K.B. Kelarestaghi, W. Zhang, Y. Wang, L. Xiao, K. Hancock, K.P. Heaslip
A traffic crash is the primary negative consequence resulting from repetitive interactions between four factors: road users, vehicles, roads, and weather conditions. Each factor possesses multiple attributes, some of which play bigger roles than others in causing a crash, particularly when they exceed certain threshold values or states. The values or qualitative states of some attributes are recorded in crash reports which enable statistical studies to investigate their collective impacts on crash frequency and severity. This paper presents the effect of adverse weather on crash severity through a macroscopic analysis. The data used comes from Pennsylvania’s 8-year statewide crash data from 2006 to 2013. On average, Pennsylvania has about 120,000 traffic crashes annually, of which 21.5% occur under adverse weather conditions. The objective is to understand the probability of having a severe crash when adverse weather combines with other causation factors. Crash records from counties that historically have higher likelihood of severe weather-related crashes were selected for model development. All attributes in the crash database were reviewed and the ones clearly related to the four causation factors were considered as possible variables in model development. Spearman correlation test was used to test the significance of each variable and the strength of correlation between different pairs of variables, and logistic regression model were then developed for each group of independent variables that enable assessing the impact to crash severity by different causal factors acting alone or in tandem. Results indicate that factors such as adverse weather conditions and young drivers reduce the severity of a crash; While involvement of motorcycles, pedestrians, unbelted passengers and heavy trucks dramatically increase the likelihood of having a severe crash.
Keywords: crash severity; crash causation; logistic regression; adverse weather; SHRP2 RID