AN EXPLORATORY STUDY ON DYNAMIC FLIGHT TRAJECTORY DEVIATION FROM CONVICTIVE WEATHER HAZARD VIA DEEP LEARNING APPROACHES

Air traffic safety has always faced serious challenges from various environmental hazards. Meteorological phenomena like wind shear, thunderstorms, typhoons, and cumulonimbus, may result in air traffic accidents. Unlike geographical features such as terrains, the intensity and the timing of aviation weather hazards are dynamic and uncontrollable, resulting in uncertainties about the potential consequences to aviation safety. Currently, weather forecasts for aviation purposes are primarily considered in pre-tactical flight planning. Such plans are conservative to avoid accidents arising from the rapid changes of potential hazards. Thus, pilots are placed at an essential role in-flight to avoid unfavourable weather, which can cause hundreds of casualties and injuries. Such practices might result in extensive deadweight loss as diverted paths with longer travelling time and more fuel required were adopted, which might even cause flight delays. Taking direct paths might not necessarily be unsafe as weather features are dynamic: the position and the corresponding affected area will evolve with time. Therefore, this research aims to reveal the underlying patterns of flight trajectory deviation due to weather hazards from the historical data and recommends formulating a dynamic flight trajectory deviation model with the least impact on the delay time.