Providing a set of Python modules for supporting processing and analysis of eddy-related data.
The module addresses oceanic mesoscale eddy analysis by providing the tools for pre-processing FESOM2 data and training DL models. It builds on top of a collaboration with AWI, which provided the necessary data and core software for the use case.
Release Notes
The development of the Python packages is currently underway. In this version, a pre-processing pipeline has been developed, divided in a series of scripts launching different blocks of code. They allow interpolations from unstructured meshes to regular grids and vice versa to prepare the dataset for the training, and all parts of the code can be run with the Slurm Workload Manager on the chosen cluster. The DL framework in use is Tensorflow, and the GPU support is included, if the underlying system offers the proper hardware. A custom loss function and a Convolutional Neural Network based on the U-Net architecture were customised for this task.
Future Plans
The next versions will improve the prediction capabilities of the trained network. It will be possible to have a neural model that can predict eddy segmentation masks much faster than the classic physics-based algorithms. Additional usage examples will be included in the documentation.