Statistical Modeling of Arrival Performance in Terminal Airspace

Operational performance improvements in Air Traffic Management are key to enhancing the economic and environmental performance of aviation. This requires a solid understanding of performance drivers across multiple scales. The terminal area arrival represents one of the most complex flight phases, being often subject to inefficiencies that increase flight costs, fuel burn and emissions. This paper investigates the main causal factors for the performance of arrival operations in terminal airspace through statistical modeling. For this, we leverage large-scale historical data about flight tracks, weather and traffic flow management reports. We study the case of Sao Paulo/Guarulhos International Airport and its most recent terminal area redesign that implemented a Point Merge system. Machine learning techniques are first applied to identify the arrival trajectory patterns actually flown and analyze changes in airspace utilization with the redesign. We then estimate regression models of both temporal and vertical efficiency. The results reveal statistically significant impacts of the structural changes, with the magnitude of such effect varying significantly by arrival pattern. In general, the airspace redesign is observed to reduce vertical inefficiencies and attenuate the impacts of traffic flow management restrictions on delays. Among the operational factors, convective weather is identified as the most important source of inefficiencies. The work contributes to a better knowledge of how different structural and operational factors affect trajectory efficiency in terminal airspace and brings novel insights that highlight opportunities for system improvement.