Crash Prediction Models for Roads Including Rainfall and Hazardous Points

Predictive models for multilane roads including also the effects of rainfall and hazardous points on the crash frequency are presented in this paper. Accident data observed on a specific motorway during a period of eight years were used in statistical analysis. Negative Binomial Distribution, was applied for modelling the random variation of the crash data. Model parameters were estimated by Maximum Likelihood Method, and the Generalised Likelihood Ratio Test was used to detect the significant variables to be included in the final model. Separate prediction models for curves and tangents are proposed. For curves, it is found that the crash frequency is positively associated with the following variables: length (L), curvature (1/R), and the presence of hazardous points such as junctions (J) or tunnels (T); whereas for tangents with L and T. The effect of rain precipitation shows that with a wet pavement significant increases in crash frequency are expected, more especially on curves compared to tangents. This might be due to the combined effect of rain and centrifugal force in curve