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.

Appendix A Detailed Literature Review

A-1 This appendix provides a detailed summary of completed research on shoulder and centerline rumble strips. The information is organized in the following manner: • Safety impacts of shoulder rumble strips • Safety impacts of centerline rumble strips • Operational impacts of centerline rumble strips • Vehicle dynamics related to vibration and noise stimuli • Effects of rumble strips on specific types of highway users (i.e., motorists, motorcyclists, and bicyclists) • Pavement performance issues • Other potential adverse concerns A.1. Safety Effects of Shoulder Rumble Strips This section presents results from studies on the safety effectiveness of shoulder rumble strip applications. Arizona Study In 1973, the Arizona DOT completed a study of the safety effectiveness of several shoulder treatments installed at locations that did not necessarily show either a high accident count or an above-average number of SVROR accidents (16). The analysis was performed using 1970 to 1972 accident data for several shoulder treatments along Interstate Routes 8 and 10. Results of the analysis indicated that sections with shoulder grooving had the fewest SVROR crashes, with 61 percent fewer SVROR crashes than other shoulder treatments. Additional analyses, using 1973 to 1976 accident data on the same 10-mi (16.1-km) test section of I-8 and an adjacent 16-mi (25.7-km) control section, concluded that the grooved shoulder sections had 80 percent fewer SVROR crashes per mile and 80 percent fewer SVROR crashes per million vehicle-mi traveled (MVMT) than the control section with standard shoulders. California Study California DOT (Caltrans) initiated a study in the early 1970s to develop and evaluate rumble strip installations that would alert motorists and prevent SVROR crashes (49). The first phase of the research consisted of developing shoulder rumble strip patterns that would alert motorists. The second phase involved installing trial installations at four locations and conducting a before-after crash analysis. After conducting sound, vibrational, and controllability studies on 57 rumble strip patterns, the most effective patterns, in terms of alerting the inattentive/drowsy motorist,

A-2 were selected for trial installations. Rumble strips made from ribs of asphalt and aggregate, rows of raised ceramic pavement markers, or incised (i.e., milled) slots were placed along selected sections of I-5, I-15, I-80, and SR 99 (a non-Interstate freeway). Criteria for site selection were adequate geometrics, a history of potentially correctable accidents, and a long period of time with no major changes to the roadway. In addition, a minimum right shoulder width of 8 ft (2.4 m) was selected since this would allow sufficient space for construction of the rumble strips and provide a substantial hazard-free runout area for vehicles that cross the rumble strips. The total length of the test installations was 36 mi (60 km). An analysis of accident data (one year before and one year after) at the treatment sites and control sites was performed to evaluate the safety effectiveness of the rumble strip treatments. The control sections consisted of highway segments equal in length and adjacent to each end of the test installations. These control sections were selected to check on the validity of the before-after data and to evaluate whether SVROR accidents were moved downstream past the treatment sections. The analysis investigated the impact on both total accidents and SVROR accidents. Table A-1 provides a summary of the accident data at the treatment sites. Results from this safety evaluation did not provide statistically significant proof that rumble strips are an effective means of reducing SVROR crashes. Table A-1. Accident Summary for Shoulder Rumble Strip Treatments in California (49) SVROR (Right Side) All Other Total Location Before After % Change Before After % Change Before After % Change I-5 Seal Coat 21 15 –28 70 80 +14 91 94 +3 SR 99 (NB) Pavement Markers 8 7 –12 22 15 –32 30 22 –27 SR 99 (SB) Seal Coat 2 9 +350 26 12 –54 28 21 –25 I-15 Seal Coat 2 0 –100 6 3 –50 8 3 –63 I-80 Milled 5 5 0 14 10 –29 19 15 –21 Total 38 36 –5 138 120 –13 176 155 –11 In the later half of the 1970s and early part of the 1980s, Chaudoin and Nelson (17) conducted another study on shoulder rumble strip applications for Caltrans. To reduce the number of SVROR crashes, rolled rumble strips were installed continuously along the shoulders of 158.5 mi (255 km) of freeways in the Mojave Desert on I-15 and I-40. The rumble strips were installed during resurfacing operations. Along most of the treatment sections, the rumble strips were placed along both the right (outside) and left (median) shoulders, but on several treatment sections the rumble strips were only installed on the right (outside) shoulder. The rolled rumble strips were 3 ft (0.9 m) in length and were spaced 8 in (203 mm) apart. The depth varied according to the thickness of asphalt concrete surfacing being placed, but the preferred and average depth was 1.5 in (38 mm).

A-3 A one-year before-after study was performed to evaluate the safety effectiveness of the rolled rumble strips. Nearby comparison sections of approximate length to the treatment sections were also identified for evaluation. In the analysis, SVROR accidents were tabulated for sections where rumble strips were installed. If rumble strips were installed only along the right (outside) shoulder, then only right-side SVROR accidents were included in the analysis for the given site. Truck accidents were not included because it was believed that rumble strips would have a lesser effect on truck drivers. Similarly, motorcycle accidents were excluded from the analysis for the same reason. The accident analysis indicated a significant decrease in SVROR accidents as a result of the rumble strip installations. On the roadway sections where the rumble strips were installed, 194 SVROR crashes occurred in the before period, versus 100 SVROR crashes in the after period. On the control sections, 272 SVROR crashes occurred in the before period, while 326 SVROR crashes were reported in the after period. Thus, SVROR crashes were reduced by 49 percent on the freeway sections where continuous shoulder rumble strips were installed, while SVROR crashes increased by 20 percent on the comparison sites. The decrease in SVROR crashes was found to be statistically significant at the 99 percent confidence level. In addition, right-shoulder rumble strips proved more effective than left-shoulder rumble strips. Right-shoulder rumble strips reduced SVROR crashes by 63 percent as compared to a reduction of 18 percent by left- shoulder rumble strips. The 18 percent reduction in median SVROR crashes was not significant at the 90 percent confidence level. A 19 percent reduction in total accidents was also attributed to the shoulder rumble strip installations. This decrease in total crashes was found to be statistically significant at the 99 percent confidence level. Finally, it was concluded that shoulder rumble strips may be effective in reducing SVROR accidents on similar monotonous routes where there is a history of this type of accident and should only be installed at locations where the application is practical and justified. The authors cautioned that this conclusion should not be construed to apply to urban freeways nor necessarily those considered rural in nature without unusual problems. Connecticut Study Between the years of 1996 and 2000, the Connecticut Department of Transportation (ConnDOT) installed 1,020 shoulder-mi (1,640 shoulder-km) of rumble strips (18). Rumble strips were installed on limited-access roadways with a minimum of a 3-ft (0.9-m) shoulder. ConnDOT exclusively installed milled rumble strips placed in a continuous pattern. The dimensions of the rumble strips are 7 in (178 mm) wide, 16 in (406 mm) long, and 0.5 in (13 mm) deep, spaced 12 in (305 mm) on center. Initially, the rumble strips were placed 6 in (152 mm) from the edge of the travel way on the outside shoulder, but after complaints of noise from residents, ConnDOT changed their policy and placed the rumble strips 12 in (305 mm) from the edge of the travel way on the outside shoulder. The rumble strip placement on the left (i.e., inside) shoulder is 6 in (152 mm) from the left edge of the travel way and has not changed.

A-4 A before-after study was conducted by Annino (18) to evaluate the safety effectiveness of the shoulder rumble strips in reducing single-vehicle, fixed object, run- off-the-road accidents. Three years of before-accident data were collected (1993-1995), and three years of after-accident data were collected (1996-1998), at 11 locations where rumble strips were installed. Accident data from these 11 sites were analyzed with accident data from 11 comparison segments that did not have rumble strips. Table A-2 summarizes the accident data from the 11 treatment sites and comparison sites. Annino used a before-after with comparison sites to evaluate the safety benefits of the shoulder rumble strips. Annino concluded there was a 32 percent reduction of rumble- strip-related accidents (defined as single-vehicle: fixed-object, off-shoulder accidents). Table A-2. Summary of Accident Data for Treatment and Comparison Sites for Shoulder Rumble Strips in Connecticut (18) Section ID Rumble or comparison section Route Dir Start MP End MP Total before accidents Total after accidents Percent change (%) 1R Rumble 8 NB 19.28 25.14 23 20 –13.04 1C Comparison 8 NB 13.42 19.28 37 59 59.46 2R Rumble 8 NB 42.64 50.11 36 36 0.00 2C Comparison 8 NB 35.17 42.64 31 35 12.90 3R Rumble 8 SB 19.28 25.14 20 18 –10.00 3C Comparison 8 SB 13.42 19.28 20 28 40.00 4R Rumble 9 NB 0.23 3.91 7 7 0.00 4C Comparison 9 NB 3.91 7.59 15 16 6.66 5R Rumble 9 NB 24.47 27.43 26 29 11.53 5C Comparison 9 NB 27.43 30.39 16 9 –43.75 6R Rumble 9 NB 37.49 39.93 9 4 –55.56 6C Comparison 9 NB 35.05 37.49 15 14 -6.67 7R Rumble 9 SB 37.49 40.71 11 9 –18.18 7C Comparison 9 SB 34.27 37.49 25 22 –12.00 8R Rumble 9 SB 24.47 29.10 36 35 –2.77 8C Comparison 9 SB 19.84 24.47 17 13 –23.52 9R Rumble 9 SB 0.23 3.91 10 13 30.00 9C Comparison 9 SB 3.91 7.59 21 24 14.29 10R Rumble 15 NB 50.20 59.72 58 24 –58.62 10C Comparison 15 NB 37.62 47.14 39 36 –7.69 11R Rumble 15 SB 50.20 59.72 34 15 –55.88 11C Comparison 15 SB 37.62 47.14 29 38 31.03 Total Rumble 270 210 –22.22 Total Comparison 265 294 10.94

A-5 Florida Study To reduce SVROR crashes along a 19-mi (31-km) section of U.S. 1 in Florida, the main highway to Key West, raised pavement markers were installed four abreast across the shoulder at a 45-degree angle (16). In its 1980 to 1981 annual report, the Florida DOT reported a decrease in fixed object crashes from 19.5 to 11.5 per year due to the raised rumble strip treatment. The Florida DOT also reported a decrease in ran-into-water accidents from 8 to 5.5 per year. Thus, the Florida DOT concluded the special raised rumble strip treatment achieved their goal in preventing the inattentive/drowsy motorists from leaving the shoulder. Illinois and California Study Griffith (1) examined the safety effects of continuous shoulder rumble strip applications on rural and urban freeways in Illinois and California. Before-after evaluations were conducted on projects that involved the installation of rolled rumble strips as part of a resurfacing project. In Illinois, the standard depth of rolled rumble strips is 0.75 in (19 mm) with a length of 3 ft (0.9 m) and a spacing of 8 in (203 mm). The rumble strips are installed 12 in (304 mm) from the edge of pavement. In California, shoulder rumble strips are 0.75 in (19 mm) or less in height if raised or 1 in (25 mm) or less in depth if indented and extend along the highway shoulder. The maximum length of the shoulder rumble strip did not exceed 3 ft (0.9 m). Data were extracted from the Highway Safety Information System (HSIS) databases to perform the safety evaluation. The Illinois data included 63 projects totaling 284 mi (457 km) that were completed between 1990 and 1993. The California data included 28 projects totaling 122 mi (197 km) that were completed between 1998 and 1993. Griffith analyzed the data using matched-pair comparisons (i.e., one-to-one matching between treatment sites) and comparison groups (i.e., one-to-many matching between treatment sites). In general, Griffith concluded continuous shoulder rumble strips provide a safety benefit to motorists on freeways. Because larger sample sizes were obtained from Illinois, more weight was given to the Illinois findings, so based on the Illinois data, Griffith estimated the average safety effectiveness of continuous shoulder rumble strips to be: • On all freeways, an 18.3 percent reduction in total SVROR crashes with a standard deviation of 6.8 percent. • On all freeways, a 13 percent reduction in injury SVROR crashes with a standard deviation of 11.7 percent. • On rural freeways, a 21 percent reduction in total SVROR crashes with a standard deviation of 10.2 percent. Griffith also evaluated two types of potential adverse effects related to the safety of continuous shoulder rumble strips. One potential adverse effect pertains to the possibility

A-6 that some motorists may overreact to the stimulation generated by the rumble strips resulting in loss of control of their motor vehicles. The second potential adverse effect is crash migration, which occurs when a motorist is temporarily saved from a crash at a treated site but crashes downstream of the treated area or in another location on the network. Griffith found these potential adverse effects to be insignificant. Kentucky Study Kirk (101) conducted at study of continuous shoulder rumble strips in Kentucky with the primary intention to answer the following questions: • Do continuous shoulder rumble strips reduce crash frequency on rural two-lane roads with little or no shoulder? • When limited pavement width is available, should shoulder width be increased to provide continuous shoulder rumble strips or should lane width be maximized? A total of 162 unique sections were identified for crash analysis, including 109 sections with rumble strips and 53 sections without rumble strips. Three years of crash data were analyzed. The crash data were analyzed using regression analysis. Based upon this analysis, the following conclusions were made: • Rural two-lane roads with continuous shoulder rumble strips have a statistically significant lower total crash rate than roads without continuous shoulder rumble strips. • Rural two-lane roads with continuous shoulder rumble strips have a statistically significant lower crash rate resulting from inattention/drowsiness than roads without continuous shoulder rumble strips. • Rural two-lane roads exhibit a statistically significant decrease in SVROR crash rates as lane width increases. • Crash rates on rural two-lane roads are generally lower when shoulder width is maximized and lane width is minimized. Maine Study Maine DOT began installing continuous shoulder rumble strips along rural freeways between 1994 and 1995 (20). Prior to these installations, Gårder and Alexander (102) assessed the safety effectiveness of continuous shoulder rumble strips for Maine DOT based upon previous studies and estimated a 50 percent reduction in sleep related accidents due to the installation of shoulder rumble strips. Approximately, 37 mi (59 km) of rumble strips were installed on the Interstate system in Maine. The dimensions of the rumble strips were as follows:

A-7 Construction—milled Length—16 in (406 m) Width—7 in (178 mm) Depth—0.5 to 0.75 in (13 to 19 mm) Spacing—12 in (305 mm) on centers A preliminary analysis of accident data was performed (20). The analysis included accident data from 1991 to 1994. Accident reports from 1995 were not available for the evaluation. Table A-3 provides a summary of the accident data. Based upon the limited data, it was not possible to draw any conclusions on the safety effects of continuous shoulder rumble strips installed along selected sections of Maine’s Interstate highways. Table A-3. Accident Counts Before and After Installation of Shoulder Rumble Strips in Maine (20) Analysis period Treatment sections Comparison sections Before Period (44 months) 53 accidents 56 accidents After Period (3 months) 3 accidents 3 accidents Michigan Study Morena (21) evaluated the safety effectiveness of different shoulder rumble strip types/designs in reducing SVROR accidents. Morena compared the safety of roads with milled, rolled, and formed (intermittent) rumble strips. Milled shoulder rumble strips in Michigan are 7 in (178 mm) wide, 16 in (406 mm) long, and 0.5 in (13 mm) deep; they are placed 12 in (305 mm) from the edgeline on the left shoulder and 12 in (305 mm) or 24 in (610 mm) from the edgeline on the right shoulder. No design criteria were given for Michigan’s rolled or formed concrete intermittent rumble strips. The safety analysis included accident data from 984 mi (1,584 km) of roadway with rumble strips installed from 1996-2001. From these 984 mi (1,584 km) of roadway, over 3,000 accident reports were reviewed to determine whether they should be considered SVROR accidents. A total of 1,887 of the 3,000 accidents were classified as SVROR. Morena reasoned that SVROR accidents were the accidents types most likely to be preventable by shoulder rumble strips and include drowsy or distracted drivers. The safety performance of the 984 mi (1,584 km) of roadway with rumble strips was compared to the safety performance of the 454 mi (730 km) of roadway without rumble strips. Morena found an almost equal number of SVROR accidents and similar severity distributions on each side of the roadway. The severity of SVROR accidents was extremely high with 17 percent of the accidents resulting in a fatal or incapacitating injury. Morena also reported a noticeable decline in SVROR accidents with the increase of ADT.

A-8 Morena found that roads with milled rumble strips had a 38 to 39 percent reduction in SVROR accidents in an ADT range of 5,000-60,000 vpd. Both formed intermittent and rolled rumble strips reduced SVROR accidents by 25 percent. Minnesota Studies Carrasco et al. (3) examined the safety effect of shoulder rumble strips constructed on rural multilane highways in Minnesota. Highway Safety Information System (HSIS) data from Minnesota were used in the evaluation along with additional data provided by the Minnesota Department of Transportation (MnDOT). The dimensions of the rumble strips installed along the rural multilane highways in Minnesota were as follows: Construction: milled Pattern: continuous Length: 24 in (610 mm) Width: 2 to 4 in (51 to 102 mm) Depth: 0.375 to 0.625 in (10 to 16 mm) Spacing: 8 to 12 in (203 to 305 mm) on centers Lateral Offset: 6 and 12 in (152 and 305 mm) from outside edge of travel lane Data from 23 treatment sites were included in the before-after evaluation. The section lengths ranged from 2.7 to 12.9 mi (15.9 to 20.8 km), encompassing a total of 163 mi (262 km) of highway. The shoulder rumble strips were constructed on the divided multilane highways between 1991 and 1998. The rumble strips were installed on the inside and outside shoulders for both directions of travel. The speed limits of the treatment sites ranged from 55 to 70 mph (89 to 113 km/h). Data from 11 comparison sites were also collected. The section lengths of the comparison sites ranged from 4.45 to 11.5 mi (7.16 to 18.5 km), encompassing a total of 83 mi (134 km) of highway. Carrasco et al. analyzed the data using both a naïve before-after study approach and a before-after analysis with comparison sites. The results of the naïve before-after analysis indicated the average reduction in total crashes and injuries at the treatment sites were 16 and 17 percent, respectively. The naïve before-after analysis also found a 10- and 22 percent reduction in total and injury SVROR crashes, respectively. Based upon the analysis with comparison sites, total crashes and injuries were reduced by 21 and 26 percent, respectively. Total and injury SVROR crashes were reduced by 22 and 51 percent, respectively. Similar to the Carrasco et al. (3) study, Patel et al. (2) conducted research on the safety effects of shoulder rumble strips in Minnesota utilizing available HSIS data. However, the focus of this evaluation was on estimating the reduction in SVROR crashes along rural two-lane roads. All sites had lane widths of 12 ft (3.7 m) and right shoulders that varied in width from 1-2 ft (0.3-0.6 m) to 12 ft (3.7 m). An initial pool of 36 sites in Minnesota were

