This paper investigates the crash-rate distribution of the freeway weaving segments with a buffer-separated high-occupancy-vehicle lane. For this purpose, crash, traffic and geometry data were extracted from various sources. These extracted data were either spatially or spatiotemporally matched with one another, and used for both descriptive and model-based analyses. The descriptive analysis revealed that crash-rate distributions of the weaving segments depend not only on the class of the subject highway but also on the presence of an access point. This finding was statistically confirmed by crash frequency models. Notably, both descriptive and model-based analyses showed that weaving segments with an access point tend to show lower crash-rates than the counterparts without one. One might attribute this counterintuitive phenomenon to a certain underlying tendency by which traffic engineers or planners generally follow in selecting locations of an access point along the freeway. Further research is required to resolve such a conjecture.
Keywords: High-occupancy-vehicle lane; access point; freeway weaving; modelling crash frequency.