Citation

Das S, Brimley BK, Lindheimer TE, Zupancich M. IATSS Res. 2018; 42(3): 143-151.

Copyright

(Copyright © 2018, International Association of Traffic and Safety Sciences, Publisher Elsevier Publishing)

DOI

10.1016/j.iatssr.2017.10.003

PMID

unavailable

Abstract

Most of the information necessary for driving a vehicle is regarded as visual information. In spite of its importance, visibility conditions at the time of a crash are often not documented at a high level of detail. Past studies have not examined the quantified values of visibility and its association with crashes. The current study merged data collected from the National Oceanic and Atmospheric Administration (NOAA) with 2010-2012 Florida crash data. From the thousands of logged weather events compiled by the NOAA, the researchers isolated periods of normal visibility and comparable periods of reduced visibility in a matched-pairs study. The NOAA data provided real visibility score based on the spatiotemporal data of the crashes. Additionally, the crash data, obtained from Roadway Information Database (RID), contains several geometric and traffic variables that allow for effects of factors and visibility. The study aims to associate crash occurrence under different levels of visibility with factors included in the crash database by developing ordinal logistic regression. The intent is to observe how different visibility conditions contribute to a crash occurrence. The findings indicate that the likelihood of a crash increase during periods of low visibility, despite the tendency for less traffic and for lower speeds to prevail during these times. The findings of this study will add valuable knowledge to the realm of the impact of visibility in the way of using and designing appropriate countermeasures.

Language: en

Keywords

Inclement weather; Ordinal logistic; Random forest; Variable importance