Michigan study finds that the most severe run-off crash is the drift-off and that rumble strip design and placement significantly reduce these crashes.

Single-vehiclerun-off-road crashes represent a large share of the Nation's crashes. The category is large in part because it covers such a broad range of roadway departures. Included are intentional departures to avoid other vehicles or objects; involuntary departures due to tire blowouts, ice, hydroplaning, or trailer sway; and driver errors such as distractions or drowsiness.

Within this category is a much smaller, more lethal crash type that is responsible for a disproportionate share of the severe run-off-road crashes. That lethal subcategory is the drift-off-road crash, defined as drivers who drift off the road due to drowsiness, inattention, or distraction. In the broader world of run-off crashes, these crashes are three-to-five times more severe than other run-off-road crashes, and they are highly correctible.

"Here in Michigan, we're focusing on reducing serious crashes, and countermeasures targeting the drift-off crashes are of prime importance in this effort," says James D. Culp, traffic and safety engineer, Michigan Department of Transportation (DOT). In fact, a recent study in Michigan indicates that a 40 to 80 percent reduction in drift-off crashes is possible with proper design and installation of rumble strips.

A Close Look at Drift-Offs

Michigan researchers analyzed 1,887 reported drift-off crashes to examine the nature of this type of freeway incident. In compiling the database, they only reviewed crashes during rainy and dry weather and excluded crashes on icy or snowy roads. The study produced a number of findings regarding drift-off crashes.

First, in Michigan, the wide range of crashes that are classified as run-off-road produce a crash data set of low to moderate severity. For the years 1996-2001, on 875 kilometers (544 miles) of freeway, only 3.4 percent of all wet road run-off crashes resulted in a severe outcome (fatal or incapacitating injury). Traffic friction run-off crashes—where the driver was changing lanes, passing, or avoiding another vehicle—resulted in severe outcomes 6.1 percent of the time. The corresponding severity percentage for winter weather run-off crashes (snowy, icy, slushy road) is 2.9 percent; for vehicle defect run-off crashes, 6.0 percent.

Drift-off crashes are extremely severe, in comparison to these other run-off crashes. For the same Michigan roadways, where rumble strips were not present on the shoulders, 17 percent of drift-off crashes included at least one fatal or incapacitating injury. Even where shoulder rumble strips are present, 12 percent of drift-off crashes result in fatal or incapacitating injury.

The only crash groups that can get close to the severity of freeway drift-offs (12 to 17 percent severe outcome) are the behavioral groups (risky driving) usually addressed by nonengineering safety agencies. Even these crash groups, many of which are the subject of continuing national and local safety efforts, fall short when compared on the severity scale. In 1999 Michigan data, for instance, approximately 14 percent of crashes involving nonseatbelt usage resulted in death or incapacitating injury; alcohol 13 percent; red light running 7 percent; and speeding 6 percent.

Meet the Drift-Off Driver

In a majority of the Michigan drift-off crashes, the driver stated to the officer that he or she was distracted, drowsy, "must have been asleep," "looking at a map," or "can't remember what happened." These statements enabled the corresponding crash reports to be classified specifically as either "drowsy" or "distracted."

Of these drift-off crashes with known causes, 82 percent of the drifting drivers were drowsy or asleep. Driver distraction, though topical in the safety community, is only a small percentage of the freeway drift-off crash problem.

Because of the national interest in drivers using cell phones, researchers further reviewed the 55 distracted crashes that occurred during 2000 and 2001 to identify the cause of the distraction. Cell phone usage in this study accounted for only 6 of these crashes (11 percent), most of which involved physical interaction with the phone, not the distraction of conversation.

The most common distraction in this group of crashes was drivers looking for, handling, or reaching for something, such as a CD, tape, or radio (9 crashes); food and drink (8); children in the back seat (3); pets (3); cigarettes or lighter (2).

In the overall study, 66 percent of all drift vehicle drivers were male, which is consistent with the overall Michigan crash data: 58 percent of all crash drivers and 73 percent of all fatal crash drivers are male.

An age profile of drift-off drivers indicates a trend similar to all Michigan crashes: high crash involvement of younger drivers, with declines in crash involvement as age progresses. Researchers found a slight shift toward the younger driver and noted no particular difference in the percentage of elderly drivers involved in drift-off-road crashes versus all crashes.

