Neural Network Prediction of Airport Arrival and Departure Capacity

Predicting the capacity of an airport to accommodate incoming or outgoing aircraft contributes to efficient air traffic management and optimization of an airline’s flight operations. This talk describes a method for mapping weather forecasts to anticipated Airport Acceptance Rate (AAR) and Departure Rate (ADR) capacity and anticipated demand for several dozen major airports in the US, Mexico and Canada. The neural network models are trained using historical weather and observed arrival and departure data, then run in predictive mode using both deterministic and ensemble forecast input. Their predictions identify periods in which demand may outstrip capacity, allowing air traffic managers and airlines to anticipate potential disruptions to their scheduled operations and act to mitigate them via flight modifications or cancelations. The AAR and ADR capacity and demand predictions are run operationally and presented in a graphical user interface that allows airline dispatchers to view and interrogate predictions for their flights’ departure and arrival airports. The talk includes case studies, verification results, plans for future improvements and a discussion of applications to improved airline decision making.