Waves

The analysis of combinations of subaperture SAR images to resolve the propagation direction ambiguity of ocean wave signatures has been discussed in the literature since the early 1990s. We will demonstrate in this presentation how spotlight-mode products of contemporary satellite SAR systems such as TerraSAR-X and COSMO-SkyMed enable us to take this concept to a new level. The long SAR integration times and small pixel sizes of these products make it possible to generate a short "movie" of subaperture images, in which wave motions become clearly visible even to the naked eye. By applying a filter based on the theoretical dispersion relation of ocean waves, we can then separate ocean wave signatures from other signatures to obtain a very clean, virtually noise-free picture of the moving ocean wave signatures before inverting them into surface slope and elevation fields. This does not only make the inversion easier, but it also eliminates the usual need to average image power spectra over several subwindows for noise reduction. As a result, the phases of the wave signatures can be taken into account in the inversion process, such that the retrieved wave field can be expected to be correct in its spatio-temporal representation (showing wave crests and troughs and their evolution during the SAR integration time at the correct locations), not just in the spectral representation that would be obtained from conventional wave retrieval techniques for SAR images. We will present several examples of TerraSAR-X and COSMO-SkyMed images and the retrieved wave fields and discuss the quality of the results and the capabilities and limitations of the proposed technique. In addition to the use of subaperture images and the dispersion relation filter, a novel element of our wave retrieval algorithm is the processing of logarithmically-scaled intensity images (dB images) instead of linear images. This approach can be demonstrated to reduce contributions of higher harmonics in the Fourier analysis of the images, and it is expected to improve the inversion to surface slope and elevation fields as well, although this is still a matter of ongoing tests at the time of writing of this abstract.

The SAR image cross-spectrum between a pair of sub-looks separated in time has been widely used to help remove 180° ambiguity of detected ocean swell systems. Ocean swells move along the waves' propagation direction during this offset time of the order of SAR integration time, therefore, co- and cross-spectra can help to consistently estimate velocity characteristics of the randomly moving sea surface scatterers related to scales larger than SAR spatial resolution. The intermediate radial waves longer than SAR spatial resolution but shorter than ocean swells are, to first order, driven by surface winds. Thus, a new parameter MACS, averaged complex cross-spectra over intermediate radial waves domain, is proposed to manifest impacts of wind speed and direction on these moving scatterers.

Sentinel-1 wave mode operates in novel ‘leap frog’ acquisitionmode. A vignette is acquired every 100 km at two alternate incidence angles (23° and 36.5° respectively ), withtwo images at the same incidence 200km apart. Large data sets have been constituted systematically by co-locating Sentinel-1A/Bwave mode acquisitions and ECMWF forecast winds. Efficiently, distinctive features are revealedfor two incidence angles as well as dual-polarizations.

Phase of MACS consistently show temporal evolution of sea surface scatterers under various wind speeds. This relationship is distinct from incidence angle of 23° to that of 36.5°. In specific, variations of phase at 23° agrees well with the present nonlinear SAR imaging mechanism, being a approximate constant relative to wind speed. On the contrary, phase of 36.5° shows increasing trend with wind speed with different slopes at upwind and downwind. This may lead to a revisit on applicability of the present SAR imaging mechanism in reproducing the phase shift at incidence angle of 36.5°.

More encouragingly, imaginary component of MACS, to first order, show a linear variation relative to wind speed. It is a signed quantity with negative at upwind and positive at downwind with greater values in HH polarization than in VV polarization for given wind speed and incidence angle. This characteristics is similar to that of Doppler Centroid Anomaly (DCA). As such, preliminary investigation is carried out, demonstrating its potential in capturing upper ocean surface current.

The capability of estimating ocean wave spectrum and some of its statistics using synthetic SAR images has been demonstrated since the 1970s. Later algorithm developments using polarimetric SAR (PolSAR) data have indicated that fairly good estimation of some ocean wave parameters is possible without the need of a first guess spectrum, normally obtained from numerical wave models, and of complex computational iterative schemes needed to converge to a best solution.  More recently, a C-band PolSAR ocean wave algorithm was presented [1] and validated for Radarsat-2 (RSAT-2) polarimetric imagery [2]. This algorithm estimates the long-wave slopes (in range and azimuth directions) using a linear polarization orientation modulation transfer function (MTF) developed by [3] in addition to the tilt, hydrodynamics and velocity bunching MTFs. Despite the relative good results obtained, the authors [2] acknowledged that further research was needed to determine the extent to which the algorithm is capable of handling a more general wind‐driven wave spectrum, particularly for cases of multimodal wave systems.

Motivated by these initial results and with a goal of advancing the polarimetric algorithm for general wave systems, we propose a modification of the original algorithm which was tested using polarimetric Radarsat-2 imagery. To separate the wave systems in multimodal cases and evaluate their individual statistics, the new approach uses a multiscale processing scheme together with the Mean Square Slope statistics theory of ocean waves, which has been originally developed from optical and radar altimetry sensors.

