UXarray: Extending Xarray for Enhanced Support of Unstructured Grids

Over the past decade, weather and climate models have rapidly adopted unstructured meshes to better leverage high-performance computing systems and approach kilometer-scale resolutions. Output from this new generation of models presents many challenges for their subsequent analysis, largely due to a lack of community tools supporting unstructured grid data. Last year, we introduced UXarray, a class extension of Xarray that provides native support for unstructured meshes. UXarray readily runs in a Jupyter Notebook and offers parallelized execution through its compatibility with Dask, demonstrating its flexibility as both a tool for lightweight exploration and communication, and for supporting intensive calculations applied to vast data volumes. Over the past year, UXarray has matured significantly and is now capable of supporting many real-world analysis workflows applied to outputs from a growing number of high-resolution models and dynamical cores, including ICOsahedral Non-hydrostatic (ICON) atmosphere model, the Finite-Element/volumE Sea ice-Ocean Model (FESOM), NSF NCAR’s Model for Prediction Across Scales (MPAS), and the U.S. DOE’s Energy Exascale Earth System Model (E3SM). This presentation will provide an overview of the UXarray’s current capabilities, which include extensive support for plotting and many foundational analysis operators; demonstrate examples in Jupyter Notebooks; present plans for the future; and discuss ways for Pangeo and the broader earth system science community to help guide new developments.