About one-fifth of the crashes in the Michigan drift-off database involved a driver under the influence of alcohol or drugs at the time of the crash. The driver action in many of these crashes was slightly erratic or excessively slow, due often to drowsiness. In this study, crashes related to alcohol had comparable severity levels to those not related to alcohol.

Trucks and buses are worth noting, first, because they are generally driven by professional drivers and, second, because they have large tires susceptible to vibration by milled rumble strips, but not by the other rumble designs used in Michigan. The Michigan study defined trucks as vehicles with a Gross Vehicle Weight Rating (GVWR) over 4,540 kilograms (10,000 pounds). The crashes reviewed included only one bus—a school bus. Although truck traffic accounted for 11 percent of total vehicle-miles driven in this study, only 4.4 percent of the drift-off vehicles were trucks. This finding supports the common assertion that truck drivers are indeed more alert and drive more professionally than the average driver.

Time of Day and Day of Week

Throughout the entire 24-hour day, the number of drift-off crashes remains fairly level. A slight peak occurs in the early morning hours and again in the early afternoon. Both of these time periods correspond to the cycle of sleepiness-alertness set by the human body's circadian rhythm.

If the numbers were adjusted to reflect that a large majority of travel is during the daytime, however, the rate of drift-off crashes would be seen to be "sky-high" for the early morning hours and very low for daytime travel. The interpretation can be expressed this way: If an agency manages a section of freeway, expect a drift-off crash at any time. If a son or daughter is driving back to college, 1:00 a.m. is a very dangerous time to make that trip.

Thirty-eight percent of the crashes in the Michigan study occurred on Saturdays and Sundays—about one-third higher than would be expected if the crashes were to occur randomly throughout the week. By comparison, only 25 percent of all Michigan freeway mainline (non-drift-off) crashes occur on a Saturday or Sunday.

"Drivers making longer trips, such as those occurring on weekends, are more susceptible to drowsiness and drift-off crashes," says Culp.

What Can Happen to a Drift-Off Vehicle?

The researchers analyzed the Michigan crash reports to determine the side of the road from which the vehicles first exited the roadway. They defined "exiting the roadway" as touching a wheel beyond the paved shoulder, except where a drift vehicle actually contacted a parked vehicle while still on the shoulder.

Nearly half of all drift-off vehicles exited in each direction (53 percent right, 47 percent left)—defying the common perception that this type of incident is primarily a right-side crash. Even more illuminating were the 289 crashes that resulted in death or incapacitating injury: 53 percent of those drift vehicles exited to the left.

In Michigan, most of the paved right shoulders are 2.7-3 meters (9-10 feet) wide, and most of the paved left shoulders are 1.2 meters (4 feet) wide. Also, most Michigan freeways are crowned at or near the middle of the pavement. Because the crown is the high point in a cross-section of pavement, an "unsteered" (or drifting) tire will flow in a direction away from the crown.

That crowning may partially explain the equality in drift direction. However, the implication for road agencies remains the same: When applying a rumble strip countermeasure, treating the left shoulder is as important as treating the right.

What Percent of Run-off Crashes Are Drift-offs? Drift-offs are the most pertinent crash type when considering rumble strip treatment, but the identification of these crashes usually requires individual review of each suspected crash report. The larger category of run-off crashes often can be reviewed through computer screening, which is much less time-consuming than case-by-case reviews. So the question is, "Is it reasonable to evaluate the subset (drift-off) by looking at the total set (run-off)?" The answer is, "You have to be very careful." In the Michigan study, large variation was noted in the ratio of drift-off to run-off for individual freeway sections. The percent of drift-offs varied widely from 2 to 57 percent, depending not on the drift-off crashes themselves, but on the number and nature of the other run-off crashes. For instance, 48 percent of all run-off crashes reviewed were due to winter weather. For the most part, these snow and ice crashes are run-off, but not drift-off, crashes. The presence of these crashes drives down the drift-off/run-off ratio fairly evenly on all the Michigan sections. Other factors drive down the ratio on a less uniform basis. Polished or rutted bituminous pavement can produce hydroplaning or standing water crashes that will reduce the drift-off/run-off ratio—but only on pavement sections that exhibit that wear. Run-off crashes due to traffic friction are relatively few on lower-volume roads, but as the average daily traffic reaches 40,000 and above, these crashes (lane change, passing, swerve to avoid) tend to dominate the run-off category—again driving down the drift-off percentage for those roads. What should the individual State agency expect on its roads? When weather-related crashes are removed from the Michigan data, a modified drift-off/run-off ratio ranges from 40 to 71 percent for the rural study sections with traffic from 5,000 to 11,000 vehicles. On road sections with average daily traffic of 40,000 to 100,000, the modified drift-off/run-off ratio ranged from 5 to 40 percent.

