Skip to Main content Skip to Navigation
Conference papers

Partitioning ocean wave spectra obtained from SAR observations or from the future real aperture wave scatterometer SWIM on CFOSAT

Abstract : In oceanography, one of the important products that can be provided from satellites is the directional spectrum of ocean waves (also called 2D spectrum), which describes the energy of the surface ocean waves as a function of wave number and direction of propagation. This information is essential for wave forecast systems, improvement of wave models, wave climate studies and scientific studies related to the ocean/atmosphere interface, coastal and marginal ice zone processes. This has motivated the specification of SAR instruments to deliver 2D wave spectra (ERS, ENVISAT, SENTINEL-1, RADARSAT, etc.), and the future CFOSAT mission [1] which will embark the wave scatterometer SWIM (low incidence real aperture scanning scatterometer) to provide directional wave spectra with wavelengths from 70 to 600 m. Because directional spectra contain a large amount of information, a way to use this information is usually to decompose the two-dimensional spectra into several partitions, each of them characterizing a “wave component”. For example partitioning wave spectra is a way to distinguish swell components from wind sea components that have different spectral characteristics and correspond to different wave evolutions (swell waves, in opposite to wind sea waves, re defined as ocean waves which are no longer under the influence of the local wind). Partitioning is a key issue for a good retrieval of wave parameters such as significant wave height, peak direction and wavelength, of the different wave components in a 2D wave spectrum. Partitioning methods of wave spectra have originally been developed for analysing outputs of wave forecasting numerical models. The most famous method is the watershed method proposed by Hanson and Phillips [2]. Extending the application of this method to satellite data raises several difficulties because of the presence of noise in the data, which makes the result of the partitioning rather sensitive to the noise. Our presentation will address this issue in the context of 2D wave spectra observed from i) SAR (Sentinel-1), ii) simulated wave spectra expected from SWIM observations (including noise effects) and iii) observed wave spectra from the airborne wave scatterometer KuROS [3]. We will compare different adaptations of the watershed principle to account for noise effects. Furthermore, we will present results obtained from an alternative method, namely a method based on a Bayesian approach. In opposite to the watershed method, the bayesian method presents the advantage of taking explicitly into account noise into the partitioning process and can even be used to estimate the noise floor. After explaining both methods and showing comparative results, we will discuss the pros, cons and limits of both partitioning techniques.
Complete list of metadatas
Contributor : Catherine Cardon <>
Submitted on : Saturday, May 7, 2016 - 8:53:27 PM
Last modification on : Friday, January 10, 2020 - 3:42:28 PM



Lauriane Delaye, Jean-Luc Vergely, Danièle Hauser, Gilles Guitton, Alexis Mouche, et al.. Partitioning ocean wave spectra obtained from SAR observations or from the future real aperture wave scatterometer SWIM on CFOSAT. Living Planet Symposium 2016. Proceedings of the Conference ESA-SP 740, May 2016, Prague, Czech Republic. pp.60. ⟨insu-01312619⟩



Record views