Ocean Surface Currents from Envisat-ERS2 30-minute Lag Sequential SAR Images
Qazi, Waqas A.1; Emery, William J.1; Fox-Kemper, Baylor2
1University of Colorado, Boulder, UNITED STATES; 2Brown University, UNITED STATES
Synthetic Aperture Radar (SAR), with its cloud-penetrating, high resolution day-night operational capability, provides an attractive option for surface current measurement. At intermediate incidence angles (20°-75°), the primary scattering mechanism is Bragg scattering through small-scale surface capillary waves; however long-scale waves also affect the backscatter by influencing the capillary waves through the processes of tilt and hydrodynamic modulation. Biogenic surface slicks associated with fish, upwelling, and phytoplankton blooms generally form monomolecular layers on the sea surface. The effect of these monomolecular slicks is to dampen the surface capillary waves by "Marangoni damping", which in turn reduces the Bragg backscatter. Biogenic surface slicks thus appear in SAR imagery as lower backscatter regions, as compared to the surrounding ocean surface. A few exploratory studies have shown that surface slicks act as passive tracers for advection due to surface currents. Previous work on surface slick feature tracking in SAR imagery for estimation of surface currents has been limited to processing and analysis of relatively small and sporadic datasets, and a concise general technique for the same has not been reported. Comparisons with other datasets have also been scarce.
For most of their mission coincidence time frame (2002 - 2011), ESA's Envisat and ERS-2 remote sensing satellites with onboard C-band SAR instruments remained in a specific orbit configuration where ERS-2 followed Envisat in the same orbit with a time-lag of 30 minutes. This small time-lag with their high spatial resolution makes these SAR sensors a unique resource for measurement of ocean surface currents at the submesoscale to mesoscale time and length scales. In our talk, we'll report the use of the Maximum Cross-Correlation (MCC) method in a semi-automated algorithm with minimal user input to process a large dataset of Envisat-ERS2 30-min lag pairs over coastal regions and generation of surface currents at resolutions of ≈2 km.
The primary geographical area of study is the coastal California Current System (CCS) which is relatively well understood, and its oceanographic characteristics at the meso- and larger scales are well-known. However, to date, many of the developments in submesoscale oceanography in this region have been led by theory and modeling without observational validation. Recent submesoscale simulations in the CCS show that the surface velocity field follows a spectral slope of k-2 that is shallower than that predicted by ordinary quasi-geostrophic (QG) theory (k-3), but steeper than that predicted for surface quasi-geostrophic (SQG) theory (k-5/3). HF radar surface current observations also show agreement with the k-2 spectral slope at submesoscale wavenumbers. The velocity spectra calculated from SAR MCC currents, shown in Fig. 1, agree with this important theoretical and modeled result and follow the the k-2 power law in the wavenumber range of 10-4 rad/m to 10-3 rad/m. This shows that the SAR MCC derived currents depict actual geophysical signals in the CCS.
MCC currents derived from SAR are compared with HF radar currents dataset over the US California Coast. Some sample plots of re-gridded MCC SAR and HF radar vectors are shown in Fig. 2. Both MCC SAR and HF radar vectors show the same general circulation, however some disagreements can also be seen in these sample plots. Fig. 3 shows the histograms of residuals for magnitude and direction (anti-clockwise from East) for HF radar currents subtracted from MCC SAR currents. The residuals are quite symmetric about the mean, are unimodal, and show some agreement to the estimated normal histograms. The MCC SAR currents agree overall with the HF radar currents in identifying current vector direction, but have higher magnitudes, by ≈11.5 cm/s, than HF radar currents, which may be due to the fact that SAR penetrates only a few cm into the ocean surface while HF radar currents are averaged over the top 1 m of the ocean surface. The histograms of the residuals for the along- and cross-shore components of the MCC SAR and HF radar vector fields are shown in Fig. 4, calculated in the same way as Fig. 3. The histograms are again unimodal and symmetrical about the mean, and show some agreement to the estimated normal histograms. However, a marked difference can be observed between the means: the means for the cross-shore components are close to 0 while the means for the along-shore components have a significant positive value (≈6 cm/s). It seems that the cross-shore component has very good agreement while the along-shore component has higher magnitudes for MCC SAR. The appearance of most of this magnitude difference in the along-shore component is probably due to higher HF radar accuracy in the direct radial cross-shore measurements as compared to lower accuracy in the along-shore components derived from interpolation of multiple cross-shore radial measurements.
Further MCC SAR currents are being processed over the US East Coast and results from them will be reported also.