NCHRP 3-65: Applying Roundabouts in the United States Preliminary Findings Lee A. Rodegerdts, P.E. AASHTO Subcommittee on Design, Chicago, IL June 27, 2005 Bhagwant Persaud, Canada David Harkey, USA George Mason University, USA McDonald & Partners, UK Rensselaer Polytechnic University, USA Rod Troutbeck, Australia Ruhr-University Bochum, Germany University of Idaho, USA
Topics of Discussion Project panel and team Project need and objective Preliminary findings Safety Operations Design Anticipated products
Project panel Beatriz Caicedo-Maddison, Florida DOT (chair) Maria Burke, Texas DOT Jerry Champa, California DOT Leonard Evans, Science Serving Society Steve King, Kansas DOT Robert Limoges, New York State DOT Richard Long, Western Michigan University Ron Pfefer, HSM liaison Brian Walsh, Washington State DOT Mohsin Zaidi, City of Kansas City, MO Joe Bared, FHWA Hari Kalla, FHWA Rich Cunard, TRB Ray Derr, NCHRP
Project team P.I.: Lee Rodegerdts (KAI) (Bruce Robinson, Co-P.I. Emeritus) USA Kittelson & Associates, Inc. University of Idaho Rensselaer Polytechnic Institute George Mason University David Harkey John Mason Australia Rod Troutbeck Canada Bhagwant Persaud Germany Werner Brilon United Kingdom Richard Hall
U.S. practice relies heavily on the experience from other countries. Current U.S. procedures depend on international methods without having U.S. data for calibration Use of roundabouts in the U.S. may differ from that experienced in other countries
Overview of research tasks 1. Summarize Existing Relationships 2. Model Formulation 3. Data Collection Plan 4. Interim Report 5. Execute the approved data-collection plan 6. Inventory U.S. Roundabout Sites 7. Operational Performance Methods 8. Safety Performance Methods 9. Design Criteria 10. Final Report 11. Prepare marketing materials
Preliminary Safety Findings Roundabout-level accident models Approach-level accident models Before-after study of intersections converted to roundabouts
Roundabout-Level Crash Prediction Used for comparing performance to other intersection types Baseline prediction on which approach-level CMFs could be applied Form: Crashes = α(aadt) β Factors affecting coefficients: Number of lanes Number of approaches
Approach-Level Crash Data (139 approaches) Total Number of Approach Crashes Total Number of Crashes 200 150 100 50 0 Single Lane Multi-Lane Entering Circulating Exiting/Circulating Rear End on Approach Loss of Control Pedestrian Bicycle Crash Type
% change in crashes from candidate approach level models per unit change in variable Variable Entering/ Circulating Exiting/ Circulating Approach Entry Radius (ft.) 1% reduction Entry Width (ft.) 5% increase Approach Half Width (ft.) 3% increase Inscribed Circle Diameter (ft.) 2.2% increase Central Island Diameter (ft.) 0.5 to 0.8% reduction 1.4% increase Circulating Width (ft.) 12% increase Angle To Next Leg (degree) 3% reduction
Before-After Results All Sites (55) All Injury Crashes recorded in after period 726 72 EB estimate of accidents expected after without roundabouts 1122 296 Reduction (Standard error) 35.4 % (3.4) 75.8 % (3.2) CONTROL BEFORE All Injury SIGNALS (9) 48% 78% TWO WAY STOP (34) 44% 82% ALL-WAY STOP (10) Insignificant increase
Preliminary Operations Findings Analysis of existing models Driver behavior and effect of geometry HCM recommendations
Analysis of Existing Models All international models (including SIDRA and RODEL) predict capacities higher than observed 2000 WA04-N (Port Orchard) Max Entering Flow (pcus/hr) 1500 1000 500 Raw Data Akcelik Uncalibrated Kimber Uncalibrated Linear (Raw Data) 0 0 250 500 750 Conflicting Flow (pcus/hr)
Influence of Flow & Geometry on Driver Behavior Entry lane width = entry width / # lanes 7.0 6.0 All tf All tc Linear (All Gap Parameters 5.0 4.0 3.0 y = -0.0079x + 4.8345 R 2 = 0.0003 y = -0.047x + 3.408 R 2 = 0.0336 2.0 1.0 3.0 5.0 7.0 9.0 Entry Width
Multilane Modeling Issues Several sites exhibit strong queuing in only one lane Possible causes: Turning movement effects Lane use assignment (or lack thereof) Geometric effects (vehicle path overlap) Driver unfamiliarity Model intended to allow designer to capture these first-order effects apparent in U.S. data
Proposed HCM Capacity Models Single-lane: Current HCM model with new t c, t f Multi-Lane: Exp. regression model for critical lane Entry capacity (veh/h) 1400 1200 1000 800 600 400 200 Roundabout Entry Capacity 0 0 500 1000 1500 2000 Conflicting flow (veh/h) Single-Lane Capacity Multilane Critical Lane Capacity
Preliminary Design Findings Design speed modeling Other design findings for motor vehicles Pedestrian and bicycle observations
Current FHWA speed prediction method is based on AASHTO speed-radius function. 40 35 30 Speed (mph) 25 20 15 10 5 0 0 50 100 150 200 250 300 350 400 Radius (ft) e=+0.02 e=-0.02
Design speed modeling: V4, Left-turn turn circulating speed (all sites) 35 Actual Speed, V4a (mph), All Sites 30 25 20 15 10 5 y = 1.1041x - 1.8409 R 2 = 0.6483 0 0 5 10 15 20 25 30 35 Predicted Speed, V4p (mph), All Sites Data Match Line 85th %ile (15+ obs.) Linear (85th %ile (15+ obs.))
