Characterizing wake vortex pairs using nonsupervised machine learning.

Area of Science:

• Atmospheric science
• Aerodynamics
• Remote sensing

Background:

• Aircraft generate persistent wake vortices that pose a collision risk.
• Accurate detection and characterization of wake vortices are crucial for air traffic management and safety.

Purpose of the Study:

• To present a novel method for detecting and characterizing aircraft wake vortices using Doppler lidar data.
• To demonstrate the efficacy of sequential cluster analysis for wake vortex identification.

Main Methods:

• Sequential Doppler lidar observations of a Boeing C-17 Globemaster III.
• Calculation of weighted gradients for Doppler data products.
• Application of Density-Based Spatial Clustering of Applications with Noise (DBSCAN) for vortex isolation.
• K-means clustering for partitioning vortex pairs into distinct regions.

Main Results:

• Successful detection and tracking of wake vortex pairs.
• Identification of prototypical regions within the vortex pairs.
• Demonstration of DBSCAN's capability to isolate vortices from background noise.
• Validation of k-means clustering for vortex pair characterization.

Conclusions:

• Cluster analysis is a powerful tool for processing sequential Doppler lidar data.
• The proposed method offers a promising approach for automated wake vortex detection.
• This technique can significantly advance aviation safety and efficiency.