Impact of Shoulder Width and Median Width on Safety (2009)

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.

S U M M A R Y The objectives of this research were to quantify the safety and operational impacts of design element trade-offs and to develop guidelines to assist designers in making reasonable choices when applying context-sensitive solutions and design exceptions. Existing research results were combined with recent practical field experience to provide a guide for planners and designers to understand relationships and quantify the trade-offs for selected design elements. This research provides the highway design community with information resources and decision tools for designing roadways where design flexibility may be appropriate to the roadway context. The research was completed in two phases. The first phase was a literature review and the development of a methodology for data collection and analysis to be used in the second phase. In the second phase, data were collected and analyzed to develop the resources and tools needed for understanding the safety and operational impacts from design element trade-offs. This report documents the findings of the research. The literature review determined that a significant amount of research had been undertaken in an attempt to quantify the relationships between safety and roadway design elements, but that these relationships were not available for cross-section elements on multilane rural roads. Therefore, in an investigation to determine the safety impacts of design flexibility on rural multilane highways, the NCHRP project panel recommended that the second phase of the research focus on three geometric elements: lane width, shoulder width, and median type and width. This decision allowed the development of useful models compatible with the current efforts in the development of the Highway Safety Manual (HSM) (1). The HSM is planned as a comprehensive compendium of current knowledge related to roadway safety treatments and a collection of tools for predicting the safety effects of different roadway design alternatives for various classes of roadways. The design elements that were examined in this research have the potential to affect safety. The degrees of influence vary by design element and application and, often, are specific to a set of roadway conditions. Parallel efforts are currently underway to address the quantification of the safety and operational impacts from design element trade-off for two-lane rural highways and, in the near future, for multilane highways. The key lesson from the literature is that values for design elements can be varied. Most research has been directed to the task of evaluating specific design elements, without con- sidering the effects when multiple elements are varied in combination. An additional issue that has not been discussed extensively is the potential for creating the opposite effect intended by the selected values for design elements. For example, wider shoulders have shown the potential to improve safety. On the other hand, they also have the potential to present conditions that result in increased operating speeds and increased crash severity. A similar counterbalancing potential was noted for the presence and type of barrier in medians. Impact of Shoulder Width and Median Width on Safety 1

2Therefore, design decisions and countermeasure applications should consider the types of associated crashes for modification and then determine the appropriate design element. The research was aimed at developing a set of recommendations to be used in evaluating safety implications from design element trade-offs. Data from three states were used to develop prediction models that could be used for this purpose, with an emphasis on developing crash prediction models and Accident Modification Factors (AMFs) for multilane rural roads with respect to lane width, shoulder width, and median width and type. The available data limited these models to four-lane roadways with 12-ft lanes. Separate models were developed for divided and undivided facilities as well as for both total crashes and injury crashes, each including single-vehicle, multi-vehicle, and all crashes. The research employed an expert- panel approach where prior research was reviewed and discussed along with the models developed herein. In this way, the research compared past results with those obtained to recommend a set of AMFs that may be used to determine the safety effects from the change in the values of a design element. Final recommendations are provided for shoulder width and median width for four-lane roads with 12-ft lanes. The available data did not permit the development of additional recommendations even though the presence of median barrier was also considered. The AMF values recommended are higher than those proposed in the HSM mainly because they address all crashes rather than only crashes related to the specific element. This fact explains the larger magnitude of these AMFs because they capture the effect of a larger number of crashes. A short literature review follows, accompanied by the research findings and the rationale for the recommended values for shoulder and median width. Shoulder Width Past Research Shoulders placed adjacent to travel lanes accomplish several functions including emergency stop and pull off, recovery area for driver error, and pavement edge support (2). However, the use of shoulders to provide an area for a stopped vehicle poses a hazard since past research has shown that 11% of fatal freeway crashes are related to vehicles stopped on shoulders (3). There is also some evidence that wider shoulders may encourage higher operating speeds because they may communicate to the driver the presence of wider space for correcting errors. Finally, number of lanes, lane width, and shoulder width are all somewhat interrelated, and the geometric value choice for any of these elements typically has an effect on the other elements. Most of the research completed to date has focused on two-lane, two-way rural roads (4) or, more recently, on urban or suburban multilane highways (rather than rural roads), further reduc- ing the number of relevant references. Hadi et al. (5) examined the effect of shoulder width on crashes on multilane rural highways. They found that for four-lane rural divided roads, a small reduction in crashes (1% to 3%) could be attained if the unpaved shoulder is widened by 1 ft. These authors also found that roads with shoulders between 10 and 12 ft have the lowest crash rates. This relationship is present only for unpaved shoulders, and the reduction factor should be used cautiously. Harwood et al. (6) produced AMFs for multilane highways. An expert panel then considered an adjustment to the AMF for two-lane rural roads. The panel determined that the AMF could remain the same for both situations based on the determination that shoulder width has a similar effect on multilane and two-lane rural roads. A recent study by Harkey et al. (7) also evaluated traffic engineering and ITS improvements to develop AMFs for rural multilane roadways. The study considered undivided roads with greater than 2,000 vehicles per day, and the AMFs developed were for roadways where the shoulder-related

