Activities

In the past, the tandem phases between 2 altimeter missions (TOPEX-Poseidon/Jason-1, Jason-1/Jason-2 and Jason-2/Jason-3) were very useful to compare all instrumental and radiometric parameters as well as the sea level height content. Thanks to this kind of comparisons, most of the anomalies present on the new mission were detected quickly, and then corrected. Furthermore, the former mission used as the reference, was also improved by these kinds of comparisons. For instance, the TOPEX and Jason-1 tandem phase revealed strong geographical correlated discrepancies on the sea level height due to error on orbit calculation and on altimeter TOPEX range (Dorandeu et al., 2004).  Therefore, to build a Sentinel-3 altimeter climate data record over ocean, it is of great importance to perform exhaustive comparisons between S3-A and S3-B during the tandem phase. All the instrumental parameters and geophysical corrections shall be compared, including the sea level height.

Several validation diagnoses shall be performed as the monitoring of global and regional differences in terms of mean or variance. Cross- comparisons with other altimeter missions can also to discriminate potential discrepancies observed between S3-A and S3-B.

Thanks to satellite altimetry, the Global Mean Sea Level (GMSL) continuous record is maintained since January 1993 using TOPEX/Poseidon data followed on the same orbit by Jason-1, Jason-2 and Jason-3 records (Ablain et al., 2009, 2015, 2017). In Zawadzki & Ablain, 2016, the importance of calibration phasesto derive a continuous and accurate MSL with respect to the GCOS requirements was shown. The Global MSL bias uncertainty is about 1 mm (in a confidence interval of 95%) between two recent missions benefiting from a tandem phase (e.g. Jason-2 / Jason-3) whereas it increases to 2.5 mm when it is not the case (e.g. Jason-3, Sentinel3-a). However, these uncertainties are obviously correlated with the method used to estimate the inter-mission relative bias.

GMSL biases for SAR and PLRM modes are estimated at, respectively, 8.32+/-0.62 mm and 0.13+/-0.57 mm (2-sigmas) but the short period of the tandem phase (4 cycles) creates strong limitations on the robustness of the results (which are based on cycle weighted mean). The orbit change of S3B after the tandem phase impacts the true GMSL bias and questions the tandem phase configuration for GMSL perspective.

The Sentinel-3 Synthetic Aperture Radar (SAR) mode enables simultaneous SAR and Pseudo-Low-Resolution Mode (PLRM) measurements from the same SAR echoes, allowing quantitative cross-comparison between high- and low-resolution mode data over the same surface samples. In this way SAR altimetry can be efficiently validated with respect to the PLRM reference. This is crucial for the calibration of long climate series gathering LRM and SAR measurements. However PLRM is operating at a pulse repetition frequency (PRF) above the 2-kHz limit at which individual echoes are de-correlated between each other [Walsh, 1982]. At high PRF (18-kHz), pulse-to-pulse correlation effects occur, leading to possible biases in the estimation of geophysical parameters.

This activity has provided the first clear evidence and quantification of the P-LRM residual correlated error related to the pulse correlation, This is a timely topic as Sentinel-3 missions, and the future Sentinel-6/Jason-CS mission generate PLRM data at different PRFs compared to conventional Jason altimeter series.

Different studies converged to say that swell impacts SAR-mode altimetry [Aouf et Phalippou, 2015; Moreau et al., 2016; Abdalla et al., 2016]. This further raises concern about potential impact of such ocean wave conditions on the sea level time-series when data from the SAR-mode alitmeter missions (the two Sentinel-3 and the future Sentinel-6) will be incorporated. Recently, Moreau et al. [2016] have shown that the reduction of the SAR altimeter footprint compared to conventional missions (typically of ~300 m for Sentinel-3A) brings it close to the wavelength scales of ocean swell. This impacts parameter retrievals, and generates a higher dispersion in consecutive measurements that increases with the wavelength of ocean swells. It is as if the elevation range estimates were influenced by the short-scale undulating surface.

