Convection indicator for pre-tactical air traffic flow management using neural networks

Abstract

Convective weather is a large source of disruption for air traffic management operations. Being able to predict thunderstorms the day before operations can help traffic managers plan around weather and improve air traffic flow management operations. In this paper, machine learning is applied on data from satellite storm observations and ensemble numerical weather prediction products to detect convective weather 36 h in advance. The learning task is formulated as a binary classification problem and a neural network is trained to predict the occurrence of storms. The neural network results are used to develop a probabilistic based convection indicator capable of outperforming existing convection indicators found in the literature. Lastly, applications of the neural network based indicator in an air traffic management setting are presented.