Assessment of Relative Risks in Flight Data - Theory and Applications

Comparing relative risks (RR) associated with diverse air traffic conditions is a common practice, often involving assessments between flights conducted in favorable flying conditions and those in adverse flying conditions. In our study, we adopt a practical approach by examining Aviation Risk Identification and Assessment (ARIA) airborne encounter events and evaluating the risk of mid-air collisions, measured by ARIA encounter scores. It's crucial to note that a lower encounter score implies a higher level of risk, with event scores ranging from 0 to 100. Our analysis utilizes binary data modeling, categorizing all events into two general categories: those with ARIA scores exceeding the median or 50 (considered relatively safer, or '0' group) and those with scores equal to or less than the median (relatively riskier, or '1' group). This latter group is the "risk target" for the purpose of conducting relative risk (RR) analysis. The ARIA data are usually merged with other specific application data by the unique event identification numbers. Two specific applications of relative risk analysis are of particular interest. The first application explores the risk involved in closely spaced parallel runway operations, as measured by ARIA encounter scores and runway approaching operations. Through a case-control study, for example, we find that narrower spacing (case group, or distance less than 706 ft., the median) results in approximately 19.5% higher odds (Odds Ratio = 1.195) of collision risk compared to airports with wider spacing (control group of more than 706 ft spacing), if without considering other multiple factors. Similarly, the aircraft pairs of "Overshoot-Overshoot" (aircraft flying towards each other) are compared with "Undershoot-Undershoot" pairs (aircraft flying away from each other); the "flying towards each other" group has approximately 20% higher odds of collision (Odds Ratio = 1.20). Furthermore, we examine the impacts on ARIA scores from multiple factors of overshoot versus undershoot, approaching time difference between paired aircraft, and the time difference from the ARIA event to the overshoot simultaneously. The second application of relative risk involves assessing the impact of Class B excursions on air traffic safety. This study explores the correlation between Class B excursions and collision risk measured by ARIA scores, investigating whether different types and lengths of excursions lead to varying rates of collision risk. The results indicate that flights with excursions (i.e., ARIA encounters within excursion duration) would not necessarily introduce additional traffic risk, as excursion flights collectively have a relatively higher mean ARIA event score compared to non-excursion flights outside the excursion duration. Moreover, excursion flights have a lower chance of falling into the riskier category of ARIA event scores (<=50), as evidenced by the Cox Proportional Hazard (CPH) model's hazard ratio or the multiple logit model's odds ratio for excursion versus non-excursion being less than 1.0. Additionally, the comparison between visual flight rules (VFR) and instrumental flight rules (IFR), as well as the investigation into various types of excursions, are conducted. The relative risk models, including case-control study, multiple logistic regression, and Cox proportional hazard model, are discussed with practical examples.