3crashes were 35% of the total. Additional procedures are available for roadways with lower volumes or different percentages. For divided highways, the current draft of the HSM uses the recommended values from NCHRP Project 17-29 (8), which developed AMFs for paved shoulder width for rural multilane segments. NCHRP Project 17-29 research results are published as NCHRP Web-Only Document 126 (www.trb.org/news/blurb_detail.asp?id=9099). NCHRP Project 15-27 The models developed in this research demonstrated that there is a relationship between shoulder width and crashes. The predictive models developed in the research support general trends observed in previous studies for two-lane, two-way rural roads. The current study distinguished between divided and undivided highways and between single- and multi-vehicle crashes. This classification allowed for the development of four distinct models to address the particular issues relative to crash types and the influence of the presence of the median. Aggregate models were also developed for all crashes to permit a comprehensive approach for determining overall effects of shoulder width. It should be noted that the shoulder width used is the average total width for the left and right shoulders (i.e., the sum of right and left shoulders divided by two) in the same direction for divided roads and the average width of right shoulders for undivided segments. For undivided, four-lane highways, the shoulder width was a significant predictive variable for multi-vehicle and all crashes. The coefficient in the model for multi-vehicle crashes is −0.11 and for all crashes is −0.07. The negative sign is indicative of the beneficial influence of the shoulder width. These values are indicative of the relative safety gains from a 1-ft increase in shoulder width. However, the magnitude of these values seems high, and it is likely that such large reductions may not be reachable. For divided highways, shoulder width was included in all three models. The coefficients were −0.05 for single-vehicle, −0.14 for multi-vehicle, and −0.12 for all crashes. The negative sign again demonstrates the reduction of crashes associated with the increase of the shoulder width. The magnitude of the coefficients for the multi-vehicle and all crashes again seems to be excessive. The similar analysis for injury-only crashes did not produce significant changes in the coefficients noted here. The variable was significant only for divided highways, and the coefficients were practically the same as those noted for all crashes. The AMFs developed for each condition based on the models developed are summarized in Table S-1. It should be noted that these factors are for the total number of crashes and for all severities (KABCO). Based on the project team’s review of past literature, the recommended values for the HSM, and the AMF from NCHRP Project 15-27, the presence of shoulders appears to influence crash occurrence, and the values noted for all crashes for undivided highways seem reasonable Average shoulder width (ft)2 Category 0 3 4 5 6 7 8 10 Undivided, multi-vehicle 1.39 1.00 0.90 0.80 0.72 0.64 0.58 0.46 Undivided, all crashes 1.22 1.00 0.94 0.87 0.82 0.76 0.71 0.63 Divided, single-vehicle 1.17 1.00 0.95 0.90 0.85 0.81 0.77 0.69 Divided, multi-vehicle 1.51 1.00 0.87 0.76 0.66 0.58 0.50 0.38 Divided, all crashes 1.43 1.00 0.89 0.79 0.70 0.62 0.55 0.44 1 The AMFs are for all crashes and all severities. 2 The average shoulder width for undivided is the average of the right shoulders; for divided, it is the average of left and right shoulder in the same direction. Table S-1. AMFs based on prediction models for average shoulder width.1