For the first time, the impact of swell on SAR altimetry has been quantified through cross-spectrum analysis of the signals.

The use of along-track Doppler processing in Synthetic Apertrure Radar Mode (SARM) leads to much finer spatial resolution in the altimeter along-track direction. In theory, the along-track resolution could be as fine as 300 metres and this has raised questions about the possible impact of long ocean surface waves on SARM measurements. The concern is that, if such an additional sea state effect exists in SARM, this could introduce regional offsets and spurious trends in the altimeter climate record as altimeter missions increasingly migrate from LRM to SARM. So far, studies of this problem by different investigators have led to conflicting conclusions about the impact of swell and high sea state on SARM (e.g. Aouf et al., 2015; Bellingham et al., 2016; Moreau et al., 2016). This activity proposes a new exploratory investigation based on S3B SRAL that takes advantage of the availability of quasi-coincident observations from S3A over the full range of sea state conditions. The investigation focuses on lower level data products (L1BS Stack) to check for possible additional sea state signatures in SARM that could degrade performance and/or yield new geophysical information about ocean wave conditions.

Fully-Focused Synthetic Aperture Radar (FF-SAR) is a new sophisticated SAR processing requiring a very precise correction of the range and phase history during the illumination time of a scatterrer, in order to achieve a very fine along-track resolution (half the length of the antenna in the along-track direction)

However the first implementations of FF-SAR processing with Sentinel-3A data has shown unexplained phase residuals.

Thanks to the tandem configuration, we have shown that empircal phase correction in FF-SAR works consistently for both satellites.

Successful identification, and harmonisation, of any differences in S3A and S3B elevation retrievals over ice sheets is critical for determining consistent, long-term climate records over ice sheets, and for robust determination of ice sheet mass loss and associated sea level rise.

The S3 Tandem Phase offers a unique and one-off opportunity to systematically inter-calibrate S3A and S3B SRAL SAR measurements over ice sheet surfaces.

Given that the tandem phase will allow detailed investigation and resolution of L1 inter-sensor biases, it also provides the unique opportunity to isolate specific discrepancies in the ice sheet L2 processing, for example in the slope correction.

This acvtivity provided the first quantitative demonstration that there is no significance difference between S3A and S3B ice sheet elevation measurements. It also demonstrated that complex waveform shapes acquired over the ice margin reflect real, repeatable geophysical signals.

A 30 s time lag for the tandem phase means that, at first order, the same atmosphere state is observed by both S3-A and S3-B MWR. So the main output for the tandem phase will consist on a robust inter-calibration relation between the two instruments based a direct comparison of the brightness temperatures (TB) measured by the two radiometers, channel by channel.

Since no instrumental drift is demonstrated for S3-A MWR so far, the monitoring of the inter-calibration during the tandem phase will also provide valuable information on a potential thermal stabilization of S3-B MWR. If such a drift on TB is observed, this will also allow quantifying the uncertainty of long-term trends estimation using classical vicarious monitoring (hottest TB over Amazonian forest and coldest ocean TB statistical selection).

We take advantage of the tandem configuration to assess the consistency of geometry and radiometry between OLCI-A and OLCI-B. First, the L1 data from OLCI A and B is ortho-rectified on a common gird and converted to TOA reflectances. Smile correction is applied to remove the impact of known spectral response differences on the atmospheric Rayleigh signal. Comparison statitstics are then computed according to the class of scene: cloud, land, water.

The observed differences are analysed in terms of spatial (in the field of view) and spectral characteristics, as well as temporal stability. Dependence on signal level (non-linearity) can be assessed as well.

Intercalibration of OLCI-A and OLCI-B using tandem data is shown to provide a performance better than 0.5%, which is an unpreceded performance

Climate assessment of Ocean Colour time series rely on multi-sensor datasets. The differences between sensors, algorithms and merging approach represent a challenge regarding the temporal consistency of the dataset. The S3 tandem provides the opportunity to understand the impact of inter-sensor uncertainty, and its impact on climate assessment.

