摘要: |
Accidents may cause nonrecurring congestion and result in additional delays, vehicle emissions, and other negative effects. Studies have been conducted on the modeling of the accident frequency, duration and even lane blockage. However, there has been less attention paid to the timely updating of the distributions of these variables. This report presents an application of the Bayesian approach to continuously update the distribution of the accident frequency, duration, and affected lanes based on new, available information. Data for the accident frequency in Harris County, Texas, from 1992 through mid-2000 and data for the accident duration and lanes affected in Houston, Texas, from 2000 to mid-2002 were used to construct the historical distributions, while information in recent 6 to 7 months were used to update the distributions. Poisson, lognormal, and binomial distributions were modeled for the accident frequency (including variations for time of day and day of week), duration, and lanes affected, respectively. The updating was realized through renewing the conjugate distributions of these modeled distributions, therefore updating key parameters inside the original modeled distributions. The entire process, though incorporating complex probability analysis, was implemented in Microsoft Excel in a rather simple form that practical engineers and researchers can use easily. The methods of microscopic and macroscopic estimation of emissions caused by nonrecurring congestion due to accidents were proposed and case studies in the Houston area show that extra emissions could be generated that would impact emissions in the entire area. |