Updated L4 products
In November 2024, the NRT product evolves in order to homogenise the NRT and MY processing. While various changes are implemented in MY series (see November 2024 MY changes), corresponding to the DT-2024 product version, only some of them are implemented in NRT. Indeed, other changes are already implemented in NRT production (see July 2024 NRT changes ). The different improvements in the NRT product are summarized in Table 21 and described below.
Table : Overview of the changes introduced in each DT-2024 NRT product version
| Evolution description | Description of the impact |
| L3 & L4: Use new MDT | Improved geostrophic current; regional biases |
| L3: Use new methodology for SLA, ERR and LWE correction computation | Improved LWE correction |
| L4: Use new methodology for geostrophic current computation | Improved mesoscale signal on L4 products |
| L4: Grid spatial resolution change | Finner spatial grid available |
| L3 & L4: CMEMS catalogue changes | Change in products/datasets names |
New MDT
The MDT used in the DUACS is described in section DUACS NRT altimeter standards and November 2024 MY changes
New MIOST mapping used for SLA and ERR
The methods used to process L4 gridded products from the merging of L3 along-track products have evolved from the Optimal Interpolation (OI) method, as detailed by Le Traon et al., (1998a) and Pujol et al., (2016), to the Multi-Scale Inversion of Ocean Surface Topography (MIOST) approach . While the MIOST system can also be considered a form of optimal interpolation, it’s important to note that OI maps are constrained by a single-scale covariance function Arhan and De Verdiére, (1985); Le Traon et al., (1998a) and primarily focus on geostrophic circulation (processes with typical space and timescales greater than 100 km and 10 days). The multiscale approach of the MIOST method allows for the resolution of some of the surface variabilities missing in OI by incorporating covariances of various surface processes within a single inversion. In MIOST, the covariance functions are represented as wavelet modes, and inversion is performed using a variational approach Ubelmann et al., (2021). In the new reprocessing, our focus was solely on the geostrophic component, as explored in Ubelmann et al., (2021) and Ballarotta et al., (2023). These studies have compared the performance of the MIOST system with the current DUACS OI mapping system in DT mode, demonstrating that while MIOST is globally comparable to OI, it shows regional improvements, particularly in mid-latitudes where complex turbulent systems, such as the Gulf Stream and Kuroshio current, are better mapped by MIOST than OI.
In the context of Near Real-Time (NRT) applications, the MIOST mapping approach has been evaluated using Observing System Simulation Experiments (OSSEs). These OSSEs are designed to test new altimeter missions, different constellation scenarios, or novel types of observations and mapping systems. The OSSEs rely on models as their input; in our case, the input data come from a state-of-the-art ocean model based on primitive equations, which describes ocean dynamics. The complexity of the ocean model can vary according to the dynamics of interest, incorporating factors such as sub-grid scale parameterizations, spatial grid resolution (ranging from coarse to eddy-permitting, eddy-resolving, sub-mesoscale-resolving, and tidal-resolving regimes), and other aspects. Currently, we are using an eddy-resolving model based on the NEMO system (GLORYS12v1[1], eNATL60[2]). This model provides three-dimensional outputs, including sea surface height (SSH), temperature, salinity, and ocean currents, among others. In our study, we primarily utilize the SSH variable. Pseudo-observations, or synthetic observations, are generated by interpolating the SSH model outputs onto the theoretical tracks of the altimeter mission. To enhance realism, simulated instrumental and geophysical errors can be added to these pseudo-observations. Finally, the gridded data are compared to the reference “truth,” which in this context is the raw model output considered as the benchmark for validation.
The MIOST NRT mapping solution has been compared to the current operational NRT OI system. Pseudo-observations were derived from the sea surface height (SSH) field of a 1/60° North Atlantic simulation without explicit tidal forcing (eNATL60-BLB002)[3]. These pseudo-observations involve a constellation of five nadir altimeters (Jason-3, Sentinel-3A, Sentinel-3B, Sentinel-6, and HY-2B), with SSH data corrected for Dynamic Atmospheric Correction (DAC) and atmospheric tide signatures (S1 and S2). These five altimeters are then used in both the MIOST and OI mapping algorithms (in NRT mode) to reconstruct the SSH scene from the nature run. The performance of the mapping solutions is evaluated using the Root Mean Square Error (RMSE) metric. The differences in RMSE between MIOST and OI shows that the MIOST NRT mapping errors are, on average, 20% smaller than those of OI.
The MIOST NRT mapping solution has been compared to the MIOST DT mapping solution to evaluate the differences in mapping errors between DT and NRT modes using the MIOST system. This assessment was conducted using a free 1/12° global simulation from Mercator Ocean (ORCA12-TRBB36). The pseudo-observations include a constellation of five nadir altimeters (Jason-3, Sentinel-3A, Sentinel-3B, Sentinel-6, and HY-2B), with data partially corrected for Dynamic Atmospheric Correction (DAC) using only the Inverse Barometer effect. These five altimeters were then used within the MIOST system, in both DT and NRT contexts, to reconstruct the SSH scene from the nature run. As expected, the DT mapping error is more than twice as small as the NRT mapping error and exhibits greater temporal stability. Additionally, by varying the number of altimeters in the constellation, we found that the NRT reconstruction with 4-5 altimeters using MIOST is comparable to the DT reconstruction with just 2 altimeters.
The previous OI method provided an estimation of the consistency of the gridded solution using a formal mapping error indicator. However, due to the computational cost and the wavelet-based formulation of the MIOST mapping approach, the error estimation method had to be revised for MIOST to deliver a global estimation of the formal error. In the MIOST products, the formal mapping error is determined using an ensemble approach. Specifically, an ensemble of 20 gridded product members is generated, each with a perturbed covariance model and observations. The standard deviation of the sea level anomaly (SLA) estimations across these ensemble members represents the formal mapping error.
[1] Global Ocean Physics Reanalysis | Copernicus Marine Service
[2] eNATL60/02_experiment-setup.md at master · ocean-next/eNATL60 · GitHub
[3] eNATL60-BLBT02 | MEOM catalog
Change of the L4 grid spatial sampling
The new products are provided on a finer spatial grid compared to the previous reprocessing. The new global products have a spatial resolution of 1/8°, while the regional products offer even finer spatial sampling at 1/16°. The temporal sampling remains unchanged, with data available on a daily basis.
New geostrophic current computation methodology
see details in section November 2024 MY changes
CMEMS catalogue changes
The new NRT DT-2024 L4 “all-sat-merged” series are accessible through new datasets reflecting the change in grid spatial sampling. Part of previous datasets have been removed. Details are given in the following table.
Table: Updated DT2024 of GLObal and EURopean Sea L4 NRT Product classification: Product and Dataset Identification.
| Product | Datasets for DT-2021 series (will be frozen after DT-2024 release) | Datasets for DT-2024 series |
| SEALEVEL_EUR_PHY_L4_NRT_008_046 | cmems_obs-sl_eur_phy-ssh_nrt_allsat-l4-duacs-0.125deg_P1D | cmems_obs-sl_eur_phy-ssh_nrt_allsat-l4-duacs-0.0625deg_P1D |
| SEALEVEL_GLO_PHY_L4_NRT_008_060 | cmems_obs-sl_glo_phy-ssh_nrt_allsat-l4-duacs-0.25deg_P1D | cmems_obs-sl_glo_phy-ssh_nrt_allsat-l4-duacs-0.125deg_P1D |