Advancing machine learning education with open software and data: A case study in global climate projections of snow

We present a machine learning tutorial that we developed to train Earth scientists to train and deploy a model to make climate projections of future snowpack. We designed our tutorial to use best practices and builds on a wide range of open source software, including many tools from the Pangeo community.  We will walk through the main components of our tutorial, namely how to: 1) Prepare the data processing pipeline, 2) Implement a pytorch model and training workflow, 3) How to evaluate the trained model, and 4) Run the trained model on future scenarios under climate change. Following this short demo/walkthrough we will discuss ongoing challenges, lessons learned, and key takeaways. This project and tutorial is part the GeoSMART project, a broader effort to improve machine learning education in the Earth sciences (https://geo-smart.github.io/ 8)