A-9 reduced to 23, representing 183 mi (294 km) of roadway, after eliminating sites where additional geometric changes might have occurred at the same time the rumble strips were installed. After an effort was made to remove non-conforming sites from the study, safety performance functions (SPFs) were developed for all crashes and injury crashes. The effect of installing shoulder rumble strips was estimated to be a 13 percent reduction for all SVROR crashes and an 18 percent reduction for injury SVROR crashes. This effect was found to be statistically insignificant at the 95th percentile confidence level; Patel et al. postulated that this was due to the relatively small sample size and restated their confidence in the positive effect of shoulder rumble strips in this highway class. Patel et al. recommended, but was unsuccessful at finding any correlation between rumble strip effectiveness and horizontal curvature, night-time crashes, shoulder clear- zone width, or clear-zone nature (i.e., side slopes). Montana Study Marvin and Clark (22) conducted a study to evaluate the safety effectiveness of shoulder rumble strips in preventing single vehicle off-road and rollover crashes (under wet or dry pavement conditions only) on Interstate and primary highways in Montana. The rumble strips in this study were all consistent with Montana’s 1996 design polices for rumble strips: Concrete: Construction—formed continuous corrugation Length—12 to 16 in (300 to 400 mm) Radius—1 in (25 mm) Depth—1 in (25 mm) Spacing—4.5 in (114 mm) on centers Lateral Offset—6 in (150 mm) outside edgeline Asphalt: Construction—milled Length—12 to 16 in (300 to 400 mm) Radius—12 in (300 mm) Depth—0.5 to 0.75 in (13 to 19 mm) Spacing—4.5 in (114 mm) on centers Lateral Offset—6 in (150 mm) outside edgeline The study investigated 393 mi (632 km) of Interstate (35 percent of Montana’s total Interstate system) and 213 mi (343 km) of primary roadway (4 percent of Montana’s total primary routes). Accident data from three years before the rumble strips were installed were compared with accident data for three years after the installation. In most cases, the study periods were 1992-1994 (before) and 1997-1999 (after). The analysis also

A-10 considered data from comparison sections. Several conclusions drawn from this evaluation are as follows: • In the case of both Interstate and primary highways, rumble strips appear to lessen the number of crashes occurring during hours of darkness. It follows that through times of compromised visibility, the ability of rumble strips to offer warning that appeals to senses other than sight would decrease the probability of being involved in a SVROR accident. • On Interstate roadways, rumble strips may have contributed to reducing the number of crashes experienced by drivers under the age of 21 and older drivers over the age of 50. No significant benefits were evident regarding primary roadways concerning age. The differences from Interstate highways likely stem from the fact that primary routes often have narrower shoulders than Interstates. • Limited data suggest that motorcycles may be impacted by rumble strip installations on Interstates. • No conclusions were drawn concerning rumble strips and bicycles because there were no records of bicycle accidents on any of the study segments. • Roadway widths were investigated for primary highways due to their high variability on such roads throughout Montana. Crash rates before and after rumble strip installation did not show any appreciable dependence upon roadway widths, however, behaviors regarding severity were found to be counterintuitive without explanation. • Statistical analyses indicated that the reduction of Interstate off-road crash rates attributable to shoulder rumble strips was 14 percent with a corresponding reduction of 23.5 percent in severity rates. In conjunction, the benefit/cost ratio for construction of shoulder rumble strips on Interstate highways was 19.5. However, the safety benefits of shoulder rumble strips on primary highways was uncertain. • In general, rumble strips seem to be moderately successful in reducing the occurrence of various situation crashes, most notably those caused by drowsy or inattentive driving. As they pertain to the roadway system, the effect of shoulder rumble strips on crash experience was not statistically significant at the confidence level investigated. In fact, while Interstate and primary system analyses shows some benefit for off-road crash rate and severity in certain situations, rumble strip performance for rollover crashes showed that severity may well have increased through rumble strip deployment or other undefined factors. No specific rationale for determining the reason behind this increase is evident. Nevada Study Nambisan et al. (103) conducted an evaluation of the effectiveness of continuous shoulder rumbles in reducing SVROR crashes in Nevada. A total of 370 roadway segments were analyzed. The segments consisted of interstates, U.S. routes, and state

A-11 routes. Crash data from 1995 to 2003 were analyzed. The analysis was based on before- after comparisons of SVROR crash frequencies and crashes rates. It was concluded from the analysis that continuous shoulder rumble strips on roads in Nevada were effective in reducing the frequency of SVROR crashes and the corresponding crash rates. The analyses did not include information related to traffic volumes or vehicle miles of travel. New York Study In New York, continuous shoulder rumble strips have been installed along Interstate highways and parkways maintained by the New York State Department of Transportation (NYSDOT) and along the New York State Thruway which is owned and operated by the New York State Thruway Authority (NYSTA) (23). In 1993, NYSDOT installed only 91 shoulder-mi (148 shoulder-km) of continuous rumble strips, as compared to 1,725 shoulder-mi (2,777 shoulder-km) in 1995 and 3,150 shoulder-mi (5,017 shoulder-km) in 1998. Continuous shoulder rumble strips were installed on 1,945 shoulder-mi (3,131 shoulder-km) of the New York State Thruway by 1996. Both agencies, NYSDOT and NYSTA, collected before and after data to evaluate the safety effectiveness of the continuous shoulder rumble strips installed within the state. A sample of the data collected by NYSTA is presented in Table A-4, summarizing the number of SVROR crashes and the vehicle-miles traveled during the before and after study periods. The crashes represented in Table A-4 are due to the following causes: • Alcohol involvement • Driver inattention • Driver inexperience • Drugs (illegal) • Fell asleep • Illness • Passenger distraction • Prescription medication • Fatigue, drowsiness • Glare Both agencies have noted a reduction in SVROR crashes of at least 65 to 70 percent based, apparently, upon a naïve before-after approach.

A-12 Table A-4. Before-After Comparison for Shoulder Rumble Strips Along the New York State Thruway (23) Year Total SVROR crashes Total injuries Total fatalities Vehicle-miles traveled (millions) Before and During Rumble Strip Installation 1991 557 358 17 6,744 1992 566 407 17 7,612 1993 588 328 8 7,792 After Rumble Strip Installation Completed (Percent Reduction from 1991) 1996 161 (74) 113 (72) 4 (75) 8,512 1997 74 (88) 54 (87) 1 (95) 8,692 Pennsylvania Study In 1984, 48 percent of the crashes along the Pennsylvania Turnpike were SVROR crashes (104). The percentage of SVROR crashes continued to increase over the next two years to 51 percent in 1985 and 57 percent in 1986. During this same time period, the Pennsylvania Turnpike Commission was developing the Sonic Nap Alert Pattern or SNAP, a narrow rumble strip to be located continuously along the right shoulder, just outside of the edgeline of the pavement. After testing five different patterns, a standard rumble strip pattern was selected: Construction—milled Length—16 in (406 mm) Width—7 in (178 mm) Depth—0.5 (13 mm) Spacing—12 in (305 mm) on centers Based upon an initial experience at a single site approximately 7 mi (11 km) in length, by May 1992 rumble strips were installed along five more sections of the Turnpike totaling 31 mi (50 km). A safety evaluation in 1993 confirmed the effectiveness of SNAP. The Turnpike experienced a 70 percent reduction in SVROR crashes over substantial time periods. As a result of their experiences, the Turnpike Commission initiated a program to have over 80 percent of the turnpike system protected with SNAP by the end of 1994. In a follow-up study, Hickey (24) reviewed the initial results presented by Wood (104) and further accounted for traffic exposure and a decline in all accidents during the years considered. Furthermore, Hickey excluded from the analysis several single-vehicle accident types considered non susceptible to SNAP such as weather (snow, ice, slippery, wet, and spun out), blow out, flat, mechanical defect, improper towing, forced movement, evading object, animal, work zone, blackout, and inside vehicle event. Thus, Hickey revised the initial reported accident reduction attributable to SNAP to a 65 percent reduction in SVROR rates. Hickey also expanded the study to consider all the sections of the turnpike where SNAP were installed, and based upon data from 1990 to 1995 Hickey

A-13 found a 60 percent reduction in SVROR crashes over 53 sections of the turnpike totaling 348 mi (560 km). Tennessee Study The Tennessee Department of Transportation (TDOT) (25) has been installing milled shoulder rumble strips on all of its Interstate resurfacing projects since 1996. In 2001, TDOT began a statewide project to install shoulder rumble strips along the approximately 315 mi (507 km) of Tennessee’s Interstate system where rumble strips had not yet been installed. TDOT reported a 31 percent reduction in SVROR crashes on the portions of the Interstate system that did not previously have rumble strips. Texas Study The effect of edgeline rumble strips on erratic driving behaviors was evaluated by Miles et al. (105) at the request of TxDOT. Using a before-after video analysis of sites where edgeline rumble strips were installed, Miles et al. attempted to document events in which the driver’s reaction to entering the shoulder was effected in such a way as to cause an unsafe situation. The erratic behaviors in question included (but were not limited to) hard braking, swerving, rapid alignment or lane shifting, correcting the trajectory in the wrong direction or losing control of the vehicle. After reviewing 120 hours of video for the before and after period (cumulative) at sights along a rural highway, Miles et al. were unable to find any events that were considered erratic. They did, however, find a reduction in passenger vehicles encroaching upon the shoulder of approximately 35 percent. Thus, it was concluded that while the installation of edgeline rumble strips did increase the number of corrective actions by drivers to remain in-lane, it did not create an unsafe environment. Utah Study The Utah DOT compared the accident rate experience along Interstate highways with and without rumble strips (26). In Utah, rolled rumble strips were installed along interstates with asphalt shoulder pavements, and formed rumble strips were installed along interstates with concrete shoulder pavements. The dimensions of the rumble strips were as follows: Asphalt: Construction—rolled Length—2 ft (0.6 m) Width—1.5 in (38 mm) Depth—1 in (25 mm) Spacing—8 to 9 in (203 to 229 mm) on centers Lateral Offset—12 in (305 mm) from outside edgeline

A-14 Concrete: Construction—formed skip pattern Length—6 ft (1.8 m) for 10 ft (3.0 m) shoulders 4 ft (1.2 m) for 6 ft (1.8 m) shoulders 3 ft (0.9 m) for 4 ft (1.2 m) shoulders Width—4.5 in (114 mm) Depth—0.75 in (19 mm) Spacing—4.5 in (114 mm) on centers Lateral Offset—6 in (152 mm) to 2 ft (0.6 m) from outside edgeline Skip Pattern—6 ft (1.8 m) rumble strips and 4 ft (1.2 m) clear space or 6 ft (1.8 m) rumble strip patterns on 10 ft (3.0 m) centers The analysis included 41 segments (30 asphalt and 11 concrete pavements) with rumble strips totaling 186 mi (299 km), and 35 segments without rumble strips totaling 110 mi (177 km). Among the segments with rumble strips, there were 111 mi (179 km) of asphalt pavement and 75 mi (121 km) of concrete pavement. Accident data for the years 1990 through 1992 were used in the evaluation. Cheng et al. compared the accidents rates of Intestate segments with and without rumble strips. Both overall accident rates and SVROR type accident rates were considered in the analysis. The results showed overall crash rates were 33.4 percent higher on the control sections as compared to the sections where rumble strips were installed. Similarly, SVROR crash rates were 26.9 percent higher on the sections without rumble strips. Statistical tests conducted during the analysis could not verify the significance level of the difference in the accident rates. Cheng et al. also compared the accident rates of rumble strip on asphalt shoulders and rumble strips on concrete shoulders. The analysis revealed the overall accident rate for concrete pavement was 16.9 percent higher than that for asphalt pavement. Similarly, the accident rate for SVROR crashes on concrete pavement was 23.8 percent higher than that for asphalt pavement. Statistical tests could not verify the significance level of difference in the accident rates. The study also showed that rumble strips were effective in reducing crash severity. In general, Cheng et al. concluded the following: • Freeways without shoulder rumble strips experience a higher rate of accidents over those highways with shoulder rumble strips. • Highway segments with rumble strips on asphalt shoulders (continuous and near travel lane design) have lower accident rates than highway segments with rumble strips on concrete shoulders (discontinuous and offset from travel lane design). • The discontinuous design proved to be less effective in alerting drivers to potentially dangerous driving patterns.

A-15 • Rumble strips are effective in lowering accident severity, and furthermore, the continuous design proved to be even more effective over the discontinuous design. Based on the results of the safety evaluation, Cheng et al. recommended the following points when considering installation of rumble strips along highway shoulders: 1. Shoulder rumble strips should be installed on highway and freeway shoulders. 2. Rumble strips should be as wide as possible. 3. Highway segments with high SVROR accident rates, such as rural areas should receive highest priorities. 4. Detours during road construction should be considered when determining the method of installation. Generally, milled rumble strips will be an effective method during any phase of construction. 5. The continuous design is preferred over the discontinuous design. 6. Rumble strips should be placed as close to the travel lanes as possible. Such placement not only provides advance warning to drivers but also provides a buffer zone between traffic and bicyclists on the shoulder. Virginia Study Chen et al. (27) performed a three-year before-after safety evaluation of continuous shoulder rumble strips installed along rural freeways in Virginia. Crash data were obtained from the Highway Traffic Records Inventory System (HTRIS) database maintained by Virginia DOT. Crash data from June 1994 through October 2000 were used in the analysis. The safety evaluation was based on a before-after analysis with comparison sites using data from 9 treated sites representing 285 mi (459 km) of roadway, and 9 comparison sites of similar length. Table A-5 summarizes the accident data and analysis results. The detailed statistical analysis indicated an overall reduction rate in SVROR crashes on Virginia’s rural highway was 51.5 percent. Washington Study The Washington State DOT installed shoulder rumble strips at six locations between 1986 and 1990 [unpublished; cited in Harwood (15)]. Three types of rumble strips were installed at the various locations. Raised pavement markers were installed on the shoulder at one location; raised 12-in (305-mm) wide rumble strips were installed at a second location; and rolled rumble strips were installed at four other locations. A safety evaluation for five of the six locations revealed a statistically significant reduction in crash frequency at only one of the five locations. However, when considering the five sites collectively, the overall crash frequency decreased by 18 percent from before to after rumble strip installation. The report does not indicate whether this overall reduction in accident experience was statistically significant.

A-16 Table A-5. Accident Summary of SVROR Crashes and Statistical Significance Before and After Installation of Shoulder Rumble Strips in Virginia (27) Treated Sites Comparison Sites Segment Before After Before After Percent change Statistically significant (95%) 1 65 46 69 73 –49.50 Yes 2 157 89 170 149 –54.60 Yes 3 52 37 47 62 –85.40 Yes 4 52 31 49 54 –84.80 Yes 5 41 25 55 35 –4.40 No 6 33 48 77 99 11.60 No 7 64 67 66 93 –34.60 Yes 8 48 45 59 103 –86.00 Yes 9 79 70 72 50 21.60 No All 591 485 654 718 –51.50 Yes Multistate Study Ligon et al. (16) conducted a before-after safety evaluation of shoulder texture treatments on rural freeways based on data from 11 states (Arizona, California, Florida, Georgia, Mississippi, Nevada, North Carolina, South Carolina, South Dakota, Utah, and Wisconsin). The various shoulder texture treatments installed in the 11 states included: Concrete Shoulder Treatments: • Concrete corrugated panels Bituminous Shoulder Treatments: • Bituminous single groove • Bituminous indented strip • Bituminous surface treatment Miscellaneous Textured Shoulder Treatments: • Raised Shoulder Treatments − Jiggle bars − Raised circular pavement markers − Bituminous ribbed panels • Indented Shoulder Treatments − Bituminous corrugated panels − Cold planing The analysis included comparing accident rates at treatment sites and control sites. The control sites were chosen to be as similar as possible as the treatment sites except that the shoulders had nontextured surfaces. In almost all cases, the comparison sites were highway sections on the same Interstate route, either adjacent to or near the treatment section. Tables A-6 and A-7 provide a summary of the accident data at the

A-17 treatment sites and control sites. About two-thirds of the sites had accident data for two years and one-third of the sites had accident data for one year. Table A-6. SVROR Accidents (right and left) Before and After Textured Shoulder Treatments in 11 States (treatment sections) (16) Test Section Accident rate (per 106 veh-mi) Number of accidents ADT Site No. Texture type Number of years Test length (mi) Before After Before After Before After Signif. change 1 BIS 1 14.7 9 9 2,715 2,915 0.309 0.288 N 2a BSG 1 12.5 10 7 2,315 2,640 0.473 0.291 N 3a,b BSG 2 18.5 38 36 4,425 7,750 0.318 0.172 Y 4a,b BSG 2 9.7 30 36 4,175 4,750 0.507 0.535 N 52 BST 2 15.3 59 46 3,750 4,075 0.706 0.507 N 6 BIS 1 17.3 26 14 10,200 10,600 0.202 0.105 Y 7 BIS 1 9.2 16 7 10,400 10,150 0.229 0.103 Y 8 BIS 1 20.7 20 16 7,750 7,400 0.171 0.143 N 9 CCP 1 13.0 NAd 1 10,473 10,000 NA4 0.011 NAd 10 CCP 2 14.2 NAd 9 2,252 2,150 NA4 0.202 NAd 11 CCP 2 23.8 NAd 8 6,834 6,525 NA4 0.035 NAd 12 CCP 2 22.2 NAd 14 6,038 5,765 NA4 0.075 NAd 13b BST 2 9.7 1 3 2,620 3,100 0.025 0.069 N 14 CCP 2 6.7 7 6 2,820 3,050 0.254 0.201 N 15 BIS 2 16.6 46 43 3,140 3,585 0.604 0.495 N 16 BIS 2 26.2 56 63 2,345 2,415 0.624 0.682 N 17c BST 2 13.5 34 25 9,125 7,625 0.189 0.166 N 18 BST 2 7.9 24 9 8,375 9,525 0.248 0.082 Y 19 CCP 2 12.0 NAd 7 2,128 2,420 NA4 0.165 NAd 20 CCP 2 10.0 13 25 3,630 4,240 0.245 0.404 N 21 CCP 2 23.0 NAd 9 3,724 4,350 NA4 0.062 NAd a Texture on right shoulder only. b Test section had textured shoulders during both before and after periods. c Test section before was textured and after was smooth shoulder. “After” data is for textured condition. d “Before” accident data not available with new construction sites. N—Not significant at 0.05 level Y—Significant at 0.05 level BIS—Bituminous indented strip BSG—Bituminous single groove BST—Bituminous surface treatment CCP—Concrete corrugated panel

A-18 Table A-7. SVROR Accidents (right and left) Before and After Textured Shoulder Treatments in 11 States (comparison sections) (16) Test Section Accident rate (per 106 veh-mi) Number of accidents ADT Site No. Texture type Number of years Test length (mi) Before After Before After Acc. rate before Acc. rate after Signif. change 1 BIS 1 20.0 12 8 3,450 3,500 0.238 0.157 N 2a BSG 1 20.0 9 13 3,400 3,450 0.181 0.258 N 3a,b BSG 2 20.0 26 22 4,200 4,625 0.212 0.163 N 4a,b BSG 2 20.0 26 22 4,200 4,625 0.212 0.163 N 5b BST 2 20.0 26 22 4,200 4,625 0.212 0.163 N 6 BIS 1 37.9 46 49 7,350 7,150 0.226 0.248 N 7 BIS 1 37.9 39 61 7,350 7,150 0.192 0.308 Y 8 BIS 1 37.9 45 61 7,350 7,150 0.221 0.308 N 9 CCP 1 13.0 7 16 9,600 11,500 0.077 0.147 N 10 CCP 2 22.9 5 7 4,550 4,570 0.033 0.046 N 11 CCP 2 24.5 10 17 4,880 5,675 0.057 0.064 N 12 CCP 2 21.4 3 9 8,715 9,600 0.011 0.029 N 13b BST 2 7.0 1 2 2,815 2,945 0.035 0.067 N 14 CCP 2 7.0 1 2 2,815 2,945 0.035 0.067 N 15 BIS 2 10.0 18 15 3,000 3,390 0.411 0.303 N 16 BIS 2 11.5 13 9 2,215 2,290 0.350 0.235 N 17c BST 2 18.7 79 59 10,525 6,750 0.275 0.247 N 18 BST 2 18.7 60 75 8,500 9,850 0.259 0.279 N 19 CCP 2 12.0 13 7 2,550 2,670 0.291 0.150 N 20 CCP 2 10.0 18 22 3,600 4,230 0.342 0.356 N 21 CCP 2 14.0 10 4 4,315 4,985 0.113 0.039 N a Texture on right shoulder only. b Test section had textured shoulders during both before and after periods. c Test section before was textured and after was smooth shoulder. “After” data are for textured condition. N—Not significant at 0.05 level Y—Significant at 0.05 level BIS—Bituminous indented strip BSG—Bituminous single groove BST—Bituminous surface treatment CCP—Concrete corrugated panel The following conclusions were drawn from a detailed analysis of the accident data: • There is insufficient evidence to indicate any significant differences in accident reduction when comparing: − Types of textured shoulder treatments − High ADT vs. low ADT sites − Day vs. night reduction in accidents − Normal vs. other driver conditions − Wide vs. narrow shoulder textured treatments − Spaced vs. continuous shoulder textured treatments

A-19 • Least squares regression analysis of accident rates before texturing compared to accident rates after texturing at the test sites showed a significant 9 percent reduction in SVROR accidents. • Chi-squared tests of before-after accident data at treatment sites and control sites indicated in a significant reduction in accidents at sites with textured treatments. Treatment sites exhibited a 19.8 percent decrease in SVROR accidents compared to a 9.3 percent increase at the comparison sites. A.2. Safety Effects of Centerline Rumble Strips This section presents results from studies on the safety effectiveness of centerline rumble strip applications. California Study A state route in California had a high number of fatal crashes that generated concerns from the local community and elected officials (29). In 1995 the location experienced six fatal crashes resulting in 14 deaths while historic data showed 2.7 fatal crashes per year for the previous nine years. Motivated by a high number of fatal crashes, the California DOT (Caltrans) conducted a demonstration project on a 23-mi (37-km) two-lane road with three passing lane sections in California. The project attempted to improve the visibility of the centerline markings and thereby reduce fatal head on crashes. The demonstration program consisted of installing the following treatments on the test section: • Profiled thermoplastic centerline markings • Milled rumble strips on the centerline to replace the double yellow strips • Raised yellow retroreflective pavement markers along the rumble strips spaced 28 in (711 mm) apart • Shoulder rumble strips Crash data for 34 months before the change and 25 months after the change were analyzed. All crash types were included in the analysis. The results showed there were 4.5 crashes per month in the before period; the crash frequency was decreased to 1.9 crashes per month after applying the aforementioned treatments. This included a 90 percent reduction in fatal crashes (10 fatal crashes before versus 1 fatal crash after). The study concluded that the centerline treatment used in the project was effective in reducing head on fatal crashes.

