EVALUATION OF RUMBLE STRIPES ON LOW-VOLUME RURAL 1 ROADS IN IOWA 2 3 Shauna Hallmark1, Tom McDonald2, and Robert Sperry3 4 5 Center For Transportation Research and Education at Iowa State University 6 2711 S. Loop Drive, Suite 4700, Ames, Iowa 50010-8664 7 Phone: (515) 294-8103; Fax: (515) 294-0467 8 9 1: corresponding author, Associate Professor Department of Civil, Construction, and 10 Environmental Engineering, email: shallmar@iastate.edu 11 2: research engineer, email: tmcdonal@iastate.edu 12 3: research engineer, email: rsperry@iastate.edu 13 14 Text: 4,472 words, 9 tables and figures = 7,403 words total 15 16 17 ABSTRACT 18 Single-vehicle run-off-road crashes are the most common crash type on rural two-lane 19 Iowa roads. Rumble strips have been proven effective in mitigating these crashes, but 20 these strips are commonly installed in paved shoulders adjacent to higher-volume roads 21 in Iowa. Lower-volume paved rural roads owned by local agencies do not commonly 22 feature paved shoulders but frequently experience run-off-road crashes. For roads where 23 paved shoulders are not a viable option due to cost, narrow shoulders, and right-of-way 24 restrictions, an alternative process has been devised, which involves milling narrow width 25 rumble strips directly along the existing pavement edge, followed by placement of 26 standard edge line pavement markings over the milled areas. 27

28 Although it is believed that using rumble strips/stripes can decrease lane departures, 29 limited information is available that demonstrates the effectiveness of the treatment. The 30 research described in this paper evaluated the effectiveness of edge line rumble strips in 31 Iowa. The project identified and selected candidate locations on rural two-lane rural roads 32 with no paved shoulders. Rumble stripes were installed and evaluated for six locations. 33 Lane position was evaluated for several sites as a surrogate measure of safety since only a 34 short after period was available for a crash analysis. Pavement marking wear was also 35 qualitatively assessed. 36 37 38 INTRODUCTION 39 40 Background 41 Run-off-road (ROR) crashes are a serious traffic safety concern. ROR crashes annually 42 account for 38% of U.S. highway fatalities and 1 million injuries. It is also estimated that 43 24% of highway fatalities occur on two-lane undivided rural roads (Taylor and 44 Meczkowski 2003). Neuman et al. (2003) also estimated that 39% of national fatal 45 crashes are single-vehicle ROR crashes. 46

TRB 2012 Annual Meeting Original paper submittal - not revised by author.

47 Lane departure crashes are the single largest category of fatal and major injury crashes in 48 Iowa. Iowa crash data show that 52% of roadway-related fatal crashes are lane departures 49 and that 39% of Iowa’s fatal crashes are single-vehicle ROR. Addressing roadway 50 departure is also one of the key emphasis areas for the American Association of State 51 Highway and Transportation Officials (2005). A number of strategies have been 52 employed to address roadway departure, including installation of rumble strips on paved 53 shoulders, which has been proven very effective in reducing the incidence of roadway 54 departure crashes. However, the strategy can only be used on roadways with paved 55 shoulders. 56 57 For roads where paved shoulders are not a viable option due to cost, narrow shoulders, 58 and right-of-way restrictions, an alternative process has been devised, which involves 59 milling narrow width rumble strips directly along the existing pavement edge, followed 60 by placement of standard edge line pavement markings over the milled areas. This type 61 of edge line rumble strips has been referred to as “rumble stripes”. Some agencies are 62 using edge line rumble strips on two-lane paved roadways with unpaved shoulders. 63 Although they do not provide the full benefit of traditional rumble strips, rumble stripes 64 can provide some alert to drivers crossing the edge line. In addition, since the edge line 65 pavement marking is painted through the rumble stripe, the grooved surface of the rumble 66 strip facing the driver can provide a near-vertical surface, which enhances edge line 67 pavement marking visibility at night and during rainy conditions. Edge line shoulder 68 rumble strips/stripes increase edge line marking visibility and longevity because part of 69 the line paint is located within the rumble strip/stripe depression. This feature is 70 particularly advantageous in climates where ice and snow are present, since raised 71 pavement markers cannot be used due to potential snowplow damage. 72 73 Effectiveness of Rumble Stripes 74 Although it is believed that using rumble strips/stripes can decrease lane departures, 75 limited information is available that demonstrates the effectiveness of the treatment. The 76 Mississippi Department of Transportation (as reported by ATTSA in 2006) installed edge 77 line rumble stripes on a two-lane roadway and conducted a before and after crash study. 78 They found that right-side ROR crashes were reduced by 25% after installing the rumble 79 stripes. The Texas Transportation Institute (TTI) evaluated the impact of edge line 80 rumble stripes on traffic operations and found that shoulder encroachment decreased by 81 46.7% after installation of edge line rumble stripes (Miles et al. 2005). 82 83 Pratt et al (2006) evaluated centerline and edgeline rumble strips (ERS) where the RS 84 was placed directly on the marked edgeline along a 5-mile segment. The rumble stripes 85 were 0.5 inches deep with a 7-inch length and 12-inch width at 12-inch spacing. A 4-86 inch edgeline was painted through the rumble stripes. They evaluated shoulder usage and 87 lateral position within the lane line before and after installation of rumble stripes using a 88 Z-configuration of road tubes and video. Encroachments were only included for daytime 89 hours. Shoulder encroachments were classified as emergency vehicles, turning, passing, 90 and inadvertent. Encroachments were assessed on both curved and tangent sections. 91 They found a 46.7% reduction for all encroachments. Inadvertent shoulder 92

