4DMedSea-ESA Data Challenge


_images/dc_2024_4dmedsea_banner1.jpg


So far, the github page visits amount to:

The 4DMedSea Data Challenge

The 4DMedSea data challenge is a comparison platform that allows a sound and fair assessment of several mapping methods in the West Mediterranean region. This platform contains an input dataset (altimetric data), for anyone to create sea level and surface current maps, as well as a reference dataset (observations independant from the input dataset) in order to evaluate the methods. The data challenge is hosted in a github repository and its documentation and results are presented on this readthedocs website.




Objectives

The goal of the data challenge is to evaluate the methods ability to map the sea level anomaly in the Western Mediterranean during the year 2017 (evaluation period). The input observations used to generate the maps are conventional altimeters (Saral/Altika, Jason-2, Jason-3, Sentinel-3A, Sentinel-3B, Sentinel-6, Haiyang-2B and Cryosat-2). The evaluation observations are from an independent conventional altimeter (Haiyang-2a) and drifters. The methods evaluated, so far, are MIOST and 4DVarNet. The DUACS-CMEMS product is also provided as a reference, however the DUACS product uses the independent altimeter data so the results provided should not be looked at as an assessment of the DUACS product performance.

A dataset of input observations, evaluation observations and evaluated maps is provided (see Download the data).

_images/DC_4DMedSea_IllustrationMap.png




Scoreboard


You can check the overall comparison for more details or click on the methods name to see their specific performances.


Statistical and spectral scores

[Alongtrack metrics details]


4DMedSea methods

RMSE(SLA) RMSE score(SLA) Eff. res.(SLA)
MIOST 4.6 cm 55 % 121.3 km
4DVarNet 4.5 cm 56 % 144.1 km

DUACS-CMEMS product (using all sat)

RMSE(SLA) RMSE score(SLA) Eff. res.(SLA)
DUACS 4.1 cm 63 % 103.0 km

Other methods

RMSE(SLA) RMSE score(SLA) Eff. res.(SLA)
--- cm % km


Statistical scores

[Along drifter metrics details]


4DMedSea methods

RMSE(u) RMSE score(u) RMSE(v) RMSE score(v)
MIOST 12.8 cm/s 31 % 12.7 cm/s 29 %
4DVarNet 13.2 cm/s 30 % 13.5 cm/s 25 %

DUACS-CMEMS product (using all sat)

RMSE(u) RMSE score(u) RMSE(v) RMSE score(v)
DUACS 12.1 cm/s 35 % 12.5 cm/s 31 %

Other methods

RMSE(u) RMSE score(u) RMSE(v) RMSE score(v)
--- cm/s % cm/s %


Averaged trajectory deviation

[Drifter trajectory metrics details]


4DMedSea methods

horizon 1d horizon 2d horizon 3d horizon 4d horizon 5d
MIOST 14.9 km 26.9 km 39.3 km 50.2 km 59.4 km
4DVarNet 15.1 km 27.3 km 39.2 km 49.8 km 59.4 km

DUACS-CMEMS product (using all sat)

horizon 1d horizon 2d horizon 3d horizon 4d horizon 5d
DUACS 13.9 km 25.6 km 36.0 km 46.0 km 54.0 km

Other methods

horizon 1d horizon 2d horizon 3d horizon 4d horizon 5d
--- km km km km km





Reconstruction

The reconstruction covers the period from January 1st, 2017 to December 31st, 2017 in the Mediterranean basin from 6°W to 36°E and 30°N to 46°N.




Observations

The SSH observations used in this study comprise data from a nadir altimeter constellation that includes:

  • Saral/Altika (alg),

  • Jason-2 (j2, j2g, j2n),

  • Jason-3 (j3, j3n),

  • Sentinel-3A (s3a),

  • Sentinel-3B (s3b),

  • Sentinel-6 (s6),

  • Haiyang-2B (h2b),

  • Cryosat-2 (c2, c2n).

The Haiyang-2A (H2A) altimeter data are excluded from the mapping process to enable an independent assessment of the different reconstructions.

These data are distributed by the Copernicus Marine Service https://doi.org/10.48670/moi-0014 .




Evaluation

Data

  • Independant nadir: Haiyang-2A (H2A)

  • Independant drifters

Drifter data are also distributed by CMEMS https://doi.org/10.17882/86236 .

Metrics

Check Metrics Details: