TOOL

shutterstock/LeManna Developing, Using, and Improving Tables Showing the Safest Feasible Design By Joseph E. Hummer, P.E., Ph.D. (M)

n recent years, the Federal Administration (FHWA) and the states have invested millions of dollars in research on the safety performance of different measures at intersections. Much of this research has been cataloged in an easy-to-use website called the “Crash Modification Factors Clearinghouse”1 maintained by the FHWA. The ClearinghouseI contains thousands of crash modification factors (CMFs), which are defined as the ratio of the estimated crash frequency after an intervention to the crash frequency before the intervention. A CMF below one thus means the intervention helped. Happily, researchers have rated the quality of each CMF in the Clearinghouse, on a scale of zero stars (poor or unknown quality, result should not be trusted) to five stars (excellent quality, trustworthy result), so that consumers of the information do not have to make that judgment. The library of CMFs, each with a quality rating, is a tremendous resource.

www.ite.org May 2020 45 However, the library of safety research results is rarely used during also known as a restricted crossing U-turn (RCUT) intersection, transportation improvement program (TIP) projects to improve superstreet, or J-turn. intersections. During those projects, teams typically use site crash The author considered two sources of CMFs not in the data to justify the project and then use physics-based nominal Clearinghouse in constructing the SaFID tables. First, the FHWA design standards. Only in a few states with comprehensive intersec- median U-turn (MUT) guidebook (20) contains a review of the tion control evaluation (ICE) policies do project teams typically use safety research on that design and implies average CMFs for the the available CMFs. conversion from a conventional signal to a MUT of 0.85 for all To help intersection project teams use the available CMFs more crashes and 0.7 for injury crashes. Second, a 2015 research report often and effectively, the author assembled tables showing the safest for the Utah DOT (21) showed a CMF of 0.88 for all crashes feasible intersection design (SaFID) for each combination of size for the conversion of a conventional intersection to a partial and demand on the major and minor . The tables should continuous flow intersection (CFI). The report did not provide be easy to use. Project teams should start their investigations of a result for injury crashes. The analysis looks to be of relatively alternatives with the design that the research shows to be the safest, good quality. The partial CFIs examined in Utah had two left turn and then examine other factors that are meaningful in a design crossovers at each site. decision. If project teams end up choosing an alternative that is not The available set of CMFs described above captures most the safest according to the research, they should have to document four-legged intersection designs used in the US as of 2020. In the why. Starting with consideration of the SaFID should mean that FHWA CAP-X tool, the only other four-legged intersection designs agencies end up building safer intersections. The objectives of this listed are full CFI (four left turn crossovers), quadrant, , paper are to show the SaFID tables, provide background on how and split.22 None of these is common. The only other common they were developed, and discuss how they should be used. intersection types that the author could think of are and offset intersections. While are common in a few states, in Sources they are not considered to be a competitive design as With two notable exceptions, the CMFs in the SaFID tables are they require more right-of-way than a partial CFI while delivering from the Clearinghouse.1 The author used only CMFs with three only a fraction of the delay-saving benefits. Meanwhile, on offset stars or better. The documentation in the Clearinghouse had to be intersections a recent literature review conducted by the NCDOT clear on the before condition, the after condition, and the context did not provide any studies with trustworthy CMF values, and the in which the crash data were collected. This effort used CMFs Clearinghouse does not mention them. Overall, with the possible for a generic four-legged intersection. In some cases, the author exception of offset intersections, it looks like we have a pretty full set averaged CMFs to create an overall CMF. For example, if a study of CMFs for common and feasible intersection designs. had separate CMFs for urban and for rural roads, the author To construct the SaFID tables, the author also considered the averaged those to get an overall CMF for that study, and if two feasibility of the various designs with the following rules: studies provided CMFs for the same design the author averaged ƒ All-way stop control (AWSC) is viable on two- roads with those to get an overall CMF for that design. demands less than 7,500 vehicles per day (VPD) on each . Table 1 shows the references from the Clearinghouse used ƒ Based on the latest national guide, a single-lane to assemble the SaFID tables and the corresponding average can handle up to 25,000 VPD total and a two-lane CMF values. Note that a reduced conflict intersection (RCI) is roundabout can handle up to 45,000 VPD total.23

Table 1. CMF values and references.

