Roundabouts in the United States (2007)

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.

4This report summarizes the findings of NCHRP 3-65, “Applying Roundabouts in the United States.” The intended audience for this report is researchers, practitioners, and pol- icy makers who establish federal, state, and local guidelines for roundabouts. Although the content of this document is directly relevant to practitioners, the document is not organ- ized as a guide for easy practitioner use. Once these findings are incorporated into the next edition of FHWA’s Round- abouts: An Informational Guide (1) and other key guidance documents, they should be more accessible to practitioners. This introductory chapter presents the problem statement and research objective, the scope of study, research approach, and a summary of literature review conducted for this project. Problem Statement and Research Objective Although traffic circles have been used in the United States since 1905, their use has been limited since the 1950s because the designs of that era were found to work neither efficiently nor safely. The modern roundabout was developed in the United Kingdom (UK) in the 1960s to address these prob- lems. Two key characteristics of the modern roundabout include (1) a requirement for entering traffic to yield to circulating traffic and (2) geometric constraints that slow entering vehicles. Many studies have shown that modern roundabouts (hereafter referred to simply as roundabouts) can be safe and effective, and they are now widely used inter- nationally. Use in the United States began in 1990 and has been increasing exponentially since that time. Due to this increased interest in roundabouts, continued demand exists for more information regarding appropriate physical locations, design parameters, and their performance relative to alternative control schemes, with a particular need for that information to be based on U.S. performance rather than simple continued reliance on international experience. The lack of comprehensive and objective U.S. field data on safety and operational performance and design of round- abouts has contributed to this demand for information, as per- ceived differences in driver behavior raise questions about how appropriate some international research and practices are for the United States. Therefore, additional information on the safety and operation of roundabouts in the United States will be very helpful to planners and designers in determining where roundabouts would reduce intersection crashes and conges- tion and in refining the design criteria currently being used. NCHRP and FHWA have identified the need to develop tools based on actual U.S. roundabout performance, rather than using foreign procedures as surrogates. Hence, the pri- mary objective of this research is to produce a set of opera- tional, safety, and design tools, calibrated to U.S. roundabout field data. These tools will enable a person who is already competent in analysis or geometric design of typical at-grade intersections to be able to specify a roundabout that is safe and performs well. Scope of Study The scope of this study includes the development of the following work products: • An updated site inventory of known roundabouts that is accessible to the transportation profession • A comprehensive database of safety, operational, and design data of selected existing roundabouts for use in future research • Planning-level safety prediction models to predict the overall safety performance of roundabouts • Design-level safety prediction models for individual roundabout approaches • An expanded comparison of safety performance before and after installation of a roundabout • An updated operational analysis procedure for the High- way Capacity Manual (HCM) (2), including capacity, C H A P T E R 1 Introduction and Research Approach

delay, and queue estimates for single-lane and multilane roundabouts • New speed prediction tools for use in design development • A comprehensive study of pedestrian and bicyclist behavior at roundabouts • Updated design guidance incorporating the results from the various studies identified above. It became clear during the early stages of this research effort that the proposed scope of work and associated data collection effort would be insufficient to address issues related to the accommodation of visually impaired pedestrians at roundabouts. As a result, issues specifically related to visually impaired pedestrians were removed from this scope and com- bined with channelized right turns at conventional intersec- tions as part of a new problem statement spawned from this project: NCHRP 3-78, “Crossing Solutions for Visually Impaired Pedestrians at Roundabouts and Channelized Right Turns.” Research Approach The detailed approaches for each of the major components of this research are described in the following sections. Summarize Existing Relationships Existing models in use around the world for roundabout safety and operational analysis were described, analyzed, and critiqued to understand the current state of practice in roundabout safety and operational modeling. This literature review is presented later in this chapter. Site Inventory and Data Collection A major element of the study included updating and expanding the inventory of U.S. roundabouts compiled dur- ing recent research and volunteer efforts and making the inventory available to transportation professionals. The prod- ucts of this task include an updated database that includes information and data on as many roundabouts in the United States that the team could locate and retrieve information about, including components that are available on line. At