT target crashes, respectively. The statistics shown for each state SPF are as follows: • Number of nontreatment sites; • Intercept: estimate and standard error; • lnADT coefficient: estimate, standard error, and p-value (or significance level); for example, a p-value of 0.05 or less indi- cates that the coefficient is significantly different from zero at the 0.05 significance (or 95 percent confidence) level; • Model dispersion parameter: estimate and standard error; and • Model R2LR value: the likelihood ratio R 2 LR, a measure of model fit between 0 and 1. The closer the value is to 1, the better the fit of the model is to the data. The SPF is represented by the following general equation: Expected crashes mi yr blnADT cRHR dWid = + +( +exp 1 th eSPD+ ) ( )3 where RHR = the average roadside hazard rating, SPD = 1 if the posted speed is less than 55 mph and 0 otherwise, Width = the roadway width in feet, and a, b, c, d and e = coefficients whose estimates are shown in Tables H-1 through H-3 in Appendix H. The SPFs were recalibrated to provide a yearly factor for each year to account for time trends in crash counts. These factors were based on total crashes. Safety effectiveness results. For each crash type, roadway type, and state, the safety effectiveness of centerline rumble strips was estimated. The final results are shown in Tables 64 through 67 for TOT, FI, TOT target, and FI target crashes, respectively. For each crash type, separate analyses were per- formed across the two roadway types of interest, based on data for individual states and combined across states. The statistics shown for each crash type, roadway type, and state (single or combined) are as follows: • Number of treatment sites, • Total site length, • Percent change due to centerline rumble strips: estimate and standard error, • Test statistic, and • An indication of whether rumble strips had a significant effect on the crash type of interest. Several relevant findings from the EB analyses are as follows: • Of the five analyses based on TOT crashes (Table 64), one yields statistically significant results at the 90 or 95 percent Percent change in crash frequency from before to after rumble strip installation (%) Roadway type State Number of sites Total length (mi) Estimatea SEb Test statisticc Significance Urban two–lane roads PA 74 25.50 1.5 8.0 0.19 Not significant at 90% CL Combined 962 416.06 –4.1 2.6 1.58 Not significant at 90% CL MN 301 181.84 –11.1 5.8 1.91 Significant at 90% CL PA 526 180.78 –1.6 3.3 0.48 Not significant at 90% CL Rural two–lane roads WA 135 53.44 2.3 8.1 0.28 Not significant at 90% CL a A negative percent change indicates a decrease in crash frequency while a positive percent change indicates an increase in crash frequency. b SE: standard error of estimate. c Test statistic = abs(Estimate/SE); not significant at 90% CL if < 1.7; significant at 90% CL if ≥ 1.7; significant at 95% CL if ≥ 2. Table 64. Safety effectiveness of centerline rumble strips on TOT crashes using the EB method.

