Integrating Terminal Airspace Weather Profiles Into Machine Learning Models for ETA Prediction Within the Terminal Maneuvering Area (TMA)

Accurate prediction of the estimated time of arrival (ETA) in the Terminal Maneuvering Area (TMA) is important for safe and efficient air traffic management. Many existing ETA prediction models for terminal airspace are based primarily on scheduled arrival times, simple distance to landing measures, and traffic demand indicators. Weather effects are frequently neglected or simplified to surface observations, and there has been limited research on how to accurately depict upper-level winds within the terminal area for ETA predictions. This study develops a baseline ETA model for arrivals to Chicago O’Hare International Airport (ORD) using aircraft characteristics, temporal and congestion indicators, TMA entry direction, and surface weather. Then it augments this baseline with upper-level wind features derived from High-Resolution Rapid Refresh (HRRR) data. We examine four representations of upper-level winds in the TMA: TMA-entry winds, airport-level winds, sector-based TMA winds, and trajectory-informed polygon winds. All cases are evaluated with a common time-based train-test split and a consistent set of data-driven regression models. The results show that adding upper level wind information yields small but repeatable improvements over the surface only baseline. An airport level representation that incorporates mid altitude (U, V) wind components aligned with typical approach altitudes provides the strongest gains relative to its feature complexity. Sector based and trajectory informed representations help identify where upper level winds exert the greatest influence along the approach path, but the specific structures tested in this study do not produce substantial improvements in overall ETA accuracy. These findings indicate that upper air information contributes to ETA prediction when represented with compact and physically interpretable features. They also suggest that alternative definitions of sector boundaries or trajectory based spatial regions should be evaluated at other airports to determine whether they offer additional performance gains.