selected sites, the research team collected and summa- rized extensive data on operational performance, safety per- formance, geometric parameters, and speeds. Specific data collection methods included assembly of crash reports, crash summaries, and plans; extensive video recording during peak and off-peak periods; and spot speed measurements using radar guns. These methods are described in detail in subse- quent chapters of this report. Operational Model Development Operational model development included the following tasks: • Evaluation of existing models and software. This task consisted of comparing the field data collected for each roundabout to the predictions from a wide range of exist- ing roundabout capacity models, plus two major software implementations in use in the United States (RODEL and aaSIDRA). The evaluation focused primarily on the ability of each model to predict capacities, delays, and/or queues under the geometric and traffic flow conditions observed at U.S. roundabouts. • Development of two operational models that attempt to best fit the U.S. data and explain the performance of U.S. roundabouts. The capacity models considered comprise the full range of potential formulations, including empiri- cal regression and analytical formulations (gap accep- tance). Delay and queuing models are based on those currently in use in the HCM for predicting performance at other unsignalized intersections. • Development of a draft revised HCM procedure that incorporates the findings from this research. Safety Model Development Unlike operational model development in the United States, where the HCM has been a definitive reference for more than 50 years, safety model development in the United States is in its infancy, with the first edition of the Highway Safety Manual still in development at the time of this research. Other countries have successfully developed safety models, but it has been unclear if these models are directly transferable to the United States. Using this international experience to guide the selection of variables and model forms, the research team performed a considerable amount of safety model development in this research: • Development of intersection-level safety performance functions (SPFs) that can be applied at a planning level for estimating the incidence of crashes. This develop- ment involved testing existing models and (1) recalibrat- ing them if feasible or (2) developing new models if the existing models were determined to be an inadequate fit to U.S. data. This latter step capitalized on insights gained from previous modeling experiences to identify model deficiencies resulting from omitted variables, incorrect functional forms, overfit models, and lack of causal variables. • Development and evaluation of approach-level SPFs to explore design relationships and assessment of the 5

value of approach-level predictive models. This devel- opment involved the same sequence of steps used for the intersection-level model development. • Exploration of the potential for speed-based safety models. The concept of a speed-based model that relates safety performance to absolute speeds and/or relative speeds (speed consistency) was pursued with the hope of providing an intermediate link to both safety and opera- tional performance. The rationale is that speed profiles are a manifestation of the driver’s response to a design. The work included testing and calibrating models for linking crashes to the speed profile and the speed profile to round- about characteristics. • Development of an expanded comparison of intersection safety before and after installation of roundabouts. This comparison between roundabouts and the form of control that preceded their installation was disaggregated as much as possible (e.g., urban versus rural, one lanes versus two lanes) to improve its utility to practitioners. Motorized Design Criteria The approach to developing updated design criteria had two major components: speed estimation, and incorporation of safety and operational findings into design criteria. Both efforts were conducted under the premise that the overall design methods currently in use in the United States are sound and that any findings from this study would supple- ment and augment those procedures, not completely replace them (unless findings suggested otherwise). The approach used for speed estimation in roundabouts involved collecting and comparing spot speed data in the field for various movements through the roundabout at key points along their paths. These speeds were then compared to cur- rent prediction techniques presented in FHWA’s Round- abouts: An Informational Guide (1) to test the overall veracity of the current methods and to propose alterations as needed to improve the fit of the models to the field data. For overall assessments of the effects of safety and opera- tional findings on design, three approaches were used. First, an overall set of descriptive statistics for the sites in the study were prepared to assess the overall safety performance of each roundabout by its general configuration (e.g., single-lane ver- sus multilane). Second, the safety and operational prediction models developed in their respective modeling efforts were examined for the relevance of various geometric parameters useful in design (e.g., entry width). Third, anecdotal evidence was used where modeling efforts were not sufficient to pro- vide insight on potential relationships between the design of the roundabout and its potential safety and/or operational performance. Pedestrian and Bicyclist Analysis Approach The approach to this study was to collect data related to pedestrian, bicyclist, and motorist behaviors from