TRB 2012 Annual Meeting Original paper submittal - not revised by author.

encroachments decreased from 616 to 359 from the before to after period. Lateral 93 encroachment onto the shoulder was also recorded and a decrease from 10.6 to 18.5 94 inches was noted. A 71.8% decrease in number of vehicles striking the right edge line 95 was also noted. 96 97 A recent study of Missouri’s Smooth Roads Initiative (SRI) included 61 sites and over 98 320 miles of both edge line rumble stripes and shoulder rumble strips. The authors 99 conducted a before and after analysis using an empirical Bayesian analysis. Overall, they 100 found that the SRI program showed a statistically significant 8% decrease in fatal and 101 disabling injury crashes and a 6% decrease in fatal and all injury crashes. However, the 102 analysis only included one year of after data (Potts et al. 2008). 103 104 Anund et al (2008) studied the effect of 4 types of rumble strips on sleepy drivers in an 105 advanced moving driving simulator in Sweden and Finland. One set of rumble strips was 106 roughly similar to what is used as edgeline rumble stripes with dimensions of 7 inches 107 wide by 0.8 inches long at a spacing of 11.2 inches apart and a depth of 0.6 inches. 108 Thirty five subjects who had worked the night shift before participating in the study were 109 evaluated over a straight section of road alternating a particular type of rumble strips. 110 Lateral position, eye blinks, and other sleepiness metrics were measured. They reported 111 that all of the rumble strips evaluated including the edgeline rumble stripes exhibited 112 similar alerting effects as measured by reduced variability in lateral position and a 113 decreased physiological sleepiness level. However the effect in all cases was only 114 moderate and the decrease in sleepiness was only sustained for 2 to 3 minutes after the 115 rumble strip was encountered. 116 117 Project Scope 118 Edge line rumble stripes may be a potentially effective strategy for keeping vehicles on 119 the roadway. However, the effectiveness of the strategy is not well documented. 120 Installing edge line rumble stripes on sections with narrow or no paved shoulders, while 121 listed in the National Cooperative Highway Research Program’s Report 500, Volume 6: A 122 Guide for Addressing Run-Off-Road Collisions, as a strategy to keep vehicles on the road, 123 is considered an experimental strategy (Neuman et al. 2003). As a result, additional 124 information on its effectiveness is necessary. 125 126 The research described in this paper evaluated the effectiveness of edge line rumble strips 127 in Iowa. The project identified and selected candidate locations on rural two-lane rural 128 roads with no paved shoulders. Rumble stripes, which are a combination of conventional 129 rumble strips with a painted edge line placed on the surface of the milled area, along the 130 edge of the travel lanes but at a narrow width to avoid possible intrusion into the normal 131 vehicle travel paths, were installed and evaluated. 132 133 Lane position was evaluated for several sites as a surrogate measure of safety since only a 134 short after period was available for a crash analysis. Pavement marking wear was also 135 qualitatively assessed. 136 137

TRB 2012 Annual Meeting Original paper submittal - not revised by author.

