Session: Methodology I

In the last two decades, satellite altimetry has emerged as one of the main tools for the investigation of surface ocean dynamics. Multiple altimeter observations can be combined to produce global maps of surface velocities capable of resolving structures of the order of 100 km. However, they cannot provide information at the smaller scales (mesoscale and sub-mesoscale), which in recent years have been recognised to be critical for the ocean energy budget and global biogeochemical cycles. The signatures of (sub)mesoscale processes can be retrieved from along track observations, characterized by higher resolution than the gridded products. However, they can only provide estimates of the velocity component perpendicular to the track. Methods based on multi-sensor satellite synergies represent a promising approach for retrieving the full two dimensional velocities.

Here, we propose an approach that combines ocean colour, sea surface temperature and sea level altimetry to improve the accuracy of along-track surface velocities. In particular, maps of ocean chlorophyll and sea surface temperature are used to identify the location and direction of the main ocean fronts. As these features are primarily stirred by horizontal surface dynamics, their direction provides a good indication of the direction of the main currents. Therefore, information on front directions is combined with altimeter information on the flow perpendicular to the track to produce a section of 2-D velocities.

This approach has been initially developed and implemented in the southern Madagascar region. The area is particularly favourable for its application because of the strong visible and infra-red signatures of the East Madagascar Current which flows poleward along the eastern edge of the Madagascar and crosses TOPEX/Jason track 196 tracks almost perpendicularly. The marked gradients in both ocean colour and sea surface temperature favour the formation of strong frontal features, providing a good test location for this method. Our results are compared to the velocity time series of three moorings deployed along this track from February 2005 to April 2006.

The general concepts at the base of the proposed approach can be easily extended to other ocean regions and would be especially applicable when ocean colour and SST sensors are on the same satellite as the altimeter, as is the case for Envisat. Co-localized multi-sensor observations from Sentinel-3 will thus provide an ideal dataset for further development and applications. Furthermore, combined with automated tracking methods, the approach could provide regional maps of surface velocities at higher-resolution than standard products.

The Delay-Doppler Altimetry (DDA) offers many advantages compared to conventional altimetry. However, only part of its potential has been exploited so far.

Within the L1 chain, some parameters can be tuned (or options selected) in order to adapt the processing to the observed scenario or to the user needs. This represents an important change and a big advantage.

Some of these processing options are the following:

Surface focusing
The L1 processor performs a sampling of the surface according to Doppler resolution. However, the sampling can be modified. This feature is very useful when inland waters are to be monitored or calibrations over transponder are to be taken place.

Adaptable delay-Doppler processing
Over low-variability surfaces like the ocean, an approximate DD map can be computed in order to minimise the processing time. On the other hand, when higher variability surfaces like coastal areas or inland waters are observed, an exact DD map is computed, prioritizing accuracy and precision over processing time.

Stack masking
Stack data can be highly contaminated. Interference can appear due to ambiguities, aliasing, land contamination, ships, etc. In order to have cleaner data, a 2-D mask can be applied.

Weightings
Different weightings can be applied both along- and across-track, that can improve the geophysical retrievals. Examples are: the antenna pattern to remove the antenna contribution both along- and across-track; a Hamming (or Hanning) weighing to reduce the undesired secondary lobes; etc.

High resolution Sigma-0 retrieval
The Level 1b provides the correction factor, derived from the radar equations, that is later used to compute the Sigma-0.

Instead of using only one value for a single L1B waveform, it is possible to correct each beam that conforms a stack by its adequate factor, prior to average the beams. Thus, a much more precise result is obtained.

These are some of the options that have been developed in isardSAT within the scope of the Sentinel-6 P4 GPP project, under an ESA/ESTEC contract, and that can be configured when processing altimetry data. Since new concepts or ideas arise every day, it is important to be able to easily tailor our processing chain to new options, as we have demonstrated the impact on the geophysical retrievals of interest.

The Cryosat-2 SAR mode ground processing has been initially designed to enable the full exploitation of the capabilities of the SIRAL altimeter and ensure that the scientific returns of this mission are maximized. The development strategy of the on-ground processing aimed at both reducing the noise speckle of the radar altimetric measurements and achieving high along-track resolution over relatively flat surfaces (initially dedicated to sea ice surfaces but this mode is also interesting in ocean domains). The operating system allows today to resolve Doppler bins with non-overlapping segments on the surface enabling to mitigate the contamination between adjacent Doppler bins. In addition, the actual method allows a better measurement precision compared to the conventional altimetry thanks to the higher number of statistically independent looks contributing to the measurement average.

Today all SAR altimetry community agreed that the range noise reduction is not as high as expected (as stated in the Raney paper considering the number of beams averaged in a stack). R&D studies also demonstrated that this limited improvement is due to the fact that in a stack, the contributing beams to the final averaged waveform have different mean shapes related to the looking angle of measurement, leading to an effective number of looks that is lower than the number of beams. This number has been computed theoretically and is consistent with the actual SAR retracking estimation noise. It is thus essential to develop alternative methods allowing a better processing to take maximum advantage from the Doppler processing. It is of major interest for the Sentinel-3 mission (and the upcoming Jason-CS mission).