enough locations and for enough pedestrians and bicyclists to answer the questions posed in the introduction of this report. The analysis produced a series of descriptive statis- tics from the acquired and derived data for each site, which defined the actions and behaviors of pedestrians, bicyclists, and motorists. These behaviors were then compared across sites to determine which locations should be reviewed more closely. Those sites that appeared to produce behaviors substantially different from the mean values for like sites were reviewed to determine if there were geometric or oper- ational features at those locations that may have contributed to the observed behaviors. In addition to comparing the roundabout sites, the research team also compared the results from the pedestrian analysis in this study to those of a study being conducted for FHWA, titled “Safety Index for Assessing Pedestrian and Bicyclist Safety at Intersections” (3). Specifically, the pedestrian time and behavior results from the roundabout approaches in this study were compared to similar data that were collected for two-way-stop–controlled, all-way-stop–controlled, and sig- nalized intersections within the FHWA research study. The goal of this supplemental analysis was to shed light on any dif- ferences or similarities among these types of intersections with respect to pedestrian behaviors. The objectives of the observational analysis were to char- acterize how pedestrians and bicyclists interact with motor vehicles at roundabouts, assess safety on the basis of these observations, and determine if there is an association between the observed behaviors and the geometric and/or operational characteristics. The following specific questions were addressed for pedestrians: • How do pedestrians behave when crossing the leg of a roundabout? How do they respond to vehicles when preparing to cross or crossing the street? Do they cross within the crosswalk? Do they cross in one stage or two stages (using the splitter island as a refuge area)? • What is the yielding behavior of motorists when they encounter a pedestrian who is crossing or waiting to cross? • Did the behaviors of motorists and pedestrians create unsafe situations? Are there conflicts between motorists and pedestrians, and what are the underlying causes? • How do the behaviors of pedestrians and motorists at roundabouts, which are yield controlled, compare to the behaviors of pedestrians and motorists at other types of crossings, including those with no control, stop control, or signal control? 6

• What are the geometric or operational characteristics that tend to cause problems for pedestrians or that tend to result in safer and more accessible designs? Are there dif- ferences in behaviors between the entry side and exit side of a leg? Are there differences in behaviors between one- lane and two-lane legs? • Do any of the characteristics differ by region of the country? The questions are similar for bicyclists crossing a leg. However, for bicyclists entering the roundabout, additional questions were addressed: • How do motorists and bicyclists interact on the approach to the roundabout and within the circulating lane? Where do bicyclists position themselves; does the bicyclist “take the lane”? • Are there conflicts or avoidance maneuvers on the approach or within the circulating lane? • What types of behaviors do bicyclists exhibit that raise safety concerns (e.g., wrong-way riding, incorrect left turns)? Marketing Materials In addition to funding the scope of work, the AASHTO Standing Committee on Research approved additional fund- ing to develop marketing materials for roundabouts. These materials consist of a series of self-guided Microsoft® Power- Point™ presentations that can be used as is or adapted as needed to assist in communicating roundabout concepts to political and technical decision-makers. These presentations are available from the TRB website (http://www.trb.org/ news/blurb_detail.asp?id=7086). Literature Review To support the research approach for this project, this report presents an extensive literature review that was con- ducted to support the two major efforts to model safety and operational performance. In addition, this report includes a brief summary of current design guidance in use in the United States to provide background for the design recom- mendations. Additional supplemental literature for other components of this study is referenced in their respective discussions of findings. Safety Prediction Models To date, most of the research and literature dealing with the safety of roundabouts has focussed on the relative change in safety following the conversion of conventional stop- and/or signal-controlled intersections to roundabouts. The explicit quantification of roundabout safety, measured in terms of expected crash frequency, has thus far not received equal attention. However, for the designer, understanding the relationships between roundabout design features and crash frequency is imperative. This report reviews, by country of origin, published models that address the relationships between roundabout geometry and other factors, and safety. These models originate from the United Kingdom,Australia, France, and Sweden.Also reviewed are studies on the safety effect of converting conventional inter- sections to roundabouts. A summary of the safety models included in the review is provided in Table 1. Appendix A contains a comprehensive review of each source, by country of origin, followed by a summary indicating how useful the insights from this review were in guiding the research effort. (All appendixes have been published as NCHRP Web-Only Document 94 available on the TRB