by Robert Jnglin Wills
Modeling centers around the world are now releasing data from simulations with the next generation of climate models, the Coupled Model Intercomparison Project Phase 6 (CMIP6). For three days in October, thirty UW climate science graduate students and postdocs got together to see what they could learn about future climate change from these new simulations. We combined efforts with CMIP6 hackathons at two other institutes, the National Center for Atmospheric Research (NCAR) in Boulder, Colorado and the Lamont-Doherty Earth Observatory in Palisades, New York. It was a test in working together remotely, which is increasingly a focus for climate scientists trying to limit their CO2 emissions from travel.
Besides studying the new climate projections, an overarching goal was for all participants to use a common set of Python tools for accessing and analyzing the CMIP6 data. These tools allow users to quickly access and analyze CMIP6 data on cloud storage, without downloading all of the data themselves. This facilitates reuse of analysis code across projects and makes the resulting science more reproducible. In particular, we worked with xarray, a Python package for working with labeled multi-dimensional datasets, dask, a Python package for parallelizing analysis, and intake-esm, a Python package developed by Pangeo to streamline access to CMIP6 data that is stored on NCAR’s supercomputer and on the Pangeo Cloud Data Catalog. Pangeo is a community promoting open, reproducible, and scalable science that also develops some Python software.
Highlights from the UW group include a public-facing web tool for visualizing local CMIP6 climate projections, an investigation of how much of Southern Ocean surface waters are sinking into the deep ocean in different CMIP6 models and how this affects climate, an analysis of how well CMIP6 models can simulate observed Pacific Ocean temperature changes over the last 50 years, an investigation of projected changes in the position of the Gulf Stream, and an investigation of projected changes in Antarctic sea ice. While these projects, as expected, were not completed after just 3 days, there is now momentum amongst the participants to continue working on them, to continue learning Python, and to work towards making our science more open and reproducible.
Public code repositories for the Hackathon projects can be found on Github:
Anyone interested in trying out these CMIP6 analysis tools for themselves can access the Pangeo hackathon template on which these projects are based or contact Robert Jnglin Wills for more information.
The UW CMIP6 Hackathon was made possible by funding from the University of Washington Program on Climate Change and the Department of Atmospheric Sciences, as well as by support from the NCAR and Lamont CMIP6 Hackathons, which were funded by US CLIVAR, Ocean Carbon & Biogeochemistry, the National Science Foundation, NASA, and the National Ocean and Atmospheric Administration Modeling, Analysis, Predictions and Projections Program.
Robert Jnglin Wills is a UW Data Science Postdoctoral Fellow in the Department of Atmospheric Sciences and the eScience Institute. He uses climate models and statistical analysis to understand the physical processes governing the spatial pattern of global warming and its changes over time. He organized the UW CMIP6 Hackathon to work together with other CMIP6 hackathons and foster a Python programming community at UW.