Low-level Wind Shear Prediction based on Machine Learning Techniques: a Case Study of Palermo-Punta Raisi International Airport

Low-level wind shear (LLWS) is one of the most prominent aviation hazards impacting safety, punctuality, and the environment. To mitigate its effects, several aerodromes have been equipped with dedicated systems capable of recognizing the presence of LLWS in the proximity of a runway. These systems usually comprise a collection of different devices, including a Terminal Doppler Weather Radar, a Doppler Light Detection and Ranging, and a network of anemometers spread along the airport grounds. The LLWS recognition technique is based on the measurement of the vertical wind profile, issuing a warning when a rapid change in wind direction or intensity is detected. Since this methodology is based on real-time data, no useful prediction is provided regarding the possibility of upcoming LLWS events. Furthermore, the costs associated with an LLWS detection system, in terms of purchase and maintenance, are very high making its installation quite prohibitive. In this study, we investigated the development of a new methodology for the prediction of LLWS events, based on the use of Machine Learning (ML) techniques applied to wind data obtained from ground station observations and pressure-level Numerical Weather models. The study is carried out considering the site of Palermo-Punta Raisi International Airport since it is the Italian airport most subject to LLWS phenomena. Historical data series from 2007 to 2022, extracted from the Era-5 reanalysis and Enav’s meteorological and aeronautical databases, were used to train and test different ML classification models, searching for the best-performing one through the analysis of specific evaluation metrics. The results we obtained are very encouraging and we are confident that our work could be very useful in developing a new generation of low-cost and high-efficiency ML-based LLWS prediction tools.