Design speed modeling: Exit speed (all sites), unadjusted Actual Exit Speed, V3a or V6a (mph), All Sites 50 45 40 35 30 25 20 15 10 5 0 y = 0.2513x + 13.834 R 2 = 0.2933 0 5 10 15 20 25 30 35 40 45 50 Unadjusted Predicted Exit Speed, V3pbase or V6pbase (mph), All Sites V3 Data Match Line 85th %ile (15+ obs.) V6 Data Linear (85th %ile (15+ obs.))
Proposed exit speed equation V 3 where: = min 1 1.47 V 3 = V 3 speed, in mph V (1.47V 3 pbase 2 ) 2 + V 3pbase = V 3 speed predicted based on path radius, in mph V 2 = V 2 speed predicted based on path radius, in mph 2a 23 d a 23 = acceleration along the length between the midpoint of V 2 path and the point of interest along V 3 path = 6.9 ft/s 2 d 23 = distance between midpoint of V 2 path and point of interest along V 3 path, in ft 23 Speed where exit radius is limiting factor Speed where circulating speed and acceleration distance is limiting factor
Design speed modeling: Exit speed (all sites), adjusted Actual Speed, V3a or V6a (mph), All Sites 50 45 40 35 30 25 20 15 10 5 0 y = 0.6694x + 5.9115 R 2 = 0.5156 0 5 10 15 20 25 30 35 40 45 50 Adjusted Predicted Speed, V3p2 or V6p2 (mph), All Sites Match Line 85th %ile (15+ obs.) Linear (85th %ile (15+ obs.))
Implications for design Tangential or nearly tangential exits do not appear to cause excessive vehicle exit speeds if the following conditions are met: The speed of circulation (V2 and V4) is kept low The distance between the start of the exit path and the point of interest (e.g., crosswalk) is kept short Similar prediction adjustment for entry speeds Entry speed appears to be limited by drivers anticipation of the speed needed for circulation However, recommend continued reliance on entry path curvature as a primary method to control entry speed
Entry width and lane width Narrow lane widths (entry and circulating) at multilane roundabouts appear to have a detrimental effect on safety Entry width: Aggregated entry width (number of lanes) has a clear safety and operational effect Variations of lane width appear to be second-order effects
Multilane roundabout issues Higher crash frequencies and crash rates than single-lane roundabouts Vehicle path overlap, poor striping apparent contributors Anecdotal evidence suggests that their correction can substantially improve safety performance
Example: Clearwater Beach, FL, before and after design modifications Before (2001) Photo: Bruce Robinson After (2005) Photo: Lee Rodegerdts
Non-motorized users Examination of observed field behaviors for two groups: Pedestrians Bicyclists Pedestrian data: 10 approaches at 7 sites; 769 events Bicyclist data: 14 approaches at 7 sites; 690 events Geographic diversity: California, Florida, Maryland, Nevada, Oregon, Utah, Vermont, Washington
How do motorists behave when encountering pedestrians? Motorists failing to yield to pedestrians All sites: 30 percent Entry leg: 23 percent Exit leg: 38 percent 1-lane approaches: 17 percent 2-lane approaches: 43 percent
How do pedestrian behaviors at roundabouts compare to other forms of control? Crossing control Uncontrolled Percent of normal crossings 70% Percent of nonyielding vehicles 48% Roundabout 85% 32% Stopcontrolled Signalcontrolled 90% 100% 15% 4%
Anticipated products Final report Draft Highway Capacity Manual procedure Components compatible with a possible Highway Safety Manual procedure Updated design research for use in future updates to FHWA Roundabout Guide, AASHTO Green Book Data that is accessible for future research Problem statement(s) for continued research Anticipated completion: December 2005
Questions? (503) 228-5230 or (800) 878-5230 Lrodegerdts@kittelson.com Photo: Lee Rodegerdts