Changing from… Changing to… All crashes Injury crashes Average CMF References Average CMF References in this Article Two-way stop control All-way stop control 0.32 2 0.28 2, 3 Conventional signal 0.81 4-8 0.74 6-9 One-lane roundabout 0.51 10-13 0.16 10 Unsignalized RCI 0.58 14-16 0.42 14, 16 Conventional signal One-lane roundabout 0.74 17 0.45 17 Two-lane roundabout 0.89 12 and 17 0.54 17, 18 Signalized RCI 0.85 19 0.78 19

46 May 2020 ite journal ƒ Based on the FHWA guidebook a signalized RCI can handle Entering the project development process with the SaFID as the up to 25,000 VPD on the minor .24 default should shift the burden of proof to advocates of generally ƒ Two-lane minor streets should be signalized in RCIs at less-safe designs, where the burden should lie. demands ranging from 3,000 to 11,000 VPD based on Project teams can use Table 2 or Table 3 or both in their NCDOT research.25 processes. In some cells the tables have the same entry, indicating ƒ Because RCIs have superior signal progression and are not as that a design is generally safest using either all crashes or injury vulnerable to driver confusion, MUTs and CFIs only become crashes. However, in some cells the two tables have different entries, feasible at minor street demands above 25,000 VPD. so agencies and project teams will have to choose which type of Agencies often make exceptions to these rules, but they should crash is more important. serve well to start. One of the reasons not to choose the SaFID during a project is One other technique needed to construct the SaFID tables is that the research justifying the design as the safest does not apply the ability to chain CMFs. If we have a CMF for the conversion of to the case in question. Indeed, there are many places where the condition a to b, and a CMF for the conversion of b to c, multiplying existing research reflected in Tables 2 and 3 is out of scope. Tables the CMF for a to b by the CMF for b to c should provide the CMF 2 and 3 apply to four-legged intersections, for example, and may for a to c without losing much accuracy. For example, for all crashes not apply to a project improving a three-legged junction. Those the conversion of an intersection from a signal to a two-way stop claiming that their project site is out of scope of the research has an average CMF of 1/0.81 = 1.23 and from a two-way stop to underlying Tables 2 and 3 ought to be careful not to stretch that all-way stop has an average CMF of 0.32, so the CMF for converting argument too far. Just because the research has not been done on from a signal to an all-way stop should be 1.23 * 0.32 = 0.4. three-legged RCIs, for example, does not mean that those designs are less safe than conventional designs. SaFID Tables Project teams should use Tables 2 and 3 early in the process Table 2 shows the SaFID table based on all crashes, while Table 3 in conjunction with a couple other tools that support quick shows the SaFID table based on injury crashes. Both tables show the judgments on different designs. For capacity, CAP-X helps analysts SaFID for any combination of major street and minor street number quickly judge which alternatives promise greater efficiency.22 For of through and vehicle demand. The demands are in terms of the quality of pedestrian and bicyclist service at intersection alter- average annual daily , or AADT, in VPD. The CMFs displayed natives, the guidebook from NCHRP project 7-25, to be published are for the conversion of a conventional signalized intersection to in 2020, should provide a quick but helpful look at any intersection the design named in the cell. design. Together, the SaFID tables, CAP-X, and the forthcoming Tables 2 and 3 are dominated by four design types, including NCHRP 7-25 guidebook provide a powerful suite of early intersec- AWSC, , RCIs, and MUTs. A TWSC is never the tion design filters. generally safest choice. A conventional signal only shows up in a small sliver of each table, at the highest demand levels handled Follow-Up Work with two-lane major and minor streets where a roundabout is not Tables 2 and 3 will hopefully prove helpful to intersection design feasible. Though the partial CFI did not appear in Table 2, it did not teams, but they could be improved with several types of additional miss by much. The CMF for all crashes for a MUT is 0.85,20 while research. First, more research on some designs already in the tables the CMF for a partial CFI is 0.88.21 Especially at high demands with would be welcome. Second, we could use research on designs not six-lane or eight-lane roads, a partial CFI may be a worthy choice in these tables, including offset and quadrant intersections, that for capacity without losing too much safety benefit. likely will be considered more in the next few years. Third, we need research on the validity of combining CMFs, so that project teams Using the Tables can evaluate the safety of hybrid designs. Fourth, we have no CMFs In view of the importance of safety, the author urges highway and almost no safety research on grade-separated intersections agencies to adopt the SaFID as the default choice during inter- (the junction of two non-freeways using a ), while many section improvement projects. Conventional TWSC and signal of these relatively high cost solutions are being proposed. Fifth, intersections are not generally the safest feasible options and should researchers should begin to derive CMFs for interchanges so an therefore not be the default designs during projects. There are many SaFID table may be assembled. Finally, work should reasons why an agency may not be able to build the SaFID in any continue on adequate crash surrogates to help designers estimate particular project, including cost, impacts, delays, and effects on the crash potential of designs that have not been built yet. A recent non-motorized travelers. However, in all cases agencies ought to paper based on research sponsored by the NCDOT (26) provided a be prepared to say why they did not end up building the SaFID. start in this direction. itej