Different processing approaches have been then recently studied with Cryosat-2 data showing potential improvement of the SAR altimetric measurements’ precision over ocean, at the expense of a slightly lower spatial resolution for some of the proposed solutions. For each SAR/Doppler processing scheme, a data set of at least one-month duration has been produced over open ocean and their performances assessed through the use of common validation tools and protocols via comparison with operational-like data products (generated with CNES processor CPP). The aim is to determine whether the new processing schemes have a potential impact in operational use or not.

We will present the principle of these methods, and their benefits will be displayed.

SAR mode altimetry represents an intrinsic technical improvement over conventional pulse-limited (LRM) altimetry allowing better signal-to-noise ratio and true 350-m resolution along the ground track of the satellite, and the satellite oceanography community is eager with anticipation for the full SAR mode coverage over the oceans that will be provided by Sentinel-3.

While we wait for Sentinel-3, measurements being gathered by CryoSat-2 over a few SAR mode oceanic patches are already allowing us to convincingly demonstrate those improvements. Over the open ocean these improvements translate into a better view especially of the sub-mesoscale features that are aliased by conventional altimetry. This is clearly shown by the comparison of spectra from CryoSat-2 with those from other missions. The advantage of the SAR mode over the LRM is potentially even more visible over the coastal zone due to the possibility of a better screening of the waveforms against land contamination. Over this domain the orientation of the track with respect to the coastline is another important factor.

In this contribution we will first review various activities for the verification and validation of Cryosat-2 SAR mode data over the open ocean. These include looking at the histograms and spectra of the parameters in comparison with other missions, analysis of the LRM/SAR mode transition regions and comparison with open ocean buoys for SWH. On the basis of those results we discuss some the improvements expected with the global SAR mode view from Sentinel-3, especially in terms of monitoring of mesoscale eddies and sub-mesoscale features.

We will then discuss the performance of the instrument in the coastal zone, with an analysis of the instrumental noise w.r.t. distance from the coastline, effects of the angle of approach to the coastline and comparison of derived Sea Level with Tide Gauges at the coast, and we discuss the implications for the observations of coastal currents and extreme events, which are illustrated by recent examples of storm surges captured by CryoSat-2.

This presentation is one of a three-part concerted effort to illustrate the technical advances and product and applications development from CryoSat through S3 to S6. The three coordinated presentations are:
1. CryoSat to Sentinel-6: Processing Steps and Technology Leaps (Martin-Puig et al.)
2. CryoSat to Sentinel-6: Oceanographic Advances and Coastal Incursions (Cipollini et al.)
3. CryoSat to Sentinel-6: Product Evolutions and Quality Revolutions (Scharroo et al.)

The activities of the Sentinel-3 Validation Team includes the German Bight, where we will calibrate and validate the altimeter data against geodetic referenced tide gauge stations.  As preparatory work, we have used CryoSat-2 data to retrieve the three geophysical parameters: sea surface height above the reference ellipsoid (SSH), significant wave height (SWH) and wind speed at 10 meter above the sea surface (U10).

 Synthetic Aperture Radar (SAR) altimetry is expected to provide improved precision and along-track resolution compared to the conventional low-resolution mode (LRM) radar altimetry. CryoSat-2 enables a quantitative comparison of SAR and Pseudo-LRM (PLRM) data derived respectively from a coherent and an incoherent processing of the same SAR echoes. In this respect CryoSat-2 is very similar to Sentinel-3.

 We cross-validated the different modes of measurements and validated them against in-situ and model data to derive precision and accuracy at 1 Hz in open ocean, at distances larger than 10 kilometers from the coast. The analysis is performed during the two years 2011 and 2012.

 Both the PLRM and the SAR scheme include waveform zero-padding and identical environmental, geophysical, and atmospheric attenuation corrections. A Look Up Table is additionally used in SAR to correct for approximations of the Point Target Response (PTR) applied in the retracking procedure.

 The regional cross-validation analysis proves the good consistency between PLRM and SAR data, with no bias and rms differences of 3 cm, 21 cm, and 0.26 m/s for SSH, SWH, and U10, respectively. The precision of SSH and SWH is higher in SAR than in PLRM (by a factor of 2), while the precision of U10 is 1.4 times better in PLRM than in SAR. At 2 m waveheight, the SAR precision is 0.9 cm for SSH, 6.6 cm for SWH. and 5.8 cm/s for U10.

 The in-situ analysis shows that SSH and U10 have comparable accuracy in SAR and PLRM, while SWH has a significantly higher accuracy in SAR. With a maximum distance of 20 km between altimeter and in-situ data, the minimum values obtained for their rms differences are 7 cm, 14 cm, and 1.3 m/s for SAR and 6 cm, 29 cm, and 1.4 m/s for PLRM.

 The regional comparison between altimetry and models shows that the altimeter products agree best with the ECMWF for wind speed and with DWD/LSM for wave height.

 We attribute part of the residual differences between SAR and PLRM to the LUT table applied in SAR and not in PLRM. We finally evaluate the correctness of this attribution by computing and applying a LUT table also to PLRM data.