The L3 products are along-track subsampled and cross-calibrated products.

SLA computation

The Sea Level Anomalies (SLA) are used in oceanographic studies. They are computed from the difference of the instantaneous SSH minus a temporal reference. This temporal reference can be either a Mean Profile (MP), in the case of repeat track, or a gridded Mean Sea Surface (MSS), when the repeat track cannot be used. Table gives the MPs and MSS used for the MY 1 Hz SLA processing. The NRT 1Hz and 5 Hz SLA processing is based on a gridded MSS only.

The errors affecting the SLAs, MPs and MSS have different magnitudes and wavelengths. The computation of the SLAs and their associated errors are detailed in Pujol et al. (2016) and Taburet et al. (2019). Errors of the MSS are described in Pujol et al. (2018). Both MP and MSS are referenced to the same reference period as specified in Table 1. The methodology applied to shift the reference period is presented in Pujol et al 2016.

Table 1: Mean Profiles (MPs) and Mean Sea Surface (MSS) used for the 1 Hz SLA computation along the different altimeters.

Altimeter mission	1 Hz MP description
Topex/Poseidon, Jason-1, OSTM/Jason-2, Jason-3 Sentinel-6A	MY: 1 Hz MP computed with Topex/Poseidon (January 1993, April 2002; cycles 11 to 353), Jason-1 (April 2002, October 2008; cycles 10 to 249); Jason-2 (October 2008, May 2016; cycles 10 to 290), Jason-3 (May 2016, December 2021; cycles 11 to 217) and Sentinnel-6A (December 2021, August 2023; cycles 42 to 103) measurements. Referenced to the (1993, 2012) period. NRT before 2022/11/10: 1 Hz MP computed with Topex/Poseidon (January 1993, April 2002; cycles 11 to 353), Jason-1 (April 2002, October 2008; cycles 10 to 249); Jason-2 (October 2008, May 2016; cycles 10 to 290) and Jason-3 (May 2016, March 2020; cycles 11 to 150) measurements. Referenced to the (1993, 2012) period. NRT after: gridded MSS is used as described in Section II.4.5, corrected with long wavelengths of the 1 Hz MP
ERS-1/ERS-2, Envisat,SARAL/AltiKa	1 Hz MP computed with ERS-2 (May 1995, January 2000; cycles 1 to 49), Envisat (May 2002, October 2010; cycles 6 to 94) and SARAL/Altika (Mars 2013, Mars 2015; cycles 1 to 22) measurements. Referenced to the 1993-2012 period.
Topex/Poseidon interleaved, Jason-1 interleaved, OSTM/Jason-2 interleaved, Jason-3 interleaved	MY:1 Hz MP computed with Topex/Poseidon interleaved (September 2002, October 2005; cycles 368 to 481), Jason-1 interleaved (February 2009, March 2012; cycles 262 to 374), Jason-2 interleaved (October-2016; March-2017; cycles 305 to 322).and Jason-3 interleaved (April 2024, August 2023; cycles 300 to 350) measurements. Referenced to the (1993, 2012) period NRT before 2022/11/10: 1 Hz MP computed with Topex/Poseidon interleaved (September 2002, October 2005; cycles 368 to 481) and Jason-1 interleaved (February 2009, March 2012; cycles 262 to 374), Jason-2 interleaved (October-2016; March-2017; cycles 305 to 322). measurements. Referenced to the (1993, 2012) period NRT after: gridded MSS is used as described in Section II.4.5, corrected with long wavelengths of the 1 Hz MP
Geosat Follow-On	1 Hz MP computed with Geosat Follow On (January 2000, September 2008; cycles 37 to 222) measurements. Referenced to the (1993, 2012) period.
Sentinel-3A	MY: 1 Hz MP computed with Sentinel-3A (June 2016, August 2023; cycles 6 to 103). NRT before 2022/11/10: 1 Hz MP computed with Sentinel-3A (March 2016, August 2020; cycles 1 to 61). NRT after: gridded MSS is used as described in Section II.4.5, corrected with long wavelengths of the 1 Hz MP
Sentinel-3B	MY: 1 Hz MP computed with Sentinel-3B (November 2018, August 2023; cycles 19 to 83). NRT: MSS is used as described in Section II.4.5, corrected with long wavelengths of the 1 Hz MP
ERS-1 geodetic phase	No MP available for these missions. A gridded MSS is used as described in Section II.4.5
Envisat new orbit
Jason-1 geodetic, Jason-2 Long Repeat Orbit Phase
Cryosat-2
SARAL-DP/AltiKa
HaiYang-2A
HaiYang-2B
SWOT-nadir

Along track (L3) noise filtering

The filtering processing consists in removing from along-track measurements the noise signal and short wavelength affected by this noise.