www.ite.org May 2020 47 Table 2. Safest feasible intersection design (SaFID) based on all crashes. Minor street Number through 2468 lanes: Major street Low AADT: 0 5,000 7,500 10,000 10,000 15,000 20,000 25,000 and Number Low High Any Any above through lanes AADT AADT High AADT: 5,000 7,500 10,000 15,000 15,000 20,000 25,000 2 0 7,500 Safest All-way stopAll-way stop n/a n/a n/a n/a n/a n/a n/a n/a

CMF 0.4 0.4 7,500 10,000 Safest One-lane One-lane One-lane n/a n/a n/a n/a n/a n/a n/a roundabout roundabout roundabout CMF 0.7 0.7 0.7 10,000 15,000 Safest One-lane One-lane One-lane One-lane n/a n/a n/a n/a n/a n/a roundabout roundabout roundabout roundabout* CMF 0.7 0.7 0.7 0.7 4 10,000 15,000 Safest Unsignalized Unsignalized Unsignalized Signalized RCI Signalized RCI n/a n/a n/a n/a n/a RCI RCI RCI CMF 0.7 0.7 0.7 0.8 0.8 15,000 20,000 Safest Unsignalized Unsignalized Signalized RCI Signalized RCI Signalized RCI Signalized RCI n/a n/a n/a n/a RCI RCI CMF 0.7 0.7 0.8 0.8 0.8 0.8 20,000 25,000 Safest Unsignalized Signalized RCI Signalized RCI Signalized RCI Signalized RCI Signalized RCI Signalized RCI n/a n/a n/a RCI CMF 0.7 0.8 0.8 0.8 0.8 0.8 0.8 25,000 30,000 Safest Unsignalized Signalized RCI Signalized RCI Signalized RCI Signalized RCI Signalized RCI Signalized RCI Median u- n/a n/a RCI turn CMF 0.7 0.8 0.8 0.8 0.8 0.8 0.8 0.8 30,000 and Safest Unsignalized Signalized RCI Signalized RCI Signalized RCI Signalized RCI Signalized RCI Signalized RCI Median u- n/a n/a above RCI turn CMF 0.7 0.8 0.8 0.8 0.8 0.8 0.8 0.8 6 Any Safest Unsignalized Signalized RCI Signalized RCI Signalized RCI Signalized RCI Signalized RCI Signalized RCI Median u- Median u- n/a RCI turn turn CMF 0.7 0.8 0.8 0.8 0.8 0.8 0.8 0.8 0.8 8 Any Safest Unsignalized Signalized RCI Signalized RCI Signalized RCI Signalized RCI Signalized RCI Signalized RCI Median u- Median u- Median u- RCI turn turn turn CMF 0.7 0.8 0.8 0.8 0.8 0.8 0.8 0.8 0.8 0.8

* One-lane roundabouts are generally feasible if the combined AADT is less than 25,000. If a one-lane roundabout is infeasible a signal is the safest feasible design.

Table 3. Safest feasible intersection design (SaFID) based on injury crashes. Minor street Number through 2 4 68 lanes: Major street Low AADT: 0 5,000 7,500 10,000 10,000 15,000 20,000 25,000 and Number Low High Any Any above through lanes AADT AADT High AADT: 5,000 7,500 10,000 15,000 15,000 20,000 25,000 2 0 7,500 Safest All-way stopAll-way stop n/a n/a n/a n/a n/a n/a n/a n/a