A-20 Colorado Study A 17-mi (27-km) section of a winding, two-lane mountain highway in Colorado with centerline rumble strips was evaluated for potential safety effects by comparing 44 months of crash histories both before and after the installation of centerline rumble strips (30). The resulting crash data and associated percent change are shown in Table A- 8. The overall study found benefits of installing centerline rumble strips from reductions in head on and side-swipe collisions in spite of increases in traffic volume after the installation. Outcalt also noted that centerline rumble strips might increase the danger to motorcyclists and bicyclists. Table A-8. Before-After Crash Analysis Summary for Installation of Centerline Rumble Strips in Colorado (30) Average number of crashes per year Crash type Before period After period Percent change Head-on 18 14 Head-on collisions/million vehicles 2.91 1.92 –34% Sideswipe opposite direction 24 18 Sideswipe collisions/million vehicles 3.88 2.46 –36.5% Average ADT 4,628 5,463 +18% Delaware Study A simple before-after study was performed to assess the effectiveness of centerline rumble strips on US Route 301 in Delaware (31). Average yearly accident data for three years prior to installation of the rumble strips were compared to eight years of data after installation. The results of the crash study are shown in Table A-9 and indicate that centerline rumble strips are a very effective method of reducing head on crashes. A benefit-cost ratio of 110 to 1 was also reported. The Delaware study reported the following advantages of installing centerline rumble strips: • Reduces the number of head on collisions owing to driver in-attention, error, and fatigue • Installation costs are low • No noticeable degradation to pavement • Requires little or no maintenance • Installation is not a function of pavement age • Novelty effects and consequent decrease in performance are not expected The disadvantages of CRS were as follows: • Noise effects

A-21 • Could transfer head on collisions to locations without centerline rumble strips Table A-9. Before-After Crash Analysis Summary for Installation of Centerline Rumble Strips in Delaware (31) Average number of crashes per year Crash type Before period After period Percent change Head-on 2 0.1 –95% Drove left of center 2 0.8 –60% Property damage only 6.3 7.1 +13% Injury 4.7 4.9 +4% Fatal 2 0 N/A Total 13 12 –8% Average daily traffic 16,500 22,472 +4% yearly Massachusetts Study Noyce and Elango (32) conducted a before-after crash analysis with comparison sites to evaluate the effectiveness of centerline rumble strips in Massachusetts. The data used for the analysis included target crash types such as head on, opposite direction angle, opposite direction sideswipe, and run-off-the-road crashes with centerline encounters. One of the three two-lane rural highways treated with centerline rumble strips experienced a slight decrease in the targeted crash types, while the other two sites experienced an increase in the targeted crash frequency. None of the results, however, were statistically significant. As such, there was no statistical evidence that centerline rumble strips decrease the frequency of targeted crash types for three two-lane rural highways in Massachusetts. Minnesota Study Briese (33) conducted a cross-sectional analysis to determine the safety effectiveness of centerline rumble strips on two-lane rural highways. Data from two-lane rural highways in Minnesota were used for the analysis. The analysis included data from 109 mi (175 km) of treatment sites and 215 mi (346 km) of nontreatment sites. Based upon total crashes, Briese found: • 73 percent lower crash rate of fatal and severe injury crashes on sections with centerline rumble • 42 percent lower crash rate on sections with centerline rumble strips • 37 percent lower severity rate on sections with centerline rumble strips • 19 percent lower crash density on sections with centerline rumble strips

A-22 When analyzing target crashes, defined to be head on, opposite direction sideswipe, and SVROR-to-the-left crashes, Briese found: • 13 percent higher crash rate of fatal and severe injury crashes on sections with centerline rumble • 43 percent lower crash rate on sections with centerline rumble strips • 37 percent lower severity rate on sections with centerline rumble strips • 20 percent lower crash density on sections with centerline rumble strips The analyses of both total crashes and target crashes produced very similar results except for the analyses based on fatal and severe injury crashes. Briese indicated that he had very little confidence in the analysis of target crashes based upon fatal and severe injury crashes because the number of crashes was so low. Briese stated that the analysis of target crashes based upon fatal and severe injury crashes should not be considered evidence that centerline rumble strips are not effective. Missouri Study Unpublished results of a study conducted by the traffic safety section of the Missouri Department of Transportation (34) found a general reduction in crossover crashes after the installation of centerline rumble strips along State Route MO 21. The analysis included 2 years of before data and 2 years of after data. Table A-10 displays crash data during the before and after periods by severity level, collision type and light condition. No statistical analysis was performed on the data. Table A-10. Before-After Crash Analysis Summary for Centerline Rumble Strips Installation Along MO 21 (34) Before: 2001 to 2003 After: 2003 to 2005 Percent change Severity Level Fatal crashes 1 0 –100 Disabling injury crashes 5 1 –80 Minor injury crashes 2 0 –100 Property damage only 2 3 +50 Total 10 4 –60 Collision Type Head-on 2 0 –100 Out of control 7 2 –71 Sideswipe 1 2 +100 Light Condition Daylight 6 3 –50 Dark 4 1 –75

A-23 Nebraska Study Unpublished results of a Nebraska Department of Roads (35) study showed a reduction in cross-center-line crashes of 64 percent along two-lane rural roads. The positive effect was found using before and after data for three years around the installation date of 28 mi (45 km) of centerline rumble strips. A statistically significant decrease in the fatal and injury crash rate of 44 percent was accompanied by a 90 percent decrease in property damage only crashes. Oregon Study The Oregon Department of Transportation installed and evaluated the safety effects of noncontinuous centerline rumble strips on crossover crashes at two rural highway locations (36). One of the sites was an 8.7-mi (14.0-km) long, four-lane highway with a posted speed limit of 55 mph (88 km/h). The average daily traffic (ADT) ranged from 12,000 to 14,000 vehicles per day along the segment with centerline rumble strips installed on a 4-ft (1.2-m) median channelizing device. The other section was an 8.4 mi (13.5 km) rural two-lane road with periodic, alternating passing zones. The posted speed limit was 55 mph (88 km/h), and the ADT was 18,000 veh/day with centerline rumble strips installed intermittently within the evaluation section. The two sections were divided into segments resulting in six sites for the safety analysis. Crossover crashes were the measure evaluated at each study site. Simple before-after and Yoked comparison analyses were used to perform the safety assessment. The simple before-after method showed a 13 to 100 percent reduction in five of the six sites within the two study sections. Overall, there was a 69.5 percent reduction in crossover crashes after the installation of centerline rumble strips. The summary of simple before-after analysis results are presented in Table A-11. The matched-pair method, which involves one-to-one matching of treatment sites to comparison sites, showed a 79.6 percent reduction in fatal crashes as a result of crossover collisions at the 95 percent confidence level. Table A-11. Simple Before-After Analysis Summary for Installation of Centerline Rumble Strips in Oregon (36) Before After Statistical analysis Site Years Target crashes Years Target crashes Adjusted crashes % reduction Z -statistic 1 6.1 33 2.7 4 9.2 –72.0 –3.42 2 6.1 2 2.7 1 2.3 15.4 0.15 3 7.4 5 1.4 0 0.0 –100.0 – 4 7.4 6 1.4 1 5.2 –13.0 –0.23 5 7.4 8 1.4 0 0.0 –100.0 – 6 7.4 1 1.4 0 0.0 –100.0 – Total 55 N/A 6 16.8 –69.5 –4.26

A-24 Texas Study The effect of centerline rumble strips on erratic driving behaviors was evaluated by Miles et al. (105) at the request of TxDOT. Using a before-after video analysis of sites where centerline rumble strips were installed, Miles et al. documented events in which drivers’ reactions to consciously crossing the centerline were effected in such a way as to cause an unsafe situation (e.g., during passing maneuvers). The erratic behaviors in question included (but were not limited to) hard braking, swerving, rapid alignment or lane shifting, correcting the trajectory in the wrong direction or losing control of the vehicle. After reviewing 479 passing maneuvers during 50 hours of video (18 before, 32 after) at sites along a rural highway, Miles et al. were unable to find any events that were considered erratic. It was concluded that centerline rumble strips do not create an unsafe environment, nor did they contribute to any irregular behaviors by drivers. Multistate Study Persaud et al. (4) conducted a before-after study of the safety effectiveness of centerline rumble strips on two-lane rural highways using data from seven states, including California, Colorado, Delaware, Maryland, Minnesota, Oregon, and Washington. The study included 98 treatment sites totaling 210 mi (338 km) in length. Table A-12 provides a summary of the treatment data for each state. The EB methodology was used in the analysis. After the installation of centerline rumble strips, total frontal/opposite direction sideswipe crashes were reduced by 21 percent, while injury crashes caused by frontal/opposite direction sideswipe collisions were reduced by 25 percent. For all crash types, the frequency was reduced by 14 percent and injury crash frequency was reduced by 15 percent after treating road sections with centerline rumble strips. These respective percent reductions were statistically significant at the 95 percent confidence interval. Persaud et al. also investigated the difference in crashes between daytime and nighttime hours. Although the percent reduction was somewhat greater at night than during the day (19 versus 9 percent), this difference was not significant at the 95 percent confidence level. Table A-12. Summary of Before and After Data for Installation of Centerline Rumble Strips in Seven States (4) Before period After period Crash count Crash count State Miles Sites Mile- years Avg AADT Total Injury Mile- years Avg AADT Total Injury California 47.8 29 206.5 9,235 679 257 112.5 10,430 351 144 Colorado 16.9 10 118.4 5,000 551 262 84.6 6,154 415 187 Delaware 2.9 1 8.4 16,500 34 16 21.3 21,685 82 38 Maryland 30.4 11 91.4 11,680 156 55 42.5 12,991 55 14 Minnesota 66.2 24 508.6 9,305 751 156 158.6 10,315 275 41 Oregon 3.1 2 22.8 11,400 31 20 4.6 11,150 6 3 Washington 43.5 21 166.5 7,290 308 116 173.3 7,963 297 109 Total 210.8 98 1,122.6 8,829 2,510 882 597.3 9,668 1,481 536

A-25 Study in an Unspecified Location Fitzpatrick et al. (29) also described two other studies that used centerline rumble strips as safety treatments in unspecified locations. One treatment consisted of 15 mi (24 km) of centerline rumble strips installed along a principal arterial that connected an Interstate highway to a rural community. Selection of this treatment site was biased as the centerline rumble strips were installed at a location with a crash frequency and severity rate above the statewide average for similar roadway types. The other segment was 10 mi (16 km) and connected a major northwestern city to a nearby suburban city and served as a high volume commuter route. The commonality among these two sites was that they were both considered high crash locations. Many improvements were made to these sites including installation of centerline rumble strips, guardrail, raised pavement markers on horizontal curves, traffic signalization and channelization, inclusion of right-turn or bypass lanes, and exclusive left-turn lanes. After the improvements were made to each location, a before-after crash analysis was conducted. The analysis indicated a reduction in various crash types. However, the study did not attribute the reductions to centerline rumble strips. This is because one of the sites did not show a statistically significant reduction, while the other site showed only a statistically significant reduction in read-end crashes which were not considered to have a direct relationship to centerline rumble strips. Because several treatments were applied at the same instance at both sites, drawing conclusions about the effects of centerline rumble strips was not possible. A.3. Operational Effects of Centerline Rumble Strips While the safety effectiveness of centerline rumble strips is fairly well documented, the effect on traffic operational characteristics has not been extensively researched. This section describes the findings from several recent research studies relating traffic operation performance changes attributed to centerline rumble strip installation. Minnesota Study Briese (33) conducted a before-after study to determine the effect that centerline rumble strips have on (1) travel speed on horizontal curve and tangent sections, (2) lateral placement of vehicles on tangent sections, and (3) centerline encroachment on horizontal curves. Data from two-lane rural highways in Minnesota were used for the analysis. The results indicate that centerline rumble strips have no impact on travel speed. Similarly, it was determined that the presence of centerline rumble strips do not impact the lateral placement of vehicles. Finally, it was determined that centerline rumble strips have a large effect on centerline encroachments and crossings along horizontal curves. Reductions in encroachments ranged from 40 to 76 percent.

A-26 Pennsylvania Study Porter et al. (37) conducted a before-after study with comparison sites to determine the effect that rumble strips have on lateral vehicle placement and vehicle speed. Data from two-lane rural highways in Pennsylvania were used for the analysis. The results indicate that there is a statistically significant difference in the mean lateral placement when comparing the before and after periods at centerline rumble strip treatment sites. The mean lateral vehicle shift ranged from 3.0 to 5.5 in (76 to 140 mm) away from the roadway centerline for 11- and 12-ft (3.3- and 3.6 m) lanes, respectively. There was a very small change in the mean lateral placement at comparison sites, but it was not statistically significant. Therefore, it is reasonable to conclude that the change in vehicle location was due to the presence of the centerline rumble strips. In addition, no statistically significant relationship between vehicle speed and the placement of centerline rumble strips was found. Texas Study Miles et al. (38) investigated how centerline rumble strips affect driver behavior in Texas. The specific measures of effectiveness that were used to quantify changes in driver behavior were: 1. Number and type of erratic driving movements during the initial stage of a passing maneuver. 2. Gap distance between the front end of a passing vehicle and the rear end of a vehicle being passed, prior to completing a passing maneuver. 3. Centerline crossing time during the initial stage of a passing maneuver. 4. Passing opportunity. 5. Percentage of traffic conducting passes along rural two-lane roads marked for passing. These measures were collected along a 15-mi (24-km) section of rural two-lane road on US 67 in Comanche County, Texas. Field data were videotaped from a data recording vehicle that was driven at 5-, 10-, and 15-mph (8-, 16-, and 24-km/h) increments below the posted daytime speed limit of 70 mph (113 km/h) to induce passing maneuvers. Analysis of the videotapes revealed the following: • Installation of centerline rumble strips did not induce erratic movements. • Installation of centerline rumble strips did not change driver behavior with respect to encroaching on the centerline prior to initiating a passing maneuver. • Passing drivers initiated their passes closer to a vehicle that they were passing after the installation of centerline rumble strips. This difference may be a

A-27 combination of the variation in the weather and the installation of the centerline rumble strips. • Driver behavior with respect to centerline crossing time while initiating a passing maneuver changed with drivers taking more time to cross the centerline after the installation of centerline rumble strips. This difference may be a combination of the differences in the weather, the part of the week that the data were collected, and the installation of the centerline rumble strips. • Passing opportunity was observed by measuring how long a passing vehicle was queued immediately behind a passed vehicle while in passing zones, no-passing zones, and when opposing vehicles were present. After installation of centerline rumble strips, drivers appear to be waiting longer before passing a vehicle traveling at 55 mph (88 km/h). • Centerline rumble strips do not appear to affect (i.e., decrease or increase) the number of passes made by drivers. Miles et al. also investigated how centerline rumble strips affect lateral positioning of vehicles within the travel lane. Field data were collected at eight locations using a camera trailer. Tape markers were placed on the pavement at 6-in (152-mm) intervals from centerline markings to determine lateral positioning of vehicles. The project sites, their geometric configurations, and the centerline rumble strip pattern are summarized in Table A-13. Miles et al. found that at all eight data collection locations, vehicular placement changed. The majority of drivers in the after period (i.e., after installation of centerline rumble strips) moved away from the centerline. Overall, Miles et al. concluded that: • No erratic movements were recorded either before or after the installation of centerline rumble strips. • None of the changes in the various measures of effectiveness were considered to be increases or decreases of a magnitude that merited a practical change in driving characteristics. • None of the changes in the various measures of effectiveness were considered to either affect the driving environment adversely or to induce unsafe driving practices.

A-28 Table A-13. Centerline Rumble Strip Lateral Position Project Site Characteristics in Texas (38) Roadway Alignment No. of lanes Shoulders Centerline rumble strip design FM 195 Curve 1 2 Yes Yellow, 4 ft (1.2 m) spacing, on each side of centerline markinga FM 195 Curve 2 2 Yes Yellow, 4 ft (1.2 m) spacing, on each side of centerline markinga FM 969 Tangent 2 No Black, 4 ft (1.2 m) spacing, staggered inside centerline marking FM 969 Curve 2 No Yellow, 4 ft (1.2 m) spacing, on each side of centerline marking FM 1431 Tangent 4 No Yellow, 4 ft (1.2 m) spacing, on each side of centerline marking FM 1431 Curve 4 No Yellow, 4 ft (1.2 m) spacing, on each side of centerline marking FM 2222 Tangent 4 No Yellow, 4 ft (1.2 m) spacing, on each side of centerline marking FM 2222 Curve 4 No Yellow, 4 ft (1.2 m) spacing, on each side of centerline marking a This site also included white pavement buttons spaced at 4 ft (1.2 m) spacing adjacent to the outside of the edgeline, and are therefore raised edgeline rumble strips. Finland Studies Räsänen (39) conducted a before-after observational study to evaluate the effectiveness of centerline rumble strips in a horizontal curve section on a rural two-lane undivided highway facility in Finland. The main objective of this research was to study the changes in lane keeping as a result of centerline rumble strips on horizontal curves. Field observations were made on a curve during four different conditions: 1. Worn-out painted centerlines 2. Freshly painted centerlines without centerline rumble strips 3. Resurfaced and freshly painted centerline with centerline rumble strips 4. One-year after installing centerline rumble strips The field observations consisted of measuring the number of encroachments, vehicle lateral positions, and vehicle speeds traveling in both directions of travel for various traffic situations. The traffic situations included free flow conditions, flow conditions less than five seconds headway, and flow conditions during the presence of on-coming traffic. The results of the field experiment indicated that the number of centerline encroachments along the curve decreased when the centerline was freshly painted than compared to the worn-out painted condition. However, there was no difference in the centerline encroachments after centerline rumble strips were installed when compared with the freshly painted condition. This suggested that the reason for reduction in centerline encroachments are due to enhanced visibility by the freshly applied pavement marking and not necessarily because of centerline rumble strips.

