3Drec provides the implementation of two deep learning networks designed to reconstruct the 3D ocean structure from 2D satellite data: one based on a deep feed-forward network (FFNN3D) and one based on a stacked Long-Short Term Memory network (LSTM3D).

The two networks and the training/test data used for their development are fully described in the paper:

• Buongiorno Nardelli, B., A Deep Learning Network to Retrieve Ocean Hydrographic Profiles from Combined Satellite and In Situ Measurements, Remote Sensing, 2020.

The code is written in Python 3

These are the python packages required (tested versions):

• keras 2.2.4
• numpy 1.18.1
• netcdf4 1.5.3
• pandas 1.0.3
• seawater 3.3.4

The data used to develop the models can be found at

Bruno Buongiorno Nardelli, Consiglio Nazionale delle Ricerche - Istituto di Scienze Marine

https://orcid.org/0000-0002-3416-7189

The development of this code was partly funded by the European Space Agency through the World Ocean Current project (ESA Contract No. 4000130730/20/I-NB).