SITE SELECTION AND INSTALLATION 138 The Iowa DOT crash database was used to identify sections of two-lane paved roadways 139 with unpaved shoulders in 10 counties in Iowa that had a large number of ROR crashes. 140 The top 5% of all locations that had the largest number of ROR crashes were selected for 141 further analysis, resulting in a list of 11 initial sites. Individual sections varied in length 142 from less than 3 miles to more than 11 miles. The 10 counties were considered because 143 county engineers from the majority of these counties indicated interest in participating in 144 the project. 145 146 The sites were reviewed by an evaluation team and site visits were made to each of the 147 initial sites. Information such as pavement condition, relevant surrounding features 148 (intersections, roadside objects, etc.), presence of horizontal or vertical curves, etc. was 149 gathered during the site visits. A final list of feasible locations was selected based on 150 several factors. The first criterion was characteristics of the test site. Sections where a 151 major intersection, a railroad crossing, or some other feature was present that would have 152 made installation and evaluation difficult were removed from further analysis. The 153 second criterion was the pavement condition and type. Rumble stripes are easier to cut 154 into hot mix asphalt (HMA) than Portland cement concrete (PCC) pavement, however it 155 was also desired to have a site for each pavement type. Locations where the pavement 156 edge had any amount of deterioration were also not included because milling the rumble 157 strips in this situation may have further compromised the pavement quality. This resulted 158 in a list of seven feasible sites in six counties. 159 160 Once a final list was developed, sites were ranked based on agreed participation of local 161 agencies, crash history, site characteristics, and potential for improvement. Sites were 162 carefully selected to maximize potential effectiveness. The installation cost was based on 163 linear foot, and there was not enough funding to complete all of the projects selected. As 164 a result, the sites were ranked by number of crashes. Sites were selected from the list in 165 descending order until funds were expended. A total of five sites had rumble stripes 166 installed. One additional site was added to the list for evaluation. Linn County had 167 independently installed edge line rumble stripes on County Road E-16 under a separate 168 contract. They had used the same rumble stripe design, so the project was consistent with 169 those installed as part of this research. The final sites included: 170

County Road W-13 in Buchanan County (PCC pavement; project length: 171 7,400 linear feet; 840 vehicles per day {vpd}) 172

County Road P-53 in Dallas and Madison Counties (PCC pavement; project 173 length: 8,600 linear feet; 1,140 vpd) 174

County Road F-70 in Polk County (HMA pavement; was installed in 2 175 segments with a total length of 32,737 linear feet; 1,317 vpd) 176

County Road F-29 in Poweshiek County (HMA pavement; project length: 177 9,560 linear feet; 1,100 vpd) 178

County Road B-30 in Sioux County (HMA pavement; project length: 8,500 179 linear feet; 780 vpd) 180

County Road E-16 in Linn County (PCC pavement; project length: 5,500 181 linear feet; 1,000 vpd) 182

183

TRB 2012 Annual Meeting Original paper submittal - not revised by author.