website [http://www.trb.org/news/blurb_detail. asp?id=7274]). A summary of the effect of each parameter according to the models from the United Kingdom, Australia, and Sweden is provided in Table 2. The table is broken into common and unique measures for each model. The French model (SETRA), which is not tabulated here, related only one vari- able (average annual daily traffic [AADT], which had a posi- tive effect) in a single model for all crash types combined. Measures common to two or more of the models are volume, pedestrian volumes, number of lanes on approach, number of circulating lanes, radius of the central island, radius of vehicle path, and approach curvature or deflection. Most fac- tors, with the exception of the radius of the central island, were found to have similar effects on safety (i.e., same posi- tive or negative direction). The literature review provided insight on how many sites may be needed for direct calibration. For example, models for other intersection types were typically based on samples of 300 to 450 sites. Conversely, the UK model for four-leg round- abouts used only about 80 sites (4), but there was quite a large variation in some of the key variables. On this basis, the research team confirmed that the development of safety mod- els for U.S. roundabouts would be a challenging task given the relatively few U.S. installations and the low numbers of crashes at them. Therefore, the research team concluded it would need to consider alternatives to the direct calibration of models relating crashes at roundabouts to all of the geometric and operational characteristics that may affect their safety. For the direct calibration of models, it was evident from the review that, even if large sample sizes were available, the characteristics of interest would need to vary enough to allow the relationship between crashes and these variables to be modeled. These difficulties appeared to be magnified 7

in the modeling of different crash types. However, the vari- ous crash types needed to be modeled, as others have done, to guard against the opposite effects of a variable being masked. For example, increased entry deflection might reduce entering-circulating crashes but increase rear-end crashes (though to a lesser degree). Most important, through the literature review, a wide array of variables for the safety analysis was identified. This list was useful in guiding the data collection and modeling efforts. Review of Before-After Safety Studies The research team also reviewed studies on the safety effect of converting conventional intersections to roundabouts and found that the results of these studies are usually reported without indicating whether regression-to-the-mean biases were considered in the analysis. Further, in most cases, the research team was unable to determine if this bias exists. Thus, the reader is cautioned to accept the results summa- rized here in the spirit in which this section is provided—to provide a flavor for the safety benefits of roundabouts. The decision to report these results in spite of possible reserva- tions was based on a belief that, with the very large reductions that were consistently observed, the benefits of roundabouts would remain substantial if regression-to-the-mean effects were removed and any other methodological limitations were to be overcome. Details on studies of conversions from other forms of intersections can be found in Appendix A. The one definitive study of U.S. conversions conducted for the Insurance Institute for Highway Safety (IIHS) (9), and subsequently updated for the New York State Department of Transportation (NYSDOT) (10), was based on a rather small sample size. As such, only limited disaggregate analysis could be done to try to isolate the geometric factors associated with the greatest safety benefits of roundabout construction. While some of these factors have been isolated in evaluations out- side of the United States, that knowledge may not be directly transferable. In addition, several of those studies had method- ological limitations. The review of the previous studies did provide useful insights for guiding the disaggregated before- after analysis for this study. Useful lessons were learned from the pitfalls and limitations of many of those studies (e.g., small sample sizes, ignoring regression to the mean, and improperly accounting for traffic volume changes over time). These lessons emphasized the need for, and the use to be made of, recent advances in safety estimation methodology aimed at overcoming these limitations. Capacity Models Capacity is a required input to delay and queuing models. In terms of existing U.S. capacity methodologies, the HCM 8 Country and Author Sample Size Constant Features Variable Features Model Input Parameters Total crashes/ roundabout • Vehicle AADT Total crashes/ crash type • Vehicle AADT United Kingdom: Maycock & Hall (4) 84 • Four legs • Single grade • Circular island • No unusual features • Island size • Speed Total crashes/ leg/crash type (geometric) • Vehicle AADT • Pedestrian volume • Entry width • Angle • Sight distance • Approach curve • Gradient • Radius • Percentage of motorcycles Australia: Arndt (5, 6) 100 None • Number of legs • Number of lanes • Urban/Rural • Island shape • Speed Total crashes/ leg/crash type • Vehicle AADT • Number of lanes • Speed variables • Vehicle path radius • Side friction Sweden: Brüde & Larsson (7) 650 N/A • Number of legs • Number of lanes • Speed limit Crashes/million of entering vehicles • Vehicle AADT France: SETRA (8) N/A N/A N/A Total crashes/ roundabout • Vehicle AADT Legend: AADT = Average annual daily traffic; N/A = Not available Table 1. Summary of safety models.

