| dc.description.abstract | Run-off-road (ROR) crashes are usually severe and account for a significant amount of highway
fatalities. It is estimated that more than 50% of the fatal crashes in the United States and in
Alabama are ROR crashes. Rural roads are the main reason of concern; approximately 62% of
fatal crashes in Alabama occur on rural roads. The frequent occurrence of run-off-road (ROR)
crashes and the corresponding high severity of them, especially on rural roads, led to several
states starting new practices or update existing policies as a tentative to improve safety. Some
common countermeasures include paved shoulders, shoulder rumble strips, and shoulder rumble
stripes. This study developed a survey regarding state agencies’ policies, studies of treatment
effectiveness, and dimensions of paved shoulders, shoulder rumble strips, and shoulder rumble
stripes in the United States. Of all states in the country, 20 completed the survey. Dates of the
most recent policies vary significantly from one location to the other, as well as dimensions.
Only a few states developed safety effectiveness studies, but data was not sufficient, as crash
data after treatment implementation were not available, or the methods were not the most
recommended for these evaluations. This study recommends that, after states have enough
“after” crash data, statistical methods, especially the ones suggested by the Highway Safety
Manual (HSM), should be applied to verify if treatments are effective and appropriate to reduce
crashes at each state. The Alabama Department of Transportation (ALDOT) implemented a
policy in 2006 to widen pavements and install milled-in rumble strips or stripes when rural two-lane
highways with less than 28 ft of pavement width are resurfaced. In practice, this policy was
also extended to four-lane divided rural roads. This study considered data from 101 resurfacing
projects in Alabama that had 2 to 4 ft of full-depth paved shoulders added, and in some cases,
rumble strips or rumble stripes were scored into the pavement within the shoulder. This study to
address the effectiveness of the treatments implemented in Alabama after ALDOT policy
considered an EPDO analysis to express changes in crash severity, an application of the
Empirical Bayes method to quantify changes in crash frequency, and a benefit-cost analysis. For
two-lane roads, the EPDO analysis showed a reduction of EPDO scores of 3.78% for the
combined effect of paved shoulders and shoulder rumble strips, 3.51% for the combined effect of
paved shoulders and shoulder rumble stripes, and 10.67% for paved shoulder only. For four-lane
roads, there was a reduction of EPDO scores of 11.10% for the combined effect of paved
shoulders and shoulder rumble strips and a reduction of 4.01% for paved shoulder only. The EB
analysis was initially performed using the HSM safety performance function. For two-lane rural
roads, the analysis resulted in CMFs of 0.79, 0.82, and 0.72 for the combined effect of paved
shoulder and shoulder rumble strips, the combined effect of paved shoulder and shoulder rumble
stripes, and paved shoulder only respectively. For four-lane roads, CMF for the combined effect
of paved shoulder and shoulder rumble strips was 0.84, but for paved-shoulder only it was not
significant and reliable. The EB analysis was also performed applying state-specifics SPFs, but
results were not consistent with typical research findings or the literature. Benefit-cost ratios on
two-lane roads were 4.57, 3.61, and 5.76 for the combined effect of paved shoulder and shoulder
rumble strips, the combined effect of paved shoulder and shoulder rumble stripes, and paved
shoulder only respectively. On four-lane roads, B/C ratio was 4.56 for the combined effect of
paved shoulder and shoulder rumble strips and 2.03 for the paved shoulder-only treatment. This
study applied Poisson, zero-inflated Poisson (ZIP), negative binomial (NB), and zero-inflated
negative binomial (ZINB) distributions to estimate run-off-road crash frequency on 72 control
segments of two-lane rural roads and 32 control segments of four-lane divided rural roads in
Alabama, during the period from 2001 to 2010. After evaluating all distributions and comparing
the value of the Vuong’s test statistic, AIC, and BIC the ZINB was the best to model crash data
for both two and four-lane rural roads. | en_US |
