We’re excited to share highlights from our recent participation in the NCSA Open Hackathon, where we joined an international team led by Hassan Moustahfid (NOAA/US IOOS) and collaborators from NOAA, Mercator Ocean International, and GCOOS. With guidance from NVIDIA mentors, we experimented with machine learning algorithms—one to emulate numerical models for high-resolution water level prediction, and another looking at the problem of data fusion and bringing many sources of data together to produce a single representation of reality.
Our main contribution was leading the data wrangling activities. This included migrating over a terabyte of data to the NCSA Delta AI supercomputer and designing data loaders to support the ML workflows. The data fusion project brought together a diverse set of datasets sourced from various distribution channels, from the CORA unstructured reanalysis and GLORYS gridded products to NOAA Laboratory for Satellite Altimetry’s global blended altimetry, as well as time series and trajectory data from the IOOS ERRDAP server and the NOAA tides and currents API. The hackathon was a great opportunity to develop new connectors, which will soon make these valuable datasets available in the Oceanum Datamesh.