The proposed methodology relies on Mélin et al., 2016 for uncertainty aspects and Méin et al. 2017 for the evaluation of trends. The idea is to acquire an understanding of the inherent uncertainty, and from there to assess the ability to detect trends reliably.

To cope with the short period of the tandem, synthetic time series are created with an imposed trend and bootstrapping the “n” times for the inter-sensor uncertainty and noise.

Inter-calibrated OLCI-A and OLCI-B TOA radiances over the drift phase allow to consider that observations over the same pixel/target are two samples of a same observable.

Analysis of the L2 water and land processing step outputs from the drift phase allows to investigate the effect of random sources on the signal such as random distribution of waves on glint and water-leaving reflectance noise, and the effect of instrument-dependent artefacts (e.g. smile). Differences in water-leaving reflectances between co-registered and inter-calibrated OLCI-A and OLCI-B should also reflect radiometric uncertainties independently for OLCI-A and OLCI-B

Inter-calibrated OLCI-A and OLCI-B TOA radiances over the drift phase allow to consider that observations over the same pixel/target are made from two different acquisition geometries (with hopefully small time delay). L2 water processing must however retrieve the same normalized water-leaving reflectance (i.e. water-leaving reflectance corrected for acquisition geometry)

Analysis of the L2 water processing step outputs from the drift phase on common pixels/targets provides a unique opportunity to assess the angular dependencies of the atmospheric correction model

The tandem phase provides a unique opportunity to perform direct pixel-to-pixel comparisons between SLSTR A and B with negligible impact from viewing geometry and scene changes.

The drift phase provides a unique opportunity to test the sensitivity of such comparisons to temporal, angular and spatial differences. It also provides an opportunity to compare the oblique and nadir views of SLSTR.

In this activity, we compute the differences in brightness temperatures (and/or radiances) for each spectral band, over representative uniform scenes (land, ocean, ice and cloud) or high-gradient scenes (coast, ocean currents). Then we analyse the influence of such parameters as: scene radiance level, instrument temperature and operating conditions, latitude, viewing angles, etc.

Although the design of SLSTR-A and B are in principle the same, the instruments have slightly different scan cone-geometries, detector line-of-sight and pixel footprint.  This will mean that the corresponding pixels of the A and B instruments will not be exactly co-registered.  This will give rise to differences in cloud screening even where the same algorithm is used. The tandem phase will allow us to characterise the sensitivity of the instrument to differences in cloud screening. The drift phase will provide additional information.

Although the design of SLSTR-A and B are in principle the same, the instruments have slightly different scan cone-geometries, detector line-of-sight and pixel footprint.  This  means that the corresponding pixels of the A and B instruments will not be exactly co-registered.  This gives rise to differences in the radiometric signals, even if the radiometric calibration is harmonised. The tandem phase allows us to characterise the sensitivity of the instrument radiometric calibration due to differences in pointing.

This activity uses scenes with strong spatial variations (clouds, gas flares, rivers, lakes, coastlines).

Sea Surface Temperature (SST) is a core product of SLSTR. This activity determines the consistency between SLSTR-A and SLSTR-B for the SST product. It will understand which differences are explained by the L1 differences and which are due to the SST processing. The drift phase provides additional opportunity to test the SST processing’s sensitivity to temporal, spatial and angular differences.

We study the differences between the two sensors in a similar manner to that undertaken by the ARC and CCI SST products which derived estimates for corrections to the observed brightness temperatures which are a function of different instrument and meteorological parameters.

We compare the SSTs from each instrument and study any inconsistencies in their error budget if present. This is done both at swath level (direct comparison) as well as at in-situ locations which given improved knowledge about each sensors’ uncertainty will enable and estimate of in-situ uncertainties to be made.