Providentia: an evaluation software package for the atmospheric modelling community

Providentia is an evaluation software package designed for the in-depth analysis of in-situ surface observations and colocated model output, tailored specifically for the atmospheric modelling community.

Reproducibility is a key concern when performing any type of model evaluation. A variety of factors can affect reproducibility, including how observations are processed and filtered, and how statistics are calculated. Even two scientists within the same institution may obtain markedly different results depending on their methodologies. Providentia addresses this challenge by leveraging harmonised observational datasets, such as GHOST and ACTRIS, which are widely used by the community, and by allowing precise customisation of fully documented statistics. Critically, by using the same configuration file, two users can be confident that their evaluations are exactly the same.

Providentia offers a variety of use modes, these include an interactive dashboard for quick-look visualisations; a report mode designed for more exhaustive evaluations, generating PDF reports; a library mode that enables Providentia's backend functions to be used in scripts or Jupyter notebooks, for example for reading, filtering, or plotting data; a download mode that automatically retrieves and formats observational (e.g. GHOST and ACTRIS) and model datasets (e.g. CAMS model forecasts and reanalyses); and an interpolation mode that spatially colocates model output with observational stations. 

Providentia is publicly available on GitHub (https://github.com/BSC-ES/providentia), and is fully documented on a dedicated ReadTheDocs page (https://providentia.readthedocs.io/).