For the 1 Hz SLA, this processing consists in a low-pass filtering with a cut-off wavelength of 65 km over the global ocean. This cut-off wavelength comes from the study by Dufau et al. (2016) and is discussed in Pujol et al, (2016). It represents the minimum wavelength associated with the dynamical structures that altimetry would statistically be able to observe with a signal-to noise ratio greater than 1. The cut-off is reduced for regional products to preserve as much as possible the short wavelength signal. The different cut-off wavelength used are summarized in Table 2. No subsampling is applied in the SLA 1 Hz processing.

Table 2: Filtering and subsampling parameters used for L3 1 Hz products.

Region	Product considered	Filtering cut-off wavelength (km)	Distance between two points (km)
Global	Filtered and not-subsampled	~65	~7
	Not-filtered and not-subsampled	–	~7
Europe	Filtered and not subsampled	~40	~7
	Not-filtered and not-subsampled	–	~7

For the 5 Hz SLA construction, two different filtering/subsampling are applied on the upstream 20 Hz measurements:

For the raw (“unfiltered”) SLA 5 Hz computation, a filtering on the raw 20 Hz measurement is applied in order to filter the measurement noise at the 5 Hz Nyquist frequency (i.e., ~2,5 km). Then a subsampling is applied to extract the 5 Hz positions.

For the filtered SLA 5 Hz computation, as for the 1 Hz processing, the short wavelength dominated by the measurement noise is reduced by a low pass filtering. Then, a subsampling is applied to extract the 5 Hz positions. The cut-off wavelength used for the filtering varies from one altimeter to the other, according to their observing capability estimated on the upstream full rate (i.e., 20 Hz) measurement. For the processing over the European area, it was estimated over the North Atlantic area over a reduced temporal period (April-June 2022) and thus could be revised in the future. Currently, the cut of wavelength varies from 40 km for the Sentinel-6A HR mission to 55 km for the Jason mission (Table 9).

For the processing over the global ocean, the observing capability of the different altimeters was estimated over a nearly 1-year period (October 2022-August 2023), considering possible geographical variations. The Figure 1 illustrates the results obtained for Sentinel-6A (SAR mode measurement) and Jason-3 (LRM mode measurement). The cut-off wavelengths can vary from 20 km (Sentinel-6A, West Pacific, to 120 km (HaiYang-2B, South-Est Pacific area). The mean observable wavelength over the global ocean is close to the one observed in the North-East Atlantic area (see Table 3).

Table 3: Effective mean spatial resolution of the DUACS NRT L3 5 Hz: Mean value for the North-Eastern Atlantic Area (estimated over the period April-June 2022) and global ocean (estimated over the period October 2022-August 2023).

	Sentinel-6A HR	Sentinel-3A/B	Jason-3	Swot-nadir	HaiYang-2B
Spatial Wavelengths observable in the North-Eastern Atlantic Area (km)	40	48	55	55	55
Mean spatial Wavelengths observable in the global ocean (km)	40	48	57	57	63

Figure 1: Spatial distribution of the cut-off wavelength used for the 5Hz “sla filtered” processing. Sentinel-6A (left) and Jason-3 (right).

Derived variables generation (for L3 5 Hz)

The L3 derived variables consist of geostrophic currents (absolute and anomalies) in the across-track direction. This processing is applied only for the final L3 5 Hz production.

The geostrophic current is computed using a finite difference of maximal order 5 in open ocean. The order is progressively reduced when approaching the coast or a gap with no valid measurement. This processing is applied on SLA, after filtering of the short wavelengths dominated by noises (see section bellow: Along track (L3) noise filtering), to retrieve the anomaly of the geostrophic current. It is also applied on MDT, to retrieve the mean component of the geostrophic current. This processing is applied on high resolution upset (20 Hz) measurements. The estimation of the geostrophic current in the equatorial band is not available.

The MDT field is obtained by interpolating the Mean Dynamic Topography (MDT) on along-track positions. The MDT used in the DUACS reprocessing is described in section Homogenization and cross-calibration

Measurement subsampling

For the L3 5 Hz production, a subsampling is applied on along-track measurement in order to reach the final 5 Hz sampling. Indeed, the previous processing steps are applied using the upstream measurement sampling, i.e., 20 Hz sampling. This subsampling is thus applied on the different variables delivered in the L3 5 Hz products (e.g., SLA, currents). No subsampling is applied in the L3 1 Hz processing.