Automated Detection, Classification, and Tracking of Internal Wave Signatures Using X-Band Radar in the Inner Shelf
Abstract
A method based on machine learning and image processing techniques has been developed to track the surface expression of internal waves in near–real time. X-band radar scans are first preprocessed and averaged to suppress surface wave clutter and enhance the signal-to-noise ratio of persistent backscatter features driven by gradients in surface currents. A machine learning algorithm utilizing a support vector machine (SVM) model is then used to classify whether or not the image contains an internal solitary wave (ISW) or internal tide bore (bore). The use of machine learning is found to allow rapid assessment of the large dataset, and provides insight on characterizing optimal environmental conditions to allow for radar illumination and detection of ISWs and bores. Radon transforms and local maxima detections are used to locate these features within images that are determined to contain an ISW or bore. The resulting time series of locations is used to create a map of propagation speed and direction that captures the spatiotemporal variability of the ISW or bore in the coastal environment. This technique is applied to 2 months of data collected near Point Sal, California, and captures ISW and bore propagation speed and direction information that currently cannot be measured with instruments such as moorings and synthetic aperture radar (SAR).
© 2021 American Meteorological Society. For information regarding reuse of this content and general copyright information, consult the AMS Copyright Policy (www.ametsoc.org/PUBSReuseLicenses).