includes a gap acceptance model limited to single-lane roundabouts, and it does not provide any guidance on delay, queues, or level of service. The methods in FHWA’s Round- abouts: An Informational Guide (1) for one- and two-lane roundabout capacities were derived using the UK empirical model with assumed values for the six geometric input parameters. The German empirical capacity relationship was recommended for the operational analysis of an urban com- pact roundabout. These models were intended to be provi- sional until further research could be performed on U.S. roundabouts. A summary of the international capacity models is shown in Table 3. These models are either gap acceptance or linear/ exponential empirical relationships. Except for the UK model, there are few geometric parameters. Details on the types of capacity models in use, as well as a survey of international practices in estimating capacity, can be found in Appendix B. Overall Literature Review Summary The literature review provided the following useful insights that were used to guide the conduct of the NCHRP 3-65 research: • A wide array of variables for the safety and operational analyses were identified. The list was useful in guiding the data collection and modeling efforts. • While safety and operational prediction models have been developed successfully in other countries, it was unclear if 9 United Kingdom (Maycock & Hall) Australia (Arndt) Sweden (Brüde & Larsson) Measure SV A PP En t/C O th er Pe d SV R E En t/C Ex t/C SS A ll Cy cli st Pe d Measures Common to All Models1 AADT/volume + + + + + + + + + + + Pedestrian volumes + + Number of approaching lanes + + + + Number of circulating lanes + + + + Radius of central island – + See Note 2 – Radius of vehicle path – – – Approach curvature or deflection – – – Measures Specific to the United Kingdom Models (Maycock & Hall) (4) Road width at entry */+ */– +/– Percentage of motorcycles + + Angle to next leg * – Gradient +/+% +/–% Sight distance + + Weaving length between splitter islands * * Distance to first sight of roundabout * Average flare length * Measures Specific to the Australia Models (Arndt) (5, 6) Length of vehicle path + 85th percentile speeds + + + + Reduction in 85th percentile speed + Potential side friction + Measures Specific to the Sweden Models (Brüde & Larsson) (7) Three legs instead of four legs – Posted speed limit * * + Presence of bicycle crossings – Legend: SV = single vehicle; APP = approaching; Ent/C = crashes between an entering and a circulating vehicle; Other = other non- pedestrian crashes; Ped = pedestrian crashes; RE = rear-end crashes on approach; Ext/C = crashes between an exiting vehicle and a circulating vehicle at multilane roundabouts; SS = sideswipe crashes on two-lane segments. + = an increase in this measure increases crash frequency – = an increase in this measure decreases crash frequency * = the measure had a significant relationship with crash frequency but the relationship was not specified Notes: 1The French model (SETRA) (8) is inappropriate to tabulate here because it related only one variable (AADT) in a single model for all crash types combined. AADT had a positive effect. 2Optimum 10 m to 25 m Table 2. Summary of geometric, traffic, and other characteristics affecting safety.

