This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination. IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING 1 Co-Cross-Polarization Coherence Over the Sea Surface From Sentinel-1 SAR Data: Perspectives for Mission Calibration and Wind Field Retrieval Nicolas Longépé , Alexis A. Mouche , Laurent Ferro-Famil , Member, IEEE, and Romain Husson Abstract— Spaceborne synthetic aperture radar (SAR) has Index Terms— Dual-polarization, polarimetric calibration been used for years to estimate high-resolution surface wind (PolCAL), Sentinel-1 (S-1), single look complex (SLC), synthetic field from the ocean surface backscattered signal. Current aperture radar (SAR), wind field retrieval. SAR platforms have one single fixed antenna, and traditional inversion/retrieval schemes rely on one copolarized channel, I. INTRODUCTION leading to an unconstrained optimization problem for providing independent estimates of wind speed and direction. For routine YNTHETIC aperture radar (SAR) ocean surface wind application, this is generally solved with aprioriinformation Sretrieval has been originally based on a single observed from the numerical weather prediction (NWP) model, inducing quantity, the copolarized normalized radar cross section severe limitations for rapidly evolving meteorological systems where discrepancies can be significant between model and mea- (NRCS), even though many other SAR observables are poten- surements. In this study, we investigate the benefit of having tially available (e.g., cross-polarized NRCS and the Doppler two simultaneous acquisitions with phase-preserving information shift). Since the first attempts ( [1] with Seasat and [2] with in copolarization and cross polarization provided by Sentinel- ERS-1) to inverse the NRCS into wind vector, the unique 1 (S-1). A comprehensive analysis of the co-cross-polarization use of the copolarized signal seems to be the rule for routine coherence (CCPC) is performed to adequately estimate and calibrate CCPC values from S-1 interferometric wide (IW) mode operational SAR wind measurements. Similarly, the only SAR images acquired over the ocean. A new polarimetric calibration observable used to provide ocean surface wind fields in (PolCAL) methodology based on least-squares (LS) criterion and the official ESA/Copernicus Level-2 Ocean products is the direct matrix inversion is proposed yielding crosstalk estimates. copolarized NRCS. This approach is directly derived from We document CCPC odd symmetry with respect to relative wind scatterometry [3] and relies on a transfer function between the direction for light to medium wind speeds (up to 14 m/s) and incidence angle from 30◦ to 45◦. The azimuthal modulation is radar observables and the surface wind speed and direction. found to increase with both wind speed and incidence angle. An This function, also called the geophysical model function analytical model C-band polarimetric geophysical model function (GMF) or C-band MODel (CMOD) when applied to C-band (CPGMF) is provided. The synergy of the CCPC with other radar scatterometer, has been refined several times since the fourth parameters, such as backscattering coefficients or Doppler, to version (CMOD4) [4], [5]. In particular, different strategies further constrain the inversion scheme is assessed, opening new perspectives for SAR-based wind field retrieval independent of have been applied to collocate the radar NRCS with a refer- any NWP model information. ence wind given by the ECMWF analysis (CMOD4 [4], [5]), the ECMWF analysis combined with aircraft measurements Manuscript received April 22, 2020; revised August 28, 2020, October 18, for high-wind speeds (CMOD5 [6]), or in situ wind measure- 2020, and November 28, 2020; accepted January 8, 2021. This work was ments (CMODIFR2 [7]). Additional modifications have been supported by the European Space Agency (ESA) through the Scientific Exploitation of Operational Missions (SEOM) “S1 for Ocean Studies” Project. also proposed to take into account atmospheric stratification (Corresponding author: Nicolas Longépé.) (CMOD5n [8]) or extreme wind (CMOD5h [9]). However, Nicolas Longépé was with the Environment and Climate BU, Collecte the SAR instruments and missions peculiarities make the Localisation Satellites (CLS), 29280 Plouzané, France. He is now with the -Lab Explore Office, European Space Research Institute (ESRIN), European challenges of wind measurement very different than for a Space Agency (ESA), 00044 Frascati, Italy (e-mail: [email protected]). scatterometer. Alexis A. Mouche is with the Laboratoire d’Océanographie Physique et First, since existing SARs have one single antenna pointing Spatiale, Institut Francais de Recherche pour