INFLUENCE OF ROADWAY CHARACTERISTICS IN THE MODELING OF THE FREQUENCY OF ROADWAY DEPARTURE CRASHES ON TWO-LANE TWO-WAY STATE ROADS

Simple item page
dc.contributor.advisorArchilla, Adrián Ricardo
dc.contributor.authorHashemi, Mohammadreza
dc.contributor.departmentCivil Engineering
dc.date.accessioned2019-10-09T18:48:22Z
dc.date.issued2019
dc.description.degreePh.D.
dc.embargo.liftdate2021-10-04
dc.identifier.urihttp://hdl.handle.net/10125/63453
dc.subjectCivil engineering
dc.subjectTransportation
dc.subjectGeneralized Linear Modeling
dc.subjectGeometry Design
dc.subjectRoadway Characteristics
dc.subjectRoadway Departure Crashes
dc.titleINFLUENCE OF ROADWAY CHARACTERISTICS IN THE MODELING OF THE FREQUENCY OF ROADWAY DEPARTURE CRASHES ON TWO-LANE TWO-WAY STATE ROADS
dc.typeThesis
dcterms.abstractAccording to the Federal Highway Administration (FHWA), Roadway Departure (RwD) crashes account for approximately 56 percent of traffic fatalities in the United States. Likewise, FHWA’s statistics indicate that RwD crashes represent a high proportion (approximately 54 percent) of traffic fatalities in the State of Hawaii. Therefore, there is a need to study their contributing factors and to quantify their effects by developing statistical models that may provide better inferences to alleviate them. Using ten years of crash data, this research explores the effect of roadway characteristics (e.g., traffic, geometry, etc.) in the modeling of the frequency of RwD crashes on Two-Lane Two-Way (TLTW) state roads in the State of Hawaii. Specifically, the study concentrates on the effects of segment length, roadway directional attributes, and the general geometric environment of the analysis segment. These factors are evaluated with various Generalized Linear Models (GLM) such as the negative binomial regression, zero-inflated negative binomial, and mixed-effects negative binomial regression model. The results show that segment length affects the model’s estimations (total number of statistically significant parameters and their estimated values) and present the development of recommendations about an appropriate segment length for modeling the frequency of RwD crashes. Also, it confirms that the consideration of directional analysis improves the quality of the models in two ways: firstly, by assigning the head-on crashes based to the direction of the vehicles causing the crashes, and secondly by identifying the contributing factors based on the direction of vehicles causing the crashes. Moreover, the results indicate that the general geometric environment of the roadway portion where the segment was located affects the frequency of RwD crashes, which means that, for example, for two similar segments, the frequency of RwD crashes are not equal if one is located on a winding road and the other segment is located right after a tangent road. This finding is in accordance with design consistency practices. Another benefit of this study is the development of robust and realistic crash frequency models for the state of Hawaii as well as the improvement of the identification of the RwD crashes’ contributing factors. In practice, decision-makers may consider the results to prioritize the location and type of countermeasures to mitigate RwD on TLTW state roads in Hawaii effectively. Other unique features of the estimated models include: 1) using an estimate of mean friction demand as an independent variable, 2) capturing the different effects of upgrades and downgrades, and 3) using Annual Average Daily Traffic (AADT) both as a measure of exposure and separately as an independent variable affecting the rate of RwD crashes. Finally, a new approach consistent with the probabilistic nature of the estimated generalized regression models is introduced for evaluation of their goodness of fit. Its use is suggested as a complementary tool in the typical evaluation of generalized regression models.
dcterms.extent172 pages
dcterms.languageeng
dcterms.publisherUniversity of Hawai'i at Manoa
dcterms.typeText
local.identifier.alturihttp://dissertations.umi.com/hawii:10342

Files

Original bundle
Now showing 1 - 1 of 1
Thumbnail Image
Name:
Hashemi_hawii_0085A_10342.pdf
Size:
6.66 MB
Format:
Adobe Portable Document Format
Download

Collections

Ph.D. - Civil Engineering