Date of Award
Master of Science
Dr. Benjamin Ofori-Amoah
Dr. Lei Meng
Dr. Laiyin Zhu
Traffic crash, spatial analysis, Lake Effect Snow, crash prediction model, GIS
Masters Thesis-Open Access
This study aims to (i) find the spatial-temporal traffic crash pattern and (ii) develop a statistical model for predicting LES induced traffic crash counts. Three southwest Michigan counties: Allegan, Kalamazoo and Calhoun are selected to conduct this study. Snow and non-snow event based comparative analysis are conducted using Getis-Ord Gi* statistic to identify high density crash cluster locations in the study area. While several new and oscillating hot spots are detected in Allegan during snow, no hot spot is detected during non-snow events based analysis. In Kalamazoo, traffic crashes do not exhibit much difference in spatial-temporal trend during snow or non-snow weather. Some new hot spots and numerous sporadic and oscillating hot spots are observed in Calhoun County roadways during snow, which are found as persistent hot spot locations from non-snow event based analysis. Negative binomial regression models with temporal random effects are fitted to the data treating daily average traffic crash counts as response variable; and temperature difference between Lake surface and overlaying air, wind speed, and wind direction as explanatory variables. All the variables exhibit statistically significant positive estimates to predict LES induced traffic crash counts for Allegan and Kalamazoo. However, wind direction is found statistically insignificant for Calhoun County model. These results will help further researches to explore LES induced traffic crashes and devise seasonal countermeasures.
Ayon, Bandhan Dutta, "Snow and Non-Snow Events Based Winter Traffic Crash Pattern Analysis and Developing Lake Effect Snow Induced Crash Count Prediction Model" (2017). Masters Theses. 1133.