The project involved milling narrow width (4–6 inch) rumble strips that were a maximum 184 of approximately 5/8 inch deep along the pavement edge of selected roadway sections. A 185 narrow width was necessary because none of the sections had paved shoulders. Following 186 placement of the rumble strips, painted edge lines were applied directly over the milled 187 areas, resulting in finished “rumble stripes.” 188 189 190 METHODOLOGY 191 The ideal method to evaluate the effectiveness of a countermeasure is to conduct a crash 192 analysis since reduction in crashes and crash severity is the objective of a safety treatment. 193 However, only 6 sites were tested and only a short after period (2 years) was available. 194 Additionally even though the locations were considered to be “high crash”, since the 195 majority of the sections were less than two miles in length and volumes were less than 196 1,500 vpd, the actual number of crashes was low. Consequently, due to the low sample 197 and crashes size it was difficult to conduct a statistically valid crash analysis. 198 199 As a result, lane position was used as a crash surrogate. Lane position or deviation from 200 the lane center has been used to assess the effectiveness of different lane departure 201 countermeasures such as wider edgelines (Donnel et al, 2006), post-mounted delineators 202 (Zador et al, 1987; Chrysler et al 2009), advance warning (Charlton, 2007), and rumble 203 strips or rumble stripes (Finley et al, 2009; Rasanen, 2005). Stimpson et al. (1977) 204 identified lateral placement and speed as the best indicators for assessing driver behavior 205 on horizontal curves. Use of lateral position, assumes that there is a correlation between 206 lane position and likelihood of running off the roadway. It is assumed that with edge line 207 rumble stripes, drivers will be more likely to lane keep. Several studies have used lane 208 keeping as a measure of effectiveness, including Porter et al. (2006), Pratt et al (2006), 209 Taylor et al. (2005), and TTI (2007). 210 211 In order to collect lateral position, the team followed a methodology used by Finley et al 212 (2009), Porter et al. (2006) and TTI (2007) that uses pneumatic road tubes set up in a “Z” 213 configuration, as shown in Figure 1. Using the time stamp of when each tire strikes a 214 particular road tube and geometric relationships, the distance that the vehicle is from the 215 edge of the roadway (Ox) can be determined. The tube configuration is set up so that the 216 right edge of the right lane line for tube 1 is the reference point (0,0). 217 218 219 220

TRB 2012 Annual Meeting Original paper submittal - not revised by author.

Figure 1: Layout configuration of road tubes to measure lateral displacement 221 222 Data were collected over several days before and then approximately 1 month after 223 installation of the edge line rumble stripes. Data were collected at one or two locations 224 at 4 sites (6 locations total). Data from the counters is output as raw data into a 225 spreadsheet with date, time (to the millisecond), and sensor (tube) number. Each row 226 represents one tire strike on a particular sensor and data are ordered sequentially. Ideally 227 tubes 1 and 2 receive one strike for each vehicle axle while tube 3 could receive one or 228 two strikes depending on whether the outside tire crosses tube 3. As a result, a particular 229 tube order indicates number of axles and direction. Other scenarios such as a vehicle 230 crossing the centerlane from the other direction, a vehicle crossing the tubes at an unusual 231 angle, and air backwash result in tube configurations that are illogical. This information 232 was used to remove erroneous data from the datasets. Vehicle speed was calculated 233 using time from when the first axle of tires strikes tube 1 and tubes 2: 234 235

v = (L1 + L2 + L3) (eq 1) 236 (t2 – t1) 237 238 When speed is known and the time the first tire strikes tube 3, the distance from tube 1 to 239 tube 3 can be calculated using: 240 241 dx = v (t3 – t1) (eq. 2) 242 243 The variable dx is different for each vehicle since the vehicle strikes tube 3 at a different 244 location depending on its distance from the edgeline. Using properties of similar 245 triangles, the lateral distance the wheel is from edge of the right lane (Oy) is calculated. 246 A 6-foot track width and known lane width were used to convert lateral position to reflect 247 distance of vehicle center in relationship to lane center. 248 249 Since data reduction was mostly a manual process and large trucks only made up a small 250 percentage of the traffic stream, only passenger cars were included in the analysis. 251 252

TRB 2012 Annual Meeting Original paper submittal - not revised by author.

