Mesoscale eddy detection in satellite imagery of the oceans using the Radon transform

The present study proposes a new methodology to identify mesoscale eddies that are ubiquitous in the ocean and can be observed in a variety of satellite images. Our focus is on eddies with mean diameter of 100-500 km. The proposed methodology uses an algorithm based on the Radon transform. The basic premise is that the inverse Radon transform of a circularly symmetric feature is angle-independent. Therefore, the more constant the Radon transform of an image is with respect to the rotation angle, the more circularly symmetric the image is. Although our method is based on circular symmetry it is able to detect non-symmetric eddies using a sensitivity parameter that quantifies the deformation. Our method is effective to determine how circular a feature observed in a satellite image is, and suggest whether it is an eddy or not, based on geometric criteria. A series of tests with synthetic data were performed to assess the sensitivity to spectral content and aspect ratio detectable by the method. Another sequence of tests was carried out with images of chlorophyll-a concentration, sea surface temperature and sea surface height. In the three cases five areas of 100 × 50 pixels were selected and filtered with exactly the same algorithm and parameters. The two sets of tests indicate that the proposed method was able to identify eddies in a single time series of satellite images and in multiple series of different variables. The main advantage of this method in comparison with either automatic ones based on dynamics and geometry or manual methods is its ability to observe eddies in multiple data sets of different variables.