Mediterranean Sea Super Resolved Geostrophic Currents

<p>Super Resolved Geostrophic Surface Currents for the Mediterranean Sea (2008-01-02 to 2019-12-31) computed by means of Convolutional Neural Networks (CNNs) and Generated using E.U. Copernicus Marine Service Information. The generation algorithm is described in Ciani et al. 2024 (https://egusphere.copernicus.org/preprints/2024/egusphere-2024-1164/) and Buongiorno Nardelli et al. 2022.&nbsp;</p> <p>The main directory contains 5 subfolders:</p> <ol> <li>ADT_MFSeas4_Copernicus. Daily data for the year 2017: please refer to Clementi et al. 2019;</li> <li>ADT_MFSeas4_OI_synth_MAP_MED_DT2018_4SAT (Satellite Equivalent Absolute Dynamic Topography, SE-ADT). Daily data for the year 2017: the generation is detailed in Ciani et al. 2021 (section 2.1, item #3);</li> <li>SST_MFSeas4_Copernicus. Daily data for the year 2017: please refer to Clementi et al. 2019;</li> <li>SST_MFSeas4_OI_synth_MAP_MED_HR (Satellite Equivalent Sea Surface Temperature, SE-SST), Daily data for the year 2017: the SE-SSTs merge infromation from SST_MFSeas4_Copernicus &nbsp;with the multi-sensor L3S satellite SSTs for the Mediterranean Area (product ID: SST_MED_SST_L3S_NRT_OBSERVATIONS_010_012). Such L3S SSTs have gaps where infrared SST retrieval is impossible (e.g., due to cloud cover) or where single-sensor satellite SSTs are deemed of poor quality.&nbsp; Choosing the data contained in subfolder #3, a synthetic model-derived L3S SST time series was generated introducing synthetic gaps in the original modelled SSTs. Subsequently, gap-free SSTs, along with an estimate of uncertainty, are generated using standard Optimal Interpolation (OI) via a dedicated algorithm, following methods outlined in Buongiorno Nardelli et al. (2013);</li> <li>dADR-SR_ADT_SST_dtSST_MED: super-resolved geostrophic currents. daily data (2008-01-02 to 2019-12-31). This dataset is generated using E.U. Copernicus Marine Service data of gridded gap-free (Level 4) ADTs and SSTs over the Mediterranean Area and employing a CNN trained by means of the dataset contained in the subfolders 1-4. The CNN architecture/algorithm is described in Ciani et al. (submitted) and Buongiorno Nardelli et al. 2022.</li> </ol> <p>All data are provided in NetCDF format</p>