10 Country Author Type Applicability Input Parameters Comments Wu (11) Gap Acceptance One to three lanes • Circulating flow • Number of lanes • Critical headway • Follow-up headway • Minimum gap Recommended model in Germany. Based on Tanner (12) Brilon et al. (13) Linear Regression One to three lanes • Circulating flow Refined for one lane Brilon et al. (14) Linear Regression One to three lanes • Circulating flow No longer applicable for one lane Germany Stuwe (15) Exponential Regression One to three lanes • Circulating flow • Number of lanes • Number of legs • Diameter • Travel distance Limited geometric range applicable Simon (16) Linear Regression One lane, bus lane • Circulating flow Not applicable to two or more lanes Switzerland Lausanne, as reported in Bovy et al. (17) Linear Regression One to three lanes • Circulating flow • Entering flow • Conflict length Three unique formulas; one lane limited to D = 22-32 m HCM (2) Gap Acceptance One lane • Circulating flow • Critical headway • Follow-up headway Provisional method. Based on Harders (18) Linear Relationship Urban Compact • Circulating flow See Brilon et al. (13) Linear Relationship One lane, Diameter = 30-40 m • Circulating flow See Kimber (19) USA Robinson et al. (1) Linear Relationship Two lanes, Diameter = 55-60 m • Circulating flow See Kimber (19) UK Kimber (19) Linear Regression All • Circulating flow • Entry width • Approach half- width • Effective flare length • Entry angle • Entry radius • Diameter Large sample of observed capacities CETE Quest (20) Exponential Regression All • Circulating flow • Exiting flow on leg • Entry width • Width of splitter • Width of circulatory lane Girabase method. Most widely used in France Louah (21) Linear Relationship N/A • Circulating flow • Exiting flow on leg France CETUR (22) Linear Relationship One lane • Circulating flow Adjustments have been developed for different geometric factors Table 3. Summary of operational models.

these models were directly transferable to the United States. Therefore, a considerable amount of model testing and subsequent model development was required for this research. However, direct transfer of the models did appear feasible and the experience from elsewhere could be used, at least, to guide the selection of variables and model forms. • New safety and operational models should be sensitive to the volume and variation of data acquired by past studies, recognizing that the U.S. database is inherently less rich at this stage in U.S. roundabout development. • Few before-after safety studies of roundabout installations have been methodologically sound. Lessons were learned from the pitfalls and limitations of these studies (e.g., small sample sizes, ignoring regression to the mean, improperly accounting for traffic volume changes over time). Even the later before-after studies that learned some of these lessons suffered from small sample sizes that limited the disaggre- gate analysis aimed at identifying the factors associated with the safety benefits of roundabouts. However, it is fea- sible and useful to capitalize on the recent advances in safety estimation methodology and a now rich sample of U.S. conversions to do, as part of NCHRP 3-65, a before- after study that would in a disaggregate analysis identify a larger number of factors associated with the safety benefits of roundabouts than was possible before. 11 Country Author Type Applicability Input Parameters Comments CROW (23) Range from macro to micro models N/A N/A Approximate and calculation methods CROW (23), Botma (24) One lane • Circulating flow • Exiting flow on leg • Number of bicycles Netherlands Arem & Kneepkens (25) Gap Acceptance • Circulating flow • Exiting flow on leg • Critical headway • Follow-up headway • Minimum gap Believed to be poorly researched. Based on Tanner (12) Sweden CAPCAL2 (26) Gap Acceptance One to two lanes • Percentage of heavy vehicles • Critical headway • Follow-up headway • Minimum gap • Proportion of random arrivals • Length of weave area • Width of weave area Guidebook based on Australian methods; Critical headway - based on geometry Israel Polus & Shmueli (27) Exponential Regression One lane • Number of legs • Number of lanes • Speed limit Units are not specified Australia Troutbeck (28) Gap Acceptance One to three lanes • Circulating flow • Turning flow • Entry flow • Number of lanes • Entry width • Diameter • Critical headway • Follow-up headway Separate equations for left and right lanes. Insufficient sites to develop linear regression equations Austria Fischer (29) Linear Regression One lane, Diameter = 23-40 m • Circulating flow Similar to Swiss method Table 3. (Continued).