l’Exploitation de la Mer, 29280 Plouzané, France. in the satellite across-track direction, there is only one view Laurent Ferro-Famil is with the Remote Sensing Department, Institute of angle per wind vector cell. This particularity yields to an Electronics and Telecommunications of Rennes, University of Rennes 1, underconstrained inverse problem to retrieve both wind speed 35000 Rennes, France. Romain Husson is with the Space Observation Division, Environment and and direction. The most common method for SAR wind Climate BU, Collecte Localisation Satellites (CLS), 29280 Plouzané, France retrieval is to combine copolarized VV NRCS and comple- (e-mail: [email protected]). mentary information on the wind direction as the input of Color versions of one or more figures in this article are available at https://doi.org/10.1109/TGRS.2021.3055979. empirical GMFs using a Bayesian scheme [3]. The simplest Digital Object Identifier 10.1109/TGRS.2021.3055979 and probably most efficient method for operational purposes is . ACCEPTED MANUSCRIPT This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination. 2 IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING to take the wind direction from a numerical weather prediction relying on Doppler and/or VH-NRCS, such as the next MetOp (NWP) model. This works generally well for meteorological scatterometer [24], [25]. situations without sharp wind gradients but tends to fail for Third, due to the on-board memory limitations, the acquisi- rapidly evolving systems, such as atmospheric fronts or low- tion strategy, and the multiple SAR acquisition modes, the pressure systems, including polar lows and hurricanes. Indeed, number of large swath products acquired over the ocean for these situations, the global operational NWP models may from SAR sensors is far less than for scatterometers. The be too coarse in space and time or may have phasing issues. scatterometer-derived CMOD-based solutions have been sys- The direction of the wind (with a 180◦ ambiguity) can be also tematically applied to C-band SAR: GMFs for VV and a so- obtained from streak-like features visible in NRCS images and called polarization ratio for HH polarization [26] However, measured using wavelet analysis [10], [11] or local gradients since the launch of ENVISAT/ASAR and Radarsat-2, the estimation at different spatial scales [12], [13]. This approach amount of data available has significantly increased. The mainly relies on the fact that organized large eddies (OLE), direct consequences are the recent attempts to derive a also referred to as roll vortices, in the marine atmospheric GMF directly based on SAR measurements (CSARMOD) boundary layer impact the centimeter ocean waves and so for both VV- and HH-NRCSs [27]. In addition, a GMF for the C-band backscattering. However, the occurrence (seasonal HH-polarized SAR data, so-called CMODH, has also been and spatial distribution) of the OLE at the global scale as developed using collocated ENVISAT ASAR backscatter mea- well as the sensitivity of the radar to them are not well surements and ASCAT winds. CMODH is validated by a large characterized yet, leading to questions regarding the feasibil- number of Radarsat-2 and S-1A/B HH-polarized acquisitions ity of estimating their directions for all SAR observations. under different wind speeds and buoy observations [28]. Moreover, although very relevant as an indication of the wind Beyond intensities, SAR sensors have also the capabili- flow, the OLE orientation is not strictly aligned with the ties to perform spectral analysis for each polarization chan- ocean surface wind direction. Typically, the measurements nel and even to provide relative phase information between from Weather Surveillance Radar by Morrison et al. [14] in polarization channels with dual- or quad-polarization acqui- the case of hurricane-generated wind rolls indicate that their sition modes. Indeed, due to the improved spatial resolution, most probable orientation was tilted by 10◦ from the mean S-1wave mode measurements can further be extended toward wind toward the center. This value is further confirmed by the shorter scale waves, i.e., within the surface wave equilibrium model experiments from [15]. range. This allows SAR image cross-spectra estimates, includ- Second, SAR can provide other observables than a scat- ing range-traveling intermediate wind waves. The statistical terometer, which can help to overcome the limitation of hav- analysis of the spectral energy confirms its sensitivity to both ing one single antenna for geophysical applications over the wind speed and wind direction. Comparable to the Doppler ocean. In particular, the launch of ENVISAT/ASAR in 2002 estimate,