Wrong-Way Driving Solutions, Policy, and Guidelines (2023)

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.

5 Chapter 5: Effects of Geometric Design Elements and Access Control Techniques Previous studies and available federal and state guidelines identified specific types of interchanges and some geometric design elements at ramp terminals that may contribute to WWD (Atiquzzaman et al., 2018, 2020, 2022; Chang et al., 2022; Hosseini et al., 2021a; Jalayer et al., 2021, 2018a, 2018b, 2017, 2016; Zhang et al., 2019). For instance, a study funded by IDOT reported that parclo interchanges, trumpet interchanges, and half- and full-diamond interchanges are more likely to cause driver confusion than other interchange types (Zhou et al., 2015). Several past studies investigated the effect of geometric design elements on the probability of WWD crashes, such as control radius from the crossroad, type of median on the crossroad, limited visibility of on-ramps, left-side off-ramps, and one-way streets connected with freeways, etc. (Pour-Rouholamin et al., 2016; Conner et al., 2004; Copelan, 1989). Recent studies also found that an intersection balance of less than 60% can provide better sight distance and reduce the risk of WWD (Wang et al., 2018; Wang et al., 2017). Off-ramp terminals with access points within 600 ft have a higher chance of WWD crashes (Pour-Rouholamin et al., 2016, 2015; Das et al., 2018). There are also some national and state guidelines for geometric design to deter WWD at interchange areas. The AASHTO Green Book suggests that channelization of three-leg intersections is often desirable for discouraging WWD (AASHTO, 2011). The IDOT design manual recommends that the median width between the two-way ramps should be more than 50 ft (IDOT, 2010). However, there is a scarcity of research evaluating the effectiveness of geometric design elements and access control techniques in reducing WWD events. The main reason is that WWD crashes are sporadic and random; collecting WWD crash data before and after the implementation of design elements can be challenging. To fill this gap, this project evaluated the effects of different interchange types on WWD crashes using the weighted score method. Additionally, the effects of geometric design elements on WWD incidents at parclo interchange terminals were evaluated using case studies and the Multiple Correspondence Analysis (MCA) method. Detailed analysis results are summarized in the following sections, including (1) a comparative analysis of different interchange types in four states (Alabama, Florida, Illinois, and Michigan) to identify which interchange types are relatively more prone to WWD; (2) a comparative analysis of ramp terminal design practices in three different states (Alabama, Florida, Illinois) to determine which geometric design elements can effectively deter WWD at the parclo ramp terminals; (3) case studies on effects of geometric design elements on WWD crashes or incidents in different states; (4) an MCA analysis of the effects of a combination of design features on WWD incidents; and (5) an analysis of WWD incident videos to identify key design features that might cause WWE at parclo interchanges. 5.1 Comparative Analysis of Different Types of Interchanges Past studies suggest that certain interchange types are relatively more prone to WWD crashes than others (Zhou et al., 2017a, Zhou & Pour-Rouholamin, 2014a). Agencies often used a weighted score to find the importance of different interchange types in their contribution to WWD. A 40 

weighted score is typically calculated as the percentage of WWD crashes originating from a particular type of interchange over the portion of that type of interchange in the state. Therefore, a higher weighted score indicates that the respective interchange type is relatively more prone to WWD. In this project, the team collected the WWD crash distribution at different interchanges in the four states (Alabama, Florida, Illinois, and Michigan). Table 5-1 shows the weighted scores for various interchange types. Results indicate that urban (modified, compressed) diamond, parclo, freeway feeder, and trumpet interchanges generally have a higher weighted score. Although a diamond interchange has the most significant proportion of interchanges in these states, it typically has a weighted score of less than one. Parclo interchanges account for the second-largest portion but have a relatively higher weighted score (1.2-2.9), with Florida being an exception (a weighted score of 0.9). Table 5-1. Weighted Scores for Different Interchange Types in Four States Interchange Type Statewide Distribution WWD Crash Distribution Weighted Score AL (Zhou et al., 2017) Diamond 63.3% 58.3% 0.9 Modified Diamond 1.6% 12.5% 7.6 Partial Cloverleaf 13.4% 20.8% 1.5 Cloverleaf 1.6% 0.0% 0 Rest Area 6.2% 0.0% 0 Directional 5.2% 0.0% 0 Trumpet 2.3% 0.0% 0 Freeway Feeder 6.2% 8.3% 1.3 FL (Kittelson & Associates, Inc. 2015) Diamond/Partial Diamond 55.7% 30.8% 0.6 Partial Cloverleaf 25.5% 23.1% 0.9 Trumpet 6.0% 15.4% 2.6 Direct Connection Design 5.7% 0.0% 0 Y Intersection 3.0% 7.7% 2.6 Other 4.1% 23.1% 5.6 IL (Zhou et al., 2012) Diamond 44.3% 44.30% 1 Compressed Diamond 6.0% 17.10% 2.9 Modified Diamond 8.1% 7.10% 0.9 Partial Cloverleaf 10.5% 12.90% 1.2 Cloverleaf 7.9% 5.70% 0.7 Rest Area 8.50% 2.90% 0.3 Semi-Directional 2.50% 0.00% 0 Directional 3.20% 1.40% 0.4 SPUI 1.10% 1.40% 1.3 Trumpet 3.30% 0.00% 0 Freeway Feeder 4.00% 7.10% 1.8 Unknown 0.50% 0.00% 0 MI (Morena and Leix, 2012) Directional 26.0% 5.70% 0.2 Partial Cloverleaf 20.6% 60.00% 2.9 Modified Diamond 19.5% 8.60% 0.4 Diamond 17.20% 0.00% 0 Urban Diamond 6.30% 8.60% 1.4 Trumpet 2.90% 11.40% 3.9 Full Cloverleaf 2.50% 2.90% 1.1 Others 4.90% 2.90% 0.6 41 

To better understand why parclo interchanges in Florida have relatively lower weighted scores, an additional comparative analysis was conducted to compare the design practices of parclo interchange terminals between Florida and Illinois/Alabama. Detailed analysis results are included in Section 5.2. Michigan was not included in this comparative analysis because MDOT made significant improvements to all parclo interchanges within its jurisdiction in 2012 due to its high weighted scores. Therefore, existing designs may not represent past conditions when the respective locations experienced crashes. 5.2 Comparative Analysis of Parclo Interchange Ramp Terminal Design Practices To understand the best practices of geometric design by different states, design features of 38, 78, and 114 off-ramp terminals of parclo interchanges in Alabama, Florida, and Illinois, respectively, were collected and analyzed. The geometric design features included in the analysis were median types and extension, channelizing island, left-turn radius from the crossroad, median width between ramps, distance to the nearest access point, and intersection balance. Chi-square tests were conducted to determine if the geometric design features are statistically different among these three states. Table 5-2. Comparison of Geometric Design Characteristics for Parclo Interchanges Alabama Florida Illinois Chi-square Test Geometric Features Total obs.= 38 Total obs.= 78 Total obs.= 114 χ2 df P-value Freq. % Freq. % Freq. % Median on Crossroad 29.65 2 <0.01 Traversable 19 50% 6 8% 20 18% Non-traversable 19 50% 72 92% 94 82% Median on Crossroad Extended Yes 3 8% 40 51% 38 33% 21.44 2 <0.01 No 35 92% 38 49% 76 67% Channelizing Island (off-ramp) None 3 8% 6 8% 6 5% 3.28 4 0.51 Traversable 4 11% 3 4% 5 4% Non-traversable 31 82% 69 88% 103 90% Left-Turn Radius from Crossroad < 60 ft 15 39% 7 9% 10 9% 61–80 ft 7 18% 28 36% 64 56% 38.47 6 <0.01 81–100 ft 10 26% 30 38% 32 28% > 100 ft 6 16% 13 17% 8 7% Width of the Median between Ramps < 30 ft 24 63% 43 55% 47 41% <0.01 20.70 4 30–50 ft 6 16% 26 33% 60 53% > 50 ft 8 21% 9 12% 7 6% Distance from Nearest Access Point < 300 ft 8 21% 30 38% 10 9% 301–600 ft 10 26% 13 17% 27 24% 25.88 6 <0.01 601–900 ft 10 26% 18 23% 44 39% > 900 ft 10 26% 17 22% 33 29% Intersection Balance Less than 60% 35 92% 70 90% 99 87% 0.92 2 0.63 More than 60% 3 8% 8 10% 15 13% 42 