RESULTS 253 Changes in mean speed, 85th percentile speed, and lateral lane position were calculated 254 and compared for the before and after period. It was not known if the treatment was 255 likely to affect vehicle speeds. During daytime conditions, driver may not notice the 256 rumble strips unless they stray near the lane edge. During nighttime and lower visibility 257 conditions, the rumble strip is expected to provide improved delineation. On one hand 258 this may cause drivers to slow since they are better able to judge the sharpness of the 259 curve and respond appropriately. On the other hand, better visibility may encourage 260 drivers to drive faster. As result, mean and 85th percentile speeds were evaluated to 261 determine if the is any evidence to suggest either might be true. Results are presented for 262 nighttime versus daytime conditions. Statistical significance of changes from the before 263 to after period were tested using a standard t-test. 264 265 Speed results are shown in Tables 1 and 2. As indicated, daytime mean speed increased 266 slightly in most cases from the before to after period but the changes were either small or 267 not statistically significant. A similar situation was observed for mean nighttime speeds. 268 Differences were smaller and in half of the cases, the differences were not statistically 269 significant. Eighty-fifth percentile speeds were also typically higher for the after period 270 for both daytime and nighttime with smaller differences occurring for the nighttime 271 conditions. It is not likely that the treatment influences speeds during daytime normal 272 conditions so the speed increases may be attributed to seasonal changes from the before 273 to after period. 274 275 Changes in the average position of vehicles in relationship to the center of the lane for 276 daytime conditions are shown numerically in Tables 1 and 2. Since lateral vehicle 277 position can be to the right or left of the lane center, the absolute difference in change 278 from the lane center was calculated. Average offset from the lane center decreased by 279 over 1 foot for both of the Vandalia sites studied. Average offset improved by 0.2 to 0.6 280 feet for the remaining sites except for P-53 where deviation from the lane center 281 increased by 0.4 feet. Differences were evaluated using a t-test and were all statistically 282 significant at the 95% level of significance. The information is shown graphically in 283 Figure 2. As noted wheel path is centered closer to the lane center in the after period for 284 all but one site. 285 286 Table 1: Speed and Position from Lane Center for Daytime Conditions 287

Location

Average Position from Lane Center (feet) Mean Speed (mph) 85th Percentile Speed (mph)

Before After Absolute

Difference Before After Difference Before After Difference

Vandalia 2 -1.40 1.10 0.30 50.6 50.7 0.1 57.7 57.1 -0.6

Vandalia 1 -1.70 0.60 1.10 41.7 44.9 3.2 49.4 51.2 1.8

F-29 -2.92 -1.37 1.55 52.8 57.3 4.5 59.3 64.6 5.3 W-13 South -1.70 -1.90 0.20 48.6 51.1 2.5 41.3 45.3 4.0 W-13 North -2.30 -1.70 0.60 51.1 51.8 0.7 56.8 58.7 1.9

P-53 -0.70 -1.10 -0.40 42.9 49.4 6.5 34.6 41.5 6.9 Values in bold are not statistically significant at the 5% level of significance 288

TRB 2012 Annual Meeting Original paper submittal - not revised by author.

289 Table 2 provides average position from the lane center for nighttime conditions. As 290 indicated, the vehicle wheel path moved closer to the lane center for all sites studied but 291 was not statistically significant at the 95% level of significance for the W-13 south and P-292 53 locations. The change was around 1.5 feet for three of the sites. Vehicle wheel path 293 in relationship to the lane center from the before to after period is shown graphically in 294 Figure 3. On average, improvement in lane position was higher for the nighttime period 295 than for the daytime period. 296 297 298 Figure 2: Speed and Position from Lane Center for Nighttime Conditions 299

Location

Average Position from Lane Center (feet) Mean Speed (mph)

85th Percentile Speed (mph)

Before After Absolute Difference Before After Difference Before After Difference

Vandalia 2 -2.2 0.7 1.5 50.4 50.6 0.2 56.5 54.0 -2.5

Vandalia 1 -2.0 0.4 1.6 42.0 43.4 1.4 49.6 49.4 -0.2

F-29 -3.2 -1.8 1.4 54.4 56.4 2.0 59.6 63.4 3.8 W-13 South -2.1 -2.2 0.1 50.5 51.7 1.2 50.5 51.7 1.2 W-13 North -2.8 -2.0 0.8 52.4 54.1 1.7 58.3 59.6 1.3

P-53 -1.3 -1.2 0.1 44.1 49.7 5.6 37.8 41.5 3.7 Values in bold are not statistically significant at the 5% level of significance 300 301

302

TRB 2012 Annual Meeting Original paper submittal - not revised by author.

303

Figure 2: Distance from Lane Center for Daytime Conditions 304 305

TRB 2012 Annual Meeting Original paper submittal - not revised by author.

