auto-icon: seamless management of ICON model runs with Autosubmit

Weather and climate modeling usually entails a large set of tasks, such as the retrieval of input data from repositories, the actual simulation or standard post-processing and visualization of the output data. Most of these tasks recur for each experiment and require only minor user input and no user intervention, allowing for a high degree of automation. In the community of the ICOsahedral Nonhydrostatic (ICON) weather and climate model, there is still no commonly established tool for workflow automation. Thus, many researchers perform pre- and post-processing steps by hand and run the model itself with a so-called runscript, a self-contained bash script that controls some model preparation and executes the run but not covering tasks surrounding it. Thus, there is apparent demand for a solution to increase automation of the entire workflow as well as to improve on the sustainability and reproducibility of the modeling workflow. The latter point is especially important as nowadays many journals require to publish all the accompanying data along with information how to reproduce the results.The workflow manager Autosubmit is specifically designed to automate weather and climate modeling workflows. Furthermore, it facilitates the reproducibility of simulation results by creating archives of the entire workflow configuration after successful runs, thus supporting the FAIR principles. As Autosubmit is model agnostic, there is the need for an application interface for use with a certain model. auto-icon provides this interface for the ICON model, thus allowing operation with Autosubmit as backend. Hereby, auto-icon consists of an extensive set of YAML configuration files to adapt the tasks to the individual demands and a set of Bash and Python scripts to execute those.In this presentation, we will showcase the auto-icon software, highlighting key points relevant for the development and usage, such as sustainability for model runs and reduction of personal hands-on time needed. We will also highlight how auto-icon can improve the performance for everyday workflows and illustrate our efforts to identifying, engaging and onboarding new users. While auto-icon itself is only relevant for the ICON community, many of the core features are provided by the model-independent backend Autosubmit, and thus can be adapted with similar tools to other models in the field. Furthermore, we show how workflow management in general can increase performance and sustainability through automation, which can act as a role model for other fields of research.