4and in accordance with current trends and literature. The AMF for all crashes for undivided highways is recommended for use since shoulder width was not a significant variable in the single-vehicle models. The project team considered the values provided for all three models for divided highways and recommended using the values from single-vehicle crashes because the values for multi-vehicles and all crashes were high and probably reflect other influences, such as volume. This adjustment is considered justifiable based on previous work by Harwood et al. (6) and the recommended values in the HSM (8). It should be noted that different parts of the HSM provide different AMFs for the same changes in design or operation; these differences are currently being reconciled. The recommended values are summarized in Table S-2. These modification factors are for all crashes and not for specific types of crashes that could relate to shoulder width issues. The recommended values are similar to those proposed in the HSM, as noted above, and those of the divided highways are comparable for almost all categories with the exception of the 8-ft shoulder AMF. For undivided highways, the differences between the NCHRP Project 15-27 and HSM-recommended AMFs were larger. These differences are attributed to the fact that the AMFs in the HSM are developed for shoulder-related crashes while the AMFs from NCHRP Project 15-27 were developed for all crashes. Even though a comparison to the HSM values is not strictly appropriate because of the difference in crashes used in each model, the comparison is meaningful in showing similarities in trends and agreement of findings. Another issue that should be addressed in future research is the lack of AMFs for shoulder width greater than 8 ft since the literature indicates that the safety effects for such shoulder widths are unknown. Median Width Past Research The most important objective for the presence of medians is traffic separation. Additional benefits from medians include the provision of recovery area for errant drivers, accommodation of left-turn movements, and the provision for emergency stopping. Median design issues typically address the presence of the median, along with type and width. There is some research on these issues and their implications on safety. A review by Hauer (9) indicated that it was not possible to identify AMFs for median width but rather noted three safety trends: (1) cross-median crashes (i.e., opposing vehicles) are reduced with wider medians; (2) median-related crashes increase as the median width increases with a peak at about 30 ft and then decrease as the median becomes wider than 30 ft; and (3) the effect of median width on total crashes is questionable. The study conducted by Hadi et al. (5) using negative binomial models showed that the median width has an influence on multilane roadways, and they produced two models based on the traffic volume range and number of lanes. This is the only study that has examined the effect of median width on safety for rural, multilane roads since the several studies reviewed by Hauer (9) and the NCHRP Project 17-27 Interim Report (10) deal with freeway median width. Average shoulder width (ft)2 Category 0 3 4 5 6 7 8 Undivided 1.22 1.00 0.94 0.87 0.82 0.76 0.71 Divided 1.17 1.00 0.95 0.90 0.85 0.81 0.77 1 The AMFs are for all crashes and all severities. 2 The average shoulder width for undivided is the average of the right shoulders; for divided, it is the average of left and right shoulder in the same direction. Table S-2. Recommend AMFs for average shoulder width (ft).1

The interim report for NCHRP Project 17-27 described development of a set of AMFs for the effect of median width on crashes for four-lane rural roadways (see Table S-3). The HSM section on multilane rural roads developed through NCHRP Project 17-29 (8) has also proposed AMF values for rural multilane highways. Two sets of values were developed based on whether a median barrier was present from the studies of Miaou et al. (11) and Harkey et al. (7). These values accounted for the total number of crashes while considering median-related crashes. The rec- ommended values are summarized in Table S-3 and have been adjusted from the normal baseline of 30-ft medians presented in the report. It should be pointed out that these AMFs are used for evaluating changes in median width for an already existing divided facility—they are not used for estimating the safety performance of highways when an undivided highway is converted to a divided facility. The models developed in this research determined that median width had an effect on multi- vehicle crashes for divided highways and distinguished between divided and undivided highways as well as between single- and multi-vehicle crashes. The effect of median width was only evaluated for the divided highways. This classification allowed for the development of two distinct models to address the particular issues relative to crash types. Aggregate models were also developed for all crashes to allow for a comprehensive approach and determination of potential overall effects of the median barrier presence. The only model where median width was significant was that for multi-vehicle crashes, and it had a positive effect—crashes reduce with wider medians. This trend is supported by the general observation that roadways with wider medians will exhibit lower crash rates than will roads with more narrow medians. The model coefficient was −0.010. The analysis of the injury-only crashes included this variable again only in multi-vehicle crashes models with a similar coefficient (−0.009). The project team reviewed past literature, the recommended values for HSM, and the AMF from NCHRP Project 15-27 and concluded that median width does have an influence on crash occurrence. The team determined that the values noted for the only model with median width influence are reasonable and in accordance with current trends and literature. The only avail- able AMF based on the models developed in this research is for multi-vehicle crashes; there is a 1% reduction for every additional foot of median width added. The values obtained from the models for multi-vehicle crashes are reasonable and agree with the previous research. The rec- ommended values are summarized in Table S-4. These AMFs are for all crashes and not for specific types of crashes that could relate to median width issues. The recommended values are greater than those proposed in the HSM. The difference could be attributed to the fact that the HSM values specifically account for median-related crashes. 5 Median width (ft) Barrier 15 20 30 40 50 60 70 80 90 With 1.000 0.997 0.990 0.984 0.977 0.971 0.964 0.958 0.951 Without 1.000 0.994 0.981 0.969 0.957 0.945 0.933 0.922 0.910 Table S-3. AMFs for median width in rural multilane roadways (7). Median width (ft) Category 10 20 30 40 50 60 70 80 Multi-vehicle 1.00 0.91 0.83 0.75 0.68 0.62 0.57 0.51 Table S-4. Recommended AMFs for median width, divided roadways.

