U.S. Department of Transportation
Federal Highway Administration
1200 New Jersey Avenue, SE
Washington, DC 20590
2023664000
Federal Highway Administration Research and Technology
Coordinating, Developing, and Delivering Highway Transportation Innovations
This report is an archived publication and may contain dated technical, contact, and link information 

Publication Number: FHWARD03037
Date: May 2005 

Validation of Accident Models for IntersectionsFHWA Contact: John Doremi, PDF Version (1.61 MB)
PDF files can be viewed with the Acrobat® Reader® 1. INTRODUCTIONEffective safety management of a highway system requires that engineers know the present safety performance of a roadway and how it will perform if contemplated actions are taken. In effect, a reliable method of estimating safety performance is required. To this end, FHWA and its contractors have developed a new approach that combines historical accident data, regression analysis, beforeandafter studies, and expert judgment to make safety performance predictions that are expected to be better than those obtained by any of the individual approaches. A recent report documents an accident prediction algorithm for implementing the new approach for twolane rural highway sections that include road segments and five types of intersections.^{(3)} Ongoing efforts aim to produce similar documents for other types of facilities. The accident prediction algorithm has been developed for incorporation in the IHSDM as the Crash Prediction Module, but is suitable for standalone applications. The structure of the accident prediction algorithm for the five types of rural atgrade intersections is as follows:
where N_{int} = predicted number of total intersectionrelated accidents per year after application of AMFs; N_{b} = predicted number of total intersectionrelated accidents per year for base conditions; and AMF_{1} AMF_{2} ... AMF_{n} = AMFs for various intersection features. Harwood et al. presented base models and AMFs for three and fourlegged intersections of twolane rural roads with STOP control, and fourlegged signalized intersections of twolane roads.^{(3)} These base models were the best of available accident prediction models developed in earlier FHWA projects and retained model variables that were statistically significant at the 15 percent level.^{(1,2)} Those projects also resulted in regression models with additional variables indented for use as AMFs in IHSDM. The full models, along with several variants, are presented in two FHWA reports: Vogt and Bared present models for three and fourlegged intersections of twolane roads, while Vogt documents models for three other types of rural intersections: three and fourlegged stop controlled with four lanes on the major and two on the minor; and signalized intersections of twolane roads.^{ (1,2)} In summary, there are five types of intersection accident prediction models pertaining to the research efforts described:
The models were developed using data that were limited in terms of geographical diversity and, in the case of Types III, IV, and V, sample size. Thus, validation of these models across both space and time has become of paramount importance. It was also of interest to validate the accident prediction algorithm as a whole, given its novelty and the fact that it relies on expert judgment for deriving the AMFs. An additional report provides additional AMFs for left and rightturn lanes for atgrade intersections at type I, II and V sites.^{(5) }These AMFs are also included in the validation effort. A natural followon to model validation is model recalibration using validation data and findings to improve the specification of the intersection models. This report presents recalibration results for the five types of rural intersections that were the subject of the validation exercise that was undertaken in the first part of the project. This model recalibration effort complemented the comprehensive model validation conducted as part of a larger technical evaluation of crash prediction models. AADT models and fully parameterized models were recalibrated, and their results are discussed in the recalibration chapter. This report consists of three chapters. This first chapter provides an introduction and presents the description of the variables used in this research. The second chapter describes four different sets of activities conducted to assess the validity of prediction models for the five types of rural intersections. The third chapter presents the recalibration efforts for the five types of intersections. The model recalibration efforts described in this chapter complement the comprehensive model validation exercise presented in the previous chapter as a part of larger technical evaluation of the models. 1.1 VARIABLE ABBREVIATIONSThis section provides, for ease of reference, the definitions of model types, accident types, and variables applied in this research investigation. 1.1.1 Model Types