The first results obtained for a multimodal wave system present off Southeastern Brazil on August 12^th, 2016, and using a C-Band full polarimetric Radarsat-2 image in Wide Fine Quad-Pol beam mode (Single Look Complex product) show a reasonable agreement when compared to data from a nearby meteocean wave buoy.

More C-band PolSAR imagery together with well collocated wave buoys data is required to better evaluate under several environmental conditions and to validate the methodology. We reckon that this new approach, being a fast and independent satellite algorithm to assess multimodal ocean wave systems, could contribute to several coastal/oceanic engineering applications and monitoring systems.

References:

[1] He, Y.; Shen, H.; Perrie, W. Remote sensing of ocean waves by polarimetric SAR. Journal of Atmospheric and Oceanic Technology, v. 23, n. 12, p. 1768–1773, 2006.

[2] Zhang, B.; Perrie, W.; He, Y. Validation of RADARSAT-2 fully polarimetric SAR measurements of ocean surface waves. Journal of Geophysical Research: Oceans, v.115, n. 6, p.1–11, 2010.

[3] He, Y. et al. Ocean wave spectra from a linear polarimetric SAR. IEEE Transactions on Geoscience and Remote Sensing, v. 42, n. 11, p. 2623–2631, 2004.

Since the launch of Sentinel-1A and -1B missions in April 2014 and April 2016, respectively, SAR images are acquired massively and increasing in worldwide coastal regions using the new TOPS (Terrain Observation by Progressive Scans) acquisition mode. Such wide swath SAR images are a crucial source of information for the observation of wave in coastal areas, where complex interactions with currents occur. Besides, these wide swath images are the only acquisition mode over European seas and the Eastern Atlantic Ocean. However, there is currently no swell spectra product available based on this acquisition mode.

TOPSAR mode is intended to replace the conventional ScanSAR mode, used before with ENVISAT/ASAR, as it achieves the same coverage and resolution as ScanSAR, but with a uniform SNR (Signal-to-Noise Ratio). In TOPS mode, the ocean surface is sampled via successive bursts during which the antenna is electronically steered in the azimuth direction. This reduces the dwell time which directly impacts the capability to extract the swell spectra: as classical wave-spectra estimation methods using SAR SLC images are based on cross-spectra estimation, reducing the time separation between cross-looks hampers the ability to remove the ambiguity on the swell propagation direction.

This paper presents an alternative technique, taking advantage of the TOPS mode scanning and exploiting two complementary approaches based on the spectral information available inside each burst and in the overlapping region between two successive bursts.

For each subswath, the overlapping areas located at burst’s edges spans over only a few kilometers in azimuth and the look separation time reaches ~2s which enables resolving the propagation direction ambiguity of the swell spectrum even over a small area (to be compared with the 0.4s of S1 Wave mode, 20x20km). The possibility to estimate the swell phase velocity is demonstrated and compares well with wave dispersion relation for noise-free images. However, the swell spectra energy is noisier than when using the overall burst area due to the small overlapping area. As a complementary information, the swell spectral energy is estimated over the overall coverage of each burst and the previously removed ambiguity information is used by continuity of the swell field over the SAR image. The impact of proposing a modified Level-0 to Level-1 processing with wider inter-burst coverage is also presented.

This study has been partly funded by the ESA SEOM Ocean studies

Sentinel-1A and -1B Level-2 products acquired in a specific Wave Mode are acquired routinely worldwide over open ocean basins. They contain partitioned directional swell spectra and integral swell parameters for each partition, i.e. significant wave height (SWH), peak period and peak direction. Based on these L2 SAR observations, Level-3 products are estimated and are delivered in the Copernicus Marine Environment Monitoring Services since 2018. These products, delivered on a daily basis, describe the swell integral parameters along its propagation in deep waters, from their storm source until the land away from islands and in the absence of currents. They are often referred to as “fireworks” due to their visual aspect when represented on a world map: they evidence swell generated in localized regions coinciding with strong storms, mostly corresponding to extra-tropical events, from which they propagate over thousands of kilometers across entire ocean basins.

First and using great circle theory, Level-2 observations with SWH larger than 30cm and peak wavelength larger than 200m are back-propagated in space and time. Converging swell observations, which belong to the same storm, are then gathered in swell field groups. This first step ensures the overall consistency of the selected converging observations in terms of peak direction and peak wavelength. Once this generation source is known, the significant height decay can be estimated on top of the swell propagation path. For each swell field group of propagated observations, the spatial structure of the 3 integral parameters can be analyzed along propagation to further filter out swell observations with SWH values appearing as outliers. After applying these successive filters based on the consistency of each Level-2 integral parameter with the overall swell field distribution described in the Level-3, about 18% of all the available Sentinel-1 Level-2 partitions remain.

The accuracy of the Level-3 propagated integral parameters is estimated over a 4 month dataset in 2016 co-located with numerical wave model outputs and quality controlled in situ measurements. The accuracy of the corresponding Level-2 is also estimated and compared with that of the rejected partitions, showing a great improvement of the integral parameter estimation for the selected observations with respect to the rejected ones.

Application of the Level-3 products are various, from assimilation in numerical wave models to study of swell decaying processes.