The results in Table 5-2 suggest that five of the seven study design features are significantly different among the three states, including non-traversable median, median extensions, left- turning radius from the crossroad, median width, and distance to the nearest access points. Two key design features that may have made parclo interchanges in Florida less prone to WWD are non-traversable median and median extensions. A detailed explanation of the various design features is provided in the following paragraphs. A non-traversable median on a crossroad is an effective design feature to reduce WW entries by making the left-turning WW maneuver difficult (AASHTO, 2011). Florida has a 92% of parclo interchanges with non-traversable medians on crossroads, which is much higher than 50% and 82% in the other two states. This difference was also found to be statistically significant at 99% confidence (χ2 =29.649, p < 0.001). A median extension on a crossroad is one of the most effective features to prevent WW left-turn movements. A median extension on a crossroad can reduce the turning radius and prevent an early left turn. Data analysis indicated that more than 50% of medians on crossroads in Florida extend into the intersection, which is more than 8% and 33% in the two other states. The significance test results suggest a statistically significant difference at 99% confidence in using median extensions among the three states. Using a non-traversable channelizing island reduces the width of an off-ramp throat and decreases the occurrence of WWD (WSDOT, 2013; Pour-Rouholamin et al., 2016b). For all three study states, more than 80% of off-ramps implemented a non-traversable channelizing island. The statistical significance test showed no significant difference among the three states on channelizing islands. The IDOT design manual recommends a left-turn radius with a maximum of 80 ft from a crossroad to an on-ramp to deter WWD (IDOT, 2010). The significance test found that distributions of left-turn radius were statistically significant based on the Chi-square test results (χ2 =38.76, p < 0.001). However, this result does not directly support the maximum 80 ft radius, as the number of lanes on the crossroad can significantly affect the left-turn radius. Among the study states, Florida has more parclo interchange terminals with multilane crossroads; as a result, the radius at the Florida locations was relatively higher than Alabama and Illinois. The minimum width of a median between the on- and off-ramps is recommended to be at least 50 ft (IDOT, 2010). Most off-ramp terminals in the three study states do not meet this design criterion. The Chi-square test showed that the width of medians between on- and off-ramps is significantly different (χ2 =20.696, p < 0.001) among the three states. However, this result does not directly support the 50 ft. minimum width. Illinois has the lowest percentage of locations with access points near ramp terminals(less than 300 ft). Overall, the difference in distance to the nearest access point among the three states was found to be statistically significant (χ2 =25.884, p = 0.006827). Florida has more interchanges in urban areas, contributing to a more significant percentage of access points closer to the ramp terminals. 43 

According to the recommendation provided by WSDOT, the intersection balance should be less than 60% at off-ramp terminals (WSDOT, 2013). When intersection balance is less than 60%, left-turning drivers will have a better view of the on-ramp from the crossroad, which could reduce the possibility of WWD caused by inadequate sight distance. As shown in Table 5-2, only 10%, 13%, and 8% of off-ramp terminals in Florida, Illinois, and Alabama did not meet this requirement. Overall, this difference was not found to be statistically significant (χ2 =0.91666, p = 0.6323) among the three states. 5.2.1 Case Example of Using Median Extension in Florida The off-ramp at westbound (WB) I-4 and N Alexander St. in Florida was found to have a high number of WWD incidents. To overcome this problem, longitudinal channelizing devices were applied to extend the median on the crossroad to prevent WWD, as shown in Figure 5-1. A wigwag flashing beacon was installed around the WW signs to warn WW drivers. Figure 5-1. Median extension at N Alexander St and I-4 WB (Source: H. Zhou) Table 5-3 summarizes before-and-after statistics of WWD incidents at this location. WWD incidents were reduced from an average of 16 per month to less than 1 per month after the median extension. The WWD turnaround rate was used to evaluate the effectiveness of the wigwag flashing beacons enhanced WW signs. The turnaround rate (86%) in the after period was slightly higher than the before period (84%). It should be noted that there should have been more random WWD incidents by impaired drivers in the after period because the improvement in median extension likely reduced most recurring incidents. 44 

Table 5-3. Before-and-After WWD Statistics at I-4 and N Alexander St. Before Period (5 mos) After Period (26 mos) #Turned Turn- #Turned Turn- Month Year #WWD Around around Month Year #WWD Around around Rate Rate Oct 2015 14 13 93% Mar 2016 1 1 100% Nov 2015 18 16 89% Apr 2016 1 1 100% Dec 2015 12 11 92% May 2016–Feb 2018 Jan 2016 18 15 83% Mar 2018 5 3 60% Feb 2016 15 10 67% Apr 2018 1 1 100% Avg. 15.4 13.0 Avg. 0.8 0.7 84% 86% Total 77 65 Total 22 19 5.2.2 Case Example for Median Barrier Extension Design FHWA and MDOT analyzed 110 WWD crashes on the Michigan freeway system for five years, 2005–2009, to determine contributing factors in WWD crashes in Michigan. Findings from the study confirmed that the parclo interchanges were more prone to WWD. Of the 35 crashes in Michigan for which the WWE points were known, 21 occurred at parclo interchange terminals. The interchange of I-94 at Gratiot Avenue in Detroit, a parclo interchange, contributed to 10 of the 21 WWD crashes. Unlike WWD crashes at other locations in Michigan, the crashes at this interchange were not primarily at night and did not involve impaired drivers. Thus, MDOT conducted a small-scale road safety audit at both ramp terminals to determine the deficiencies at this parclo interchange. The audit group visited the interchange ramps and found that the median guardrail extended nearly entirely to the curb lane. As shown in Figure 5-2, the extended median guardrail blocks the left-turn driver's view of the on-ramp and distorts the view of the median. Without a clear view of the on-ramp, drivers may have confused the off-ramp with an on-ramp, leading to the wrong direction of travel. MDOT removed some portion of the guardrail in 2014 to ensure a clear view of the on-ramp to fix the problem. Figure 5-2. Street view of ramp terminal at I-94 and Gratiot Ave in Detroit (Source: Morena et al., 2012) 45 