Percent change in crash frequency from before to after rumble strip installation (%) Roadway type State Number of sites Total length (mi) Estimatea SEb Test statisticc Significance Urban two–lane roads PA 74 25.50 –9.3 9.5 0.98 Not significant at 90% CL Combined 962 416.06 –9.4 3.5 2.57 Significant at 95% CL MN 301 181.84 –21.8 6.6 3.30 Significant at 95% CL PA 526 180.78 –6.2 4.2 1.48 Not significant at 90% CL Rural two–lane roads WA 135 53.44 4.1 14.6 0.28 Not significant at 90% CL a A negative percent change indicates a decrease in crash frequency while a positive percent change indicates an increase in crash frequency. b SE: standard error of estimate. c Test statistic = abs(Estimate/SE); not significant at 90% CL if < 1.7; significant at 90% CL if ≥ 1.7; significant at 95% CL if ≥ 2. Percent change in crash frequency from before to after rumble strip installation (%) Roadway type State Number of sites Total length (mi) Estimatea SEb Test statisticc Significance Urban two–lane roads PA 74 25.50 –40.2 17.0 2.36 Significant at 95% CL Combined 962 416.06 –37.0 5.3 6.98 Significant at 95% CL MN 301 181.84 –48.9 7.3 6.69 Significant at 95% CL PA 526 180.78 –25.8 17.9 1.44 Not significant at 90% CL Rural two–lane roads WA 135 53.44 –35.4 29.2 1.21 Not significant at 90% CL a A negative percent change indicates a decrease in crash frequency while a positive percent change indicates an increase in crash frequency. b SE: standard error of estimate. c Test statistic = abs(Estimate/SE); not significant at 90% CL if < 1.7; significant at 90% CL if ≥ 1.7; significant at 95% CL if ≥ 2. Percent change in crash frequency from before to after rumble strip installation (%) Roadway type State Number of sites Total length (mi) Estimatea SEb Test statisticc Significance Urban two–lane roads PA 74 25.50 –63.7 26.9 2.36 Significant at 95% CL Combined 962 416.06 –44.5 6.4 6.98 Significant at 95% CL MN 301 181.84 –44.7 6.7 6.69 Significant at 95% CL PA 526 180.78 –44.4 30.8 1.44 Not significant at 90% CL Rural two–lane roads WA 135 53.44 –35.4 29.2 1.21 Not significant at 90% CL a A negative percent change indicates a decrease in crash frequency while a positive percent change indicates an increase in crash frequency. b SE: standard error of estimate. c Test statistic = abs(Estimate/SE); not significant at 90% CL if < 1.7; significant at 90% CL if ≥ 1.7; significant at 95% CL if ≥ 2. Table 65. Safety effectiveness of centerline rumble strips on FI crashes using the EB method. Table 66. Safety effectiveness of centerline rumble strips on TOT target crashes using the EB method. Table 67. Safety effectiveness of centerline rumble strips on FI target crashes using the EB method.

109 confidence level. This single significant result for rural two- lane roads in Minnesota indicates a decrease in TOT crashes when centerline rumble strips are installed. • Of the five separate analyses based on FI crashes (Table 65), two yield statistically significant results at the 90 or 95 per- cent confidence level. Each significant result for rural two- lane roads indicates a decrease in FI crashes when centerline rumble strips are installed. • Of the five separate analyses based on TOT target crashes (Table 66), three yield statistically significant results at the 90 or 95 percent confidence level. The significant results for both urban and rural two-lane roads indicate a decrease in TOT target crashes when centerline rumble strips are installed. • Of the five separate analyses based on FI target crashes (Table 67), three yield statistically significant results at the 90 or 95 percent confidence level. The significant results for both urban and rural two-lane roads indicate a decrease in FI target crashes when centerline rumble strips are installed. Combined results from this research and the IIHS study. For rural two-lane roads, Table 68 presents results on the safety effectiveness of centerline rumble strips from this research, the IIHS study (4), and both combined. The IIHS study was an EB evaluation of 98 rural sites in 6 states with a total of 211 mi (340 km) of centerline rumble strip installa- tions. Because comparisons between the two studies appear favorable, at least in terms of direction and general order of magnitude of the effects, the two sets of results were com- bined. The results of this research are the combined results from all three states. It should also be made clear that the com- bined results are based upon raw data from both studies, and the combined results are based upon the EB methodology. Because the raw data from the IIHS study were available, procedures for combining study results for incorporation in the HSM (65) were not used. The combined results estimate reductions of 8.7, 11.7, and 30.2 percent are expected for TOT, FI, and TOT target crashes, respectively, with the installation of centerline rumble strips. Estimating the Safety Effectiveness of Centerline Rumble Strips Along Varying Roadway Geometry The same SPFs used to estimate the safety effectiveness of centerline rumble strips on different roadway types were used to estimate the safety effectiveness of centerline rumble strips