Estimating Wind Shear Magnitude Near Runways at Hong Kong International Airport Using an Interpretable Local Cascade Ensemble Strategy

Abstract

The elevated occurrence rate of wind shear (WS) events near airport runways presents one of the major hazards to the safe and efficient operation of landing and takeoff procedures. As a consequence of this, aircraft are more likely to experience the possibility of losing control or encountering hindrances. Hence, it is crucial to assess the factors influencing wind shear occurrence. Previous studies extensively reported the susceptibility of the runways at Hong Kong International Airport (HKIA) to significant wind shear events. Therefore, in order to estimate WS magnitude near runways at HKIA and assess various contributing factors, this study presents a novel Local Cascade Ensemble (LCE) model with its hyperparameters optimized via a Tree-Structured Parzen Estimator (TPE) to estimate the wind shear magnitude. The pilot report data obtained from HKIA between 2017 and 2021 was employed for the training and evaluation of the TPE-tuned LCE model. The outcomes of the TPE-tuned LCE model were also compared to those of other contemporary machine learning (ML) models. The findings indicated that the TPE-tuned LCE model exhibited better predictive performance in comparison to other models, as assessed by a mean absolute error (MAE) of 4.38 knots, a mean squared error (MSE) of 70.28 knots, a root mean squared error (RMSE) of 8.38 knots, and coefficient of determination (R²) value of 0.79. Subsequently, model interpretation via SHapley Additive exPlanations (SHAP) technique was performed on the outcomes of TPE-tuned LCE. It indicated that that certain runways at HKIA, such as runway 07 C, 07 L, 25 C, and 25R, had a higher likelihood of experiencing elevated wind shear conditions within 1000 ft above the runway level.