5.2.3 Case Examples for Median Barrier Traversability Design Case studies at different ramp terminals found that the traversability of the median barrier between the on- and off-ramp of parclo interchange terminals played a vital role in turnaround time for WWD incidents. As shown in Figure 5-3, a wide median with grass between on- and off-ramps may prevent WW drivers from directly switching from the off-ramp to the on-ramp when they realize they are driving the WW. Most were found to make U-turns or back up to the crossroad. When turning around on the off-ramp, some large trucks hit traffic signs on channelized islands or roadsides. Figure 5-3. Ramp terminal at I-65 exit 208 SB in Clanton, AL (Source: Q. Chang) WW drivers used median openings on the two-way ramps with a traversable median for self- correction at the two parclo interchange terminals. Case study results found it can reduce the turnaround time for WW drivers. Figure 5-4 shows an example of a median opening before the WW sign used for self-correction. Figure 5-5 shows a location with low-mounted concrete barriers between the on- and off-ramps. It was found that many drivers ran over the barrier to make corrective maneuvers after entering WW. Some states provide median openings on two- way ramps with a non-traversable median for maintenance purposes. 46 

Figure 5-4. NB ramp terminal at I-81 exit 141, VA (Source: Google Earth) Figure 5-5. SB ramp terminal at I-85 exit 147 in Commerce, GA (Source: Google Earth) 47 

5.3 Effects of Geometric Design Features on the Occurrence of WWD Incidents The multiple Correspondence Analysis (MCA) method was applied to analyze the effects of different combinations of geometric design features on the occurrence of WWD incidents based on the WWD incident data collected at 75 parclo off-ramp terminals from 13 states. The geometric design features and WW-related TCDs at the study locations were collected using Google Maps’ aerial and street views. A graphical representation of the data collected for each site is shown in Figure 5-6. Based on past studies and guidelines, a total of 20 geometric designs and TCDs features were selected, as they were reported to affect the occurrence of WWD incidents. These 20 features were classified into five categories: (1) Crossroad Design, (2) Off- Ramp Design, (3) Intersection Design, (4) Traffic Sign Condition, and (5) Pavement Marking Condition. Each category includes one or more features, as shown in Figure 5-6. Descriptive statistics of the collected data are presented in Table 5-4, and a brief description of each feature is provided in the following paragraphs. Figure 5-6. Geometric design features and WW-related TCDs at the study locations Crossroad Design 1. Median traversable – if a vehicle can pass through the median on the crossroad (e.g., double yellow lines or concrete barrier). 2. Median extended into the intersection – as shown in Figure 5-6(2), if the median on the crossroad has fully/half/not extended to cover the off-ramp. 3. Number of lanes on the ramp side – how many lanes exist on the ramp side direction for a two-way crossroad? 48 

Off-ramp Design 4. Number of lanes on off-ramp – how many lanes exist on the off-ramp? 5. Ramp type – as shown in Figure 5-6(5), the on- and off-ramps for a parclo interchange terminal can be located at the left or right side of the crossroad; as a result, the off-ramp will be looped or curved. A looped off-ramp may cause more confusion for left-turning drivers from the crossroad because it is oriented in the same direction as where the driver intends to go. 6. Channelizing island traversable – is the channelizing island painted or raised? 7. Length of channelizing island – how long is the side of the channelizing island parallel with the crossroad? This length influences the connection angle between the exclusive right-turn lane and the crossroad. 8. Median traversable – can a vehicle pass through the median on the ramp (e.g., double yellow lines or concrete barrier)? 9. Median width – how wide is the median on the ramp? Intersection Design 10. Control type –signalized, all-way stop-controlled, or uncontrolled. 11. Intersection balance – the ratio between the distance from the stop bar of the left-turn lane on the crossroad to the centerline of the median between the ramps and the entire length of the intersection. 12. Intersection angle – angle between the centerline of the off-ramp and the centerline of the crossroad. 13. Corner radius – turning radius from the left-turn lane on the crossroad to the on-ramp. 14. Distance to the nearest access point – distance from the nearest access point to the center of the ramp terminal. 15. Lighting – if streetlights are available at an intersection or if any other lighting source is available nearby. 16. Off-ramp volume higher – if the off-ramp has a higher traffic volume than the on-ramp Traffic Sign 17. Enhanced sign – if signs are low-mounted, enlarged, having retroreflective tape on the sign pole, or an extra guidance sign exists; having none was recorded as level 0, one or two as level 1, and three or four as level 2. Pavement Marking 18. Condition of pavement marking – well-maintained and visible or faded? 19. Left-turning guidance – are left-turn skip strips available to guide left-turning vehicles from the crossroad to the on-ramp? 20. Enhanced pavement marking – are there any enhancements to pavement markings (e.g., rumble strips, raised pavement markers, or bolder pavement markings)? 49 