along varying roadway geometry. Safety effectiveness results. For each horizontal align- ment category, crash type, and state, the safety effectiveness of centerline rumble strips was estimated. The final results for horizontal curves are shown in Tables 69 through 71 for TOT, FI, and TOT target crashes, respectively, and the final results for tangents are shown in Tables 72 through 74 for TOT, FI, and TOT target crashes, respectively. For each crash type, sep- arate analyses were performed based on data for individual states and combined across states. The statistics shown for each horizontal alignment category, crash type, and state (single or combined) are as follows: • Number of treatment sites; • Total site length; • Percent change due to centerline rumble strips: estimate and standard error; • Test statistic; and • An indication of whether rumble strips had a significant effect on the crash type of interest. Several relevant findings from the EB analyses are as follows: • Of the four analyses based on TOT crashes for horizontal curve sites (Table 69), two yield statistically significant results at the 90 or 95 percent confidence level. One signif- icant result indicates TOT crashes decrease when centerline rumble strips are installed along horizontal curves, while the other indicates an increase in TOT crashes when center- line rumble strips are installed along horizontal curves. • Of the four analyses based on FI crashes for horizontal curve sites (Table 70), one yields statistically significant results at the 90 or 95 percent confidence level. The one significant result indicates a decrease in FI crashes when centerline rumble strips are installed along horizontal curves. • Of the four analyses based on TOT target crashes for hori- zontal curve sites (Table 71), three yield statistically signifi- cant results at the 90 or 95 percent confidence level. All three significant results indicate a decrease in TOT target crashes when centerline rumble strips are installed along horizontal curves. • Of the four analyses based on TOT crashes for tangent sites (Table 72), two yield statistically significant results at the 90 or 95 percent confidence level. Both significant results indicate a decrease in TOT crashes when centerline rumble strips are installed along tangents. • Of the four analyses based on FI crashes for tangent sites (Table 73), three yield statistically significant results at the 90 or 95 percent confidence level. All three significant results indicate a decrease in FI crashes when centerline rumble strips are installed along tangents. • Of the four analyses based on TOT target crashes for tan- gent sites (Table 74), all four yield statistically significant results at the 90 or 95 percent confidence level. All four sig- nificant results indicate a decrease in TOT target crashes when centerline rumble strips are installed along tangents.

Table 68. Safety effectiveness of centerline rumble strips on rural two-lane roads— comparison and amalgamation with results from the IIHS EB study. Table 69. Safety effectiveness of centerline rumble strips on TOT crashes at horizontal curve sites using the EB method. Mile years (and crashes) Percent change in crash frequency from before to after rumble strip installation (%) Crash type Study Number of sites Total length (mi) Before After Estimatea SEb Test statisticc Significance Combined 1060 626.86 3239.8 (5875) 1952.1 (3393) –8.7 2.0 4.3 Significant at 95% CL 17–32 962 416.06 2117.2 (3365) 1380.1 (1912) –4.1 2.6 1.6 Not significant at 90% CL TOTAL IIHS 98 210.8 1122.6 (2510) 572.3 (1481) –14.1 3.0 4.7 Significant at 95% CL Combined 1060 626.86 3239.8 (2615) 1952.1 (1456) –11.7 2.8 4.2 Significant at 95% CL 17–32 962 416.06 2117.2 (1733) 1380.1 (920) –9.4 3.5 2.7 Significant at 95% CLFI IIHS 98 210.8 1122.6 (882) 572.3 (536) –15.5 4.5 3.4 Significant at 95% CL Combined 1060 626.86 3239.8 (733) 1952.1 (301) –30.2 4.5 6.7 Significant at 95% CL 17–32 962 416.06 2117.2 (403) 1380.1 (154) –37.0 5.3 7.0 Significant at 95% CL TOTAL TARGET IIHS 98 210.8 1122.6 (330) 572.3 (147) –21.4 7.8 2.7 Significant at 95% CL a A negative percent change indicates a decrease in crash frequency while a positive percent change indicates an increase in crash frequency. b SE: standard error of estimate. c Test statistic = abs(Estimate/SE); not significant at 90% CL if < 1.7; significant at 90% CL if ≥ 1.7; significant at 95% CL if ≥ 2. Percent change in crash frequency from before to after rumble strip installation (%) Roadway type State Number of sites Total length (mi) Estimatea SEb Test statisticc Significance Combined 331 68.15 +3.5 6.5 0.54 Not significant at 90% CL MN 135 28.41 –17.1 9.6 1.78 Significant at 90% CL PA 144 29.32 +16.0 9.2 1.74 Significant at 90% CL Rural two–lane roads WA 62 10.42 +2.7 16.0 0.17 Not significant at 90% CL a A negative percent change indicates a decrease in crash frequency while a positive percent change indicates an increase in crash frequency. b SE: standard error of estimate. c Test statistic = abs(Estimate/SE); not significant at 90% CL if < 1.7; significant at 90% CL if ≥ 1.7; significant at 95% CL if ≥ 2.

