Guidelines on the Use of Auxiliary Through Lanes at Signalized Intersections (2011)

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.

Page 29 4. SAFETY ATLs are primarily implemented as an operational treatment, but ATLs certainly have safety implications . ATLs may be expected to have a fewer number of some types of crashes when compared to a conventional intersection handling the same volume , because ATLs allow for smoother, less congested operations . On the other hand, ATLs may also cause an increase in ot her types of crashes due to the added merging area . This chapter explores these safety trade - offs based on empirical crash data and simulation - based safety models. Various design elements were observed in an analysis of crash data from 16 sites across four U.S. states over a period of 9 years . These 16 sites were also included in the data used to develop the operational models in Chapter 3 . Crash reports from within the ATL and its tapers were collected from the responsible ag encies in each of the four states . Rear - end and sideswipe crashes were the crash types thought to be most closely related to ATL operation . Overall, the average reported frequency of rear - end and sideswipe crashes was 4.5 crashes per year per site . This is a relatively low frequency when compared to intersections generally identified as potentially hazardous in safety studies . This relatively low frequency indicates that these sites were probably not unsafe as designed. This research also employed the FHWA Surrogate Safety Assessment Model (SSAM) ( 12 ), which can be used in conjunction with microsimulation programs like VISSIM ( 13 ) to record simulated traffic conflicts . SSAM has the potential to allow practitioners to quantitatively examine the safety consequ ences of an alternative like an ATL , even if no crash prediction model is available. Although the analysis objective of this research was to correlate SSAM conflicts with the crash data taken from the 16 study sites, the crash sample size was ultimately to o low to draw any significant conclusions . However, the trends in the SSAM conflict output allowed the researchers to identify several design elements that may affect an ATL’s safety . Appendix A contains guidance on how analysts could use SSAM to help exam ine the safety of an ATL in the future . SAFETY PRINCIPLES ATLs add lane - changing activity to the through - movement lanes at a signalized intersection . T his activity may lead to an increase in sideswipe crashes , especially near the downstream merge . At the s ame time, a n increased through - movement capacity may prevent some rear - end crashes on the approach by decreasing congestion . In particular, the following ATL elements are critical to its safe operation: • Downstream length . A sufficient downstream ATL length and taper is needed to allow for safe merging operation into the adjacent CTL traffic stream by providing drivers with enough distance to accelerate and find acceptable gap s in the CTL traffic . • Access control . Driveways along an ATL create potential hazards for drivers who are preoccupied with merging into the adjacent CTL traffic

Page 30 stream . During the data collection for this project, driveway - related conflicts were identified in the field . Right - turning vehicles from a shared ATL create a sim ilar hazard. • Sight distance . Sites with an adequate view of the downstream ATL from the stop bar experienced more ATL use, presumably because drivers feel more comfortable using an ATL when they can see the entire downstream merge area . In addition, with a n adequate view of the end of the ATL , drivers in the ATL can plan for their merge back into the CTL more carefully . • Queuing downstream of the ATL merge . Traffic spilling back into the ATL taper from a downstream bottleneck could create a safety issue . Ana lysts should pay particular attention to potential spillback into an upstream ATL that could occur from a downstream bottleneck. • Taper design . The length and rate of the ATL taper should conform to AASHTO ( 1 ) and MUTCD policy ( 3 ). • Signing, marking, and li ghting . An ATL should be clearly signed as a through - movement lane so that drivers are not discouraged from using it . Lighting may also promo te better nighttime operations. OBSERVED SAFETY PERFORMANCE The 16 ATL study sites produced a combined average of 4 .5 related (sideswipe plus rear - end) crashes per year on the ATL , indicating that these sites were likely not unsafe as designed . Although this research could not do so, it might be possible in the future to develop a crash modification factor (CMF) to con vert a conventional intersection approach to one with an ATL . It might also be possible to use crash prediction models from the Highway Safety Manual ( 4 ), calibrated for a particular state, to estimate the number of crashes that would have occurred at a particular site if the ATL had not been installed . Until the data are available to estimate a CMF or calibrate a crash prediction model, the best interpretation of the available evidence is that ATLs at the studied sites did not seem to add many crashes .