CMF 0.4 0.4 7,500 10,000 Safest One-lane One-lane One-lane n/a n/a n/a n/a n/a n/a n/a roundabout roundabout roundabout CMF 0.4 0.4 0.4 10,000 15,000 Safest One-lane One-lane One-lane One-lane n/a n/a n/a n/a n/a n/a roundabout roundabout roundabout roundabout* CMF 0.4 0.4 0.4 0.4 4 10,000 15,000 Safest Unsignalized Unsignalized Unsignalized Two-lane Two-lane n/a n/a n/a n/a n/a RCI RCI RCI roundabout roundabout CMF 0.5 0.5 0.5 0.5 0.5 15,000 20,000 Safest Unsignalized Unsignalized Two-lane Two-lane Two-lane Two-lane n/a n/a n/a n/a RCI RCI roundabout roundabout roundabout roundabout CMF 0.5 0.5 0.5 0.5 0.5 0.5 20,000 25,000 Safest Unsignalized Two-lane Two-lane Two-lane Two-lane Two-lane Two-lane n/a n/a n/a RCI roundabout roundabout roundabout roundabout roundabout roundabout** CMF 0.5 0.5 0.5 0.5 0.5 0.5 0.5 25,000 30,000 Safest Unsignalized Two-lane Two-lane Two-lane Two-lane Two-lane Median u-turn Median u- n/a n/a RCI roundabout roundabout roundabout roundabout roundabout** turn CMF 0.5 0.5 0.5 0.5 0.5 0.5 0.7 0.7 30,000 and Safest Unsignalized Two-lane Two-lane Two-lane Two-lane Median u- Median u-turn Median u- n/a n/a above RCI roundabout** roundabout** roundabout** roundabout** turn turn CMF 0.5 0.5 0.5 0.5 0.5 0.7 0.7 0.7 6 Any Safest Unsignalized Median u- Median u-turn Median u-turn Median u- Median u- Median u-turn Median u- Median u- n/a RCI turn turn turn turn turn CMF 0.5 0.7 0.7 0.7 0.7 0.7 0.7 0.7 0.7 8 Any Safest Unsignalized Median u- Median u-turn Median u-turn Median u- Median u- Median u-turn Median u- Median u- Median u- RCI turn turn turn turn turn turn CMF 0.5 0.7 0.7 0.7 0.7 0.7 0.7 0.7 0.7 0.7

* One-lane roundabouts are generally feasible if the combined AADT is less than 25,000. If a one-lane roundabout is infeasible a signal is the safest feasible design. ** Two-lane roundabouts are generally feasible if the combined AADT is less than 45,000. If a two-lane roundabout is infeasible a median u-turn is the safest feasible design.

