Evaluation of Transverse Rumble Strips at Stop Controlled Intersections Raghavan Srinivasan Jongdae Baek Forrest Council UNC Highway Safety Research Center
Why Evaluate this Treatment? Some studies have shown a reduction in approach speeds NCHRP Synthesis 191 (1993) found reductions in crashes in 10 before-after studies None of the studies discussed site selection (possibility of regression to the mean) Control sites were not used Only two studies accounted for differences in traffic volumes before and after the installation Evaluation using more recent methods will be helpful
Location of Rumble Strips in Minnesota Minimum of three sets of rumble strips on approaches to stop controlled intersections (can have maximum of five sets) First set located 250 feet (75 m) before the Stop Ahead sign The set closest to the intersection is about 500 feet (150 m) from the Stop Sign The set in the middle located 15 feet (4.5 m) before the set that is closest to the intersection
Rumble Strips in Minnesota
Location of Rumble Strips in Iowa Until 2006, three sets of rumble strips were required First set located 200 feet (60 m) before the Stop Ahead sign The set closest to the intersection is 300 feet (90 m) from the Stop sign The set in the middle is midway between the two
Rumble Strips in Iowa
Transverse Rumble Strip in Iowa
Data from Minnesota 20 intersections (9 4-leg and 11 3-leg) Reference sites identified from HSIS intersection database Traffic volume Lighting Number of lanes on the major road 0.58 crashes per intersection per year
Data from Iowa 134 (49 3-leg and 85 4-leg) stop controlled intersections where rumble strips were installed Unfortunately records on installations at high-volume locations were thrown out! Comparison/reference sites without rumble strips 177 4-leg and 101 3-leg Data from 1987 to 2005 0.34 crashes per intersection per year
Approach Empirical Bayes before-after study Estimate the expected crashes in the after period had rumble strips not been implemented Compare this with actual crashes in the after period Developed Safety Performance Functions (SPFs) using reference group data Negative binomial regression Major and Minor road AADT Tried to account for trends Annual factors
Results from Iowa (3 and 4 leg combined) Actual crashes in the after period EB-Expected Crashes without treatment Crash Modification Factor (S.E.) Total 300 273.0 1.095 (0.088) Crashes 128 128.5 0.991 (0.111) (KABC) 66 82.8 0.793 (0.113)* (KAB) (KA) 20 28.8 0.689 (0.165)* Ran stop sign 21 25.5 0.818 (0.191) **Statistically significant at the 5% significance level *Statistically significant at the 10% significance level
Results from Minnesota (3 and 4 leg combined) Actual crashes in the after period **Statistically significant at the 5% significance level *Statistically significant at the 10% significance level EB-Expected Crashes without treatment Crash Modification Factor (S.E.) Total 110 92.6 1.158 (0.212) Crashes 53 53.6 0.939 (0.238) (KABC) 20 26.0 0.716 (0.235) (KAB) (KA) 3 8.6 0.311 (0.187)** Right Angle 47 37.6 1.182 (0.316)
Combined Results from Iowa and Minnesota (3 and 4 leg combined) Actual crashes in the after period EB-Expected Crashes without treatment Crash Modification Factor (S.E.) Total Crashes 410 365.6 1.118 (0.086) 181 182.2 0.987 (0.109) (KABC) 86 108.8 0.785 (0.107)** (KAB) (KA) 23 37.4 0.608 (0.140)** **Statistically significant at the 5% significance level * Statistically significant at the 10% significance level
Economic Analysis of Crash Harm Crash Type Actual crashes in the after period EB-Expected Crashes without treatment Crash increase or decrease Change in crash cost (crash harm) Change in crash cost per intersection per year PDO 228 190.8 37.2 $290,160 $290 (C) (KAB) (KA) 95 73.4 21.6 $1,070,258 $1,068 86 108.8-22.8 -$8,056,585 -$8,041 23 37.4-14.4 -$9,544,565 -$9,526
Conclusions More severe injury crashes seem to have reduced significantly 21% reduction in KAB crashes and 39% reduction in KA crashes Reduction in crash harm of between $6,600 to $8,100 per intersection per year Increase in total crashes which implies an increase in PDO crashes the reason for this is not clear No definitive results regarding ran stop sign and right angle crashes These are typically installed on low volume roads. Need a larger sample for more definitive results
Further Information Results presented at the 2010 TRB Annual Meeting Acknowledgments Michael Pawlovich and Tom Welch at Iowa DOT Loren Hill and others from Minnesota DOT Yusuf Mohamedshah at Lendis Corporation