Disaggregation of Intersection Crash Data: An Approach-Based Crash Frequency and Crash Rate Analysis
General Material Designation
[Thesis]
First Statement of Responsibility
Azra, Nuzhat
Subsequent Statement of Responsibility
Medina, Juan C.
.PUBLICATION, DISTRIBUTION, ETC
Name of Publisher, Distributor, etc.
The University of Utah
Date of Publication, Distribution, etc.
2019
GENERAL NOTES
Text of Note
86 p.
DISSERTATION (THESIS) NOTE
Dissertation or thesis details and type of degree
M.S.
Body granting the degree
The University of Utah
Text preceding or following the note
2019
SUMMARY OR ABSTRACT
Text of Note
This thesis analyzes crash frequency and crash patterns using comprehensive datasets of traffic data at fine aggregation levels and detailed crash records, targeting some of the limitations related to using Average Annual Daily Traffic (AADT) for safety analysis. Being the most aggregated form of traffic volume, AADT does not capture variations of traffic volumes at different times of day but allows for developing crash frequency models and provides valuable information to understand the overall effect of conflicting demands, which in turn can be used to devise countermeasures and achieve safety improvements. However, AADT or even approach-level demands do not provide a clear picture of vehicle interactions, particularly for left-turn (LT) movements, where conflicting movements specifically relate to left-turning and opposing demands. In this research, high-resolution movement-level demands disaggregated up to 5-minute periods were collected on UDOT (Utah Department of Transportation) intersections over extended time periods. This high-resolution dataset was used to analyze conflict levels at the moment crashes occurred, and this information together with the overall exposure to such conditions was used to develop a measure of true left-turn crash risk. This unique method of estimating the risk (crash rates) in the intersections is used to quantify the effect of increasing left-turning and opposing volumes, develop a risk model, and estimate risk levels for a given approach throughout the day. Analysis shows a clear trend of increasing risk with increasing volumes, with a defined sharper increase in risk for an increase in left-turn volume versus a similar increase in through volume. This method can be used with online data streams to continuously monitor risk changes in real time and proactively identify candidate locations for safety improvements even before an increase in crash frequency is observed.