Automated retrieval of internal wave phase speed and direction from pairs of SAR images with different look directions

Synthetic aperture radar (SAR) is the premier instrument in satellite remote sensing for the detection of oceanic internal waves due to its sensitivity to changes in small-scale ocean surface roughness and relatively large spatial coverage. The satellite constellation COSMO-SkyMed offers the unique capability to acquire pairs of images of the same scene within 24 min, which is ideal for making internal wave motions visible and extracting phase speeds. Other researchers have utilized pairs of airborne SAR images or images from two different satellite systems, but to our knowledge none have developed an automated method or applied standard feature tracking techniques due to the challenges SAR data poses. Unlike optical images, SAR images suffer from speckle noise, and images in a pair can look quite different if acquired from different look directions and incidence angles, making it difficult to apply feature tracking algorithms for quantifying motions. We propose a multistep approach building on a feature tracking algorithm from the literature to overcome this issue and successfully estimate the phase speed and direction of SAR detected internal waves. Our technique is developed using COSMO-SkyMed SAR data from the Inner Shelf Departmental Research Initiative and additionally tested on an ICEYE SAR image pair from a separate experiment. The estimated internal wave speeds are shown to be comparable to results based on in situ measurements and Korteweg–de Vries (KdV) theory. This new technique paves the way for a more automated approach to derive internal wave parameters from SAR images. •Short time gaps between SAR image pairs make internal wave motions visible.•Application of feature tracking and unsupervised machine learning to SAR data.•Propose new automated technique to extract internal wave phase speed and direction.•Results agree well with in situ/theory estimated internal wave phase speeds.