Table 5-4. Summary of Variables and Categories All Locations No WWD WWD Features Categories (75) Locations (47) Locations (28) (Code Name) (Coded Value) Count % Count % Count % Median Traversable Non-traversable (0) 14 19% 10 21% 4 14% (CRMediantran) Traversable (1) 61 81% 37 79% 24 86% Crossroad Design If the median Half Covered (0) 26 35% 17 36% 9 32% extended to cover Fully covered (1) 22 29% 21 45% 1 4% off-ramp (CRmedianext) Not covered (2) 27 36% 9 19% 18 64% Number of lanes on 1 (1) 48 64% 27 57% 21 75% the ramp side 2 (2) 20 27% 16 34% 4 14% (CRlaneonrampside) 3 or more (3) 7 9% 4 9% 3 11% 50 

All Locations No WWD WWD Features Categories (75) Locations (47) Locations (28) (Code Name) (Coded Value) Count % Count % Count % Number of lanes on 1 (1) 60 80% 37 79% 23 82% off-ramp 2 (2) 6 8% 4 9% 2 7% (ERnumlane) 3 or more (3) 9 12% 6 13% 3 11% Ramp looped? No (0) 30 40% 21 45% 9 32% (ERLooped) Yes (1) 45 60% 26 55% 19 68% Channelizing island No (0) 32 43% 19 40% 13 46% traversable Yes (1) 13 17% 9 19% 4 14% Off-ramp Design (ERChantran) None (2) 30 40% 19 40% 11 39% Length of None (0) 30 40% 19 40% 11 39% channelizing Less than 50 ft (1) 17 23% 7 15% 10 36% island 51–100 ft (2) 19 25% 13 28% 6 21% (ERChanLen) More than 100 ft (3) 9 12% 8 17% 1 4% Median traversable No (0) 64 85% 39 83% 25 89% (ERMEDTran) Yes (1) 11 15% 8 17% 3 11% 0-30 (0) 56 75% 39 83% 17 61% Median width 31–60 (1) 13 17% 7 15% 6 21% (ERmedwid) More than 60 (2) 6 8% 1 2% 5 18% Control type Uncontrolled (0) 50 67% 30 64% 20 71% (INType) Controlled (1) 25 33% 17 36% 8 29% Less than 40% (0) 22 29% 16 34% 6 21% Intersection balance 40–60% (1) 47 63% 28 60% 19 68% (INIB) More than 60% (2) 6 8% 3 6% 3 11% Intersection Right (0) 62 83% 38 81% 24 86% angle Obtuse (1) 7 9% 6 13% 1 4% (INANGLE) Intersection Design Acute (2) 6 8% 3 6% 3 11% Less than 50 ft. (0) 39 52% 23 49% 16 57% Corner radius 51–100 ft (1) 31 41% 19 40% 12 43% (INRAD) More than 100 ft. (2) 5 7% 5 11% 0 0% Distance to nearest 0–50 ft (0) 24 32% 11 23% 13 46% access point 51–500 ft (1) 39 52% 26 55% 13 46% (NEARDIS) More than 500 ft (2) 12 16% 10 21% 2 7% Lighting No (0) 33 44% 17 36% 16 57% (NEARLIG) Yes (1) 42 56% 30 64% 12 43% Off-ramp volume No (0) 56 75% 31 66% 25 89% higher than on-ramp Yes (1) 19 25% 16 34% 3 11% (Ratio) Enhanced sign Level 0 (0) 23 31% 9 19% 14 50% Sig (SignEnhanced) Level 1 (1) 29 39% 17 36% 12 43% 51 

