Predicting crash severity based on recorded crash data is associated with several challenges, including limited accessibility, incomplete information, and potential reporting errors. Consequently, surrogate safety measures provide a proactive approach that reduces reliance on crash data and enables the assessment of factors influencing crash severity prior to crash occurrence. Urban roundabouts, as intersections with a circular central island, play an important role in moderating vehicle speeds, making the analysis of factors affecting interaction severity particularly relevant.

In this study, aerial image processing techniques were used to analyze rear-end interactions among passenger cars at the Roudband roundabout, located in Dezful County, Khuzestan Province, Iran. Performance-related variables were extracted using safety analysis software, and gamma regression models were employed to examine the factors influencing interaction severity. Surrogate safety measures, including time to collision (TTC), were used as the dependent variable to evaluate factors affecting the severity of rear-end and lane-change interactions. The results indicate that maximum speed has the strongest effect in the final rear-end interaction model, such that a one-unit increase in maximum speed leads to a 0.488-unit reduction in interaction severity. This finding highlights the statistically significant mitigating effect of maximum speed on the safety level of traffic interactions and demonstrates the influence of speed variations on reducing interaction severity at urban roundabouts.