Investigating the influence of contributing factors and predicting visibility at road link-level.

Data from weather stations at airports, far away locations or predictions using macro-level data may not be accurate enough to disseminate visibility related information to motorists in advance. Therefore, the objective of this research is to investigate the influence of contributing factors and develop visibility prediction models, at road link-level, by considering data from weather stations located within 1.6 km of state routes, US routes and interstates in the state of North Carolina (NC). Four years of meteorological data, from January 2011 to December 2014, were collected within NC. Ordinary least squares (OLS) and weighted least squares (WLS) regression models were developed for different visibility and elevation ranges. The results indicate that elevation and cloud cover are negatively associated with low visibility. The chances of low visibility are higher between six to twelve hours after rainfall when compared to the first six hours after rainfall. A visibility sensor was installed at four different locations in NC to compare hourly visibility from the selected regression model, High-Resolution Rapid Refresh (HRRR) data, and the nearest weather station. The results indicate that the number of samples with zero error range was higher for the selected regression model compared with the HRRR and weather station observations.

Modeling and comparing injury severity of at-fault and not at-fault drivers in crashes.

This paper examines and compares the effect of selected variables on driver injury severity of, both, at-fault and not at-fault drivers. Data from the Highway Safety Information System (HSIS) for the state of North Carolina was used for analysis and modeling. A partial proportional odds model was developed to examine the effect of each variable on injury severity of at-fault driver and not at-fault driver, and, to examine how each variable affects these two drivers' injury severity differently. Road characteristics, weather condition, and geometric characteristics were observed to have a similar effect on injury severity in a crash to at-fault and not at-fault drivers. Age of the driver, physical condition, gender, vehicle type, and, the number and type of traffic rule violations were observed to play a significant role in the injury severity of not at-fault drivers when compared to at-fault drivers in the crash. Moreover, motorcyclists and drivers 70 years or older are observed to be the most vulnerable road users.

Predictability and interpretability of hybrid link-level crash frequency models for urban arterials compared to cluster-based and general negative binomial regression models.

Machine learning (ML) techniques have higher prediction accuracy compared to conventional statistical methods for crash frequency modelling. However, their black-box nature limits the interpretability. The objective of this research is to combine both ML and statistical methods to develop hybrid link-level crash frequency models with high predictability and interpretability. For this purpose, M5' model trees method (M5') is introduced and applied to classify the crash data and then calibrate a model for each homogenous class. The data for 1134 and 345 randomly selected links on urban arterials in the city of Charlotte, North Carolina was used to develop and validate models, respectively. The outputs from the hybrid approach are compared with the outputs from cluster-based negative binomial regression (NBR) and general NBR models. Findings indicate that M5' has high predictability and is very reliable to interpret the role of different attributes on crash frequency compared to other developed models.