Level 2 (2) 23 31% 21 45% 2 7% Pavement condition Not good (0) 5 7% 1 2% 4 14% (PAVCOND) Good (1) 70 93% 46 98% 24 86% Pavement Left-turn guidance No (0) 56 75% 37 79% 19 68% (PAVLEFT) Yes (1) 19 25% 10 21% 9 32% Enhanced No (0) 67 89% 40 85% 27 96% (PAVENHANCE) Yes (1) 8 11% 7 15% 1 4% MCA is a dimension reduction method that aims to plot all variable categories onto lower- dimensional space (e.g., 2-D dimension plot). Each variable category is represented as a point on the plot. A shorter distance between two points indicates a stronger correlation between two variables. The MCA analysis was conducted to explore which design feature combinations tend to increase the risk of WWD incidents. The variable categories that had a significant contribution in the first two dimensions are presented in Figure 5-7. Two additional points, “WWD_1” and “WWD_0,” indicated whether the WWD incidents occurred at the ramp terminal or not (where 0 told that there are no recurring WWD incidents, and 1 means there are recurring WWD incidents). Clouds close to the “WWD_1” suggested that these design features will be more likely to correlate with WWD incidents. On the contrary, clouds close to “WWD_0” indicated that these design features would be less likely to be associated with WWD incidents. Three clouds near the “WWD_0” were identified, implying that the likelihood of recurring WWD incidents is relatively low when they exist at the same location. Cloud 1 means that the ramp terminals containing three or four enhanced signs are less likely to have WWD incidents even when there are no channelized islands at the off-ramp. Cloud 2 confirmed that large signalized or stop-controlled intersections are less likely to have WWD incidents. Cloud 3 suggests that non-traversable median on the crossroad and lighting on 4-lane crossroads can reduce the risk of WWD incidents. The variables in Clouds 1-3 for “WWD_0” are as follows: • Cloud 1 − Length of channelized island is zero (ERChanLen_0) − No channelized island at off-ramp (ERChantran_2) − Sign enhanced level 2 contains three or four types of enhanced signs (SignEnhanced_2) • Cloud 2 − Number of lanes on off-ramp is 3 or more (ERnumlane_3) − Number of lanes on crossroad at the ramp side is 3 or more (CRLaneonrampside_3) − Number of lanes on off-ramp is 2 (ERnumlane_2) • Cloud 3 − Intersection corner radius between 5–100 ft (INRAD_1) − Controlled intersection type at ramp terminal (INType_1) − Traversable channelized island on off-ramp (ERChantran_1) − Non-traversable median on crossroad (CRMediantran_0) − Number of lanes on crossroad at ramp side is 2 (CRLaneonrampside_2) 52 

− Lighting around ramp terminal (NEARLIG_1) Similarly, there are three clouds near the “WWD_1”, which indicates that if the geometric design features included in these clouds appeared at one parclo interchange terminal, it might tend to have recurring WWD incidents. Cloud 1 shows that the uncontrolled ramp terminal intersections with no street lighting on the two-lane crossroad are more likely to have WWD incidents. According to Cloud 2, poor pavement marking & signing conditions, a wide median between the on- and off-ramps, and an acute angle between the off-ramp and crossroad are likely to increase the risk of WWD incidents. Cloud 3 indicates that traffic signs that do not meet the minimum requirements of MUTCD will increase the risk of WWD incidents. The variables in Clouds 1-3 for “WWD_1” are as follows: • Cloud 1 − Uncontrolled intersection type at ramp terminal (INType_0) − Number of lanes on crossroad is at ramp side 1 (CRLaneonrampside_1) − Intersection corner radius is less than 50 ft (INRAD_0) − No lighting around ramp terminal (NEARLIG_0) • Cloud 2 − Pavement markings are not in good condition (PAVCOND_0) − Median width between on- and off-ramps is more than 60 ft (ERMedwid_2) − Acute angle between off-ramp and the crossroad (INANGLE_2) • Cloud 3 − Traffic signs do not meet minimum requirements of MUTCD (SignEnhanced_0) − Non-traversable channelized island at off-ramp terminal (ERChantran_0) − Length of channelized island on off-ramp is 40–100 ft (ERChanLen_2) 53 

Figure 5-7. Effects of design features on WWD incidents at parclo interchange terminals 54 