1.1.2 Accident ModelsModels I and II Total: Total number of policereported intersectionrelated accidents within 76.25 meters (m) (76.25 m (250 ft) (ft)) of the intersection. Injury: Total number of policereported intersectionrelated injury accidents within 76.25 m (250 ft) of the intersection. Models III, IV, and V TOTACC: Total number of accidents within 76.25 m (250 ft) of the intersection. TOTACCI: Only those crashes considered intersectionrelated and within 76.25 m (250 ft) of the intersection. INJACC: Total number of injury crashes within 76.25 m (250 ft) of the intersection. INJACCI: Only those injury crashes considered intersectionrelated and within 76.25 m (250 ft) of the intersection. 1.1.3 Definitions of VariablesAADT1: Average daily traffic on major road (vehicles per day). This variable is identical to ADT1 in the original (published) models. This change was made after determining that the traffic flow variables were, in fact, estimated AADT. AADT2: Average daily traffic on minor roads (vehicles per day). This variable is identical to ADT2 in the original models. COMDRWY1: Commercial driveways on major roads within 76.25 m (250 ft) of the intersection center. This variable is identical to NODRWYC1 in the original models. COMDRWY2: Commercial driveways on minor roads within 76.25 m (250 ft) of the intersection center. This variable is identical to NODRWYC2 in the original models. DRWY1: Driveways on major roads within 76.25 m (250 ft) of the intersection. This variable is identical to ND and NODRWY1 in the original models for Type III and IIIV intersections, respectively. DRWY2: Driveways on minor roads within 76.25 m (250 ft) of the intersection. This variable is identical to NODRWY2 in the original models. GRADE1: Average absolute grade on major roads within ±244 m (±800 ft) of the intersection center (percent). GRADE2: Average absolute grade on minor road within ±244 m (±800 ft) of the intersection center (percent). HAU: Intersection angle variable defined where the angle between the major and minor roads is measured from the far side of the minor road:
HAZRAT1: Roadside hazard rating on major road within 76.25 m (250 ft) of the intersection center (from 1, least hazardous case, to 7, most hazardous case).^{(6)} This variable is identical to RHRI in the original models for Type I and II intersections. HAZRAT2: Roadside hazard rating on minor road within 76.25 m (250 ft) of the intersection center (from 1, least hazardous case, to 7, most hazardous case).^{(6)} HEI1: Sum of degree of curve in degrees per hundred feet of each horizontal curve on major road, any portion of which is within ±244 m (±800 ft) of the intersection center divided by the number of such curves. HEI2: Sum of degree of curve in degrees per hundred feet of each horizontal curve on minor road, any portion of which is within ±244 m (±800 ft) of the intersection center divided by the number of such curves. HEICOM: (1 / 2) (HEI1 + HEI2). HI: Sum of degree of curve in degrees per hundred feet of each horizontal curve on major road, any portion of which is within 76.25 m (250 ft) of the intersection center divided by the number of such curves. HI1: Sum of degree of curve in degrees per hundred feet of each horizontal curve on major road, any portion of which is within 76.25 m (250 ft) of the intersection center divided by the number of such curves. This variable is identical to HI in the original models for Type I and II intersections. HI2: Sum of degree of curve in degrees per hundred feet of each horizontal curve on minor road, any portion of which is within 76.25 m (250 ft) of the intersection center divided by the number of such curves. HICOM: (1 / 2) (HI1 + HI2). L1LT (Type IIIV): Leftturn lane on major roads (0 = no, 1 for one approach, and 2 for both approaches). This variable is identical to LTLN1 in the original models for Type IIIV intersections. L1RT (Type IIIV): Rightturn lane on major roads (0 = no, 1 for one approach, and 2 for both approaches). This variable is identical to RTLN1 in the original models for Type IIIV intersections. L3LT (Type IIIV): Leftturn lane on minor roads (0 = no, 1 for one approach, and 2 for both approaches). This variable is identical to LTLN2 in the original models for Type