ACCIDENT PREDICTION MODELS FOR TWO-LANE RURAL HIGHWAYS

Approximately 2.5 million miles or 63% of the highways in the U.S. are classified as two-lane rural highways, and 50% of fatalities occur on these highways. Past studies indicate that inconsistencies in highway system characteristics such as geometry, vehicle mix, level of road usage, environment, and driver behavior are the main causes of accidents. Of these, geometry has the potential to directly and indirectly influence the other characteristics. This report is based on a study aimed at modelling the influence of the geometric design variables on traffic accidents on two-lane rural highways. The objectives were to: (1) Review previously developed relationships between accidents and geometric variables, and identify significant variables; (2) Examine the significance of the above identified variables as well as other geometric variables on accidents in selected parts of highways 89 and 91 in northern Utah; (3) Establish statistically significant relationships between accidents and the geometric variables; and (4) Examine the spatial and temporal validity of the above relationships. It was found that the exposure in terms of distance travelled (length of road section) is the most significant variable in the empirical models developed in the present case. Contrary to previous findings, horizontal curvature and cross-section were found to have rather negligible effects on accident occurrence. Other findings of the study were: (1) Disaggregation of data by type of roadway section (curved or tangent) does not improve model predictability; and (2) Models developed in the present case are temporally and spatially transferable by updating parameters using Bayesian statistics whereas previously developed models are not.

• Record URL:
  https://www.ugpti.org/resources/reports/details.php?id=318
• Supplemental Notes:
  • This study was funded through the Mountain-Plains Consortium by the University Transportation Centers Program of the U.S. Department of Transportation.
• Corporate Authors:

  Mountain-Plains Consortium

  North Dakota State University
  Fargo, ND  United States  58108
• Authors:
  • Kalokota, K R
  • Seneviratne, P N
• Publication Date: 1994-5

Language

• English

Media Info

• Media Type: Digital/other
• Features: Appendices; Figures; References; Tables;
• Pagination: 113 p.

Subject/Index Terms

• TRT Terms: Crash exposure; Fatalities; Forecasting; Geometric design; Mathematical models; Rural highways; Traffic crashes; Trip length; Two lane highways
• Subject Areas: Highways; Safety and Human Factors; I81: Accident Statistics;

Filing Info

• Accession Number: 00669027
• Record Type: Publication
• Report/Paper Numbers: MPC Rept No. 94-32
• Files: TRIS
• Created Date: Nov 23 1994 12:00AM