5.4 Effects of Parclo Interchange Configurations on WWD WWD incident data collected at 75 parclo interchange terminals were further analyzed to determine if the configuration of parclo interchange ramps impacts WWD. Two configurations were studied—circular loop off-ramp and circular loop on-ramp. Of the 75 monitored parclo interchange terminals, 45 (60%) have circular loop off-ramps, and 30 (40%) have circular loop on-ramps. The incident data showed that a much higher percentage of loop off-ramp terminals had WWD incidents than loop on-ramp terminals. As shown in Table 5-5, a total of 28 terminals out of 75 locations had at least one WWD incident per three-day period; 19 (68%) occurred at the circular loop off-ramp terminals. Additionally, among the 17 parclo interchange terminals with more than one WWD incident per day, 14 (82%) had circular loop off-ramps; overall, 68% of circular off-ramp terminals were found to have more than one WWD incident compared with 32% of circular on-ramp terminals. Table 5-5. WWD Incident Frequency at Two Types of Parclo Interchange Ramp Circular On-ramp Circular Off-ramp Total Frequency % Frequency % At least 1 WWD/3 days 9 32% 19 68% 28 At least 3 WWD/3 days 3 18% 14 82% 17 All monitored locations 30 40% 45 60% 75 Based on current geometric design guidelines, there are three types of parclo interchange design—Parclo A, Parclo B, and Parclo AB. As illustrated in Figure 5-8, the horizontal red lines represent the crossroad, and the vertical double green lines represent the freeway. The letters A, B, or AB represent the design types of loop ramps. Blue dots indicate the loop off-ramp terminals: • Parclo A has only loop on-ramp • Parclo B has only loop off-ramp • Parclo AB has both loop on-ramp and loop off-ramp In the AASHTO Green Book (7th edition) (AASHTO, 2011), the parclo A and parclo B designs are described as “major road exits on the near side” and “major road exits on the far side,” respectively. As shown in Figure 5-9, this naming policy is based on whether the ramps will be constructed near or far side to the oncoming highway drivers. Based on the WWD incident analysis, 68% of WWD incidents occurred at the terminals with circular loop off-ramps, which appear in parclo B and parclo AB interchanges. Note that this terminal design is often combined with a half-diamond interchange on the other bound. Among the 28 locations with recurring WWD incidents, 19 were parclo B or parclo AB. 55 

Figure 5-8. Parclo interchange types (Source: Zhang, 2012) Figure 5-9. Parclo A and Parclo B designs (AASHTO, 2011) Comparing parclo B terminals with loop off-ramps and traditional diamond interchanges can help explain why this type of interchange terminal is more prone to WWD. As shown in Figure 5-10, the same left turn from the crossroad should be made to enter the freeway at these two types of terminals. Due to a large proportion of interchanges being diamond interchanges, drivers may easily misinterpret this type of parclo interchange as a diamond interchange, especially when the connection angle between the off-ramp right-turn lane and crossroad is similar to the connection angle between the on-ramp at diamond interchange and crossroad. 56 

Thus, further analysis was conducted to examine if the connection angle between the off-ramp right-turn lane and crossroad impacts WWD incidents in the following section. Figure 5-10. Parclo interchange terminal with circular off-ramp vs. diamond interchange terminal (Google Images) 5.5 Connection Angle between Off-ramp Right-Turn Lane and Crossroad A further study was conducted to understand the effect of the connection angle between the off- ramp right-turn lane and the crossroad on WWD incidents. Six parclo B interchange off-ramp terminals were selected. Three had 177 WWD incidents, and the other three had no recurring WWD incidents. All incidents were analyzed to determine the WWD entry point, turning radius, and speed. The initial analysis found that approximately 90% of the WWD incidents were left- turning movements from crossroads onto off-ramp right-turn lanes. A further analysis was conducted to compare the design features between the three locations with WWD incidents and the other without WWD incidents. Geometric design features and WW-related TCDs at these six parclo interchange terminals were collected from field review and Google Earth. Detailed configurations of these six terminals are shown in Figure 5-11. The angles between an off-ramp right-turn lane and a crossroad of each location were measured based on the tangent at the end of the curve generated by a channelizing island and double yellow lines on a crossroad. Trajectories of each WWD incident were recorded from the collected videos. Figure 5-12 shows the possible trajectories of WWD incidents highlighted by the pink-shaded route (i.e., potential WWD route). The green-shaded routes show the roadway used by correct left-turning or through traffic. The radius of each pink-shaded curve route was measured from Google Earth. 57 

Figure 5-11. Geometric elements and TCDs at six study locations Figure 5-12. Potential WWD routes of six study locations 58 

Among 177 recorded WWD incidents, 81 occurred at study site A (I-65 Exit 284 SB in Alabama), 68 occurred at B (I-65 Exit 208 SB in Alabama), and 28 occurred at C (Exit 147 SB on I-85 in Georgia). No WWD incidents were found at three comparison sites D (US-94 Exit 8 WB in California), E (I-5 Exit 39 SB in California), and F (I-75 Exit 121 SB in Georgia). Table 5-6 shows the total monitored hours and the number of WWD incidents. Table 5-6. Number of WWD Incidents and Monitoring Hours for Study Locations Location A B C D E F Hours monitored 528 504 48 48 48 48 # WWD incidents 81 68 28 0 0 0 # WWD incidents entering through 74 58 26 0 0 0 off-ramp right-turn lane % WWD incidents entering 91% 85% 93% NA NA NA through off-ramp right-turn lane A common feature at the study sites A, B, C, and E is that the connection angle between an off- ramp right-turn lane and the crossroad is similar to a traditional diamond interchange. However, left turns on the crossroad at location E are under stop control, which can slow down left-turn drivers and provide them with more decision time. Locations D and F have an off-ramp right- turn lane tangent to the crossroad's centerline, making WW left turns more difficult. Figure 5-13 shows an example of a right-turn lane tangent with the centerline on the crossroad or edge of the crossroad at the right-angle intersection, as highlighted by the red line. Analysis results suggested that the off-ramp right-turn curve (as shown in red in Figure 5-13) should be at least tangent with the centerline of the crossroad (as shown in yellow in Figure 5-13) to reduce the risk of WWD. In contrast, the off-ramp right-turn lane tangent to the crossroad edge line can provide additional benefits to eliminate the potential WWD route. Figure 5-14 illustrates an off-ramp right-turn lane design tangent to the centerline of the crossroad at acute and obtuse intersections. The number of lanes and median widths on the crossroad can provide more space for an off-ramp to be tangent to a crossroad centerline to reduce the risk of WWD incidents. In addition, a wider non-traversable median on a crossroad can eliminate the potential WWD route. 59 