A-29 The means and standard deviations of lateral positions were compared among all the four conditions. The results indicated that there was a decrease in the lateral position (i.e., the drivers moved closer to the edgeline) after the installation of centerline rumble strips. This suggests that the drivers are providing more attention while negotiating the curve when centerline rumble strips are installed. Comparisons of speed on the curve did not indicate any apparent differences among the different conditions. Overall, Räsänen concluded that the effects of centerline rumble strips and freshly painted centerlines on encroachments, lateral position, and vehicle speeds were not major differences. However, it was postulated that centerline rumble strips could improve driver behavior consistency more effectively than painted lines on horizontal curves. This is because noise and vibration levels produced by the rumble strips are expected not to degrade as quickly as the visibility or retroreflectivity of painted lines. The study concluded that the presence of centerline rumble strips on curves has the potential to enhance safety by preventing unintentional and intentional centerline encroachments and by improving driver attention on horizontal curves during both free flow and oncoming traffic situations. Touvinen and Enberg (40) investigated the operation effects of centerline rumble strips on two lane highways in Finland. The operational areas that were being examined were: mean travel speeds, spot speeds, number of passes, amount of platooning, and lateral placement. The centerline rumble strips at the study site were rolled in asphalt; however, a recent review in Finland has recommended the future use of milled rumble strips. Using, what appears to be, a naïve before-after analysis, there was no statistically significant change one month and one year after the treatment installation in any of the key areas under investigation. After looking at the data and its applications from several angles, it was concluded that the installation of centerline rumble strips had no significant impact on driver behavior; at least for flow rates comparable to those observed during the study of 500 veh/hr. A driver survey accompanied the statistical analysis and revealed that 86 percent of drivers did not feel that centerline rumble strips affected their driving behavior, a result that was corroborated by the data. Japan Study Hirasawa et al. (42) investigated how centerline rumble strips influenced driving behavior in Japan relative to other treatments. Driving speeds and lateral positionings were compared along different sections of roads with median strips, center poles, chatter bars, rumble strips, and double-yellow centerlines (Figure A-1). The differences in average driving speed measured in each section were within 1.2 mph (2 km/h) so Hirasawa et al. concluded the safety measures including the rumble strips did not impact driving speeds. When measuring the lateral position of a vehicle in the lane, the location of the left front wheel relative to the outer edge of the shoulder was recorded using a video camera. Table A-14 shows the lateral positions of large vehicles along the various treatment sections. In the section with a double-yellow centerline, very few vehicles

A-30 traveled on the shoulder. In the section where center poles separated the lanes, many vehicles distanced themselves from the center poles, and some even drove on the shoulder. In the chatter bar and rumble strip sections, the vehicles distanced themselves from these treatments, but not as far as on the center pole section. Small vehicles showed the same behavior as large vehicles. Thus, Hirasawa et al. regarded the rumble strips as being effective in reducing head on collisions because they kept vehicles at a proper distance from the centerline. Figure A-1. Installation of Safety Treatments Against Head on Collisions on the Yakumo Section of National Route 5 in Japan (42) Table A-14. Lateral Position of Large Vehicles Relative to the Outer Edge of Shoulder at a Site in Japan (42) Treatment Average distance to outer edge of shoulder Median strip 26 in (671 mm) Center pole 7 in (173 mm) Chatter bar 14 in (349 mm) Rumble strip 14 in (360 mm) Yellow double centerline 23 in (576 mm) Simulation Studies Noyce and Elango (32) used a passenger car driving simulator to determine how drivers react when encountering centerline rumble strips. Although the analysis showed drivers take more time to return to their intended travel lane when encountering centerline rumble strips than compared to two-lane roadway centerline encroachment without rumble strips, the results were not statistically significant. It was also shown that drivers take longer time to return to their travel lane when centerline rumble strips are encountered as compared to centerline encroachments when no rumble strips are present. Again, the results were not statistically significant. Differences between the time to return to the intended travel lane were compared for roadway sections with and without

A-31 centerline rumble strips. The findings show a statistically significant difference in mean time to return to the travel lane on roadway sections that are curved. Thus, this result indicates that roadway geometry has an influence on the time needed to return to the travel lane when centerline rumble strips are present. Lastly, it was shown that 20 to 40 percent of drivers initially steer left when encountering centerline rumble strips on two-lane roads. This is the opposite maneuver expected and should be considered an inappropriate corrective maneuver. Harder et al. (41) also conducted a study to examine the effects of centerline treatments on driving performance using a driving simulator. Lateral position and speed were used as measures of driver performance. The following six centerline treatments were investigated: 1. The control condition: 12-ft (3.6-m) lanes and 4-in (102-mm) centerline dashes (current US standard) 2. 14-ft (4.3-m) lanes with 4-in (102-mm) centerline dashes 3. 14-ft (4.3-m) lanes with both longitudinal rumble strips and 4-in (102-mm) dashes marking the centerline 4. 12-ft (3.6-m) lanes separated by a 4-ft (1.2-m) central buffer area bounded by 4-in (102-mm) dashes 5. 12-ft (3.6-m) lanes separated by a 4-ft (1.2-m) central buffer area bounded by longitudinal rumble strips. In addition, there were 4-in (102-mm) centerline dashes 6. 12-ft (3.6-m) lanes separated by a 4-ft (1.2-m) central buffer area bounded by 8-in (203-mm) dashes Human participants drove these six test trials and in each trial, the participant faced several different driving situations, including: • Cruising with no traffic in the opposing lane • Cruising with traffic in the opposing lane • Following behavior, when the driver had to adjust to the speed of the car that it was following • Attempts to overtake a car in the same travel lane (i.e., passing maneuver) Harder et al. found that all the centerline treatments were effective when compared to the control condition based on driver performance. Participants drove significantly further away from the centerline for Conditions 2 and 3 when compared to Conditions 1, 4, 5, and 6. However, driver performance did not change when comparing Conditions 2 and 3. Other findings from the research include: • Use of 12-ft (3.6-m) lanes with a 4-ft (1.2-m) center buffer area resulted in a lateral vehicle position shift away from the centerline when compared to the use of 14-ft (4.3-m) lanes. • Subjects tended to shy away from the centerline during the presence of oncoming traffic when compared to the condition of cruising with no oncoming traffic.

A-32 • Widening of the markings from 4 to 8 in (102 to 203 mm) had similar driver performance effects. Overall, Harder et al. concluded that if any of the centerline treatments are to be implemented, it can be expected that drivers would position their vehicle further away from the centerline than they would with the control centerline marking configuration. Thus, it is expected that the alternative centerline treatments studied would reduce the likelihood of a crash with on-coming traffic. Particularly, Harder et al. supported the use of a 14-ft (4.3-m) lane with 4-in (102-mm) skip lines or a 14-ft (4.3-m) lane with 4-in (102-mm) skip lines with centerline rumble strips. A.4. Vehicle Dynamics Related to Vibration and Noise Stimuli The noise and vibration created by rumble strips is the key feature in their use. Unlike most other visual based traffic controls, rumble strips use noise and vibration to create a response from the driver. In an effort to determine optimum rumble strip dimensions, numerous studies have been conducted to determine the amount of vibration and noise generated by vehicles as they traverse different types and patterns of rumble strips. This section summarizes studies in which the vibration levels, as measured at a particular location or locations on a vehicle, were measured as the vehicle was traversing a known rumble strip pattern. Similarly, this section summarizes studies where the noise levels were measured inside the vehicle passenger compartment. Since the mid 1990s the Virginia Department of Transportation (VDOT) has been conducting research on continuous shoulder rumble strips (27). One of the primary issues that VDOT has researched is clarifying the optimal continuous shoulder rumble strip pattern based on measured and tactile levels as well as ease of construction and quality. Chen et al. (27) indicate that the overall performance quality of rumble strips can be determined using the following relationship: P = f (ad – ar; td – tr) where: P = effectiveness of rumble strip ad = mean audible index of travel way ar = mean audible index of rumble strips td = mean tactile index of travel way tr = mean tactile index of rumble strips To determine the optimal rumble strip pattern, it is necessary to find the difference between ad and ar and td and tr. Thus, the optimal rumble strip pattern is a function of the difference in the mean values of both audible and tactile indexes, not the absolute values of each.

A-33 Miles and Finley Study Finley and Miles (106,107) aimed their study at evaluating the factors related to noise generation that impact the effectiveness of rumble strips. Both passenger car and commercial truck variations were driven on milled, rolled and raised rumble strips at speeds of 55 and 70 mph. Among all of the factors investigated, including vehicle speed, vehicle type, surface type and rumble strip dimensions, the geometric dimensions of the rumble strips proved to be the most striking factor at alerting the driver by means of auditory signals. Figure A-2 shows the most effective rumble strip layouts are those that maximize the vehicle’s tire displacement into the rumble strip. Figure A-2. Effect of Rumble Strip Design (106,107)

A-34 Gardner et al. Study Researchers in Kansas conducted a study on the design of a new rumble strip pattern for centerline rumble strips (43). A new “football” shaped rumble strip pattern was installed along a Kansas highway, and several tests were conducted to evaluate the new football shaped rumble strip versus the typical rectangular rumble strip. Figure A-3 illustrates the difference between the two rumble strip patterns. The comparison consisted of water and debris collection, interior sound and vibration production, the opinions of bicyclists, and the opinions of residents in areas where the rumble strips were installed. Figure A-3. Rectangular Rumble Strip Compared to “Football” Shaped Rumble Strip (43) As part of the sound and vibration testing, noise and vibration produced from vehicle crossover of the rectangular rumble strips were compared to the noise and vibration produced from vehicle crossover of the football shaped rumble strips. Interior noise and steering wheel vibration were measured and compared. As part of the noise testing, six vehicles were included: • 1996 International 4900 DT466 Dump Truck • 1999 Chevrolet 2500 Diesel Pickup Truck • 2000 Ford Ranger XLT 2WD Pickup Truck • 2002 Dodge Caravan • 1996 Ford Taurus LX • 2005 Lexus RX 300 Sport Utility Vehicle Table A-15 shows the difference in noise levels compared to a base noise level in the travel lane for each type of rumble strip pattern and vehicle type. Gardner et al. note that a rumble strip must generate at least 9 to 10 dBA above the ambient noise level to alert a driver. Based upon this assumption, Gardner et al. concluded that each type of rumble strip pattern produced a recognizable amount of noise when crossover, and the football Rectangular Pattern Football Pattern

A-35 shaped rumble strips produced at least as much noise as the rectangular shaped rumble strips. Table A-15. Differences in Noise Levels for Each Type of Rumble Strips and the Base Level (dBA) on K-96 (43) Vehicle type Rectangular rumble strips vs. base Football rumble strips vs. base Football rumble strips vs. rectangular rumble strips 1996 International 4900 DT466 Dump Truck 23.1 31.4 8.3 1999 Chevrolet 2500 Diesel Pickup Truck 7.7 7.7 0.0 2000 Ford Ranger XLT 2WD Pickup Truck 9.3 13.7 4.4 2002 Dodge Caravan 7.8 8.5 0.7 1996 Ford Taurus LX 12.3 16.2 3.9 2005 Lexus RX 300 Sport Utility Vehicle 16.2 15.9 –0.3 As part of vibration testing, the same six vehicle were used as those in the noise study. Table A-16 shows the average resultant acceleration, f(x,y,z), for the base, football shaped rumble strips, and rectangular shaped rumble strips. Based upon an analysis of the data, Gardner et al. concluded that both types of rumble strips produce a significant tactile response; however, there is no statistical difference between the mean values of vibration for five of the six tested vehicles. Table A-16. Average f(x,y,z) for Vibration Trails (g) on K-96 (43) Vehicle type Base Football Rumble Stripsa Rectangular Rumble Stripsa 1996 International 4900 DT466 Dump Truck 1.027 1.074 (4.5 %) 1.084 (5.5 %) 1999 Chevrolet 2500 Diesel Pickup Truck 1.020 1.036 (1.6 %) 1.049 (2.8 %) 2000 Ford Ranger XLT 2WD Pickup Truck 1.003 1.089 (8.6 %) 1.027 (2.4 %) 2002 Dodge Caravan 1.015 1.053 (3.7 % 1.081 (6.5 %) 1996 Ford Taurus LX 1.001 1.043 (4.2 %) 1.021 (2.0 %) 2005 Lexus RX 300 Sport Utility Vehicle 1.012 1.115 (10.2% 1.128 (11.5 %) a The percent difference from the base condition is given in parentheses. Hirasawa et al. Study In Japan, Hirasawa et al. (42) collected vibration and sound levels generated in a passenger car while traversing three rumble strip patterns (Table A-17). The rumble strip patterns were installed at the Tomakomai Winter Test Track. The test vehicle was a station wagon. The test vehicle was driven at speeds of 25, 37, 50, and 62 mph (40, 60, 80, and 100 km/h) over each of the rumble strip patterns. Figure A-4 shows the setup for measuring the sound and vibration levels. Figures A-5 and A-6 show the average sound and vibration levels. The sound generated by all three rumble strip patterns was at least

A-36 15 dBA greater than the ambient sound generated by the pavement surface without rumble strips. According to Figure A-5, sound levels increase with groove depth. The vibration generated by each rumble strip pattern exceeded that on smooth pavement by 7 dB. The smallest vibration was measured at 37 mph (60 km/h). This was attributed to the vehicle characteristics, such as suspension performance. At each of the other speeds, the measured vibration for each groove depth exceeded that generated on the smooth pavement by 10 dB. Table A-17. Dimensions of Rumble Strips Installed at Tomakomai Winter Test Track in Japan (42) Dimension Pattern 1 Pattern 2 Pattern 3 Length 14 in (350 mm) 14 in (350 mm) 14 in (350 mm) Width 5 in (127 mm) 6 in (147 mm) 6.5 in (163 mm) Depth 0.32 in (9 mm) 0.5 in (12 mm) 0.625 in (15 mm) Spacing 12 in (302 mm) 12 in (302 mm) 12 in (302 mm) Figure A-4. Setup for Measuring Sound (left) and Vibration (right) (42) Hirasawa et al. also collected sound and vibration data from a vehicle traversing rumble strips on snow-covered roads. Figure A-7 shows the road surface condition on National Route 274 at the time of the testing, and the sound and vibration data measured inside the test vehicle. The road surface was slushy, and the centerline was not visible. Without rumble strips, the sound was 60 to 65 dB; with rumble strips the sound was 75 to 80 dB. The ambient vibration was 90 to 95 dB when not running on the rumble strips compared to 95 to 105 dB when running on the rumble strips. Hirasawa et al. concluded that the rumble strips gave sufficient warning (sound and vibration) of lane deviation on slushy winter roads, even when the centerline was not visible.

A-37 Figure A-5. Sound Versus Driving Speed (42) Figure A-6. Vibration Versus Driving Speed (42)

A-38 Figure A-7. Sound and Vibration Measured on National Route 274 During Winter Conditions (42) Bucko and Khorashadi Study Bucko and Khorashadi (14) tested a variety of rumble strip and edge stripe treatments to determine which applications are the most appropriate for bicyclists and still provide sufficient audible and vibratory sensation to alert automobile drivers. One of the primary objectives of this research was to collect and evaluate sound and vibration data from various vehicles being driven over different rumble strip patterns. The testing was conducted at the Caltrans Dynamic Test Facility in West Sacramento, California. Table A-18 summarizes the rumble strips patterns installed and tested at the Caltrans test facility. Six motor vehicles were used to collect sound and vibration data: • Light Passenger Vehicles − Chevrolet Lumina (1992) − Dodge Spirit (1993) − Dodge Ram 150 Pick-up Truck (1997) • Commercial Style Trucks − International 10-Wheel Tractor (without trailer) (1999) − Autocar 10-Yard Dump Truck (1991) − GMC Topkick Single Unit Van (1996)

A-39 Table A-18. Rumble Strip Patterns Tested by Caltrans (14) Pattern number Rumble strip type Groove length in (mm) Groove width in (mm) Groove spacing in (mm) Groove depth in (mm) Comments 1 Rolled 24 in (600 mm) 2 in (50 mm) 8 in (200 mm) 1 in (25 mm) 2 Milled 16 in (406 mm) 5 in (123 mm) 12 in (305 mm) 0.25 in (6 mm) 3 Milled 16 in (406 mm) 6 in (151 mm) 12 in (305 mm) 0.32 in (9 mm) 4 Milled 16 in (406 mm) 7 in (174 mm) 12 in (305 mm) 0.5 in (13 mm) 5 Milled 16 in (406 mm) 7.5 in (194 mm) 12 in (305 mm) 0.625 in (16 mm) 6 Chip Seal N/A N/A N/A N/A Installed using a tar epoxy and chip seal grade aggregate 7 Raised pavement marker N/A N/A 12 in (305 mm) N/A Installed using Caltrans standard Botts Dot pavement markers on 12 in (305 mm) centers 8 Raised pavement marker N/A N/A One run of raised pavement markers was installed using Caltrans standard Botts Dot pavement markers on 12-in (305-mm) centers. A second run was placed 6 in (152 mm) to the right of section one and skewed 6 in (152 mm) for two skewed runs of pavement markers. 9 Carsonite bars 24 in (600 mm) N/A 24 in (600 mm) N/A Carsonite bars placed 2 ft (0.6 m) on center and 2 ft (0.6 m) in width. 10 Raised and inverted thermoplastic stripe N/A N/A N/A N/A 11 Raised thermoplastic stripe N/A N/A N/A N/A The light passenger test vehicles were driven over the rumble strip patterns 1 through 5 at speeds of 50 and 62 mph (80 and 100 km/h), while the commercial style test vehicles were tested at 50 mph (80 km/h) only. The instrumented tests were conducted by driving each test vehicles right side tires onto and following a straight path over the series of rumble strip patterns. The steering wheel was instrumented with accelerometers to measure the vibration generated by the rumble strips (Figure A-8). The four accelerometers were positioned as such because in addition to providing direct values from each accelerometer, additional motion parameters could be calculated. Sound levels were also collected at ear level close to the center of the vehicle front passenger seat.

