Advancing Earth System Model Evaluation with AQUA: A Modular Framework for Quality Assessment

The diversity of the Earth System Model (ESM) ecosystem, with output data varying in format, grids and metadata, pose significant challenges for effective intercomparison and evaluation. Additionally, the comparison with observational dataset and the rapid development of Artificial Intelligence techniques to progress ESM representations may introduce additional complexity, with the need of a fast integration of them in a uniform and standardized evaluation pipeline.

AQUA, an Application for QUality Assessment, is a python package developed in the context of Destination Earth Climate Digital Twin, a major initiative by the European Commission. AQUA is designed to overcome these challenges by standardizing variables and metadata across different models and observations. Additionally, it provides a simple and unified interface to open and process different data formats and support different APIs (netCDF, GRIB, FDB, Zarr, Parquet) to access data. It leverages the power of Dask, Xarray and Intake libraries, for a lazy data access to all the supported formats.

AQUA offers the possibility to add any dataset in a catalog, a series of yaml files describing the path or the API needed to open data. The code can accommodate a hierarchy of needs, from simple access to complex data structure, going through extensive usage of the processing capabilities such as metadata conversion, data regrid and integrated area weighting as an extension of Xarray, or eventually by exploiting the integrated suite of diagnostics for model evaluation.

Additionally, the code provides diagnostics that can exploit lazy access to high-resolution data, a suite for uncertainty quantification of the diagnostics output and provides the backend capabilities for the development of project dashboards.

Finally, the code supports the possibility to develop other diagnostics and to have them integrated in the model evaluation along with the others.