6This means of accounting for median crashes was not possible in the current research, and similar adjustments could affect the values recommended. Another possible relationship that could influence these values is the presence of a median barrier. Roadway segments with a barrier typically have narrower medians; this could influence the AMFs as shown in the HSM values. However, the available dataset was not large enough to examine this interaction. To determine the AMFs for all crashes, it may be assumed that the median width has “no effect” on single-vehicle crashes and, therefore, the AMF for single-vehicle crashes could be considered 1.00. In this case, a weighted AMF can be estimated using the relative percentages of single- and multi-vehicle crashes for the roadway of concern. The AMF developed herein can be used to estimate the design element value’s relative impact for a rural four-lane roadway segment. The process described could be applied to determine the safety implications using different values for a single or combination of design elements. The ratio of AMFs for two different conditions can be used to establish the relative change in crashes anticipated from the change in design element values. The use of this approach was noted as a method for estimating change in crashes by using Equation S-1: where ΔN is the change in crashes and AMFi are the AMFs for the designs to be evaluated. This equation was modified from the form presented by Lord and Bonneson (12) since no base models or base estimates are available. A positive value of ΔN denotes an increase in crash frequency. Summary References 1. Highway Safety Manual Home Page. Transportation Research Board. www.highwaysafetymanual.org. Accessed May 10, 2008. 2. AASHTO. “A Policy on Geometric Design of Highways and Streets.” Washington, D.C. (2004). 3. Agent, K. R., and J. G. Pigman. Accidents Involving Vehicles Parked on Shoulders of Limited Access Highways, Report KTC-89-36. Kentucky Transportation Center, Lexington, KY (1989). 4. Hauer, E. “Shoulder Width, Shoulder Paving and Safety.” www.trafficsafetyresearch.com (2000). 5. Hadi, M. A., J. Aruldhas, L. Chow, and J. Wattleworth. “Estimating Safety Effects of Cross-Section Design for Various Highway Types Using Negative Binomial Regression,” Transportation Research Record 1500. Transportation Research Board, National Research Council, Washington, DC (1995); pp. 169–177. 6. Harwood, D. W., E. R. Rabbani, K. R. Ricard, H. W. McGee, and G. L. Gittings. NCHRP Report 486: Systemwide Impact of Safety and Traffic Operations Design Decisions for 3R Projects. Transportation Research Board of the National Academies, Washington, DC (2003). 7. Harkey, D. L., R. Srinivasan, J. Baek, F. M. Council., et al. NCHRP Report 617: Accident Modification Factors for Traffic Engineering and ITS Improvements. Transportation Research Board of the National Acade- mies, Washington, DC (2008). 8. Lord, D., B. N. Persaud, S. W. Washington, J. N. Ivan, I. van Schalkwyk, C. Lyon, T. Jonsson, and S. R. Geedipally. NCHRP Web-Only Document 126: Methodology to Predict the Safety Performance of Rural Multilane Rural Highways. Transportation Research Board of the National Academies, Washington, DC, 2008. 9. Hauer, E. “The Median and Safety,” www.trafficsafetyresearch.com (2000). 10. iTrans Consulting. “NCHRP Project 17-27 Interim Report.” NCHRP 17-27: Project Parts I and II of the Highway Safety Manual. Richmond Hill, Ont., 2005. 11. Miaou, S. P., R. P. Bligh, and D. Lord. “Developing Median Barrier Installation Guidelines: A Benefit/Cost Analysis using Texas Data,” Transportation Research Record 1904, Transportation Research Board of the National Academies, Washington, DC (2005); pp. 3–19. 12. Lord, D., and J. Bonneson. “Role and Application of Accident Modification Factors Within Highway Design Process,” Transportation Research Record 1961, Transportation Research Board of the National Academies, Washington DC (2006); pp. 65–73. ΔN AMF AMF = − 1 2 1 1( )S-