A-40 Figure A-8. Measuring the Acceleration of the Steering Wheel (14) Results of the vibration and noise tests are presented in Tables A-19 and A-20. Table A-19 shows the average vibration measurements for the light passenger vehicles and commercial vehicles. The average vibration values are the “resultant” vibrations (i.e., above the background level) calculated from the 4 accelerometers mounted on the steering wheel. Table A-20 shows the average noise measurements for the light passenger vehicles and commercial vehicles. The average noise values are the “resultant” levels above the background noise. The vibration and sound testing yielded the following results: • Vibration on light vehicles: − The vibration for rumble strip 1 (rolled) was greater than milled rumble strip 2 and less than rumble strips 3, 4, and 5 (all milled). − Vibration for rumble strips 3, 4, and 5 appeared to be linear in ascending order. Rumble strip 2 produced substantially less vibration than the other milled strips and consequently was not linear when compared to them. • Noise in light vehicles: − In relationship to the instrumented vibration tests, the noise tests followed the same trend. − The noise created by rumble strip 1 (rolled) was greater than milled rumble strip 2 and less than rumble strips 3, 4, and 5 (all milled). − Noise levels for rumble strips 3, 4, and 5 appeared to be linear in ascending order. Rumble strip 2 produced substantially less noise than the other milled rumble strips and consequently was not linear when compared to them. • Vibration on commercial vehicles: − When compared to the averages of the 50 mph (80 km/h) instrumented tests of light vehicles, the vibration averages of the commercial vehicles were less, but followed the same general trends. − For two of the trucks, vibration levels for rumble strip 1 (rolled) were greater than for rumble strips 2, 3, 4, and 5 (all milled). Significantly less

A-41 vibration was produced in the dump truck on rumble strip 1 than on the other strips. − Vibration for rumble strips 2, 3, 4, and 5 appeared to be linear in ascending order. • Noise in commercial vehicles: − When compared to the averages of the 50 mph (80 km/h) instrumented tests of light vehicles, the noise averages of the commercial vehicles were less, but followed the same general trends. − The average noise created by rumble strip 1 (rolled) was greater than for rumble strips 2 and 3 and less than rumble strips 4 and 5. − Noise averages for rumble strips 2, 3, 4, and 5 appeared to be linear in ascending order. Table A-19. Average Vibration Measurements of Light Passenger Vehicles and Commercial Vehicles (14) Rumble strip configuration 1 2 3 4 5 Resultant accelerations (g) of light passenger vehicles 50 mph (80 km/h) 0.253 0.115 0.379 0.432 0.542 62 mph (100 km/h) 0.306 0.135 0.437 0.469 0.591 Average 0.280 0.125 0.408 0.450 0.567 Resultant accelerations (g) of commercial vehicles Average 0.342 0.150 0.226 0.246 0.286 Table A-20. Average Noise Measurements of Light Passenger Vehicles and Commercial Vehicles (14) Rumble strip configuration 1 2 3 4 5 Resultant noise levels (dBA) of light passenger vehicles 50 mph (80 km/h) 14.4 11.8 18.2 19.7 21.4 62 mph (100 km/h) 12.6 10.2 15.2 16.9 18.5 Average 13.5 11.0 16.7 18.3 19.9 Resultant noise levels (dBA) of commercial vehicles Average 4.72 1.88 3.62 4.61 4.62 Outcalt Study A Colorado DOT by Outcalt (44) compared various rumble strip configurations to find a rumble strip pattern that is less disruptive to bicyclists than the standard rumble strip but still provides a safety factor to help prevent accidents caused by motorists running off the road. Twelve milled rumble strip configurations were tested along with a rolled concrete pattern. Table A-21 lists the dimensions of the rumble strip patterns included in the evaluation.

A-42 Table A-21. Rumble Strip Dimensions Tested in Colorado (44) Pattern number Rumble strip type Groove width (in) Flat width (in) Rumble strip/gap (ft) Average depth (in) Target depth (in) Max. measured (in) Min. measured (in) 1 Milled 2 10 Continuous 0.44 0.5 0.58 0.36 1A Milled 2 10 12/6 0.44 0.5 0.58 0.36 2 Milled 2 5 Continuous 0.44 0.5 0.46 0.43 2A Milled 2 5 12/6 0.44 0.5 0.46 0.43 3 Milled 2 5 12/6 0.29 0.375 0.38 0.20 4 Milled 2 3 Continuous 0.39 0.5 0.48 0.33 4A Milled 2 3 12/6 0.39 0.5 0.48 0.33 5 Milled 7.5 4.5 48/6 0.58 0.75 0.71 0.50 6 Milled 6.5 5.5 Continuous 0.49 0.5 0.59 0.35 7 Milled 6.0 6.0 Continuous 0.46 0.375 0.53 0.42 8 Milled 5.5 6.5 Continuous 0.41 0.25 0.47 0.37 9 Milled 5.0 7.0 Continuous 0.28 0.125 0.40 0.22 10 Rolled 2.375 1.625 Continuous 0.75 0.5–1.0 – – The investigation included both bicycle and motor vehicle testing. As part of the bicycle testing, bicycle vibration levels were collected. Table A-22 shows the frequency and amplitude of the vibrations measured on the test bike. The vibration maximum level is expressed in decibels (dB) (re: 1 m/s2). Table A-22 shows the frequency at which the highest level of vibration occurred. In many cases there where peak values at more than one frequency. For those cases, the highest peak is listed. It was noted that the frequency of the maximum vibration level did not necessarily increase with an increase in speed. Four vehicles were used in the motor vehicle testing. Sound levels generated by the rumble strip patterns were compared to the sound levels generated inside the vehicles on smooth pavement. The four test vehicles included: • 1994 Oldsmobile Cutlass station wagon • 1999 Dodge full sized pickup truck • 2000 GMC minivan • Unloaded tandem axle dump truck

A-43 Table A-22. Measured Vibration Levels on a Test Bicycle (44) Bicycle speed 5 mph 10 mph 15 mph 20 mph Pattern number Max (dB) Freq. (Hz) Max (dB) Freq. (Hz) Max (dB) Freq. (Hz) Max (dB) Freq. (Hz) 1, 1A 8 31.5 21 25 21 20 23 25 2, 2A 11 12.5 18 20 27 31.5 26 40 3,4,4A 10 12.5 25 31.5 34 40 21 63 5 12 20 28 12.5 35 20 N/A N/A 6 13 25 25 12.5, 25 33 20 35 25 7 11 31.5 26 25 32 20 33 25 8 10 25 24 25 31 16 33 25 9 6 31.5 21 25 26 20 31 25 10 8 31.5 18 40 15 63 12 20 The sound level in the passenger compartment was measured using an A-weighted scale (i.e., dBA). Motor vehicle tests were conducted at 55 and 65 mph (80 and 105 km/h). Figures A-9 and A-10 show the sound levels generated above the ambient noise for the respective rumble strip patterns and motor vehicles. The sound data showed that the sound levels generated by the various rumble strip configurations were different in the various test vehicles. There was considerable variation in which rumble strip was loudest in each vehicle. Also, the loudest at 55 mph (80 km/h) was not necessarily the loudest in the same vehicle at 65 mph (105 km/h). In general, rumble strips patterns 5 through 9 had the highest sound levels. Figure A-9. Increase in the Sound Level Inside the Vehicles at 55 mph (80 km/h) (44)

A-44 Figure A-10. Increase in the Sound Level Inside the Vehicles at 65 mph (105 km/h) (44) Motor vehicle vibration levels were also collected using the GMC minivan. Vibration data were collected at two locations. One accelerometer was mounted on the floor of the van just behind the driver’s seat at the spot where the floor was welded to the vehicle frame. The second accelerometer was mounted to the steering wheel. Vibrations were measured perpendicular to the plane of the steering wheel and perpendicular to the floor of the van. Vibration measurements were taken at 55 and 65 mph (80 and 105 km/h). Background measurements were taken in the travel lane at each speed to provide a comparison to the vibration measurements while traversing the rumble strips. Table A-23 presents the vibration levels in decibels and the frequency at which it occurred for each rumble strip pattern. Patterns 5 through 10 had the highest vibration levels. Table A-23. Measured Vibrations and Frequencies of Motor Vehicle Tests (44) Accelerometer Mounted to Floor Accelerometer Mounted to Steering Wheel 55 mph 65 mph 55 mph 65 mph Pattern number Max (dB)a Freq. (Hz) Max (dB) Freq. (Hz) Max (dB) Freq. (Hz) Max (dB) Freq. (Hz) 1 –6 80 –9 100 5 80 –5 40 1A –9 80 –11 200 5 80 –5 100 2 –8 125 –6 160 0 80 –6 40 & 160 2A –9 125 –8 160 –3 125 –4 160 3 –10 125 –9 160 –5 125 –6 160 4 –9 200 –1 250 –6 80 –b –b 4A –17 25 –4 250 –4 80 –b –b 5 6 80 3 100 11 80 7 100 6 8 80 3 100 8 80 2 100 7 8 80 3 100 9 80 3 100 8 5 80 2 100 5 80 & 160 5 100 9 –2 160 –1 100 2 80 & 160 7 100 10 3 630 8 630 1 63 1 63 a dB, re: 1 m/s2. b Data at or below background acceleration (as measured on smooth pavement alongside rumble strips).

A-45 Torbic Study Vehicle ride models provide guiding principles to control the vibration of vehicles for better passenger comfort. By applying the same principles to bicycles that have been developed for passenger cars and other transport vehicles, Torbic (54) evaluated the vibrational ride characteristics of bicycles. This served to identify how bicycles may be improved to provide better ride quality for bicyclists when they encounter rumble strips. Torbic utilized a system of equations to solve for the oscillation centers of several types of bicycles. The location of the oscillation centers has practical significance to ride behavior. Depending upon where the oscillation centers are located, it can be determined whether the bicycles exhibit good or poor vibrational characteristics. For good ride quality of a vehicle, it is desirable that the oscillation centers be located near the front and rear axles. As the oscillation centers move further from axles, this results in poorer ride quality. Using data from fifteen bicyclists, Torbic calculated the oscillation centers for three bicycles. In general the oscillation centers do not vary as a function of the characteristics of the bicycle or its rider. For each bicycle, the bounce oscillation center was located approximately 6.5 ft (2 m) to the left of the center of gravity, and the pitch oscillation center was located approximately 2.6 ft (0.08 m) to the right of the center of gravity. Based upon the locations of these oscillations centers, Torbic concluded each bicycle exhibited poor vibrational characteristics and that bicycles without active suspension exhibit poor vibrational characteristics. As part of the same research, Torbic examined the conditions that cause bicyclists to experience the highest levels of discomfort and control problems while traversing milled rumble strips. Torbic concluded that whole-body vibration of bicyclists increases with unit increases in groove depth and tire pressure, and whole-body vibration decreases with unit increases in groove spacing. In addition, a bicyclist experiences the highest levels of whole-body vibration while traversing rumble strip configurations at a speed of approximately 12 mph (20 km/h). Russell et al. Study In an attempt to find an appropriate pattern for centerline rumble strips, field tests of 12 rumble strip patterns were conducted in Kansas (46). The rumble strips patterns were as follows: Pattern 1: Continuous 12 in (305 mm) on center / 16 in (406 mm) in length Pattern 2: Continuous 24 in (610 mm) on center / 16 in (406 mm) in length Pattern 3: Alternating 12 and 24 in (305 and 610 mm) on center / 16 in (406 mm) in length Pattern 4: Continuous 12 in (305 mm) on center / 12 in (305 mm) in length

A-46 Pattern 5: Continuous 24 in (610 mm) on center / 12 in (305 mm) in length Pattern 6: Alternating 12 and 24 in (305 and 610 mm) on center / 12 in (305 mm) in length Pattern 7: Continuous 12 in (305 mm) on center / 8 in (203 mm) in length Pattern 8: Continuous 24 in (610 mm) on center / 8 in (203 mm) in length Pattern 9: Alternating 12 and 24 in (305 and 610 mm) on center / 8 in (203 mm) in length Pattern 10: Continuous 12 in (305 mm) on center / 5 in (127 mm) in length Pattern 11: Continuous 24 in (610 mm) on center / 5 in (127 mm) in length Pattern 12: Alternating 12 and 24 in (305 and 610 mm) on center / 5 in (127 mm) in length The cutting spindle of the milling machine used to install the rumble strip patterns had a 12-in (305-mm) milling radius and the depth of the cut was 0.5 in (13 mm) on all patterns. Seven vehicles were included in the field testing: 2 dump trucks, 1 pick-up truck, 1 full size car, 1 compact car, 1 minivan, and 1 sport utility vehicle. Interior noise level and steering wheel vibration levels associated with the centerline rumble strip patterns at vehicle speeds of 60 mph (97 km/h) were measured for each combination of pattern and vehicle type. Table A-24 shows the average decibel level inside the passenger compartment of the respective vehicles for each of the rumble strip patterns. The results of the noise level analysis indicated that the continuous 12-in (305-mm) patterns produced higher noise levels at 60 mph (97 km/h) followed by the alternating 12- and 24-in (305- and 610-mm) and continuous 24-in (610-mm) patterns. Thus, it was theorized that patterns with higher densities produce higher average decibel levels. As for trends in decibel levels owing to rumble strip length, it appeared that the longer rumble strips generally produced higher average decibel levels, but there was no consistency among the longer lengths. This could be the result of the vehicle tires not remaining in full contact with the shorter rumble strip patterns. This was reasoned because with the shorter patterns there was a greater likelihood that the vehicles left tires did not remain in full contact with the patterns. The results of the steering wheel vibrations indicated that the alternating 12- and 24-in (305- and 610-mm) patterns produced the highest vibration levels followed by continuous 12-in (305-mm) and continuous 24-in (610-mm) patterns. Based on these preliminary results, continuous 12-in (305-mm) and alternating 12- and 24-in (305- and 610-mm) patterns were selected for further research in an existing highway setting.

A-47 Table A-24. Decibel Level at Driver’s Position—60 mph (97 km/h) (46) Pattern Tested Vehicle 1 2 3 4 5 6 7 8 9 10 11 12 91.23 92.16 92.94 94.12 92.23 93.35 93.41 91.47 92.84 92.24 – – 1996 IH 4900 DT 466 Dump Truck (GW = 75,000) 0.316 0.685 0.373 0.429 0.494 0.346 0.546 0.482 0.490 0.852 – – 91.34 90.73 91.07 92.73 90.48 91.43 92.01 90.03 90.54 92.31 88.21 – 1995 Ford L8000 Dump Truck (GW = 48,000) 0.915 0.263 0.587 0.465 0.440 0.592 0.456 0.433 0.283 0.950 0.445 – 83.50 82.86 83.77 87.47 82.68 84.18 88.77 81.44 84.11 85.29 – – 1991 Chevrolet 2500 Pick-Up Truck 1.194 0.845 0.452 0.796 0.572 0.896 1.242 0.614 0.753 1.117 – – 82.89 80.01 83.48 84.24 79.61 84.65 83.59 79.46 83.75 83.32 79.01 82.86 1993 Pontiac Bonneville Full-Size Passenger Car 0.568 0.312 0.179 0.274 0.150 0.374 0.970 0.371 0.459 0.786 0.703 1.053 87.34 86.22 87.76 89.97 86.57 87.44 89.74 87.75 88.62 88.42 85.60 – 1994 Ford Escort Wagon Compact Passenger Car 0.711 0.351 0.508 0.430 0.083 0.238 0.483 0.465 0.083 0.990 0.390 – 88.33 85.89 85.59 87.77 84.97 86.12 89.49 82.83 84.09 87.83 80.62 82.56 1995 Ford Aerostar Minivan 1.146 0.904 0.612 0.600 0.530 0.668 0.692 0.851 0.604 0.437 1.083 1.255 85.63 81.24 83.80 88.65 80.48 84.22 86.76 79.87 82.82 – – – 1997 Jeep Cherokee Sport Utility Vehicle (SUV) 0.676 0.821 0.544 0.338 0.419 1.014 0.683 0.725 0.563 – – – Grand Mean 87.18 85.59 86.92 89.28 85.29 87.34 89.11 84.69 86.68 88.24 83.36 82.71 Notes: For each vehicle, the first row of numbers is the mean and the second row is the standard deviation. (–) Indicates the test results were inconclusive.

A-48 Bahar et al. Study Complaints about noise from rumble strip contact prompted the province of Alberta (CA) to commission a noise study on rumble strips to identify the optimum dimensions for rumble strips in terms of alerting drivers, as well as the noise impacts of rumble strips on the surrounding area (57). The testing involved continuous and intermittent milled rumble strips of varying depths, from 0.8 to 0.32 in (2 to 8 mm), and varying lengths, from 12 to 20 in (300 to 500 mm). The tests were conducted using three vehicles (i.e., tractor trailer, passenger vehicle, and motorcycle). In general, it was found that sound levels increase with rumble strip depth. Elefteriadou et al. Study Elefteriadou et al. (45) conducted a study for the Pennsylvania DOT (PennDOT) with the primary objective to develop new rumble strip configurations that decrease the level of vibration experienced by bicyclists when traversing the rumble strips. At the same time, an adequate amount of stimuli, both auditory and tactile, must be maintained to alert an inattentive/drowsy motorist. To achieve this objective, Elefteriadou et al. utilized a simulation model to evaluate how numerous rumble strip configurations impact the dynamics (i.e., the vertical acceleration and pitch angular acceleration) of a bicycle and its rider. By comparing the vertical and pitch angular acceleration of the bicycle/rider system traversing different simulated rumble strip configurations, the simulated configurations could be ranked as having the greatest or least potential to be “bicycle- tolerable.” Those configurations that had the greatest potential of being “bicycle- tolerable” and could also be constructed were selected for installation at the Pennsylvania Transportation Institute (PTI) test track for further evaluation (see Table A-25). Table A-25. Rumble Strip Configurations Installed at PTI’s Test Track for Further Evaluation (45) Test pattern Groove width in (mm) Flat portion between cuts in (mm) Depth in (mm) 1a 7 in (178 mm) 5 in (127 mm) 0.5 in (13 mm) 2 5 in (127 mm) 7 in (178 mm) 0.5 in (13 mm) 3 5 in (127 mm) 7 in (178 mm) 0.375 in (10 mm) 4 5 in (127 mm) 6 in (152 mm) 0.5 in (13 mm) 5 5 in (127 mm) 6 in (152 mm) 0.375 in (10 mm) 6 5 in(127 mm) 7 in (178 mm) 0.25 in (6.3 mm) a PennDOT's current standard. The test track experiments involved testing several bicycles and a motor vehicle. To measure the effects of the different configurations on bicyclists, volunteer participants rode different types of bicycles over the rumble strip configurations at different speeds and different angles. Table A-26 presents the rankings of the test configurations based upon the vertical acceleration levels measured on mountain, road, and hybrid bicycles across all speed ranges. Similarly, Table A-27 presents the rankings from the tandem

A-49 bicycle across all speed ranges. Table A-28 presents the rankings of the test configurations based upon the pitch angular acceleration levels measured on mountain, road, and hybrid bicycles across all speed ranges, and finally Table A-29 presents the rankings from the tandem bicycle across all speed ranges. Table A-26. Ranking of Test Configurations Based on Vertical Acceleration (mountain, road, and hybrid bicycles) (45) Rank Test pattern Average RMS—vertical acceleration ft/s2 (m/s2) Best 1 6 37.123 (11.315) 2 3 43.228 (13.176) 3 2 55.613 (16.951) 4 5 63.396 (19.323) 5 1 71.880 (21.909) Worst 6 4 71.988 (21.942) Table A-27. Ranking of Test Configurations Based on Vertical Acceleration (tandem bicycle) (45) Rank Test pattern Average RMS—vertical acceleration ft/s2 (m/s2) Best 1 6 29.327 (8.939) 2 3 34.003 (10.364) 3 2 40.850 (12.451) 4 1 55.131 (16.804) 5 5 58.940 (17.965) Worst 6 4 62.605 (19.082) Table A-28. Ranking of Test Configurations Based on Pitch Angular Acceleration (road, mountain, and hybrid bicycles) (45) Rank Test pattern Average RMS—Pitch angular acceleration rad/s2 Best 1 6 13.288 2 3 15.809 3 5 18.994 4 2 21.230 5 4 21.711 Worst 6 1 30.593

