FAIR Digital Objects in Autosubmit Workflows

Abstract

Climate and weather models are, respectively, complex computer programs used to better understand and predict the climate and to forecast weather. Running these programs involves a series of tasks that may require the use of other computer programs and scripts to prepare the input data, replace values in configuration files (e.g. Fortran namelists), prepare parameters for a batch server used in a high-performance computing facility (e.g. Slurm user configuration, project details, job request, etc.), among other possible variables. There are multiple challenges to make these programs and their outputs FAIR (Findable, Accessible, Interoperable, and Reusable) digital objects. The use of computational workflows executed with workflow managers eases the configuration and execution of climate and weather models, encapsulating some of its complexity, and contributes to a more uniform collection of provenance information about the execution of these models. In this article we focus on Autosubmit, an experiment and workflow manager used for running weather, air quality, and climate experiments that implements the Workflow Run RO-Crate profile to archive provenance information. We also highlight points that are not covered by RO-Crate and explain how we plan to address these shortcomings through recommendations found in FDO specifications.