306

Figure 3: Distance from Lane Center for Nighttime Conditions 307

TRB 2012 Annual Meeting Original paper submittal - not revised by author.

PAVEMENT MARKING WEAR 308 One of the main benefits of the rumble stripes is to provide additional visibility. It is 309 thought that the shape of the rumble stripe itself provides a raised (vertical) surface so 310 that the markings are more visible at night and particularly when some amount of 311 precipitation is on the pavement surface. Additionally, the depression protects part of the 312 pavement marking which can lead to improved wear. Consequently, part of the 313 evaluation was to monitor wear over time. Iowa receives a significant amount of snow 314 from December to February. Road maintenance in Iowa is aggressive and includes use of 315 salt and sand. As a result, winter maintenance is harsh on pavement markings. 316 317 Several sites were visited two years (May 2011) after application of the rumble stripes 318 and pavement markings were reviewed for wear. At all of the sites, a significant portion 319 of the regular pavement markings which were flush with the pavement surface had been 320 worn away by the snowplows. Much of the marking within the rumble stripe remained. 321 A qualitative assessment of the wear is shown in Figures 4 to 6. 322 323 Figures 4 and 5 show wear images for F-29 and Vandalia. Both are HMA pavement. As 324 noted most of the pavement marking on the regular surface has been worn away. Most of 325 the marking within the rumble stripe is still intact. Figure 6 provides images for P-53 326 which is PCC. Similar wear is noted. 327

328

TRB 2012 Annual Meeting Original paper submittal - not revised by author.

329

Figure 4: Wear at F-29 after 2 years (HMA) 330

331

TRB 2012 Annual Meeting Original paper submittal - not revised by author.

332

Figure 5: Wear at Vandalia After 2 years (HMA) 333 334

335

TRB 2012 Annual Meeting Original paper submittal - not revised by author.

336

Figure 6: Wear at P-53 After 2 years (PCC) 337

TRB 2012 Annual Meeting Original paper submittal - not revised by author.

One problem that was noted with the rumble stripes was accumulation of material (sand, 338 gravel, dirt) within the stripe as shown in Figure 7. The most likely cause is shoulder 339 maintenance. As shoulder material is pulled up it is likely that the pavement edge is 340 covered and the depression depends to trap the material. 341 342 343

Figure 7: Rumble Stripe Depression filled With Material

TRB 2012 Annual Meeting Original paper submittal - not revised by author.

344 SUMMARY 345 The research described in this paper evaluated the effectiveness of edge line rumble strips 346 in Iowa. Rumble stripes are a modified rumble strip placed along the edge of the travel 347 lanes on roadways with unpaved shoulders. The pavement marking is painted through 348 the rumble strip itself. Rumble stripes were installed at six sites on rural 2-lane roadways 349 with unpaved shoulders. 350 351 Lane position was evaluated before and after installation of edge line rumble stripes as a 352 surrogate measure of safety since only a short after period was available for a crash 353 analysis. Lateral position data were collected using a Z-configuration setup with road 354 tubes at 6 locations total. Data were collected over several days before and then 355 approximately 1 month after installation of the edge line rumble stripes. Lateral position 356 from the right roadway edge is measured using the road tubes. Distance from the lane 357 center was calculated using lateral position and lane width. Data were evaluated for day 358 and nighttime periods separately. 359 360 Average offset from the lane center decreased by over 1 foot for two locations during the 361 daytime period. Average offset decreased by 0.2 to 0.6 feet for three sites and increased 362 at one site by 0.4 feet. As indicated, the vehicle wheel path moved closer to the lane 363 center for all six sites for the nighttime period but was not statistically significant at the 364 95% level of significance for the W-13 south and P-53 locations. The change was 365 around 1.5 feet for three of the sites. On average, improvement in offset from the lane 366 center was higher for the nighttime period than for the daytime period. 367 368 Average and 85th percentile speeds were also compared from the before to after period. 369 In general both mean and 85th percentile speeds increased after installation of the 370 treatments. Average speed increases ranged from 0.1 to 6.5 mph and changes in 85th 371 percentile speeds increased from -0.6 to 6.9 mph. It is not know if speed increased due to 372 better delineation of the roadway or if the speed decrease was unrelated to the treatment. 373 374 Pavement marking wear was also qualitatively assessed at 2 years after installation of the 375 rumble stripes. Iowa winters are characterized by a number of snowfall events and 376 winter maintenance takes its toll on pavement markings. In general the pavement 377 markings which were flush with the pavement surface were highly worn while the 378 markings within the rumble stripes were largely intact. 379 380 381 REFERENCES 382 Porter et al. (2006), Pratt et al (2006), Taylor et al. (2005), and TTI (2007). 383 384 Charlton, S.G. and J.J. DePont. Curve Speed Management. Land Transport New Zealand 385 Research Report 323. July 2007. 386 387