A-50 Table A-29. Ranking of Test Configurations Based on Pitch Angular Acceleration (tandem bicycle) (45) Rank Test pattern Average RMS—Pitch angular acceleration rad/s2 Best 1 6 10.912 2 3 13.171 3 5 15.407 4 2 16.404 5 4 17.829 Worst 6 1 22.006 Elefteriadou et al. also measured the lateral stability of bicycles by measuring the ability of bicyclists to ride along a designated path along the rumble strip. Lower percentages of time spent deviating from the designated path indicate better control while traversing the rumble strip configuration. Higher percentages of time spent deviating from the designated path indicate greater loss of control. Table A-30 presents the rankings of the test configurations based on this objective control measure. Table A-30. Ranking of Test Configurations Based on Objective Control (all bicycles) (45) Percentage of Time Off the Line Rank Test pattern Average Difference (Pattern—Base) Smooth 0.0814 Best 1 3 0.1228 0.0414 2 6 0.126 0.0446 3 4 0.1644 0.0830 4 5 0.1922 0.1108 5 2 0.1956 0.1142 Worst 6 1 0.2535 0.1721 To assess the effectiveness of the various rumble strip configurations on alerting inattentive/drowsy motorists, measurements were taken of the auditory and vibrational stimuli generated by the rumble strip configurations. A 1998 Plymouth Grand Voyager was used to collect the motor vehicle data. Three objective measures were collected during the motor vehicle testing at PTI’s test track: • vertical acceleration of the body frame • pitch angular acceleration of the body frame • maximum sound level in the passenger compartment Based upon preliminary results of the vibration data, only the maximum sound level in the passenger compartment of the minivan was used to assess the relative effect of the six rumble strip patterns on the dynamics of the minivan. Table A-31 presents the rankings

A-51 of the test configurations based upon the noise level testing for the motor vehicle. One set of rankings is provided based upon the low speed testing at 45 mph (72 km/h), and a second set of rankings is provided based upon the high speed testing at 55 mph (88 km/h). Table A-31. Ranking of Test Configurations Based on Noise Level Testing (45) Rank Test pattern Speed mph (km/h) Avg. max. sound level dBA Difference (pattern—smooth) Best 1 4 45 (72) 83.6 15.2 2 1 45 (72) 80.0 11.6 3 5 45 (72) 79.3 10.9 4 2 45 (72) 78.4 10.0 5 3 45 (72) 75.2 6.8 Worst 6 6 45 (72) 74.7 6.3 Smooth 45 (72) 68.4 Best 1 1 55 (88) 88.9 23.7 2 2 55 (88) 83.7 18.5 3 3 55 (88) 81.3 16.1 4 4 55 (88) 81.2 16.0 5 5 55 (88) 79.1 13.9 Worst 6 6 55 (88) 78.2 13.0 Smooth 55 (88) 65.2 Chen Study Chen (48) studied the vibrational and noise stimuli generated by rumble strips on motor vehicles. As part of the study, Chen conducted a theoretical analysis of the tire drop to evaluate the effectiveness of rumble strips. For rumble strips to be effective, Chen determined that the width of the strip should be large enough for the tire to drop into the groove so as to generate vibration and noise. The tire drop is dependent on the tire static and dynamic deflection, which is a function of the load and inflation pressure, the speed of the motor vehicle, and the width of the rumble strip. As part of the field testing, Chen conducted pavement roughness and sound level tests on three different types of rumble strips: continuous rolled rumble strips on asphalt shoulders, continuous milled rumble strips on asphalt shoulders, and intermittent corrugated rumble strips on concrete shoulders. Data were collected at 112 locations on Interstates 85 and 295 in Virginia under the following conditions: • Testing speed: 55 and 65 mph (88 and 105 km/h) • Angle of departure: 5° • Road conditions: dry and clean

A-52 From the pavement roughness data, Chen concluded that the milled rumble strips performed better than the rolled or corrugated strips. Roughness levels measured by the International Roughness Index (IRI) were 12.6 times greater for the milled rumble strips than the rolled rumble strips. The roughness levels for the milled rumble strips were also 7.2 times greater than the corrugated rumble strips. Chen also noted rolled rumble strips have very little effect (i.e., change in vibration) on trucks. As part of the sound level testing, Chen compared the difference in sound levels between driving in the travel lane and driving over the different rumble strips. The sound levels were measured while driving at 65 mph (105 km/h). The tests showed milled rumble strips generated the largest sound excesses. The sound excesses for each type of rumble strip were as follows: • 2.5 dBA difference between rolled rumble strips and travel lane • 7.0 dBA difference between corrugated rumble strips and travel lane • 10.87 dBA difference between milled rumble strips and travel lane In a follow-up to the research that was conducted in 1994, Chen et al. (27) compared sound data collected from Virginia highways with regression models developed by Khan and Bacchus (97) that predict sound levels generated by rumble strips. The regression models developed by Khan and Bacchus are as follows: Nonlinear: dBA = e3.412C0.074V0.172 Linear: dBA = 53.636 + 0.585C + 3.28E + 0.161V where: dBA = sound level in passenger car (in decibels) C = width of rumble strip (cm) E = depth of rumble strip (cm) V = speed of test vehicle (km/h) Chen et al. (2003) concluded the field data from Virginia highways fit the models developed by Khan and Bacchus very well. The comparison reached a correlation factor of 96 percent. Pennsylvania Turnpike Study The Pennsylvania Turnpike Commission conducted tests to evaluate the level of sound generated by various milled rumble strip patterns (Figure A-11) (24,104). Tests were conducted with both a passenger car and a truck. During the testing, the motor vehicles were driven over the different designs, and the sound level was recorded inside the motor vehicles to compare their effectiveness. Several speeds were tested: 40, 50, 60, and 65 mph (truck only) (64, 80, 97, and 105 km/h). Tables A-32 and A-33 present the results of the sound measurements. Tables A-32 and A-33 do not show the mean noise level inside the sedan or truck while being driven in the travel lane. The average level of noise inside a car being driven in the travel lane was 73 dBA at 55 mph (88 km/h), and the average level of noise in the truck was around 79 dBA. Considering first the tests with the sedan, pattern 5 gave the highest dBA readings for any of the speeds. In the case

A-53 of the truck tests, only pattern 5 gave a noise level higher than the background noise in the passenger compartment. Pattern 5 had the deepest and the widest groove. Figure A-11. SNAP Test Patterns (104) Table A-32. Noise Test Results for a Passenger Car Traversing Various Rumble Strip Patterns (104) Mean noise level (dBA) Sedan speed 40 mph (64 km/h) 50 mph (80 km/h) 60 mph (97 km/h) Pattern 1 74 77 80 Pattern 2 70 75 76 Pattern 3 68 74 74 Pattern 4 71 73 74 Pattern 5 75 78 80 Table A-33. Noise Test Results for a Truck Traversing Various Rumble Strip Patterns (104) Mean noise level (dBA) Truck speed 40 mph (64 km/h) 50 mph (80 km/h) 60 mph (97 km/h) 65 mph (105 km/h) Pattern 1 – – – – Pattern 2 – – – – Pattern 3 – – – – Pattern 4 – – – – Pattern 5 – 82 82 86 Note: Not recordable, 79 dBA in the truck cab.

A-54 Chaudoin and Nelson Study In their study of rumble strips on Interstates 15 and 40, Chaudoin and Nelson (17) investigated the influence of groove shape and spacing on noise. Noise measurements were gathered from three different shapes of rumble strips: v-shape, rectangular, and rounded. According to the evaluation, the v-shaped groove gave a good sound effect, and the rounded shape gave a very good sound effect. There were no results concerning the noise generated by the rectangular groove. Chaudoin and Nelson also studied the effects of four groove spacings: 4 in (102 mm), 8 in (203 mm), variable, and 16 in (406 mm). The 4-in (102-mm) spacing gave a high-pitched sound effect. The 8-in (203-mm) spacing provided a good sound effect with a lower pitch than the 4-in (102-mm) spacing. The variable spacing provided a sound that was more of a flat tire sound than a tone. The 16-in (406-mm) spacing did not provide adequate sound to be heard. Tye Study Tye (49) evaluated raised and milled rumble strips by instrumenting a test car and driving over various configurations of rumble strips to collect data on sound, vibration, and handling. The test vehicle was instrumented with a tri-axial accelerometer mounted on the front floor over the transmission to measure the vertical, transverse, and longitudinal vibration components. The controllability of the motor vehicle was reasoned to be related to the front-wheel bounce so the magnitude of the wheel bounce was measured by a rectilinear potentiometer mounted behind the right front wheel. Using the instrumented test vehicle, Tye collected sound, vibration, and handling data for numerous rumble strip designs. Raised rumble strip patterns were tested using plywood rib rumble strips. The plywood rumble strip patterns tested had a thickness between 0.25 and 0.75 in (6 and 19 mm) and had a rib width ranging from 3 to 8 in (76 to 203 mm), in 1-in (25-mm) increments. The spacing was varied from 3 to 6 in (76 to 152 mm), also in 1-in (25-mm) increments. A total of 57 raised rumble strip patterns were tested at speeds ranging from 30 to 70 mph (48 to 113 km/h). In summary, these tests revealed: • Sound level: Rumble strip ribs 0.25-in (6-mm) thick were marginally effective, in producing a sound level that averaged 7 dBA above the background level on bare pavement. The 0.5-in (13-mm) high rumble strips produced a sound level that averaged 9 dBA above the background level, while the 0.75-in (19-mm) high strips produced an 11 dBA increase. For each rumble strip pattern, the sound level generally increased to the highest levels at the fastest speeds, 60 and 70 mph (97 and 113 km/h); however, the background sound level from the bare pavement increased with speed as well. The difference between the sound levels was greatest in the 30- to 40-mph (48-

A-55 to 64-km/h) range than at higher speeds for more than 80 percent of the plywood rib rumble strip patterns tested. This difference was lowest at the 60- to 70-mph (97- to 113-km/h) range for over 95 percent of the patterns. Varying the width of the rumble strip ribs did not produce significantly higher or lower sound levels for any given spacing of ribs. Rumble strips spacings of 3, 4, and 5 ft (0.91, 1.22, and 1.52 m) produced significantly different sound levels. Sound levels decreased with the 6 ft (1.83 m) and greater rib spacing. • Right front wheel movement: The data showed a somewhat erratic pattern of right front wheel movement. There was a tendency for the movement to decrease with increasing speed and to increase with increasing rib spacing. Many of the test patterns caused little or no difference from the background level of wheel movement on normal pavement. • Vertical acceleration: The 0.75-in (19-mm) thick rumble strip ribs produced a higher than average increase in vertical acceleration over the background level. The 0.5-in (13-mm) thick pattern, with 7-in (178-mm) wide ribs were the next best. The 0.5-in (13-mm) thick patterns, by 3- or 5-in (76- or 127-mm) wide ribs, produced vertical accelerations that were 1/3 lower in amplitude than those 0.5-in (13-mm) high by 7-in (178-mm) wide. The 0.25-in (6-mm) thick rumble strip ribs produced an even lower average level of vertical acceleration. In general, thicker rumble strips generated greater vertical acceleration. • Lateral acceleration: Regardless of rib thickness or width, the greatest increase in amplitude of lateral acceleration was produced by ribs spaced on 5-ft (1.5-m) centers. The ribs spaced on 8-ft (2.4-m) centers produced the lowest average increase in amplitude over the background level. The instrumented vehicle was also driven over a series of milled rumble strips. The grooves varied from 0.5 to 0.75 in (6 to 19 mm) in depth. The width of the grooves was either 3, 5, or 7 in (76, 127, or 178 mm). All of the grooves were spaced 5 ft (1.5 m) apart, and the sides of the grooves were vertical. The milled rumble strips, in general, produced lower average differences in sound, wheel movement, and vertical and lateral acceleration from background levels than the raised plywood rumble strips. The 7-in (178-mm) wide grooves produced slightly greater average increases in sound and vibration over the background level. Tests with the instrumented vehicle were also made on a series of rumble strips composed of rows of ceramic pavement markers. The markers were 4 in (102 mm) in diameter and approximately 0.75 in (19 mm) in height. All of the pavement marker rumble strips except one used variable row spacing. The pavement marker rumble strips as a group produced sound and vibration levels that were less effective than the 0.25-in (6-mm) high plywood rumble strips.

A-56 Franke Study Franke (108) studied the optimum spacing of shoulder rumble strips on the Interstate relative to speed. The optimum spacing was determined from vibrational measurements of a car. Heights of 0.5 in (13 mm) and 0.375 in (10 mm) were evaluated, but time limitations for the study did not allow evaluation of various rumble strip widths. Various spacings were tested: 1.25, 2, 2.5, 3.25, 3.75, 5, 7.5, 10, 10.5, and 15 ft (0.38, 0.61, 0.76, 0.99, 1.14, 1.52, 2.29, 3.05, 3.20, and 4.57 m). The test car was driven at the following speeds: 20, 30, 40, 50, 60, and 70 mph (32, 48, 64, 80, 97, and 113 km/h). Conclusions from the study were as follows: • A spacing of 2 ft (0.61 m) or less created a large amount of wheel hop and/or did not allow the tires to descend between the rumble strips, which created a situation where the vibration level increased when the speed decreased. • A 5 ft (1.52 m) spacing seemed to be the best suited for use on the shoulder of roadways. • Rumble strips should not be of a height or depth greater than 0.5 in (13 mm). A.5. Effects of Rumble Strips on Specific Types of Highway Users This section describes the effects that rumble strips have on specific types of highway users (i.e., drivers of passenger cars, drivers of trucks, motorcyclists, bicyclists, and pedestrians), primarily from a human factors perspective. In most cases, the intended effect of shoulder and centerline rumble strips is to alert inattentive or drowsy drivers of motor vehicles that their vehicles have departed from the travel lane. However, shoulder and/or centerline rumble strips may also cause unintended behaviors or may negatively impact certain types of highway users such as motorcyclists and bicyclists. This section focuses primarily on those studies in which participants subjectively rated the impact of rumble strips. To the extent possible, this section also focuses on the correlation between the alerting properties of the rumble strips (i.e., vibration and sound levels) and the reactions or behaviors of highway users to these stimuli. Drivers of Passenger Cars Rumble strips are intended to warn motorists that their vehicles have partially or completely left the travel lane and that they must correct their steering to return back within their intended travel lane. Rumble strips provide this warning through audible and tactile stimuli. As motor vehicle tires pass over rumble strips, the dynamic interaction between the vehicle tires and the pavement surface cause noise to be generated within the passenger compartment and cause vibration of the vehicle floor, seat, and steering wheel. The noise provides an audible warning to the motorist, while the vibration provides a tactile warning. Rumble strips are intended to provide a sufficient amount of both audible

A-57 and tactile stimuli to effectively alert drivers, while minimizing any adverse effects such as startling drivers to a degree that drivers overreact or over steer when trying to return to their travel lane or by causing too much vibration which could negatively impact vehicle handling capabilities. Recently, Anund et al. (77) conducted a driving simulator experiment to investigate the effects of rumble strips on fatigued drivers. The driving simulator was an advanced moving based passenger car simulator (Figure A-12). Four patterns of rumble strips (Table A-34) were simulated and placed at two locations along the shoulder (i.e., near the edgeline and near the outside edge of the shoulder) and along the centerline. Data from 40 regular shift workers driving during morning hours after a full night shift were used in the analysis. Both driving behavior and physiological data were recorded. The primary measures of interested included: • Maximum lateral position • Time from lane departure to “steady state” • Velocity at “steady state” • Time since last lane departure • The number of steering wheel corrections per minute • Time to correct action Table A-34. Patterns of Simulated Rumble Strips at VTI (77) Pattern Length Width Depth Spacing Pennsylvania 20 in (500 mm) 7 in (170 mm) 0.5 in (12 mm) 12 in (300 mm) Swedish 20 in (500 mm) 12 in (300 mm) 0.75 in (20 mm) 21 in (530 mm) Malilla 14 in (350 mm) 6 in (150 mm) 0.375 in (10 mm) 48 in (1200 mm) Finnish 7 in (175 mm) 0.75 in (20 mm) 0.625 in (15 mm) 12 in (300 mm) Figure A-12. Moving Base Driving Simulator at VTI (74)

A-58 The primary findings of interest from this study are as follows: • For departures to the right (i.e., shoulder rumble strips), the maximum excess lane departure (i.e., how far beyond the rumble strip the vehicle traveled after encountering the rumble strip) was greater when the rumble strips were placed along the outside portion of the shoulder than when placed close to the edgeline. • For departures to the right (i.e., shoulder rumble strips), there were no significant differences in other measured driver behaviors (i.e., time from lane departure to “steady state,” velocity at “steady state,” time since last lane departure, number of steering wheel corrections per minute, and time to correct action). • For departures to the left (i.e., centerline rumble strips), the results showed that the time gaps between lane departures were shortest for the Finnish rumble strips and the longest for the Swedish rumble strips. There were no other significant differences in the other measured driving behaviors. • Through a subjective questionnaire, the drivers rated the effectiveness of different levels of sound and vibration which contributed most to the alerting properties of the rumble strips. The majority of the drives indicated that both sound and vibration contributed to their impression from the rumble strip. Figure A-13 illustrates in more detail the drivers’ opinions on the effectiveness of different levels of sound and vibration. Figure A-13. Drivers’ Opinions of the Effectiveness of Different Levels of Sound and Vibrations (77) Given that the behavior data did not show any great differences between the types of rumble strips, Anund et al. concluded that the more aggressive rumble strips (i.e., Swedish and Pennsylvania style patterns) should be used. Anund et al. reasoned this because the subjects preferred them and they did not see any danger of being scared and thereby causing accidents. Anund et al. also cited that this investigation only included passenger cars and reasoned that heavy vehicles should benefit even more from the more aggressive designs. Concerning the placement of the rumble strips, Anund et al.

A-59 recommended placing the rumble strips close to the edgeline which provides a wider recovery area along the shoulder. The support for this conclusion was that drivers preferred this solution and most of them did not think that the road width was too narrow as a result of this placement. Further, there are other benefits of this placement as well, especially as it relates to bicyclists. Hirasawa et al. (42) also conducted subjective testing at the Tomakomai Winter Test Track. The vehicles tested during the subjective experiments included passenger cars, motorcycles, and bicycles. Sixty-two participants participated in the test track experiments. At the driving/riding experiment, each participant filled out a questionnaire rating the safety of the three test patterns (Table A-17). Figure A-14 shows the results of the safety questionnaire. The results are combined for all road users. The participants’ negative evaluation of the rumble strips increases with the depth of the grooves. More participants answered that they felt danger when riding on the deep grooves than on the shallow grooves. Figure A-14. Subjective Ratings of Safety (42) When Outcalt (44) conducted his research in Colorado, he assumed that if rumble strips generated noise levels 6 dBA above the ambient noise level during normal operations (i.e., while driving the in travel lane), that this change in noise level would be a “clearly noticeable change” and would be sufficient to alert an inattentive/drowsy driver. Outcalt based his assumption on how a typical person perceives different amounts of change in sound level. Outcalt never investigated or verified this assumption as part of this research. Table A-35 suggests the approximate human perception of changes in sound level.

