The Critical Need to Foster Reproducibility in Computational Geoscience

Chains of computer models translate emissions into climate signals and subsequent into impacts regarding floods, droughts, heatwaves, and other perils. While the need for computational geoscience is significant, recent publications across the geo- and environmental sciences suggest that reproducibility of computational geoscience might be limited. So far, the focus of reproducibility largely remained on discussions of this problem in the social sciencesor medicine; in this talk, we take a peek behind the curtain of everyday geoscientific research and unveil how we need to foster reproducibility in computational geoscience and what is required to do that. A poll among more than 300 geoscientists reveals that geoscientific research is currently not reproducible enough. 61% say that a lack of reproducible research is putting trust in our results at stake, and only 3% strongly agree that computational geoscientific research is reproducible. Leading causes, contrasting previous polls, are not only a lack of resources and willingness to share code and data but also a lack of knowledge about state-of-the-art software development methods and licenses among the geoscientific community. To lay a path for a future where Open Science is the norm, we let the voices of the community speak on what they think is necessary and paint a picture of a future that fosters reproducible geoscience and thus trust.