1 CRASH PREDICTION ON RURAL ROADS Cheng Zhong Research Assistant, Civil, Construction and Environmental Engineering, School of Engineering, University of Alabama at Birmingham, Birmingham AL, USA, e-mail: zhongch@uab.edu Virginia P. Sisiopiku Associate Professor, Civil, Construction and Environmental Engineering, School of Engineering, University of Alabama at Birmingham, Birmingham AL, USA, e-mail: vsisiopi@uab.edu Khaled Ksaibati Professor, Department of Civil & Architectural Engineering, College of Engineering and Applied Science, University of Wyoming, Laramie WY, USA, email: khaled@uwyo.edu Tao Zhong Transportation Analyst, Transportation Economics & Management Systems, Inc, Frederick MD, USA, email: transtao@gmail.com Submitted to the 3 rd International Conference on Road Safety and Simulation, September 14-16, 2011, Indianapolis, USA ABSTRACT Historical data confirm that rural roadways carry less than half of America’s traffic but account for the majority of the nation’s vehicular deaths. According to NHTSA, Wyoming has the highest crash fatality rate in the nation with a reported 2009 road death rate of 24.6 per 100,000 population, more than twice the national average of 11.0. High speed two-lane rural roads are believed to contribute to the fatal crash occurrence in rural states, such as Wyoming. An urgent need exists to systematically examine historical data to better understand contributing factors and develop countermeasures to improve traffic safety in rural settings. The paper discusses the development of a methodology that utilizes available data from Wyoming (crash records, traffic volume, speed, etc) for crash prediction on rural roads. Prediction models were developed by using regression analysis techniques and data from three counties. Two methods were used in the building process, namely the Negative Binomial Regression (NBR) and the Poisson regression methods. The paper describes the process for selection of candidate roads, data collection and processing, methods employed in model development, and findings and conclusions. Overall, the analysis showed that the NBR method better fitted the over-dispersed crash data available in the study. The proposed model demonstrated that high speed, in conjunction with high volume result in higher crash rates (number of crashes per mile in this study) at high risk locations. The results from the case study can be used to classify rural road segments according to crash risk as well as provide the foundation for similar crash prediction analyses in other states in the future. Keywords: Rural roads safety, Low volume roads, Crash prediction, Wyoming.