Percent change in crash frequency from before to after rumble strip installation (%) Roadway type State Number of sites Total length (mi) Estimatea SEb Test statisticc Significance Combined 331 68.15 –6.4 8.1 0.79 Not significant at 90% MN 135 28.41 –36.7 11.6 3.16 Significant at 95% CL PA 144 29.32 +9.8 11.4 0.86 Not significant at 90%Rural two–lane roads WA 62 10.42 –20.7 12.9 1.60 Not significant at 90% a A negative percent change indicates a decrease in crash frequency while a positive percent change indicates an increase in crash frequency. b SE: standard error of estimate. c Test statistic = abs(Estimate/SE); not significant at 90% CL if < 1.7; significant at 90% CL if ≥ 1.7; significant at 95% CL if ≥ 2. Percent change in crash frequency from before to after rumble strip installation (%) Roadway type State Number of sites Total length (mi) Estimatea SEb Test statisticc Significance Combined 331 68.15 –47.1 9.9 4.76 Significant at 95% CL MN 135 28.41 –52.1 13.6 3.83 Significant at 95% CL PA 144 29.32 –46.9 13.9 3.37 Significant at 95% CL Rural two–lane roads WA 62 10.42 +45.5 102.9 0.44 Not significant at 90% CL a A negative percent change indicates a decrease in crash frequency while a positive percent change indicates an increase in crash frequency. b SE: standard error of estimate. c Test statistic = abs(Estimate/SE); not significant at 90% CL if < 1.7; significant at 90% CL if ≥ 1.7; significant at 95% CL if ≥ 2. Percent change in crash frequency from before to after rumble strip installation (%) Roadway type State Number of sites Total length (mi) Estimatea SEb Test statisticc Significance Combined 312 215.24 –8.0 4.3 1.86 Significant at 90% CL MN 166 153.44 –9.7 5.5 1.76 Significant at 90% CL PA 73 23.25 –9.9 8.4 1.18 Not significant at 90% CL Rural two–lane roads WA 73 38.55 2.0 9.3 0.22 Not significant at 90% CL a A negative percent change indicates a decrease in crash frequency while a positive percent change indicates an increase in crash frequency. b SE: standard error of estimate. c Test statistic = abs(Estimate/SE); not significant at 90% CL if < 1.7; significant at 90% CL if ≥ 1.7; significant at 95% CL if ≥ 2. Table 70. Safety effectiveness of centerline rumble strips on FI crashes at horizontal curve sites using the EB method. Table 71. Safety effectiveness of centerline rumble strips on TOT target crashes at horizontal curve sites using the EB method. Table 72. Safety effectiveness of centerline rumble strips on TOT crashes at tangent sites using the EB method.

