Calibrating Highway Safety Manual Equations for Application in Florida

1. Calibrating Highway Safety Manual Equations for Application in Florida Dr. Siva Srinivasan, Phillip Haas, Nagendra Dhakar, and Ryan Hormel (UF) Doug Harwood and Darren Torbic (MRI) Funded by Florida Department of Transportation

2. Highway Safety Manual • Published by AASHTO in 2010 • Provides tools to conduct quantitative safety analyses • Methods for a roadway safety management program • Predictive methods to estimate crash frequency by severity • Crash modification factors (CMFs)

3. Rural two-way two-lane highways Undivided segments Three-leg minor stop controlled intersections Four-leg minor stop controlled intersections Four-leg signalized intersections Rural multilane highways Undivided segments Divided segments Three-leg minor stop controlled intersections Four-leg minor stop controlled intersections Four-leg signalized intersections Urban and Suburban Arterials Two-lane undivided segments Three-lane segments with two-way left-turn lanes Four-lane undivided segments Four-lane divided segments Five-lane segments with two-way left-turn lanes Three-leg minor stop controlled intersections Four-leg minor stop controlled intersections Three-leg signalized intersections Four-leg signalized intersections HSM Facility Types

4. HSM Crash Prediction • Segments • N = exp[ a + b×ln(AADT) + ln(Length) ] • Intersections • N = exp[ a + b×ln(AADTmajor) + c×ln(AADTminor) ] Npredicted= predicted average crash frequency for an individual site for a specific year CMF1x…CMFyx= crash modification factors specific to facility type x and geometric design and traffic control features y Cx= calibration factor to adjust SPF for local conditions for facility type x Nspf= predicted average crash frequency for base conditions

5. Why Calibrate? • HSM models were developed in national studies • Segments based on Washington • Intersection based on California • Calibrate for local factors • Weather • Driver behavior and population • Crash reporting methods & thresholds • Animal populations • Distinct calibration factor for each facility type and crash severity inclusion

6. HSM Calibration Procedure • Identify facility types • Collect road characteristic data • Collect crash data • Apply SPF for crash prediction • Apply CMFs • Compute calibration factor

7. Segment Calibration

8. Roadway Data • Year-end archives of FDOT RCI obtained for 2005-2008 • Roadways split into homogeneous segments • Segments with incomplete or inaccurate data removed

10. Crash Data • Crashes extracted from the Crash Analysis Reporting System (CARS) • 2005-2008 • Only includes fatal and injury crashes • CARS only includes long-form crash records • Calibration factors developed for fatal and injury crash frequency prediction • Intersection crashes removed • “occurring at an intersection” • “influenced by an intersection” • Crashes assigned to segments based on Roadway ID and milepost

11. HSM Facility Types

12. Statewide Segment Calibration Results

13. Intersection Calibration

14. Intersection Data • Listing of all Florida intersections obtained from FDOT Safety Office • Restricted to analysis of only HSM defined facility types • Rural two-lane two-way roads and multilane highways • Three-leg stop controlled • Four-leg stop controlled • Four-leg signalized • Urban arterials • Three-leg stop controlled • Three-leg signalized • Four-leg stop controlled • Four-leg signalized • Restricted to analysis of intersections of two state maintained roads • AADT and crash data not available for non-state roads

15. Intersection Characteristics • Two primary sources • RCI • Available for each year individually • Google maps • Only available for date of satellite image • Assumption made that these data were applicable for all years • Site removed if construction was present in image • Site removed if site characteristics in RCI changed

17. Google Data Collection Approaches with LT Lanes: 4 CMF1i = 0.66 Approaches with RT Lanes: 2 CMF3i = 0.92

18. Approaches with LT Lanes: 4 CMF1i = 0.66 Approaches with RT Lanes: 2 CMF3i = 0.92 LT-Protected: 4 Approaches CMF3i = 0.944 Lighting Present CMF3i = 0.91 No Red-Light Cameras RTOR Permitted

19. 3 Bus Stops: CMF1p = 4.15 1 School: CMF2p = 1.35 4 Alcohol Establishments: CMF3p = 1.12 Pedestrian Volume: Med-Low (240/day) Max Lanes Crossed: 9

20. Crash Data • Crashes extracted from the Crash Analysis Reporting System (CARS) • 2005-2009 • Fatal and injury crashes only • CARS only includes long-form crash records • Intersection crashes identified • “occurring at an intersection” • “influenced by an intersection” • Crashes assigned based on node ID

21. Intersection Facility Types

22. Intersection Calibration Results

23. Intersection Calibration Spreadsheet Sample

24. Intersection Calibration Spreadsheet • Can be used for future recalibration • Specific to each facility type • Requires user entry of AADT and crash data • Hidden cells allow for intersection characteristics to be updated as needed • Lighting, turn lanes, etc. • Calculates SPFs and CMFs to determine predicted crashes • Calculates calibration factor for both KAB and KABC severity levels

25. HSM Calibration Summary • Calibration allows for HSM implementation in Florida • Calibration factors developed for intersection and segments • Fatal and injury crashes only • Calibration benefits of segment calibration • Improved prediction over uncalibrated models (error, absolute error, variance) • Limited data availability for intersections • Significantly restricted by requiring the intersection of two state maintained roads • Significant manual effort involved for data collection

26. Future Potential HSM Research Areas • Expansion to include models and calibration for more facility types • Development of CMFs for additional geometric design, traffic control, and area attributes • Implementation of localized calibration – by region or areas with otherwise similar characteristics • Exploration of alternative crash prediction methods and modeling structures • Investigation of crash prediction validation and comparison metrics

27. Thank you!