What Is Hit?

The Michigan study found that the objects hit by drift-off vehicles included parked vehicles on the shoulder (4 percent), trees (13 percent), signposts and light poles (6 percent), guardrails and bridge rails (19 percent), and cross-medians affecting an opposite-direction vehicle (1 percent).

The data are significant not for what was hit, but for what was not hit: 47 percent of the 1,887 drift-off vehicles hit no fixed object at all. The drivers of these vehicles, once they dropped a wheel off the paved shoulder, simply had trouble steering the vehicle to safety.

A few of these vehicles hit the backside of a ditch, but most simply came to harm trying to negotiate the side slope. On all of the roads in this section of the Michigan study, the side slopes are American Association of State Highway and Transportation Officials (AASHTO) standard (4-to-1) or better.

Additional analysis tells the story: Of the 1,805 vehicles in the study that drifted beyond the paved shoulder, 35 percent of the drivers overcorrected their steering. Forty-five percent of these same 1,805 vehicles rolled over during the crash. Simply dropping the wheel off the paved shoulder was enough to start the process for many of these vehicles. In this study, 68 percent of the drivers who oversteered continued on to roll over.

Drift-off Crash Rate—A Moving Target

The researchers used 6 years of crashes on 762 kilometers (473 miles) of nonrumbled Michigan freeway to investigate the effect of traffic volume on drift-off crashes. Although the number of drift-off crashes showed little correlation with traffic volume, the rate of drift-off crashes is extremely sensitive. With increasing traffic volume, the rate of drift-off crashes decreases markedly. On nonrumbled road sections, a doubling in average daily traffic is likely to decrease the drift-off crash rate by 30 to 50 percent.

Most likely, two circumstances are driving the drift-off crash rate lower as traffic increases. First, higher traffic volumes require a higher level of alertness just for the routine task of driving. Second, a drowsy or distracted driver may indeed cause an accident on a high-volume road, but it may become a multivehicle crash that would not have been detected in the drift-off crash screening for the Michigan study.

Curves Linked to Crash Rate

A closer look revealed an unexpected piece of information about curves. When the researchers grouped the data according to road curvature (percent of road in horizontal curvature), this created a set of trendlines showing a pattern. In each successive group of data, as the percent of road in curvature increases, so does the crash rate trendline. In other words, on the Michigan freeways, more horizontal curvature is associated with higher drift-off crash rate.

"It's obvious that when drivers become drowsy, they are more prone to drift off the road on curves," says Culp. "The sharper the curve, the less time a driver has to correct the drift-off, and the crash rate goes up."

This explanation would be consistent with the concept of microsleeps, in which the brain stops processing information, even though the eyes may still be open. According to the Federal Motor Carrier Safety Administration, during a "microsleep" lasting 4 or 5 seconds, a car can travel 91.5 meters (100 yards), which is plenty of time to cause a serious crash.

Rumble Strip Comparisons

Three types of rumble strips are used on Michigan freeways: two older designs (rolled-in and concrete intermittent rumbles) and the current milled-in rumble design. The researchers matched the 1,887 Michigan crash reports to road sections of known average daily traffic and known rumble strip design, including sections with no rumbles. The drift-off crash data review corresponded to 32 billion vehicle-miles of travel.

The advantage of the milled rumble strip over the earlier designs is that the milled cross-section is designed to allow vehicle tires to partially drop into them. This effect provides a vibration to the vehicle that translates up to the steering wheel. Whereas the rolled and concrete intermittent designs can provide some outside noise to alert a drifting driver, the milled design produces a louder noise and adds a vehicle vibration that most certainly increases the potential for alerting the drowsy or distracted driver.