Percent change in crash frequency from before to after rumble strip installation (%) Roadway type State Number of sites Total length (mi) Estimatea SEb Test statisticc Significance Combined 312 215.24 –14.9 5.9 2.53 Significant at 95% CL MN 166 153.44 –17.8 7.8 2.28 Significant at 95% CL PA 73 23.25 –21.8 10.0 2.18 Significant at 95% CL Rural two–lane roads WA 73 38.55 +10.2 17.3 0.59 Not significant at 90% CL a A negative percent change indicates a decrease in crash frequency while a positive percent change indicates an increase in crash frequency. b SE: standard error of estimate. c Test statistic = abs(Estimate/SE); not significant at 90% CL if < 1.7; significant at 90% CL if ≥ 1.7; significant at 95% CL if ≥ 2. Percent change in crash frequency from before to after rumble strip installation (%) Roadway type State Number of sites Total length (mi) Estimatea SEb Test statisticc Significance Combined 312 215.24 –49.3 6.9 7.14 Significant at 95% CL MN 166 153.44 –48.8 7.9 6.18 Significant at 95% CL PA 73 23.25 –43.3 18.4 2.35 Significant at 95% CL Rural two–lane roads WA 73 38.55 – 67.3 19.0 3.54 Significant at 95% CL a A negative percent change indicates a decrease in crash frequency while a positive percent change indicates an increase in crash frequency. b SE: standard error of estimate. c Test statistic = abs(Estimate/SE); not significant at 90% CL if < 1.7; significant at 90% CL if ≥ 1.7; significant at 95% CL if ≥ 2. Table 73. Safety effectiveness of centerline rumble strips on FI crashes at tangent sites using the EB method. Table 74. Safety effectiveness of centerline rumble strips on TOT target crashes at tangent sites using the EB method.

113 Estimating the Safety Effectiveness of Dual Applications of Rumble Strips Based upon the crash statistics in Table 63, no observable trends are apparent concerning the safety effectiveness of dual applications of both centerline and shoulder rumble strips along the same roadway. In some cases, the crash frequen- cies are greater in the after period compared to the before period, and in other cases the reverse is true. This observation is very likely due to limited sample sizes of the data. Summary of Key Findings The primary objectives of the safety evaluation of center- line rumble strips are as follows: • Quantify the safety effectiveness of centerline rumble strips on specific types of roads including urban multilane un- divided highways (nonfreeways), urban two-lane roads, rural multilane undivided highways (nonfreeways), and rural two-lane roads. • Quantify the safety effectiveness of centerline rumble strips along varying roadway geometry (i.e., tangent vs. horizon- tal curve). • Quantify the safety effectiveness of dual applications of rumble strips (i.e., centerline and shoulder rumble strips installed on the same road section). Based upon the analysis results, the key findings from the safety evaluation of centerline rumble strips are as follows: • The most reliable and comprehensive estimates of the safety effectiveness of centerline rumble strips on urban and rural two-lane roads with their associated standard errors are: Urban Two-Lane Roads – Centerline rumble strips (based on results from this research):  40 percent reduction in TOT target crashes (SE = 17) and  64 percent reduction in FI target crashes (SE = 27). Rural Two-Lane Roads – Centerline rumble strips [based on combined results from this research and Persaud et al. (4)]:  9 percent reduction in TOT crashes (SE = 2),  12 percent reduction in FI crashes (SE = 3),  30 percent reduction in TOT target crashes (SE = 5), and  44 percent reduction in FI target crashes (SE = 6) (based on results from this research). • Limited mileage of centerline rumble strips along urban multilane undivided highways (nonfreeways) and rural multilane undivided highways (nonfreeways) prohibited formal evaluation of the safety effectiveness of this treat- ment along these respective roadway types. • The safety benefits of centerline rumble strips on horizontal curves and tangents, based on TOT target crashes, are remarkably similar with estimated 47 percent and 49 percent reductions in TOT target crashes, respectively. This result would indicate that the safety effectiveness of centerline rumble strips is for practical purposes the same for both curved and tangent alignments. • Limited mileage of dual applications of rumble strips (i.e., centerline and shoulder rumble strips installed on the same road section) along rural two-lane roads prohibited formal evaluation of the safety effectiveness of this treatment along this respective roadway type.