Modeling Bicycle-Vehicle Crash Frequency on Urban Roads

1 online resource (136 pages) : PDF

2017

University of North Carolina at Charlotte

Bicyclists and motorists make mistakes that contribute to traffic crashes involving bicyclists on urban roads. The likelihood of a bicyclist being severely injured or killed daily in traffic crashes is creating fear, anxiety, and becoming a potential danger to the increasing number of Americans using bicycle as a mode of transportation. It is also making bicycling to work or for other purposes less lucrative. Building bicycling friendly and safe environment is, therefore, vital to encourage and have more people use bicycle as a mode of transportation. Therefore, the main goal of this research is to improve safety of bicyclists on urban roads. The main objectives are to understand the role of explanatory variables on risk to bicyclists on urban roads and to develop macroscopic bicycle-vehicle crash frequency models (safety performance functions) for urban roads.Mecklenburg County in North Carolina was considered as the study area. Reported bicycle-vehicle crash data from 2010 to 2015 along with demographic, land use and network characteristics data was obtained from the local agencies. One-hundred and nineteen locations (intersections) were randomly selected in the study area. These locations were selected such that they are geographically distributed in the study area. Features available in Geographic Information Systems (GIS) software were used to ensure that these locations fall in high, moderate, low and no bicycle-vehicle crash areas. Data within one-mile buffer (vicinity) of 119 randomly selected locations was then captured. These 119 locations accounted for 91.8% of total bicycle-vehicle crashes observed during the study period.Data for 99 randomly selected locations was used for modeling, while data for the remaining 20 randomly selected locations was used for validating the models. Poisson and Negative Binomial log-link distribution based models were then developed using the modeling dataset. The bicycle-vehicle crash dataset used in this research was observed to be over-dispersed (variance greater than the mean). Therefore, Negative Binomial log-link distribution based models were selected and discussed in this research.Several demographic, land use and network characteristics were observed to be linearly correlated to bicycle-vehicle crash frequency at a 95% or higher confidence level. Correlations, with p-values = ~0.000, were also observed between demographic, land use and network characteristics (explanatory variables). Six alternate models were developed considering various combinations of explanatory variables, land use and network characteristics, that are not correlated to each other. Two models using all the explanatory variables by ignoring multicollinearity, one each with and without eliminating insignificant explanatory variables, were also developed. The validation dataset was used to compare the estimated bicycle-vehicle crash frequency from each model with the actual bicycle-vehicle crash frequency.The results obtained from analysis and modeling indicate that bicyclists are at a significantly higher risk of getting involved in a crash while traveling(1) on segments with no bicycle lane,(2) on segments with traffic lights,(3) on segments with 45 mph as speed limit,(4) in commercial areas,(5) in areas with research activity and institutions,(6) in areas with multi-family residential units (densely populated), and,(7) in heavy industrial areas.Overall, this dissertation explores interdisciplinary concepts related to transportation engineering, GIS, data analytics and statistical methods to develop and validate models to estimate bicycle-vehicle crash frequency.

doctoral dissertations

Civil engineering
Transportation
City planning

Ph.D.

Bicycle-Vehicle
Crash
Frequency
Modeling
Roads
Urban

Civil Engineering

Pulugurtha, Srinivas

Janardhanam, Rajaram
Kane, Martin
Fan, Wei
Delmelle, Eric

Thesis (Ph.D.)--University of North Carolina at Charlotte, 2017.

This Item is protected by copyright and/or related rights. You are free to use this Item in any way that is permitted by the copyright and related rights legislation that applies to your use. For other uses you need to obtain permission from the rights-holder(s). For additional information, see http://rightsstatements.org/page/InC/1.0/.

Copyright is held by the author unless otherwise indicated.

Mukoko_uncc_0694D_11389

http://hdl.handle.net/20.500.13093/etd:1529