Modelling the Likelihood of Air Traffic Management Regulations due to Weather at Airports

Adverse weather conditions, such as low visibility, can have a significant impact on airport capacity. When the capacity reduction is substantial and traffic demand remains high, air traffic flow management regulations are implemented to ensure that traffic demand remains below the (reduced) capacity. Traditionally, regulations are established by human operators hours in advance, relying on their subjective perception of the weather forecast and expected traffic demand. This paper introduces a machine learning model explicitly designed to capture the likelihood of air traffic flow management regulations based on weather conditions and traffic demand. To address the inherent noise in the dataset labels, stemming from decisions made in advance by operators relying on uncertain data, confident learning techniques are proposed to build a more robust and reliable model. The robustness of the model against noise is further enhanced by enforcing monotonic constraints during the training process. The experiments demonstrate satisfactory model performance for major European airports that frequently encounter adverse weather conditions. The main objective of this model is to assist operators in determining the effectiveness of implementing regulations and aid airlines in predicting potential delays or airborne holdings resulting from adverse weather.