IIIV intersections. L3RT (Type IIIV): Rightturn lane on minor roads (0 = no, 1 for one approach, and 2 for both approaches). This variable is identical to RTLN2 in the original models for Type IIIV intersections. LEGACC1: Acceleration lane on major roads (0 = no, 1 = yes). LEGACC2: Acceleration lane on minor roads (0 = no, 1 = yes). LIGHT: Light at intersection (0 = no, 1 = yes). LTLN1S (Type IIIV): Leftturn lane on major roads (0 = no, 1 = yes). LT MAJ (Type III): 1 if leftturn lane exists on at least one approach of major roads, 0 otherwise. LT MIN (Type III): 1 if leftturn lane exists on at least one approach of minor roads, 0 otherwise. MEDIAN (Type III): 1 if median exists on major roads, 0 otherwise. MEDTYPE (Type IIIV): Median type (0 = no median, 1 = painted, 2 = curbed, 3 = others). MEDWDTH1: Median width on major roads (feet). This variable is identical to MEDWIDTH1 in the original models. MEDWDTH2: Median width on minor roads (feet). PKLEFT: Peak leftturn percentage (percent). PKLEFT1: Peak leftturn percentage on major roads (percent). PKLEFT2: Peak leftturn percentage on minor roads (percent). PKTHRU1: Peak through percentage on major roads (percent). PKTHRU2: Peak through percentage on minor roads (percent). PKTRUCK: Peak truck percentage passing through the intersection (percent). PKTURN: Peak turning percentage (percent). PROT_LT: Protected left lane (0 = no, 1 = yes). RESDRWY1: Residential driveways on major roads within 76.25 m (250 ft) of the intersection center. This variable is identical to NODRWYR1 in the original models. RESDRWY2: Residential driveways on minor roads within 76.25 m (250 ft) of the intersection center. This variable is identical to NODRWYR2 in the original models. RT MAJ (Type III): 1 if rightturn lane exists on major roads, 0 otherwise. RT MIN (Type III): 1 if rightturn lane exists on minor roads, 0 otherwise. SD1: Longitudinal sight distance on major roads (feet). SDL2: Leftside sight distance on minor roads (feet). SDR2: Rightside sight distance on minor roads (feet). SHOULDER1: Shoulder width on major roads (feet). SHOULDER2: Shoulder width on minor roads (feet). SPD1: The average posted speed on major roads in vicinity of the intersection (miles per hour). This variable is identical to SPDI in the original models for Type I and II intersections. SPD2: The average posted speed on minor roads in vicinity of the intersection (miles per hour). TERRAIN (Type III): 1 if flat, 2 if rolling, or 3 if mountainous terrain. TERRAIN1 (Type IIIV): Terrain on major roads within 76.25 m (250 ft) of the intersection center (0 = flat, 1 = rolling, 2 = mountainous). TERRAIN2 (Type IIIV): Terrain on minor roads within 76.25 m (250 ft) of the intersection center (0 = flat, 1 = rolling, 2 = mountainous). VCEI1: Sum of absolute change of grade in percent per hundred feet for each crest curve on major roads, any portion of which is within 800 feet of the intersection center, divided by the number of such curves. VCEI2: Sum of absolute change of grade in percent per hundred feet for each crest curve on minor roads, any portion of which is within 800 feet of the intersection center, divided by the number of such curves. VCI1: Sum of absolute change of grade in percent per hundred feet for each crest curve on major roads, any portion of which is within 76.25 m (250 ft) of the intersection center, divided by the number of such curves. This variable is identical to VCI in the original models for Type I and II intersections. VCI2: Sum of absolute change of grade in percent per hundred feet for each crest curve on minor roads, any portion of which is within 76.25 m (250 ft) of the intersection center, divided by the number of such curves. VEI1: Sum of absolute change of grade in percent per hundred feet for each curve on major roads, any portion of which is within ±244 m (±800 ft) of the intersection center, divided by the number of such curves. VEI2: Sum of absolute change of grade in percent per hundred feet for each curve on minor roads, any portion of which is within ±244 m (±800 ft) of the intersection center, divided by the number of such curves. VEICOM: (1 / 2) (VEI1 + VEI2). VI1: Sum of absolute change of grade in percent per hundred feet for each curve on major roads, any portion of which is within 76.25 m (250 ft) of the intersection center, divided by the number of such curves. VI2: Sum of absolute change of grade in percent per hundred feet for each curve on minor roads, any portion of which is within 76.25 m (250 ft) of the intersection center, divided by the number of such curves. VICOM (Type IIIV): (1 / 2) (VI1 + VI2). 