CLIMLAB 2.0: Lessons Learned and Future Roadmap for Interactive, Process-Oriented Climate Modeling

CLIMLAB is a flexible Python-based toolkit for interactive climate modeling for use in education and research. It is motivated by the need for simpler tools and more reproducible workflows with which to "fill in the gaps" between blackboard-level theory and the results of comprehensive climate models. The project has grown tremendously in features and complexity since its first public release in 2015. This growth has revealed plenty of shortcomings in the core software design which have hindered more rapid development and community adoption, despite plenty of interest from potential user/developers. We are currently rewriting the internals and much of the public API with several goals in mind: more intuitive user interface and visualization, less reliance on class inheritance, deeper ties to the emerging xarray-based ecosystem, leveraging dask for easy parallelism, and a cleaner, leaner codebase.

I will discuss lessons learned in battle-testing the software in the classroom and research lab, and how we are using those lesson to produce a "new and improved" CLIMLAB that gets closer to the original vision for the project - a tool that empowers and enables better science with climate models.