TRB 2012 Annual Meeting Original paper submittal - not revised by author.

Chrysler, Susan T., Jon Re, Keith S. Knapp, Dillon S. Funkhouser, and Beverly T. Kuhn. 388 Driver Response to Delineation Treatments on Horizontal Curves on two-lane roads. 389 Texas Transportation Institute. FHWA/TX-09/-5772-1. May 2009. 390 391 Donnell, E.T., M.D. Gemar, and I. Cruzado. Operational Effects of Wide Edge Lines 392 Applied to Horizontal Curves on Two-Lane Rural Highways. Pennsylvania 393 Transportation Institute, University Park, PA, November 2006. 394 395 Finley, Melisa D., Dillon S. Funkhouser, and Marcus A. Brewer. Studies to Determine 396 the Operational Effects of Shoulder and Centerline Rumble Strips on Two-Lane 397 Undivided Roadways. Texas Transportation Institute. August 2009. 398 399 Rasanen, Mikko. Effects of a Rumble Strip Barrier Line on Lane Keeping in a Curve. 400 Accident Analysis and Prevention. Vol. 37. 2005. pp. 575-581. 401 402 Stimpson, W.A., H.W. McGee, W.K. Kittleson, and R.H. Ruddy. Field Evaluation of 403 Selected Delineation Treatments on Two-Lane Rural Highways. Report no. FHWA-RD-404 77-118, FHWA, U.S. Department of Transportation, Washington, D.C., 1977. 405 406 Zador, P, H.S. Stein, P. Wright, J. Hall. Effects of Chevrons, Post-Mounted Delineators, 407 and Raised Pavement Markers on Driver Behavior at Roadway Curves. Transportation 408 Research Record No. 1114. 1987. p. 1-10. 409 410 Porter, R.J., E.T. Donnel, and K.M. Mahoney. Evaluation of the Effects of Centerline 411 Rumble Stripes on Lateral Vehicle Placement and Speed. Journal of the Transportation 412 Research Record. No. 1862. Transportation Research Board, Washington, DC. 1995. 413 pp. 10-16. 414 415 Taylor, William, Ghassan Abu-Lebdeh, and Sachin Rai. Effect of Continuous Shoulder 416 Rumble Strips and Pavement Markings on Lateral Placement of Vehicles. TRR 1911. 417 2005. pp. 105-112. 418 419 420 ACKNOWLEDGMENTS 421 The authors would like to thank the Iowa Highway Research Board, the Federal Highway 422 Administration, Iowa Department of Transportation Office of Traffic and Safety, and the 423 Midwest Transportation Consortium for sponsoring a student to work on the project. We 424 would also like to thank the following county engineers who provided assistance and 425 follow-up: 426

Brian Keierleber, Buchanan County Engineer 427 Jim George, Dallas County Engineer 428 Todd Hagan, Madison County Engineer 429 Kurt Bailey, Polk County Engineer 430 Lyle Brehm, Poweshiek County Engineer 431 Doug Julius, Sioux County Engineer 432 Steve Gannon, Linn County Engineer 433

TRB 2012 Annual Meeting Original paper submittal - not revised by author.

434 The research team also thanks Scott Hanson of the Iowa DOT Office of Contracts for his 435 valuable advice, assistance, and patience during the development of the plans and 436 contract documents describing the installation of the rumble stripes. 437 438

TRB 2012 Annual Meeting Original paper submittal - not revised by author.