Estimation and tactical allocation of airport capacity in the presence of uncertainty

Abstract

Major airports in the United States and around the world have seen an increase in congestion-related delays over the past few years. Because airport congestion is caused by an imbalance between available capacity and demand, the efficient use of available capacity is critical to mitigating air traffic delays. A frequently-adopted traffic management initiative, the Ground Delay Program (GDP), is initiated when an airport expects congestion, either because of very high demand or a reduction in its capacity. The GDP is designed to efficiently allocate the limited airport capacity among the scheduled flights. However, contemporary GDP practice allocates delays to arrivals independent of departures, and relies on deterministic capacity forecasts. This thesis designs and evaluates a GDP framework that simultaneously allocates arrival and departure delays, and explicitly accounts for uncertainty in capacity forecasts. Efficient capacity allocation requires the accurate estimation of available airport capacity. The first module of this thesis focuses on the modeling of airport capacity and its dynamics. A statistical model based on quantile regression is developed to estimate airport capacity envelopes from empirical observations of airport throughput. The proposed approach is demonstrated through a case study of the New York metroplex system that estimates arrival-departure capacity tradeoffs, both at individual airports and between pairs of airports. The airport capacity envelope that is valid at any time depends on the prevailing weather (visibility) and the runway configuration. This thesis proposes a discrete choice framework for modeling the selection of airport runway configurations, given weather and demand forecasts. The model is estimated and validated for Newark (EWR) and LaGuardia (LGA) airports using archived data. The thesis also presents a methodology for quantifying the impact of configuration switches on airport capacity, and applies it to EWR and Dallas Fort Worth (DFW) airports. The second module of this thesis extends two existing stochastic ground-holding models from literature, the static and the dynamic, by incorporating departure capacity considerations to existing arrivals-only formulations. These integer stochastic formulations aim to minimize expected system delay costs, assuming uniform unit delay costs for all flights. The benefits of the integrated stochastic framework are demonstrated through representative case studies featuring real-world GDP data. During GDPs, the Collaborative Decision-Making framework provides mechanisms, termed intra-airline substitution and compression, which allow airlines to redistribute slots assigned by ground-holding models to their flights, depending on flight-specific delay costs. The final part of this dissertation considers collaborative decision-making during GDPs in stochastic settings. The analysis reveals an inherent trade-off between the delay costs achieved by the static and the dynamic stochastic models before and after the application of the CDM mechanisms. A hybrid stochastic ground-holding model that combines the desirable properties of the static and dynamic models is then proposed. The performance of the three stochastic ground-holding models under CDM are evaluated through real-world case studies, and the robustness of the final system delay cost reduction achieved by the hybrid model for a range of operating scenarios is demonstrated.