 

The Michigan crash data bear out this logical progression—that more noise and more vehicle vibration lead to increased effectiveness. Milled rumble strips in Michigan reduced drift-off crashes by 40 percent, through the entire range of traffic volumes studied. The researchers calculated that, in the Michigan data, the two older designs—rolled-in and concrete intermittent rumbles—were approximately 20 percent effective in reducing drift-off crashes.

Although Michigan attributes a 40 percent reduction in drift-off crashes to the milled rumble strips, engineers with the Pennsylvania Turnpike Commission, New York State DOT, and New York State Thruway Authority reported crash reductions of 60 to 80 percent after installation of milled-in rumble strips.

What accounts for the comparatively lower reduction in drift-off crashes in Michigan? Possibly, the lateral location of the rumble strips makes a difference. The Michigan DOT installed its rumbles at 300-millimeter (12-inch) and 600-millimeter (24-inch) offset to the pavement edge line, whereas the Pennsylvania Turnpike, for instance, installed its rumbles at 100-millimeter (4-inch) offset. New York State DOT installs its freeway rumbles at 250 millimeter (10-inch) and 100 millimeter (4-inch) offsets.This suggests that the closer the rumble is to the edge line, the more effective it might be in reducing crashes.

If Rumble Strips Were King

"As an engineering agency, we are naturally somewhat conservative and therefore cautious about new ideas," says John Friend, director of the Bureau of Highways, Michigan DOT, "but several of our staff stepped forward and said, 'We've got a problem, and if rumble strips can help, why not try them?'"

Indeed, if rumble strips are that effective, and the drift-off crashes so severe, safety advocates across the county might be expected to apply this safety feature aggressively.

In some quarters, they are. Aggressive usage can start with choice of where the rumble is located. Across the country, shoulder rumble strips are placed at anywhere from 50 millimeters (2 inches) to 750 millimeters (30 inches) offset from the edge of the travel lane, as road authorities balance the needs of safety, maintenance, and pavement integrity. Yet the effectiveness of a rumble strip may be related to how early in the process a drift-off driver can be alerted. The Pennsylvania Turnpike Commission, who pioneered the milled-in rumble strip in the early 1990s, continues to place the strip at 100 millimeters (4 inches) offset to the travel lane. Pennsylvania is joined in this practice by several other road authorities nationwide that are especially safety sensitive.

Some agencies mill on imperfect pavement. If the agency's overwhelming goal is to provide route-long rumble strip vibration to the traveling public, then some imperfect pavement condition is simply tolerated. The highway profession is accustomed to the question, "Can our pavement hold up if we install rumbles here?" Safety advocates must be sure that the following question also is addressed: "What is the safety consequence of not providing a rumble strip on this roadway, both this year and for the foreseeable future?"

As for milling over existing older rumble designs, research by several States has established that milled rumbles provide increased vibration and noise compared to the older rumble designs. So the milled rumbles should be more effective at alerting drifting drivers and preventing crashes. And, according to the recent Michigan data, they are more effective. It appears that the milled rumble strips in Michigan, even at their wide offset from the travel lane, would be likely to reduce drift-off crashes by an additional 25 percent over existing rolled or concrete intermittent designs.

Painting the Rumbles

Several States, starting with Mississippi and Pennsylvania, have applied pavement markings on top of their shoulder rumble strips. As part of this package treatment, the rumble strips are placed closer to the travel lane, a safety benefit.

In addition, the benefits to the pavement marking itself are proving to be phenomenal. In Michigan, a test marking placed on a shoulder rumble strip was compared to the existing edge line placed about the same time but on flat pavement. On a dark rainy night, when the existing marking was nearly invisible, the rumbled line was reported to have lit up "like a runway."

As an added bonus, Michigan is finding that moving the edge line from the travel lane onto the shoulder appears to protect the edge line from the greatest force of the snowplow. A test line came through the winter looking like new.

"People have to drive in adverse weather conditions and at all times of the year," says Jill Morena, pavement marking engineer, Michigan DOT. "The combination rumble strip/paint line will provide guidance in those situations since it can be seen, heard, and felt."