A-60 Table A-35. Approximate Human Perception of Changes in Sound Level (44) Change in sound level (dBA) Change in apparent loudness 1 dBA Imperceptible 3 dBA Barely noticeable 6 dBA Clearly noticeable 10 dBA About twice—or half as loud 20 dBA About four times—or one-fourth as loud As part of the investigation conducted by Bucko and Khorashadi (14) in California, two drivers provided a subjective assessment of the experimental rumble strips patterns. Based upon the subjective opinions of the two evaluators who tested several light passenger vehicles at speeds of 50 and 62 mph (80 and 100 km/h), Bucko and Khorashadi (2001) concluded the following [NOTE: The vibration levels provided in parentheses for the rumble strip patterns are the average resultant accelerations from both speed levels (Table A-19). Similarly, the sound levels provided in parentheses for the rumble strip patterns are the average resultant noise levels from both speed levels (Table A-20)]: • The vibration for rumble strips 1 (0.280 g) and 2 (0.125 g) was relatively similar to each other at 50 and 62 mph (80 and 100 km/h). • Rumble strips 3 (0.408 g), 4 (0.450 g), and 5 (0.567 g) produced a higher degree of vibration than rumble strips 1 (0.280 g) and 2 (0.125 g) at 50 and 62 mph (80 and 100 km/h). • The degree of vibration increased in ascending order with rumble strip 1 having the lowest vibration and rumble strip 5 having the highest vibration. • The noise generated by rumble strips 1 (13.5 dBA) and 2 (11.0 dBA) were relatively similar to each other at 50 and 62 mph (80 and 100 km/h) and were considered to have a low to moderate alerting value when compared to rumble strips 3 (16.7 dBA), 4 (18.3 dBA), and 5 (19.9 dBA) which were considered to have a high alerting property. • Vibrations felt through the steering wheel are negligible in their alerting properties compared to the noise level produced in the passenger compartment. • Although several rumble strip configurations required an additional amount of hand grip strength, none of the rumble strips caused any fishtailing or loss of control of light passenger vehicles. Watts (76) investigated the alerting properties of the rumble strips using a driving simulator. A stereo tape player was connected to the driving simulator, and noise pulses were triggered each time the motorist would drift from the lane. The motorist was asked to evaluate the noise patterns on a scale from one to seven, from not noticeable to very noticeable. The motorists were also asked to answer multiple choice questionnaires related to the type of noise they heard and what generated the noise. Watts concluded that rumble strips that produce 4 dBA increases or above would be readily detected by motorists if the noise level was sustained for 350 ms or longer. However if the noise

A-61 increase was only 2 dBA, a pulse length of at least 900 ms would probably be required. Also, a pattern of noise consisting of a regular series of 500 ms pulses separated by 500 ms would be suitable for alerting motorists. The noise increase in the pulses over the ambient levels should be at least 4 dBA. As part of the motor vehicle tests conducted by Tye (49), a subjective evaluation of the rumble strips patterns was performed by two California highway patrol officers and two traffic engineers. Their combined opinion was that the 0.25-in (6-mm) high plywood rumble strips did not provide adequate vibration. The 0.5-in (13-mm) high rumble strips were considered to have adequate alerting properties, and the 0.75-in (19-mm) high rumble strips were considered to have adequate to good properties. There were good correlations between the sound levels measured by the instrumentation and the loudness experienced by the evaluators. Similarly, there was a good correlation between the vertical vibration data and the shaking felt by the evaluators. However, the evaluators’ subjective opinions of motor vehicle controllability did not correlate well with the instrument data for front wheel bounce. This was understandable because the motorists’ sensation of control was related to the wheel spinning and fishtailing. The front wheel bounce data would probably bear a better relationship to vehicle control if the motorists were turning while traversing the rumble strips. All test runs, however, were made with the motor vehicle on a straight path through the rumble strip pattern. O’Hanlon and Kelly (50) conducted an experiment to empirically evaluate the effectiveness of different shoulder treatments for arousing fatigued drivers who allow their vehicle to drift from the travel way onto the shoulder. Physiological data from drivers were recorded from an instrumented vehicle in which 51 young male drivers collectively drove for 11,841 minutes on four test circuits covering 3,976 mi (6,400 km). Three shoulder treatments were installed along the test circuits: • Rib treatment which consisted of parallel raised strips of rock aggregate set in bituminous binder • Marker treatment which consisted of parallel arrays of raised circular pavement markers • Groove treatment which consisted of parallel slots cut into the shoulder surface In total the test subjects drove onto the shoulder on 229 different occasions, and these excursions resulted in 112 separate impacts with the various shoulder treatments. The subjects were monitored as to heart rate, electrocardiogram, electroencephalogram, skin conductance, and overall performance. O’Hanlon and Kelley did not measure the vibration levels experienced by the drivers so they did not investigate the correlation between vibration levels and measured physiological measures. Several relevant conclusions from this investigation are as follows: • Relative to SVROR incidents which do not involve impacts with the respective shoulder treatments, those involving impacts tended to forestall the occurrence of subsequent SVROR incidents. However, the elapsed driving time preceding the first SVROR incident is generally much longer than the time elapsed before

A-62 the next incident, regardless of whether or not the first incident involved an impact with a shoulder treatment. • Sizable percentages of drivers experiencing SVROR incidents, which do and do not involve impacts with shoulder treatments, run off the road again within the next five minutes (in this study, 18.5% and 28.6%, respectively). • The rib shoulder treatment evoked the greatest immediate increase in arousal when struck during the course of SVROR incidents. • The marker shoulder treatment was less effective in evoking arousal than the rib treatment, although probably not to any significant degree. • The groove shoulder treatment was ineffective in arousing the driver. In other words, the SVROR incidents involving this groove treatment evoked no greater arousal than those without impacts. • The persistence of arousal following impacts with every type of shoulder treatment was very brief. Little, if any, measurable residual of the immediate arousal reaction was present five minutes after any type of SVROR incident. • No hazardous behavioral overreaction occurred following the impacts with the shoulder treatments in this study. • Drivers tend to inadvertently drift to and beyond their lane boundaries with increasing frequency as a function of time on the road. Under some circumstances, drivers will allow their lane drifting frequencies to reach dangerously high levels. • Lane drift frequency is sensitive to as yet undefined road and traffic factors which vary between different highways and between dissimilar segments of the same highway. • Drivers drift more frequently from their lane to the right than to the left. • Driver arousal falls as a function of time on the road in correspondence with a deterioration in continuous road tracking performance. • In this study, the average angle of departure followed during SVROR incidents was approximately 3 degrees. Relevant recommendations made by O’Hanlon and Kelly (50) include: • The use of rib or marker shoulder treatments along highways having relatively high incidents of SVROR accidents. Further, they recommended that the treatment be placed along a considerable length of the target highway because evidence suggests that “spot” applications of 10 to 20 mi (16.1 to 32.2 km) would be ineffective in reducing the overall SVROR accident frequency on the target highway. • Place the treatments on the shoulder as close as possible to the travel way. • Install signing along the target highway to inform motorists of the presence of the treatment and of the significance of inadvertent impacts with the treatment.

A-63 • Heightened visual contrast between the travel way and the shoulder may reduce the tendency for drivers to drift onto the shoulder. In Michigan a recent community survey was conducted on the safety of rumble strips along with focus group interviews (109). The survey results showed a positive reaction to the presence of rumble strips by the majority of survey respondents. The major findings are as follows: • Roughly 84 percent of respondents either “strongly” or “somewhat” agree that rumble strips warn them when they are distracted, while 77 percent feel safer because of their presence. • 71 percent of respondents indicated that an added benefit of the rumble strips was improving the visibility of the road edge at night. • During focus group interviews, drivers felt that the best placement of rumble strips is 8 to 14 in (203 to 350 mm) from the edgeline. As part of their centerline rumble strip research, Russell and Rys (36) conducted a survey to determine the driver’s perceived effectiveness of several rumble strip patterns and the presence of centerline rumble strips. The most telling results of the survey were that 38 percent of respondents indicated that they would prefer a continuous pattern of centerline rumble strips over an alternating pattern and 96 percent felt that centerline rumble strip applications would reduce accidents. Drivers of Heavy Vehicles (i.e., Trucks) With heavy vehicles, the most relevant issue is whether a sufficient amount of stimuli is generated within the passenger compartment to alert a truck driver. Only one study was found that investigated the subjective reaction of truck drivers while traversing rumble strips. Based upon the subjective opinions of two evaluators who tested several commercial vehicles at speeds of 50 and 62 mph (80 and 100 km/h), Bucko and Khorashadi (14) concluded the following: • The vibration for rumble strips 1 (0.342 g) and 2 (0.150 g) was judged to be minimal and have a low to negligible alerting value. • Rumble strips 3 (0.226 g), 4 (0.246 g), and 5 (0.286 g) produced a higher degree of vibration than rumble strips 1 (0.342 g) and 2 (0.150 g). However, in commercial vehicles vibrations are dampened considerably because of the size and weight of the vehicles. Thus, the alerting properties of the vibration levels are essentially insignificant. • The noise generated for rumble strips 1 (4.72 dBA) and 2 (1.88 dBA) were considered to have a low alerting value when compared to noise generated by rumble strips 3 (3.62 dBA), 4 (4.61 dBA), and 5 (4.62 dBA) which were considered to have a moderate alerting value.

A-64 • The noise in the passenger compartment of a commercial vehicle, generated from rumble strips, has a greater effect in alerting a driver than the vibration produced by the same rumble strips. • The noise in the passenger compartment of a commercial vehicle, generated from rumble strips, has low-to-moderate alerting properties. • None of the rumble strips caused any fishtailing or loss of control of commercial vehicles. Motorcyclists The primary concern of motorcyclists about rumble strips relates to controllability. Only a few studies of rumble strips have included motorcycles in field experiments. The most detailed study of the interaction between motorcycles and rumble strips was performed by Miller (53). Miller investigated motorcycle rider behavior on roads with centerline rumble strips. The research involved a review of motorcycle crash records, an observational study of motorcyclists on roads with centerline rumble strips, and a closed course field study where 32 motorcyclists traversed rumble strips. Miller concluded that centerline rumble strips add no measurable risk to motorcyclists. In the research conducted by Bucko and Khorashadi (44), a limited number of field tests with motorcycles were completed by several members of the California Highway Patrol (CHP). Four members of the CHP, all considered to be advanced motorcyclists, subjectively rated all of the rumble strip treatments installed at the Dynamic Test Facility while traveling at 50 and 65 mph (80 and 105 km/h) on either a BMW R1100RTP or Harley Davidson FX motorcycle from their fleet. Although the results were statistically insignificant, Bucko and Khorashadi (14) noted that the results of the tests were quite positive. None of the treatments was found to have significant deficiencies from a safety point of view. In fact, all treatments were rated very high. The only concerns noted from the CHP participants were that the raised pavement markers and Carsonite Bars were slick when wet. The CHP also participated in earlier rumble strip research conducted in California by Tye (49). Based upon testing that involved driving a fully equipped CHP Harley- Davidson motorcycle over plywood rumble strips at speeds of 30, 50, and 60 mph (48, 80, and 97 km/h), Tye reported that control of the motorcycle was not affected by any of the rumble strips. Tye speculated that the wheelbase of the motorcycle was such that the motorcycle was affected by only one rib at a time. Kansas and Massachusetts also report testing motorcycles on rumble strips (63). While the composition of the Kansas test group was unknown, the Massachusetts test group was comprised of the police motorcycle squad. In both studies the test groups reported noticing the rumble strips, but none of the motorcyclists reported experiencing control problems.

A-65 Bicyclists Bicyclists and bicycle groups have expressed concerns about both shoulder and centerline rumble strips. Their main concerns are that a bicyclist may lose control while riding over the rumble strips and that bicyclists experience discomfort while traversing rumble strips. Several studies have investigated the subjective reactions of bicyclists as they experienced a variety of rumble strip patterns. Gardner et al. (43) conducted a survey of bicyclists to assess their like or dislike of the football shaped rumble strips compared to rectangular shaped rumble strips. The survey was distributed to a Wichita based bicycle group, and twenty-three responses were obtained. The survey revealed that the bicyclists preferred the football shaped rumble strips over the rectangular shaped rumble strips. Torbic (54) examined the fundamental relationships between rumble strip dimensions, bicyclists’ perceptions of ride comfort, and the controllability of a bicycle and investigated the causes of discomfort as well as controllability problems that bicyclists experience while traversing rumble strips. The primary objectives of this research were: • To investigate the relationship between whole-body vibration generated by milled rumble strips and bicyclists’ perceptions of comfort • To investigate the relationship between whole-body vibration generated by milled rumble strips and the controllability of a bicycle • To identify the conditions that cause bicyclists to experience the highest levels of discomfort and control problems while traversing milled rumble strips Torbic used data gathered during the PennDOT Bicycle-Tolerable Shoulder Rumble Strips project (45) and supplemented it with additional data to evaluate the ride quality of bicycles. In this research, Torbic developed a methodology for quantifying whole-body vibration of bicyclists. The methodology was based upon guidelines in International Standard (ISO) 2631 (55) for quantifying whole-body vibration to assess human response. Whole-body vibrations were calculated by combining vertical and pitch angular accelerations into one measure to assess the combined effect on comfort and controllability. Then, the relationships between whole-body vibration and bicyclists’ perceptions of comfort and the controllability of bicycles were assessed. Based upon the analysis results, Torbic concluded that the relationship between whole-body vibration and a bicyclist’s perception of comfort is linear; as vibration increases comfort decreases. The analysis also indicated there is no clear relationship between whole-body vibration and the controllability of a bicycle. As part of the bicycle testing portion of their study, Bucko and Khorashadi (14) collected subjective data from bicycle riders of all ages and experience levels while riding over eleven different rumble strip patterns (Table A-18). Participants traversed the eleven rumble strip patterns at varying speeds, angles, in groups, and as a single rider. Participants could select to ride one of the 18 bicycles available for the testing or could

A-66 use their own bicycle during the tests. After traversing the rumble strip patterns, the participants stopped and subjectively rated the comfort and control level of the rumble strips. Fifty-five bicyclists participated in the subjective testing. Figure A-15 shows the mean subjective comfort ratings for the eleven treatments. In Figure A-15, higher values of mean response represent a higher level of comfort, while lower values of mean response represent a lower level of comfort. Figure A-16 shows the mean subjective control ratings for the eleven treatments. In Figure A-16, higher values of mean response represent a higher level of control, while lower values of mean response represent a lower level of control. Figure A-15. Bicyclist Subjective Comfort Ratings Averaged Across Various Factors (14) Figure A-16. Bicyclist Subjective Control Ratings Averaged Across Various Factors (14) As part of the bicycle testing conducted in Colorado, 29 bicyclists participated in subjective testing to rate the comfort and controllability of bicycles while traversing 10 rumble strip patterns (Table A-21). Road bikes with narrow, high pressure tires were used in the testing as well as mountain bikes with fat, low pressure tires. Each bicyclist

A-67 traversed the test patterns at 5, 10, 15, and 20 mph (8, 17, 24, and 32 km/h). Each bicyclist rated the pattern for both comfort and controllability on a scale from 1 (No Effect) to 5 (Severely Uncomfortable/Uncontrollable). Some of the bicyclists were unable or unwilling to rid some of the test sections at higher speeds because their bicycles became uncontrollable. Those sections were recorded as a 5 (Severely Uncomfortable/ Uncontrollable) for that speed. Figure A-17 and Table A-36 show the average ratings from all riders for all speeds for each rumble strip pattern. As Figure A-17 and Table A- 36 illustrate, the 0.75 in (19 mm) deep rumble strip pattern (Section 5) is the most objectionable to bicyclists and the concrete strip (Section 10) is the most favorable. Outcalt notes that the trends in Figure A-17 and Table A-36 are similar to the trends in sound levels in the motor vehicles, indicating that rougher rumble strips to bicyclists are louder rumble strips with more vibration felt in a motor vehicle. Figure A-17. Average Comfort and Control Ratings of Bicyclists [scale: 1 (no effect) to 5 (severely uncomfortable/uncontrollable)] (44) Table A-36. Average Comfort and Control Ratings of Bicyclists (44) Section Average control rating Average comfort rating 1 1.8 2.3 2 2.3 2.9 3 2.1 2.6 4 2.4 3.0 5 4.4 4.7 6 4.2 4.6 7 3.9 4.3 8 3.4 4.0 9 2.9 3.5 10 1.4 1.4 Scale: 1 (No Effect) to 5 (Severely Uncomfortable/Uncontrollable).

A-68 As part of the bicycle tests conducted by Elefteriadou et al. (45), 25 volunteer riders subjectively rated the comfort and controllability of bicycles while traversing the 6 rumble strip patterns (Table A-25) over an extended distance of approximately 45 ft (14 m). The bicyclists traversed the rumble strip patterns at various speeds and after each trial, the bicyclists were asked to fill-out a questionnaire rating the comfort level of different body parts while traversing the rumble strip configurations. In addition, the bicyclists were asked to rate the overall control level while traversing the configurations. Table A-27 presents the average ranking (based on a 25-point scale) of the comfort level for the different body parts, the overall comfort level, and the overall control level across all test patterns, all bicycle types, and all speed ranges. The table indicates the body parts that are most affected while traversing rumble strips. Lower values indicate greater discomfort, while higher values indicate better comfort. Based on the subjective results, the ordered list below ranks the body parts most affected while traversing rumble strips: • Wrists, fingers, and elbows (most uncomfortable) • Seat area • Shoulders and neck • Back • Knees, ankles, and feet (most comfortable) Table A-37. Overall Average Subjective Ranking of Comfort and Control Levels for Bicyclists (45) Wrists, fingers, and elbows Shoulders and neck Back Seat area Knees, ankles, and feet Overall comfort Overall control Average Value 11.26 12.90 13.47 12.02 13.57 11.64 11.30 Comfort scale: very uncomfortable (0) and very comfortable (25). Control scale: uncontrollable (0) and no effect on handling (25). Tables A-38, A-39, and A-40 present the rankings of the test configurations based on the subjective wrist comfort level, overall comfort level, and overall control level, respectively. The average values combine the ratings across all bicycle types and all speeds. From the perceptions of the participants, test configurations 6, 3, and 5 in that order consistently ranked the best from the standpoint of comfort and control. Test pattern 1 was consistently perceived as the worst test pattern from the standpoint of comfort and control. Young (110) conducted a test with a road bicycle on a section of U.S. 191 in Teton County, Wyoming, that had milled “Pennsylvania Turnpike” style rumble strips. A test rider rode over or across the rumble strips at speeds of less than 5 mph (8 km/h), 10 mph (16 km/h), 20 mph (32 km/h), and 30 mph (48 km/h). In general, at speeds greater than 5 mph (8 km/k), the test rider found it dangerous riding over or across the rumble strips. Gårder (111) conducted tests to verify bicyclist concerns about maneuverability problems associated with rumble strips. Gårder, together with 20 students and staff at the

A-69 University of Maine, rode over two different configurations of milled rumble strips on several types of bicycles. Gårder found that, “Not a single rider reported any tendency to lose control at any speed or any angle even when not holding on to the handle bars. But every rider reported that riding on the rumble strips was annoying.” Thus, these tests did not support bicyclists’ fears that shoulder rumble strips would cause them to lose control of their bicycles. Table A-38. Ranking of Test Configurations Based on Subjective Wrist Comfort Level (all bicycles) (45) Rank Test pattern Average wrist comfort levela Best 1 6 14.3 2 3 13.7 3 5 11.9 4 4 10.5 5 2 10.1 Worst 6 1 7.1 a Comfort scale: very uncomfortable (0) and very comfortable (25). Table A-39. Ranking of Test Configurations Based on Subjective Overall Comfort Level (all bicycles) (45) Rank Test pattern Average overall comfort levela Best 1 6 14.8 2 3 14.5 3 5 12.1 4 2 11.0 5 4 10.0 Worst 6 1 7.3 a Comfort scale: very uncomfortable (0) and very comfortable (25). Table A-40. Ranking of Test Configurations Based on Subjective Overall Control Level (all bicycles) (45) Rank Test pattern Average overall control levela Best 1 6 14.3 2 3 13.4 3 5 11.5 4 2 10.8 5 4 9.5 Worst 6 1 7.4 a Control scale: uncontrollable (0) and no effect on handling (25)

