Best Practices for Publishing Reproducible Simulation Datasets in DesignSafe

A working group of simulation researchers and data curators convened to create guidelines for publishing simulation datasets in DesignSafe. The guidelines reflect the modeling community needs and are informed by the experience of the working group members and the larger DesignSafe expert cohort. These best practices focus on attaining published datasets with precise descriptions of the simulations’ design, complete publication of inputs and if possible, of all outputs, and references or access to the research software/code involved in the simulation. Tying these pieces together requires documentation to understand the research motivation, origin, processing, and functions of the simulation dataset in line with FAIR principles. In crafting these best practices, we specifically considered the characteristics of simulations conducted within the natural hazards space, as well as the data model that drives the interactive simulation curation and publication pipeline and the simulation and analysis tools available in Tools and Applications in DesignSafe. And yet, these best practices can also be used by simulation researchers in any domain to curate and publish simulation data in any repository. In this first version the report does not address Machine Learning /surrogate models. This is a working document that will evolve as we observe how best practices are implemented, and gain experience from new publications and community feedback. We expect researchers publishing simulations in DesignSafe to follow the guidelines, which are reinforced in the curation and publication pipeline.