Aircraft noise annoyance is inherently subjective, and its accurate quantification represents a challenging task. There is a lack of consensus in the scientific community regarding which metrics are best for effectively representing this type of annoyance. The present study aims at relating various sound metrics to noise annoyance ratings measured in listening experiments featuring 60 aircraft flyover recordings (30 landings and 30 takeoffs). This is done by considering different sound quality metrics (SQMs), psychoacoustic annoyance models, and more conventional noise certification metrics, such as the effective perceived noise level (EPNL) or the sound exposure level. A correlation analysis was subsequently performed on a large pool of sound metrics considering both linear and non-linear functions. The results show that, in general, metrics derived from psychoacoustic annoyance models (especially those proposed by Zwicker and Di et al.) present considerably better correlations compared to conventional metrics and most individual SQMs. The metrics of loudness, EPNL, and maximum perceived noise level (PNL) also exhibit strong correlations and capacity to predict a substantial portion of the variance observed in the reported annoyance ratings. Moreover, considering non-linear functions (e.g. logarithmic or hyperbolic tangent power) further improves the prediction performance.

In this paper, we propose open machine learning models that can provide airport delay predictions in a network with an error of around or less than five minutes. Due to the complexity of different components of air traffic networks, traditional flight performance model-based predictions fall short when dealing with numerous flights and often are not able to deal with delays that propagate among airports in a network. In this study, we employ three different machine learning models to predict delays at three different scopes: individual flights, airports, and the network of airports. Consequently, we tested three approaches with different levels of complexity, including statistical regression models, recurrent neural networks, and spatial-temporal graph attention neural networks. We conduct experiments for all three types of models using the Eurocontrol research data archive. After training and testing with two years of data covering the top 50 European airports, our models produce prediction errors of around or less than 5 minutes with look-ahead time up to 3 hours. These metrics have shown a significant advancement compared to existing prediction models. We also openly share this model to support open science in aviation.