Short-Term National Airspace System Delay Prediction Using Weather Impacted Traffic Index

Abstract

This paper describes a method to predict delay in the National Airspace System for durations of up to two hours. Various linear autoregressive model structures with exogenous inputs were implemented to perform the delay prediction. Current and forecast weather impacted traffic indices and air traffic volume were used as inputs to the system while the air traffic delay is the predicted output of the model. The refined methodology for generating the weather impacted traffic indices, together with the high-update-rate of Corridor Integrated Weather System is well suited for real-time delay prediction. An adaptive scheme was implemented to provide both weather-related and non-weather-related delay predictions depending on the weather and air traffic condition. The method will benefit air traffic management by facilitating the development of strategies to reduce delays, cancellations, and other costs during the day of operations in various weather conditions.