Electronic Theses and Dissertations


Diwas Thapa



Document Type


Degree Name

Doctor of Philosophy


Civil Engineering

Committee Chair

Sabyasachee Mishra

Committee Member

Mihalis Golias

Committee Member

Claudio Meier

Committee Member

Mohamed Osman


This research consists of four studies aimed at enhancing traffic safety. These studies employ innovative approaches and model frameworks based on duration models. The first study focuses on Work Zone Intrusion Alert Systems (WZIAS) and explores their component layout and impact on work zone safety. By analyzing the duration between work zone intrusion and crashes, this study identifies critical factors contributing to work zone crashes when WZIAS are in use. The study concludes by offering recommendations to optimize WZIAS effectiveness, including potential adjustments to existing work zone guidelines. The second study addresses the limitations of duration models in handling time-varying factors and their inability to support real-time predictive modeling. It introduces a novel duration-based model framework capable of accommodating dynamic covariates for proactive crash prediction modeling. This approach involves discretizing the time between crashes into forecasting epochs. The study explores various sampling methods to address computational challenges arising from increased data size due to forecasting epochs. A 25% epoch-level sample is found effective for computational efficiency without sacrificing accuracy. The third study extends the duration-based crash prediction framework, to develop a novel approach for predicting crash and severity simultaneously. The study further seeks to strike a balance between model performance and estimation time. Findings suggest that a 15% sample drawn at the epoch level provides an efficient compromise. Stability analysis of predictor variables sheds light on their reliability across different samples. Variables such as Time of day (Early afternoon), Weather condition, Lighting condition (Daytime), Illumination, and Volume require larger samples for more accurate coefficient estimation. Conversely, Time of day (Early morning, Late morning, Late afternoon), Lighting condition (Dark lighted), Terrain, Land use, Number of lanes, and Speed converge towards true estimates with small incremental increases in sample size. The fourth study employs the duration-based proactive crash prediction framework to showcase its versatility and identify the predictors of speeding events in work zones. The model's validity is tested, with higher accuracy observed for speeding events occurring with shorter durations between consecutive occurrences. The study demonstrates how the framework can help transportation agencies identify high-risk highway segments and implement safety measures proactively.


Data is provided by the student.

Library Comment

Dissertation or thesis originally submitted to ProQuest.


Open Access