breakout room 2

[alee]

Jonathan Donges, Gary Polhill, Volker Grimm, George Mobus, Elizabeth Blood, Manuela

Easy consensus: make all parts of the model open source where possible

We should include concrete re-use cases in this standards page to help us define what we mean by reuse.

• executing a model with different inputs, e.g., applying a model to different dataset, different map, or a new analysis / recalibration

Standards for the person reusing the model. Guidelines for someone reusing the model on how to cite the model.

Think about how to address building a model to serve an internal purpose (PhD, fixed term funded project) or intended to support the community later. Likely the model would need to be rearchitected for reuse by a larger community, but this would be facilitated by the OSS model as well.

Encourage modelers to code for open at the start. What are the things you need to do to be an open modeler? Embedding these principles in the way early career modelers develop their models. Facilitate a culture change where modelers write their models for others, not just themselves.

Jonathan: Spotlight exemplars of best practices from existing models PISM and https://www.gfdl.noaa.gov/climate-modeling/

Liz: Building training modules for the high school / undergraduate level that embed these principles would be an excellent output

Volker: reuse may have a relatively small application, taking a model that worked on the Chesapeake Bay is likely to not work on a different area due to all the assumptions embedded within the model. It can take months for someone to decipher the existing model enough to apply it meaningfully.

[platipodium]

Volker: reuse may have a relatively small application, taking a model that worked on the Chesapeake Bay is likely to not work on a different area due to all the assumptions embedded within the model. It can take months for someone to decipher the existing model enough to apply it meaningfully.

Exactly! So rthe model is not reusable because it is tied closely to the "Chesapeake Bay" setup. It helps to have modularization here: the functional model code and the "setup" including parameters, inputs, outputs ...