Technical documentation

OceanBench is a benchmarking tool to evaluate ocean forecasting systems against reference ocean analysis datasets (such as GLORYS reanalysis and GLO12 analysis) as well as observations.

This documentation describes how to prepare and evaluate a challenger with the Python package oceanbench. The definitions of the evaluation methods have their own page.

Guides

• Shape of the challenger dataset
• Input datasets for OceanBench challenger evaluation
• Units and time conventions of the input data
• 1 degree track
• Run OceanBench offline

API reference

The API reference documents the public Python API of the oceanbench package, generated from its docstrings:

• Evaluate a challenger
• Metrics
• Regions
• Datasets: challenger, input, reference and training