Connected Data Products Hackdays Overview

Trey Stafford

Robyn Marowitz

Luis López

2026-05-13

Connected Data Products Hackdays

Berkeley, CA April 13-15
Co-hosted by Englacial and Berkeley Institute for Data Science.
Focus on bringing together members of the cryospheric data science community to explore connected and living data products.
April 13: Presentations/demos and ideas formation
April 14-15: Hack time!

Image courtesy of Englacial

Open Polar Radar Tools (xopr): Python package for accessing and analyzing CReSIS radar depth sounder data.
icepyx: Python library for searching and accessing IceSat2 data.
GStatSim: Python package for geostatistical interpolation methods.
NISAR data at ASF DAAC.
Arcadia bioscience overview and demo of The Stacks scientific publishing platform.
MyST for scientific publishing with markdown.
Source Cooperative: nonprofit publishing of scientific data to cloud infrastructure.
Cloud-Native Geospatial Forum overview and data formats guide
Cryocloud: Collaborative Earth Science Platform in the cloud.
VirtualiZarr and IceChunk.
ISMIP 6 Model Viewer

earthaccess v1.0 updates
Virtualization of datasets for efficient cloud access (e.g,. radar sounder data).
Analysis of velocity changes at grounding zones.
Analysis of radar echo sounding crossovers with IceSat2 crossovers.
Geostatistics/Interpolation of data with GStatSim.

Englacial posted a nice writeup of the hackdays projects here:

Trey, Joe Kennedy (ASF), Julia Lober (graduate student at Boise State), Matt Fisher (DSE), and Jessica Scheick (U. of New Hampshire)
The project is working toward a v1 release, and there are various milestones/breaking changes that we want to complete before v1 is done.
We ended up focusing on updates to the results API, and providing convenience methods/properties to the Results class (e.g., we added the ability to export to GeoDataframe and worked toward attaching query provenance to the results).
Associated PR: https://github.com/earthaccess-dev/earthaccess/pull/1298
Lots of people expressed interest in this work, and were excited for the changes. Almost all of the projects (maybe all) used earthaccess.

Team: Robyn Marowitz, Eliza Dawson (NOAA C&GC Postdoc Fellow, Georgia Tech), Joseph Rotondo (UW grad student), Lindsay Stark (Boise State grad student), Emmy Muniz (Stanford undergraduate student)
Expanded the GStatSim Jupyter Book documentation with two new real-world application notebooks:
- HiCARS radar sounder — end-to-end workflow: downloading 860 granules from NSIDC via earthaccess, parsing flight-line text files into a single GeoParquet (9.3M points, EPSG:3031), and fitting variogram models for kriging and geostatistical simulation
- Arctic pole hole — filling the gap in satellite sea-ice coverage near the pole using geostatistical interpolation of sea-ice albedo
Together these notebooks demonstrate real-world approaches to connecting datasets through an open-source package to produce a new data product
Repo: https://github.com/elizadawson/gstatsim_largedatasets

Data archived in the cloud is not always “cloud-optimized” or “AI-ready”. Significant interest in data virtualization techniques (e.g, VirtualiZarr, IceChunk)
Importance of living/executable documentation (e.g,. via MyST).
Discussions about how to maintain citations for products that receive updates from a rotating cast of contributors.
The role of AI in scientific software/product development (e.g., Jupyter AI).
Reproducibility in science is more than just being open. Can AI be used as a “reproducibility checker”?
Challenges with scaling notebooks to production
Challenges with co-locating and co-analyzing data that were not “designed” to work together natively.