Data used in the publication "Decadal wave reconstruction in the Mediterranean Sea with graph neural networks" by Benassi et al.

Federica Benassi, A S Wadalkar, Lorenzo Mentaschi

This repository contains the data used for training and validation of the model presented in Decadal wave reconstruction in the Mediterranean Sea with graph neural networks by Benassi et al. The wave data will be published as Wadalkar et al. (2026), a bias-corrected version of the dataset by Mentaschi et al. (2023), and contain: mwd_x and mwd_y: sine and cosine of mean wave direction mwp: mean wave period swh: significant wave height We also provide 10-metre wind speed components from ERA 5 (u10 and v10), interpolated on the unstructured mesh of the wave data. These data are stored as .zarr files. To extract them type in a terminal: tar -xf swh.tar Please beware that each file weighs ~7 GB, for a total of ~42 GB. We also provide the mesh over which the graphs are built. original_mesh_gv.gr3: the original unstructured mesh over which wave fields are defined mediterranean_60km_gu: the uniform-resolution unstructured mesh used as processing layer in the graph neural network For further information, check code at https://github.com/CMCC-Foundation/WaveGraph