48 May 2020 ite journal Acknowledgements 14. Hummer, J. E., R.L. Haley, S.E. Ott, R.S. Foyle, and C.M. Cunningham, The author appreciates helpful comments from colleagues on early “Superstreet Benefits and Capacities,” Report No. FHWA/NC/2009-06, versions of the tables and support from the NCDOT administra- NCDOT, 2010. tion. The author also appreciates the engineers in North Carolina 15. Inman, V.W. and R.P. Haas, “Field Evaluation of a Restricted Crossing U-Turn and elsewhere who put their careers and reputations on the line Intersection,” Report No. FHWA-HRT-11-067, 2012. to get alternative designs built so that we might be able to develop 16. Edara, P., C. Sun, and S. Breslow, “Evaluation of J-turn Intersection Design CMFs from them. The views and opinions in this paper are the Performance in Missouri,” Missouri DOT, December 2013. author’s and do not necessarily represent the views, opinions, 17. Gross, F., C. Lyon, B. Persaud, and R. Srinivasan, “Safety Effectiveness of policies, or practices of the NCDOT. Converting Signalized Intersections to Roundabouts,” Presented at the 91st Annual Meeting of the Transportation Research Board, Paper No. References 12-1658, Washington, DC, 2012. 1. FHWA, “Crash Modification Factor Clearinghouse” website, accessed 18. Zhao, Y., J. Andrey, and P. Deadman, “Whether Conversion and Weather December 30, 2019, http://www.cmfclearinghouse.org/. Matter to Roundabout Safety,” Accident Analysis and Prevention, Vol. 2. Lovell, J. and E. Hauer, “The Safety Effect of Conversion to All-Way Stop 66, 2018, pp. 151-159. Control,” Transportation Research Record 1068, 1986, pp. 103-107. 19. Hummer, J.E., and S. Rao, “Safety Evaluation of a Signalized Restricted 3. Simpson, C.L. and J.E. Hummer, “Evaluation of the Conversion from Two- Crossing U-Turn,” Report No. FHWA-HRT-17-082, December 2017. Way Control to All-Way Stop Sign Control at 53 Locations in 20. Reid, J., L. Sutherland, B. Ray, A. Daleiden, P. Jenior, and J. Knudsen, “Median North Carolina.” Journal of Transportation Safety and Security, Vol. 2, No. 3, U-Turn Informational Guide,” Report No. FHWA-SA-14-069, August 2014. 2010, pp. 239-260. 21. Zlatkovic, M., “Development of Performance Matrices for Evaluating 4. Pernia, J.C., J.J. Lu, M.X. Weng, X. Xie, and Z. Yu, “Development of Models Innovative Intersections and Interchanges,” Report No. UT-15.13, Utah to Quantify the Impacts of Signalization on Intersection Crashes,” Florida DOT, September 2015. DOT, 2002. 22. FHWA, “CAP-X: Capacity Analysis for Planning of Junctions,” FHWA- 5. McGee, H., S. Taori, and B.N. Persaud, “NCHRP Report 491: Crash Experience SA-18-067, Version 3.0, October 2018. Warrant for Traffic Signals,” Transportation Research Board, 2003. 23. Rodegerdts, L., J. Bansen, C. Tiesler, J. Knudsen, E. Myers, M. Johnson, M. 6. Davis, G.A. and N. Aul, “Safety Effects of Left-Turn Phasing Schemes at High- Moule, B. Persaud, C. Lyon, S. Hallmark, H. Isebrands, R.B. Crown, B. Guichet, Speed Intersections,” Minnesota DOT, Report No. MN/RC-2007-03, 2007. and A. O’Brien, “NCHRP Report 672: Roundabouts: An Informational Guide, 7. Srinivasan, R., B. Lan, and D. Carter, “Safety Evaluation of Signal Installation Second Edition,” Transportation Research Board, 2010. With and Without Left Turn Lanes on Two Lane Roads in Rural and 24. Hummer, J., B. Ray, A. Daleiden, P. Jenior, and J. Knudsen, “Restricted Suburban Areas,” Report No. FHWA/NC/2013-11, NCDOT, October 2014. Crossing U-turn Informational Guide,” Report No. FHWA-SA-14-070, 8. Wang, J.H. and M.A. Abdel-Aty, “Comparison of Safety Evaluation August 2014. Approaches for Intersection Signalization in Florida,” Presented at the 25. Warchol, S., N. Rouphail, C. Vaughan, and B. Kerns, “Guidelines for 93rd Annual Meeting of the Transportation Research Board, Paper No. Signalization of Intersections with Two or Three Approaches,” Report No. 14-0374, Washington, DC, 2014. NCDOT/NC/2017-11, NCDOT, December 2017. 9. Sacchia, E., T. Sayed, and K. El-Basyouny, “A Full Bayes Before-After Study Ac- 26. Lee, T., T. Chase, C. Cunningham, and S. Warchol, “Movement-Based Safety counting for Temporal and Spatial Effects: Evaluating the Safety Impact of New Performance Functions for Signalized Intersections,” Presented at 99th Signal Installations,” Accident Analysis and Prevention, Vol. 94, 2016, pp. 52-58. Annual Meeting, Transportation Research Board, Washington, DC, 2020. 10. Persaud, B. N., R.A. Retting, P.E. Garder, and D. Lord, “Observational Before- After Study of the Safety Effect of U.S. Roundabout Conversions Using the Joseph E. Hummer, P.E., Ph.D. (M) is the state traffic Empirical Bayes Method,” Transportation Research Record 1751, 2001. management engineer with the North Carolina 11. Rodegerdts, L. A., M. Blogg, E. Wemple, E. Myers, M. Kyte, M. Dixon, G. List, Department of Transportation Mobility and Safety A. Flannery, R. Troutbeck, W. Brilon, N. Wu, B. Persaud, C. Lyon, D. Harkey, Division. He specializes in alternative intersection and and D. Carter, “NCHRP Report 572: Applying Roundabouts in the United interchange designs. Joe began researching the designs States,” Transportation Research Board, 2007. in 1990, has published numerous articles about them, and has 12. Qin, X., A. Bill, M. Chitturi, and D. Noyce, “Evaluation of Roundabout invented several new designs. His two-part series in the ITE Journal Safety,” Presented at the Transportation Research Board 92nd Annual in 1998 helped create momentum in this area. He has a B.S. and an Meeting, Paper No. 13-2060, Washington, DC, 2013. M.S. in civil engineering from Michigan State University and a 13. Abdel-Aty, M.A., C. Lee, J. Park, J.Wang, M. Abuzwidah, and S. Al-Arifi, Ph.D. in civil engineering from Purdue University. Joe currently “Validation and Application of Highway Safety Manual (Part D) in Florida,” serves as vice president of the North Carolina Section of ITE. Florida DOT, May 2014.

www.ite.org May 2020 49