A general data-driven methodology for predicting future air traffic distributions around airports

A reliable forecast of future low-altitude air traffic distributions is essential for the prediction of aircraft noise and pollutant emissions in airport areas. This work proposes a novel methodology for generating aircraft movements around airports based on historical flight tracking data. These data are exploited to derive representative flight paths via unsupervised machine learning and typical airport operations via statistical analysis. These outcomes feed a custom-made algorithm that generates future traffic distributions, requiring only traffic intensity and forecast timeframe as inputs. The methodology, tested at Stockholm Arlanda Airport, uses 2022 tracking data to predict the traffic distribution in 2024 and is validated comparing simulated and actual 2024 day-evening-night noise levels L DEN . An accurate prediction is achieved, as moderate L DEN differences involve regions under 45 dB(A) while L DEN ≥ 45 dB(A) contour areas are overestimated by less than 6 %. Errors come mostly from changes in airport operations, to be addressed in future developments.