Page 31 Proportion of ATL Crashes Although a crash reconstruction analysis was not within the scope of this research, it is generally true that the rear - end and sideswipe crashes that are the types most likely to be related to ATLs are not typically as severe as other crash types such as angle, hea d - on, and run - off - road crashes. Exhibit 4 - 1 displays a breakdown of the field crash d ata obtained for all 16 sites by crash type. The total number of crashes reported at all 16 sites was 1,0 50 — this amounts to approximately eight crashes per site per year , including both related and non - related ATL crashes . Although the majority of crashes (52 percent ) were rear - end crashes, only 10 percent were sideswipe crashes, which might be expected to be higher in ATLs . Exhibit 4 - 2 displays a summary of the crash data collected from each site. Exhibit 4-1 Breakdown of ATL Crash Types 52% 10% 13% 11% 4% 1% 9% Rear End Sideswipe Turning Angle Fixed Object Backing Other

Page 32 Approach Number of Years Analyzed Rear End Crashes Sideswipe Crashes Total Crashes SB MD-2 at Arnold Rd * 9 57 13 112 NB MD-2 at Arnold Rd * 9 45 6 99 SB La Canada Dr at Magee Rd 9 42 5 54 EB NC-54 at Fayetteville Rd 6 41 12 207 WB Walker Rd at Murray Blvd 6 34 2 45 NB La Canada Dr at Orange Grove Rd 9 33 6 44 WB Magee Rd at La Canada Dr 9 29 5 48 EB Walker Rd at 185th St 9 28 1 63 WB Walker Rd at 185th St 9 27 2 58 EB Magee Rd at La Canada Dr 9 27 3 35 SB La Canada Dr at Orange Grove Rd 9 24 2 32 EB Walker Rd at Murray Blvd 6 23 4 34 NB Garrett Rd at Old Chapel Hill Rd 6 20 12 115 SB Sunset Lake Dr at Holly Springs Rd 6 17 12 33 NB La Canada Dr at Magee Rd 9 15 0 22 SB Garrett Rd at Old Chapel Hill Rd 6 12 4 49 Total 126 474 89 1050 * Denotes 2-CTL approach Although, as noted previously , calibrated crash prediction models from the HSM were not available for the four states analyzed in this effort, the researchers employed uncalibrated models to make comparisons on the proportions of cra sh types observed . Exhibit 4 - 3 shows the p roportion of sideswipe crashes among all related crashes (sideswipe plus rear - end) for uncalibrated HSM crash models and the 16 ATL sites based on a summary of crash records . Th e e xhibit shows that the proportions generally matched well . Z - tests for proportions revealed that only the proportion from the North Carolina data had a significant difference from the HSM prediction at a 95 percent confidence level . For all other states , individually and combined , the diffe rence between the HSM prediction and the project data was not statistically significant . Exhibit 4 - 3 lends support to the idea that the crash types experienced at the ATL sites studied were not much different from crash types experienced at comparable conv entional intersections. Proportion of Sideswipe among All Related Crashes State HSM ATL Data Arizona 0.15 0.12 Maryland 0.11 0.16 North Carolina 0.14 0.28 Oregon 0.13 0.06 Combined 0.13 0.15 Distribution of Crashes Relative to Location in ATL The rear - end and sideswipe crash data were aggregated by relative location within the ATL, as shown in Exhibit 4 - 4 . The line for total crashes is simply the sum of rear - end and sideswipe crashes . Note that the distribution of sideswipe crashes is spread more evenly over the length of a typical ATL than th e distribution of rear - end crashes . This suggests that , while rear - end crashes usually occur in the queuing areas near the intersection, sideswipe crashes are Exhibit 4-2 Summary of Crash Data Exhibit 4-3 Comparison of Sideswipe Crash Data

Page 33 more likely to occur in other areas of the ATL . Also note that almost exactly half of these crash es were upstream of the intersection and half were downstream. Relationship Between Crashes and Congestion Exhibit 4 - 5 plots the number of rear - end crashes from 2006 to 2008 against the maximum X T obtained from field data collected in 2009 and 2010 . X T indicat es the level of congestion in the through - movement lanes assuming no ATL is present . The line in Exhibit 4 - 5 is the best - fit linear relationship between the maximum X T observed and rear - end crash frequency for each of the 16 sites . Only the most recent 3 years of crash data were used in order to shorten the time period between safety and operational data collection , considering that all of the operational data were obt ained in 2009 and 2010. Exhibit 4-4 Field Crash Data Distribution versus Relative ATL Position 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 -1 -0.8 -0.6 -0.4 -0.2 0 0.2 0.4 0.6 0.8 Cu m ul at iv e Pr ob ab ili ty Relative Position within Upstream / Downstream ATL Total Rear End Sideswipe 1