Figure 5-13. Relationship between right-turning radius and angle between off-ramp right-turn lane and crossroad Figure 5-14. Examples of acute and obtuse connections between ramps and crossroads Equation 5-1 illustrates a relationship between the radius of an off-ramp right-turn lane and the connection angle between the right-turn lane and the crossroad, where R represents the turning radius of the right-turn lane (can also represent the left-turning radius of the potential WWD from the crossroad to off-ramp right-turn lane); W represents the width available for WWD 60 

intrusion, and 𝜃𝜃 represents the connection angle between the off-ramp right-turn lane and the centerline of a crossroad. When 𝜃𝜃 is zero, the risk for WWD is low. When it is 45 degrees, the risk for WWD is high. 𝜃𝜃 was measured to be about 45 degrees at three study locations with WWD incidents. This equation can be applied for estimating connection angles between an off- ramp right turn lane and crossroad, as shown in Figure 5-13 to Figure 5-15. 𝑅𝑅−𝑊𝑊 𝜃𝜃 = 𝑐𝑐𝑐𝑐𝑐𝑐 −1 � � 𝑤𝑤ℎ𝑒𝑒𝑒𝑒 𝑊𝑊 > 0 � 𝑅𝑅 (5-1) 𝜃𝜃 = 0 𝑤𝑤ℎ𝑒𝑒𝑒𝑒 𝑊𝑊 = 0 Channelizing island design impacts the connection angle between the off-ramp right turn lane and crossroad. Three types of channelizing island design were found in the current practices at unsignalized parclo off-ramp terminals, as shown in Figure 5-15 A, B, and C. The black dashed lines represent their impact on the turning radius of the off-ramp right turn lane and its connection angle with the crossroad. Figure 5-15A shows a large angle (θ1) between the off- ramp right-turn lane and the crossroad, with a wider intrusion distance on the crossroad (W1). As a result, the left-turning drivers from the crossroad could make easy and comfortable WWD movements. Figure 5-15B shows an example of an off-ramp right-turn lane tangent with the centerline on the crossroad, which can be considered a relatively lower risk for WWD. The narrower width (W2) will likely reduce the risk of WWD due to human errors. Figure 5-15C shows an example of current practices, with a smaller angle (θ3) between the off-ramp right- turn lane and the crossroad. As a result, the right-turn lane tangent is nearly parallel with the crossroad making the WWD more difficult. This design practice should be considered for reducing the risk of WWD, especially at parclo B interchange terminals. In addition, the case studies and field observations found that all-way stop control or roundabout at the terminal might reduce the risk for WWD by forcing drivers to slow down or stop before turning left onto ramps. Figure 5-15. Three design practices with different intrusion distances 61 

5.6 Conclusions In this chapter, a comparative analysis of design practices was conducted for three states (Florida, Alabama, and Illinois). The results suggest that systematic application of a raised median and median extension on the crossroad can significantly reduce the risk of WWD at parclo interchanges. MCA analysis of WWD incident data indicates that (1) the ramp terminals that contain three or four types of enhanced signs are less likely to have WWD incidents; (2) the signalized or stop-controlled intersections are less likely to have WWD incidents; (3) non- traversable median on the crossroad and lighting on 4-lane crossroads can reduce the risk of WWD incidents; (4) the uncontrolled ramp terminal intersections with no street lighting on the two-lane crossroad are likely to have WWD incidents; and (5) the poor pavement marking & signing conditions are likely to increase the risk of WWD. The two key design features identified can potentially affect WWD risks at parclo interchange terminals, including (1) the loop off-ramp terminals at parclo B or AB interchanges are more prone to WWD than parclo A or other interchange types, and (2) the connection angle between off-ramp right-turn lane and crossroad has a direct impact on WWD incidents when the median on crossroads are traversable. 62