<p>This research investigates the lateral safety of vehicles navigating roundabouts by utilizing surrogate safety measures and extreme value theory. To capture vehicle interactions at the Kasna Gol Chakkar roundabout in Greater Noida, India, we employed drone technology to gather high-resolution aerial footage. Our main goal was to quantify conflict probabilities and set safety thresholds under different traffic conditions using metrics like Time to Collision (TTC) and Post Encroachment Time (PET). Traditional safety analysis methods often overlook near-misses and other critical incidents because they primarily focus on collision data. To address this, we used advanced data collection and processing techniques for a more proactive safety assessment. The data was processed using Data from Sky (DFS) software, which excels at extracting vehicle trajectories and identifying conflict points. DFS’s advanced trajectory analysis features, including TTC and PET calculations, enabled a thorough evaluation of conflict severity and frequency. We identified distinct interaction patterns during on-peak and off-peak hours, with a significant rise in rear-end conflicts during on-peak periods due to increased traffic density. Additionally, we applied the Generalized Pareto Distribution (GPD) to model TTC values and estimate critical conflict probabilities, providing a statistical framework for assessing high-risk scenarios. The study’s findings emphasize the importance of implementing targeted safety measures, such as enhanced lane markings, signage, and intelligent traffic management systems, to mitigate risks and improve overall roundabout safety.</p>