Improved Prediction of Gate Departure Times Using Pre - Departure Events

‡Traffic management decisions and automation su pporting those decisions currently lack accurate departure demand information. Future departure demand is typically predicted using the scheduled departure times of each individual flight, which are shown to be poor estimates of actual departure times. This paper describes an approach to improve the prediction of parking gate departure times for individual flights , thereby improving the knowledge of departure demand at longer prediction horizons. The approach us es air carrier -provided data about when cer tain milestones in preparing for departure are completed. Actual times for pre -departure events are compiled in advance from historical data. These statistics are then used to improve gate departure time predictions in real -time. This paper presents an algorithmic approach for applying pre -departure event times to improve gate departure time predictions for individual flights and applies the method using Aircraft Communication Addressing and Reporting System (ACARS) messages from a large domestic air car rier at a busy airport. Results show the benefit of applying this pre departure information to two existing automation systems that predict gate departure times – the Surface Management System and the Enhanced Traffic Management System . The application o f ACARS data using the proposed algorithm provides a 36% improvement to the gate departure time predictions .