Data and Model Evaluation¶
This section links climate datasets, shared frameworks, diagnostics, and visualisation tools to create workflows for model evaluation. Standardisation through the workflows will support reproducible, transparent, scalable, and community-driven evaluation. There are catalogues and tools supported on Gadi for the ACCESS community with training resources available.
If you are new to using the Gadi supercomputer, visit the Set Up your NCI Account section for instructions on how to create an account, log in, and get set up to access climate data on Gadi. ACCESS-NRI supports and maintains the conda/analysis3 Python environment which supports most of the evaluation and data tools described.
Connecting Data and Evaluation¶
Climate data is at the center of any workflow. Sources include published CMIP, observations, reanalysis products and climate simulations. They need to be found, preprocessed and standardised for evaluation.
Evaluation frameworks provide the structure for running diagnostics, benchmarking simulations, and comparing model performance across datasets. They are developed specifically for the climate community to leverage common data-processing functions. The determined structure allows for easier sharing and scalability of common diagnostics.
Recipes in this context are a set of instructions that define reusable workflows into a single reproducible analysis pipeline. Recipes are often shared in the community for visibility and reusability, helping accelerate collaborative climate science. An individual recipe may take advantage of existing evaluation frameworks (e.g ESMValTool, ILAMB) or implement its custom logic by using Python libraries such as xarray and numpy.
Training Materials¶
See the Tutorials page for material taken from previous workshops and events grouped by model/tool.
Each subcategory page has links to further documentation and tutorials.
Support and new requests¶
For infrastructure support and to request new diagnostics and features, reach out on the ACCESS-Hive Forum, create a topic and tag with #help.
See User support for more details.
