Xarray adds dimension names, coordinate labels, and arbitrary metadata on top of Dask arrays allowing scalable, convenient, and expressive analysis code (think pandas for n-dimensional arrays). The combination of Xarray and Dask for labelled distributed data analytics has seen wide adoption across scientific fields, particularly the geosciences.

This will be a two-part session. First, an hour of updates from multiple developers on exciting progress at extending Xarray's flexibility over the past year funded by the Chan Zuckerberg Initiative. Highlights include support for array objects that follow the __array_function__ protocol; a new plugin interface for ingesting data from storage backends; and ongoing work to supercharge label-based indexing through custom indexes (e.g. those that represent grid cell intervals). Each talk will be followed by a 5-minutediscussion and Q&A on how these new features work to unlock new domain-specific functionality and use-cases. Our goal is to spark a conversation in the user community on how best to adapt and use these new features for domain-specific solutions.

The second half will focus on Xarray+Dask user stories from non-geoscience domains, and associated challenges. We will reach out to speakers from domains such as condensed matter physics, radio astronomy, finance, and the life sciences to put together a compelling series of three 15 minute talks that highlight use of Xarray and Dask. Each talk will be followed by 5 minutes of moderated Q&A with the speaker.