A Methodology For The Identification Of Safety Improvement Candidate Intersections

Author : Sadia Sharmin
Publisher :
ISBN 13 : 9780438564725
Total Pages : 116 pages
Book Rating : 4.5/5 (647 download)

DOWNLOAD NOW!

Download or read book Developing Safety Performance Functions for Wyoming Urban and Suburban Four-leg Signalized Intersections written by Sadia Sharmin and published by . This book was released on 2018 with total page 116 pages. Available in PDF, EPUB and Kindle. Book excerpt: The complexity of traffic movements at signalized intersections is of a significant interest to transportation professionals due to a disproportional high amount of conflicts and crashes involving turning vehicles followed by through vehicles. Moreover, the wide variability in geometric and operational characteristics at signalized intersections may exacerbate this complexity. Intersection-related fatalities in Wyoming exhibited a sharp increase after 2012. Fatality rate in Wyoming exceeded the national rate by 12 percent in 2014 and 2015. In 2015, Wyoming’s urban intersections crashes constituted 50% of total urban crashes while that the proportion at the national level in the US was 32%. These issues could be addressed through successful engineering approaches which necessitate reliable methods to estimate intersection safety. Currently, Safety Performance Functions (SPFs) are used as the main tool to estimate traffic safety of different roadway elements including intersections. The Highway Safety Manual (HSM), published by the American Association of State Highway and Transportation Officials (AASHTO) in 2010 provides SPFs for intersections. The unique nature and distinct attributes of Wyoming also necessitates developing its own jurisdiction-specific SPFs. SPFs are needed to identify potential candidate locations for safety improvements and for implementation of various safety countermeasures. Therefore, the research in this thesis focuses on calibrating SPFs for four-leg signalized intersections in Wyoming. The thesis also utilizes advanced statistical techniques to address various issues in crash data at intersections. Crash data at four-leg signalized intersections were structured as pooled data averaged across time and panel data with respect to time. The key issues addressed in this study include: i) addressing overdispersion characteristics of crash data, ii) considering unobserved random effects of the intersections due to their heterogeneity, and iii) allowing the effects of the confounding factors on intersection crash trends over time. This study examined the effects of geometry, weather, traffic and driver characteristics on crash frequencies incorporating time trends and taking longer time frames into consideration. Standard crash models using longer time frames and aggregated data do not consider time-varying effects on crash frequency, and can introduce error due to unobserved heterogeneity. This study analyzed the findings among traditional Negative Binomial (NB) model (using pooled and panel data) which account for the overdispersion of data, Random Effect Negative Binomial (RENB) model using panel data that consider unobserved random effects of the sites due to their heterogeneity and Generalized Estimating Equation (GEE) with Negative Binomial (NB) distribution. RENB model outperformed NB models by comparing the Akaike Information Criterion (AIC) values and hence, RENB models were chosen to explain the effects of the confounding factors on the crash counts. A marginal model approach which is different from the random effect approach was attempted through applying GEE to the panel data. “Time” was included as an explanatory variable with linear and quadratic effects to investigate the crash trends across time. The interactions of variables with “time” were considered to determine the effect of fixed variables over time. GEE was also fitted using variables without incorporating “time”, however, GEE accounting for “time” performed better. Model assessment was performed using the quasi-likelihood information criterion (QIC). This method also showed that the GEE model with autoregression structure had the best model performance. Installing left-turn lanes at signalized intersections physically separates left-turning vehicles, hence can reduce the number of conflicts by removing slow or decelerating vehicles from through traffic resulting in safer operations. The safety effectiveness of left-turn lanes at four-leg signalized intersections was estimated with these approaches along with predicting and describing the confounding factors. The aforementioned objectives were attained by developing SPFs using 174 four-leg signalized intersections crash data of Wyoming from the years 2005 through 2014. Wyoming specific data, including traffic, weather and driver characteristics, were incorporated into the models along with site-specific geometric characteristics. The results of this study highlighted the advantages of predicting crash frequencies by crash type and severity. To examine the impact of weather on intersections crashes, three sets of SPFs were calibrated; winter, summer, and all year. The cross-sectional panel data analysis for evaluating the safety effectiveness of adding left-turn lanes at four-leg signalized intersections indicated a reduction of Fatal+Injury (F+I) and angle crashes by 21.6% and 30.8%, respectively. The calibrated models for the winter indicated a reduction of 19.4% and 38.8% for the aforementioned crashes, respectively for adding left-turn lanes. For the summer, crash reduction for these crashes for the same treatment were found 71% and 69%, respectively.