Expansion to Nonfreeway Usage

Shoulder rumble strips originally were designed to address drift-off crashes on long, monotonous stretches of road. Due perhaps to the availability of paved shoulders, many road authorities came to think of shoulder rumbles as strictly a freeway treatment, even though some portion of their nonfreeway system had adequate shoulders available to accommodate rumbles.

Not as common, but gaining ground, is the practice of providing shoulder rumbles on nonfreeways. Kansas, Minnesota, Oklahoma, and Pennsylvania DOTs are four agencies that have established a department policy for shoulder rumble strips on nonfreeways, when certain conditions of traffic volume and paved shoulder are met.

At least 16 States have placed centerline rumble strips in the center of 2- or 4-lane undivided roadways. Pennsylvania systematically places centerline rumble strips, according to road type and average daily traffic. Four States—California, Colorado, Oregon, and Washington—have had centerline rumble strips in place long enough to evaluate this safety feature, and they all report a substantial reduction in crossover crashes, and even larger reductions in severe crossover crashes.

Making Intelligent Compromises

The older, less vibratory shoulder rumble designs did not pose a significant problem for bicyclists in terms of stability, but the new milled rumbles are more vibratory by design and present a greater challenge to riders. Providing recurring gaps in the otherwise continuous rumble strip line is an intelligent way to provide for bicycle movement across the rumble strip without reducing the effectiveness of the safety feature. Other methods, such as reducing the depth or width of the rumble strip to be less vibratory, have the undesirable side effect of providing less vibration to alert motorists.

One factor that often inhibits the cutting of a rumble strip into existing shoulders is the deteriorating condition of the shoulder itself. To address this situation, the Michigan DOT developed a compromise design that can be used on partially deteriorated shoulders. The "mill and chip seal combo" adds safety by milling the rumbles, then increases shoulder stability by topping the entire shoulder with a layer of chip seal, which conforms to the shape of the rumbles. The net result: Shoulder life is extended, and a vital safety feature is provided on a road that otherwise might have gone without rumbles.

Finally, as with anything else, accurate field measurement is needed to ensure that any designed rumble pattern is properly milled into the pavement. When all the pieces are put together, the result is a blueprint for an aggressive safety effort. In the words of Michigan DOT's Jill Morena, "The way to get the most benefit from a rumble strip is to mill it, move it, paint it, and measure it."

---

David A. Morenais the safety and traffic operations engineer at the FHWA Michigan Division Office. Morena led the research effort reported in this article.

For more information, contact David Morena at 517-702-1836.

---

References

1. Federal Highway Administration Rumble Strip Web site, . Accessed October 22, 2002.http://safety.fhwa.dot.gov/roadway_dept/pavement/rumble_strips/
2. Hickey, John J. Jr. Shoulder Rumble Strip Effectiveness: Drift-Off-Road Accident Reductions On The Pennsylvania Turnpike. In Transportation Research Record 1573, TRB, National Research Council, Washington, DC, 1997, pp. 105-109.
3. Griffith, M. Safety Evaluation of Rolled-In Continuous Shoulder Rumble Strips Installed on Freeways. In Transportation Research Record 1665, TRB, National Research Council, Washington, DC, 1997, pp. 28-34.
4. Cheng, E., Gonzalez, E. and M. Christensen. Application and Evaluation of Rumble Strips on Highways, Utah Department of Transportation, 1994.
5. Chen, S. et al. Optimal (Milled) Continuous Shoulder Rumble Strips and the Effects on Highway Safety and Economy, Virginia Department of Transportation, April 2001.
6. Perrillo, K. The Effectiveness and Use of Continuous Shoulder Rumble Strips, FHWA, Albany, NY, August 1998.
7. Safe-Strips (Safety Shoulder Rumble Strips) NYSDOT Program, New York State Department of Transportation, April 1998.
8. Roadway Shoulder Rumble Strips. Technical Advisory T5040.35. FHWA, USDOT, Dec 20, 2001.
9. Hassan, D. Demonstration of Milled Rumble Strips on Asphalt Shoulders in the State of Kansas, FHWA, Topeka, KS, December 23, 1999.
10. MICRO-SLEEPS. In Parents Against Tired Truckers Newsletter, Issue #40, P.A.T.T., Washington, DC, 2002.