A-70 Pedestrians Very few pedestrians encounter rumble strips so for the most part rumble strips do not affect pedestrians. Shoulders are not usually appropriate as pedestrian facilities (112), particularly on facilities where vehicular traffic speeds are high, which is often the type of facility where rumble strips are installed. At intersections, rumble strips are discontinued so pedestrians do not encounter rumble strips while crossing at intersections. Several studies (113,114) have looked at the vibration levels experienced by wheelchair users while traversing obstacles similar to rumble strips, but this is viewed as a relatively low priority issue so the results of these studies are not summarized here. A.6. Pavement Performance Issues Several pavement performance concerns associated with shoulder and centerline rumble strips have been identified. Very little scientific based research has been conducted to address these concerns, but through observational reports most of the pavement performance concerns appear to be unwarranted. Several maintenance concerns associated with shoulder and centerline rumble strips have been reported. Maintenance crews reported concerns that heavy traffic would cause shoulder pavements with rumble strips to deteriorate faster and that the freeze-thaw cycle of water collecting in the grooves would crack the pavement (115). Although the literature review revealed no published, controlled studies regarding the impact of rumble strips (primarily milled rumble strips) on pavement integrity, FHWA reports that these concerns have proven to be unfounded. Rumble strips have little effect on the rate of deterioration of new pavements. Older shoulder pavements tend to degrade more quickly, but tests in several states indicate that these rumble strips continue to perform their intended function. There are also no apparent problems with installation or faster deterioration of rumble strips on open-graded pavement surfaces. Most transportation agencies do advice against installing shoulder rumble strips on pavements that are rated as deformed or show high degrees of deformation and/or cracking. Inclement weather also appears to have an insignificant impact on the durability of shoulder rumble strips. Field tests refute concerns about the effects of the freeze-thaw cycle as water collects in the grooves. In fact, field tests show that vibration and the action of wheels passing over the rumble strips knock debris, ice, and water out of the grooves. Snow plow drivers have also noted that they have come to depend on shoulder rumble strips to help them find the edge of the travel lane during heavy snow and other low visibility situations. Shoulder rumble strips may also present a challenge to maintenance and rehabilitation crews when lane closures require traffic to be diverted to the shoulder. For long-term rehabilitation projects involving asphalt shoulders, most agencies simply mill a trench around the rumble strips and fill the trench with asphalt. Once construction is

A-71 complete, the shoulder can be resurfaced and new rumble strips installed along the new asphalt overlay. Similar to the experience with shoulder rumble strips, several agencies have expressed concerns about pavement deterioration associated with the installation of centerline rumble strips (56). However, none of these concerns have been validated. The pavement performance issue that has received the most detailed investigation deals with preparation of rumble strips prior to overlayment of the shoulder surface so that rideability and/or pavement integrity are not compromised. New Hampshire DOT (NHDOT) conducted a study to develop a specification defining materials, sequences, and/or options to perform this operation successfully. In summer 2005 NHDOT prepared four test sections to evaluate how the preparation of rumble strips prior to overlayment of the shoulder surface impacts the rideability and/or pavement integrity of the shoulder. Four test sections, each 500 ft (152 m) in length, were prepared along I-89 in slightly different manners to compare the differences in preparation practices. The four preparation scenarios were performed prior to placing a 1.5 in (38 mm) bituminous overlay (116): Scenario A (Shim and Overlay): • Tacked and shimmed entire 10 ft (3.0 m) width of shoulder section • Used 0.5 in (13 mm) shim coat with 1.5 in (38 mm) overlay • 10 ton roller used to shim, both 10 ton and 30 ton rollers used on overlay Scenario B (Just Overlay): • No special treatment of rumble strip, just tack and 1.5 in (38 mm) overlay • Compressed 1.5 in (38 mm) overly with roller Scenario C (Mill, Inlay, and Overlay): • Ground out 20 in (508 mm) wide rumble strip first, 0.5 in (13 mm) deep • Tack coated ground out rumble strip portion of shoulder • Filled inlay with asphalt • Compressed rumble strip inlay with 10 ton back roller. Compacted to same level as existing pavement. • Tack coated over inlay and rest of area to be overlayed. • Overlayed entire shoulder with 1.5 in (38 mm) overlay Scenario D (Mill and Overlay): • Ground out rumble strip 0.5 in (13 mm) deep • Tack coated over inlay and rest of area to be overlayed. • Overlayed shoulder with 1.5 in (38 mm) overlay, except near rumble strip inlay which required 2 in (51 mm) of material

A-72 The resurfacing operations were performed at night. The following observations were made immediately following the preparation and resurfacing activities: • Scenarios A, C, and D seemed to be about the same resulting product with A having the rumble strip show through slightly, while C and D did not show through at all. • Scenario B resulted in the rumble strips clearly showing through the pavement, made more visible due to nighttime lighting. The following observations were reported during an inspection approximately 3 weeks after the overlay operations (117): Scenario A (Shim and Overlay): • Showed no indication of rumble strip reflection Scenario B (Just Overlay): • Showed occasional longitudinal cracks along the edge of the rumble strip, indicating movement of the mix by the roller through the affected rumble. The rumble strip reflected through the overlay along the entire length of the test section. • Additionally, a parallel line of “reflected” rumble strips was observed in this test section. It is hypothesized that the vibratory roller drum bounces due to the alternating mix thickness in the rumble strip resulted in the indentation of the surface alongside of the original rumble strips. Scenario C (Mill, Inlay, and Overlay): • Showed no reflection of the milling, which would have displayed as a rut Scenario D (Mill and Overlay): • Showed no sign of the former rumble strip The following observations were reported during an inspection in April 2006, following the first winter after the overlay activities (118,119): Scenario A (Shim and Overlay): • Mild depressions are now visible due mainly to the abrasion of the snowplows over the rumble strip area. The rumble strips are also felt when driven over. Scenario B (Just Overlay): • Continues to show a pronounced rumble strip reflection, enhanced by the abrasion of the rumble strips. • Not additional deterioration was noted.

A-73 Scenario C (Mill, Inlay, and Overlay): • Showed no reflection of the milled area • The outline of the former rumble strip area is vaguely visible on the shoulder surface. Scenario D (Mill and Overlay): • Showed no sign of reflection in the area of the former rumble strips A third inspection of the test sections was performed in June 2007, after the second winter of the overlay activities (118,119). No noticeable changes were observed since the last inspection in April 2006. This suggests that the rumble strip reflection is complete, having occurred within the first year after the overlay. A.7. Other Potential Adverse Concerns This section presents other potential issues or concerns associated with shoulder and/or centerline rumble strips that have not been discussed in Sections A.1 through A.5. A brief discussion of the following issues/concerns is presented. Impact of Noise on Nearby Residents A common problem cited by transportation agencies concerning the use of rumble strips is noise that disturbs nearby residents (15). However, noise is generated relatively infrequently by rumble strips placed on the shoulders and on the centerlines of undivided highways. For shoulder and centerline rumble strips, noise is generated only by errant motor vehicles, not by every motor vehicle. Although the noise produced by shoulder and centerline rumble strips is intermittent, transportation agencies continue to receive complaints from nearby residents. For example, when shoulder rumble strips were installed along a limited-access road in Connecticut, several noise complaints were received from residents in the near vicinity (18). As a result, the Connecticut DOT modified the offset for rumble strips in the right shoulder from 6 to 12 in (152 to 305 mm). The reason for this change was to decrease the incidence of vehicles falsely traversing the rumble strips. As a result of this offset modification, noise complaints eventually decreased. Concerning noise produced from centerline rumble strips, the Alberta Transportation Authority has received complaints about noise where the ambient noise level is very low (56). Some residents claim to be able to hear the noise generated from the centerline rumble strips from up to 1.2 mi (2 km) away. On the other hand, Gardner et al. (43) conducted a survey of residents along a Section of US Highway 40 where football shaped centerline rumble strips were installed. All of the respondents to the survey (n = 32) lived within 600 ft (183 m) of US Highway 40. The survey showed that 78 percent (n=25) of the respondents could hear noise from the rumble strips while in their homes, but only 16 percent (n=4) indicated that the noise is loud enough to cause a concern or distraction. Gardner et al. concluded

A-74 that the majority of the residents are satisfied with the centerline rumble strips along US Highway 40 because there is more potential for driver safety than the effects of external noise produced from coming in contact with the rumble strips. Noise concerns are one area where evaluations of noise generated from transverse rumble strips may be applicable to this research associated with shoulder and centerline rumble strips. Gupta (120) evaluated the effectiveness of several rumble strip designs on speed reduction in the travel lane. As part of the evaluation, noise levels associated with the rumble strip designs were measured. Noise data were gathered from four different rumble strip designs for both cars and trucks. The sound levels generated by the rumble strips were measured approximately 10 ft (3 m) from the pavement edge. Gupta determined that the rumble strips created an increase in noise level of 6 to 8 dBA. The amount of noise created by rumble strips is related to various factors such as vehicle speed, vehicle types, tire tread, pavement surface, and rumble strip dimensions. Higgins and Barbel (81) conducted a study to determine the noise levels in the surrounding neighborhood generated from transverse rumble strips. Higgins and Barbel concluded that transverse rumble strips produced a low frequency noise that can increase the noise levels by up to 6 or 7 dBA over the noise levels produced by traffic on normal pavement. These values were confirmed in a Texas study (106,107) aimed at measuring the exterior noise created by various configurations of rumble strips. Both passenger car and commercial truck variations were recorded on milled, rolled, and raised rumble strips at a distance of 50 ft (15 m) from the travel way. The average base exterior noise was measured for the passenger car at speeds of 55 and 70 mph (88 and 113 km/h) and the commercial vehicle at 55 mph (88 km/h); the respective sound levels were found to be 76, 79 and 83 dBA. The average exterior noise for the same passenger vehicles traveling over the range of rumble strip configurations was 82, 87, and 88 dBA, respectively. Finley and Miles noted, from these results, that the noise generated by the average passenger car traveling over the average rumble strip configuration is still lower than the noise impact of a commercial vehicle on a smooth surface. It also appears that the greatest noise increase (10 to 19 dBA) for milled rumble strips from the base condition came when the length was maximized (16 in [406 mm]) and the spacing was minimized (12 in [305 mm]). In a study of various milled rumble strip patterns conducted in Alberta (56), the following points were found: • A change in speed from 50 to 75 mph (80 to 120 km/h) has little effect on the outside sound level when the vehicle is traveling in the normal driving lane (e.g., from 75 dBA to 82 dBA); however, when driving on rumble strips, the sound level is greatly affected (e.g., from 88 dBA to 102 dBA). • The sound level outside the vehicle increases linearly with vehicle speed. • The majority of sound created by rumble strips dissipates at approximately 328 ft (100 m).

A-75 • For heavy vehicles (i.e., trucks), a minimum rumble strip depth of 0.32 in (8 mm) is required to produce an increase in sound within the cab. Studies show that rumble strips that are terminated 656 ft (200 m) prior to residential or urban areas produce tolerable noise impacts on nearby residents (57). At a distance of 1,640 ft (500 m), the noise generated from rumble strips is negligible. Several transportation agencies have experimented with numerous alternatives to mitigate the noise generated by rumble strips installed near residential areas. One alternative is to construct noise barriers. Some have also moved the shoulder rumble strips further away from the edge of the travel lane. This measure, however, provides less time and distance for errant motorists to recover control of their vehicles. Other Bicycle Issues Most of the studies that investigate the impact of rumble strips on bicyclists focus on the comfort and control problems that bicyclist may (or may not) experience while traversing rumble strips. In other words, most studies have been concerned whether the vibrations experienced by bicyclists when they encounter rumble strips cause discomfort to bicyclists or loss of control of bicycles. However, bicyclists have several other concerns associated with rumble strips. One concern with shoulder rumble strips is that they may encourage bicyclists to ride in the travel lane in situations where bicyclists would rather ride on the shoulder (15). Even though rumble strips are typically installed on only about half of the paved shoulder, the remaining area between the outer edge of the rumble strip and the outside edge of the shoulder is often littered with debris. This discourages bicyclists from utilizing that area. Therefore, bicyclists may prefer to ride in the travel lane. A possible solution to this dilemma is to move the rumble strip further from the travel lane to provide bicyclists with adequate room to ride between the travel lane and the rumble strip. This, however, decreases the recovery area available to errant motor vehicles. Another possibility is to make the rumble strips narrower. Yet, another possibility is to provide a gap in the rumble strip pattern to allow bicyclists to cross back and forth from the paved shoulder to the travel lane without having to encounter rumble strips. Moeur (99) conducted a study in Arizona to determine the optimum length of gaps in continuous shoulder rumble strips to accommodate bicyclists. As part of the field experiment, bicyclists traversed through gaps in rumble strips ranging from 10 to 20 ft (3.0 to 6.1 m) at various speeds. Based upon the bicyclists behaviors, Moeur recommended that rumble strips on all noncontrolled access highways include periodic gaps of 12 ft (3.7 m) in length, and that these gaps be placed at periodic intervals at a recommended spacing of 40 ft (12.2 m) or 60 ft (18.3 m). A general concern with centerline rumble strips is that motorists may not provide sufficient clearance distance between the bicyclist and the motor vehicle when passing a bicyclist on a section of roadway with centerline rumble strips (56). In other words, the

A-76 centerline rumble strips may force motorists away from the centerline closer to bicyclists riding near the outside edge of the travel lane, leaving less distance between bicyclists and motor vehicle during the actual passing maneuver. Another concern is that when motorists encounter centerline rumble strips during the passing maneuver, the noise generated by the rumble strips may startle bicyclists which could result in an undesirable maneuver by the bicyclist. Maintenance Concerns Weather does cause problems with raised rumble strips. Snow plow blades passing over the rumble strips tend to scrape them off the pavement surface, which is why raised rumble strips are usually restricted to areas that do not contend with snow removal. When raised rumble strips get scraped from the pavement surface, a secondary concern is that the material could become a projectile. Visibility/Retroreflectivity of Centerline and Edgeline Pavement Markings Some transportation agencies have reported concerns over the visibility and retroreflectivity of centerline pavement markings installed on centerline rumble strips (32,36). This could potentially be a problem under nighttime conditions especially if snow, salt, sand, or debris collect in the grooves of the rumble strips. Visibility of pavement markings can also be an issue when rumble strips are installed along the edgeline. Conflicting evidence as to whether this is an actual problem is provided in the literature. For example, Colorado reports that during winter the grooves in the strip tend to collect some of the sand that is applied during snow removal (30). The sand does not completely fill the grooves; however, it does obscure some of the paint strip at the bottom of the grooves. Saskatchewan reports a loss of retroreflectivity of the centerline markings during wet conditions (56). A focus group in Minnesota reported that some participants felt that the painted centerline markings were less visible at night, particularly under wet conditions. Conversely, Alberta indicates that they have not experienced any difficulties or adverse wear of pavement markings after the installation of centerline rumble strips (56). In fact, Alberta reports that the pavement markings in the grooves of rumble strips may actually experience less wear and tear from snowplows and other vehicles because the paint is somewhat protected from the surface. In Texas, pavement markings applied over rumble strips were found to maintain their visibility during rainy nighttime conditions. Although as indicated above Saskatchewan reports a loss of retroreflectivity of the centerline markings during wet conditions, Saskatchewan also reports no reduction in nighttime visibility of markings painted on top of rumble strips. In fact, one location was selected for installation of centerline rumble strips based upon a frequent lack of visibility due to fog, and centerline rumble strips were found to enhance the centerline delineation at this site.

A-77 In a 2003 survey transportation agencies were asked whether centerline rumble strips reduce nighttime retroreflectivity of the material (36). Fourteen of the 24 respondents answered that there was not any reduction in nighttime visibility. Four respondents answered yes, and six other respondents indicated unknown. Russell and Rys (36) also indicated that all the answers to the survey were based upon subjective evaluations (i.e., no data were mentioned). The objective of a study conducted by Filcek et al. (121) in Michigan was to discover alternative methods of increasing the durability, retroreflectivity, and wet-night retroreflectivity of pavement markings subjected to winter maintenance activity, utilizing standard Michigan DOT (MDOT) pavement marking materials. The study consisted of a side-by-side comparison of standard MDOT edgelines and standard MDOT waterborne paint and glass beads placed on milled shoulder rumble strips. The results of the study yielded the following conclusions: • Milled rumble strip edgeline pavement markings are more resilient to winter maintenance activities than standard pavement markings. • Retroreflectivity measurements for dry and wet-night conditions are significantly higher for milled rumble strip edgeline markings as compared to standard edgeline markings. Filcek et al. also reported about a study in Mississippi that indicated wet-night retroreflectivity benefits of pavement markings being placed on a profiled surface. The North Dakota DOT conducted a study to determine if placing pavement markings on a rumble strip would improve the marking’s wet-night retroreflectivity (122). North Dakota noted that the position of the markings, on the rumble strip, does not appear to greatly affect the day-time appearance of the marking. The application of marking paint on fog seal material may cause some durability problems, but so far it has only caused some limited problems in one area with unusually thick fog seal material. Wet-night retroreflectivity readings appear to show that “rumble strips” provide higher retroreflectivity readings than nearby flat markings. Results from a study in Texas (123) indicate that rumble stripes do provide at least twice the wet-night retroreflectivity compared to their equal but flat thermoplastic counterpart. Partial results are presented in Table A-41.

A-78 Table A-41. Retrorefectivity Measurements by Rainfall Rate for Pavement Markings (123) Measured retroreflectivity (mcd/m2/lux) for indicated rainfall rate in inches per hour Material, bead, color Dry Recovery 0.28 0.87 1.2 2.0 4.0 6.0 8.0 9.5 11.5 14.0 Flood Thermo, Type II, W 524 96 71 39 31 25 19 22 23 22 22 27 21 Thermo Rumble, Type II, W 503 185 144 129 99 101 70 64 64 57 61 58 49 A study comparing the service life, life-cycle costs, and wet-night visibility of edgeline rumble stripes and standard thermoplastic edge markings was conducted by the Alabama Department of Transportation (124). Data were collected over a three year period using a mobile retroreflectometer at locations where each of the treatments had been installed. The results were then extrapolated over time to determine the long term effectiveness of both types of edgeline markings. While the standard markings had initial retroreflective values approximately 25 percent higher than the rumble stripe counterparts, dry retroreflective decay curves indicated that the rumble stripes lost their visibility at a slower rate. It was postulated that this was due to less traffic driving on the marking due to the rumble effect on the vehicle. Service life values were established and are presented in Table A-42 that display a marked improvement for rumble stripes over a range of ADT values and retroreflective thresholds. Because of the inability to collect sufficient data for wet-night visibility of standard pavement markings, the retroreflectivity of the rumble stripes were incomparable; however, the Lindly and Narci concluded that the values would have been lower for flat pavement markings than for rumble stripes. Table A-42. Estimated Service Lives in Terms of Age of Markings (124) Average service life in months Threshold = 100 mcd/m2/lux Threshold = 150 mcd/m2/lux FTM Rumble stripe FTM Rumble stripe ADT per lane Average 95% C.I. Average 95% C.I. Average 95% C.I. Average 95% C.I. 2,500 60+ 60+ 60+ 60+ 60+ 45-60 60+ 45-60 5,000 46 33-60 60+ 60+ 34 23-51 48 27-60+ 7,500 31 22-48 60 42-60+ 22 15-34 32 18-54 10,000 23 16-36 45 32-60+ 17 11-26 24 14-40