Page 34 As shown in Exhibit 4 - 5 , the relationship between rear- end crashes and congestion, as represented by XT, is very weakly correlated, with little observable trend above X T = 0.8. Not surprisingly, the number of rear- e nd crashes appears to be less frequent when congestion level s are very low compared to the remaining data set. Exhibit 4 - 6 displays the trend between rear - end crashes and average ATL flow observed in the field for each of the 16 sites . This exhibit does not indicate that more crashes occur at ATLs with higher flow rates — consequently, it does not provide evidence that a well - utilized AT L is less safe than a poorly utilized ATL . T he two influential points in the far right portion of the exhibit with very high ATL flow are the sites with two CTLs. Exhibit 4-5 2006–2008 Crash Data Trends versus Maximum XT Observed from Data Exhibit 4-6 2006–2008 Rear-End Crashes versus Average ATL Hourly Flow Observed from Data R² = 0.1292 2 0 5 10 15 20 25 30 0 0.2 0.4 0.6 0.8 1 1.2 1.4 R ea r- En d Co lli si on s Max XT Observed in Field (15-min Interval) y = 11.423x0.6118 0 5 10 15 20 25 30 0 50 100 150 200 250 300 350 400 ATL Hourly Flow (vph) R ea r- En d Co lli si on s

Page 35 Relationship between ATL Crashes and Total ATL Length Exhibit 4 - 7 compares sideswipe crashes (combined over all analysis years in the dataset, unlike the preceding exhibits ) at each site to total ATL length (sum of upstream ATL length, intersection width, and downstream ATL length not including tapers) . While i t may be hypothesized that longer ATLs allow for safer merging, it is also possible that more exposure to merging areas would lead to more frequent si deswipe crashes . Based on the direction of the linear relationship shown in Exhibit 4 - 7 (again the best - fit line), it appears that the probability of sideswipe crashes increases as the length of the ATL increases . In summary, the analysis of the data from the 16 study sites showed some re lationship s between rear - end crashes and congestion , between rear - end crashes and flow rates in the ATLs, and between sideswipe cra shes and ATL length . However, t he relationships are weak and causation is unclear in all cases , so practitioners should not over - interpret the findings . In the future , perhaps calibrated crash prediction models for ATLs will be available to provide firmer guidance to practitioners considering ATLs. SAFETY EVALUATION CONSIDERATIONS The following guidance is recommended for conducting a safety evaluation of an existing ATL: • Collect crash data for the ATL approach and remove non - ATL - related crashes. • C losely examine rear - end and sideswipe crashes along the approach with the ATL, including the tapers . • C ollect crash data over as long a t ime as possible given that important safety - related conditions remained unchanged. Exhibit 4-7 2006–2008 Rear-End Sideswipe Crashes/Year versus ATL Total Length Si de sw ip e Co lli si on s y= 0.0043x - 0.0532 R2 = 0.3744 0 2 4 6 8 10 12 14 0 500 1000 1500 2000 2500 3000 Upstream + Downstream ATL Length (ft)

Page 36 • Exercise caution in observing when the ATL was constructed and not include data from prior to the ATL opening . • U s e a method similar to the safety analysis of conventional intersections as described in the HSM ( 4 ) to evaluate the crash data . • Review t he crash data from the 16 sites examined in this chapter to understand how typical ATLs perform. Evaluating the safety implications of ATL proposals or designs is currently dif ficult given the lack of crash prediction models or CMFs . Until those tools are available, practitioners should be confident that, based on the data presented in this chapter, well - designed ATLs are not likely to cause safety problems . A n SSAM analysis could also be used when a practitioner wishes to examine the potential safety effects of building an ATL or altering an ATL’s design or operational elements . As Appendix A describes, a n SSAM analysis requires a calibrated microsimulation mod el of the intersection that exhibits an appropriate estimate of the flow in the ATL . Ten or more simulation runs should be used to populate the sample size . During SSAM analysis of the trajectory files, only rear - end and lane - change conflicts should be exa mined, and a time - to - crash (TTC) threshold of 1.5 seconds is preferred to yield a larger sample size . The practitioner should then look for the relative change in conflict frequency when a design element (e.g., downstream length) or operational element (e. g., X T ) is altered to draw conclusions about the safety effects of the ATL design.