Safety modeling for two certain urban intersection types, HSM calibration, and crash severity distribution
Metadata[+] Show full item record
According to several years' official crash data, a significant proportion of crashes happened at roadway intersections in Missouri. Of those intersection crashes, a large number of crashes occurred at urban roadway intersections. Although much safety research has been performed, it is still crucial to further understand the factors that contribute to crashes at such locations. It is also noticed that adverse weather and precipitation would lead to high crash risk and more severity crashes. Therefore, it is meaningful to do safety research about urban roadway intersections considering adverse weather and precipitation conditions. Safety Performance functions (SPFs) were built for urban two-lane four-leg unsignalized intersections and urban four-leg signalized intersections considering precipitation factors. All the intersection samples were collected from Missouri. For urban two-lane four-leg unsignalized intersections, AADT and precipitation factors were considered for developing SPF. For urban four-leg signalized intersections, AADT, speed limit, and precipitation factors were considered for building SPF. SPFs were built based on negative binomial regression model. Several diagnostics methods were used in model developing processes, such as Exploratory Data Analysis (EDA), Variable Introduction Exploratory Data Analysis (VIEDA), over-dispersion, log-likelihood, and cumulative residual plots (CURE Plot). The models revealed that precipitation factors have significant impacts on the safety of urban two-lane four-leg unsignalized intersections. However, for urban four-leg signalized intersections, the effects were not obvious. Calibration factors were also developed for several unsignalized intersections types in this thesis. Calibration factors can be applied to adjust crash frequency by comparing actual roadway conditions to base roadway conditions. The new calibration factors were almost same as HSM1.