Mattia Almansi*, Aleksi Nummelin, Atousa Saberi, Thomas W. N. Haine, Renske Gelderloos, Department of Earth and Planetary Sciences, Johns Hopkins University
The dynamical and physical processes governing the ocean circulation can be investigated using different approaches such as numerical simulations, and observations (e.g., in-situ or remote sensing measurements). We have set up and run several high-resolution numerical simulations to study the three-dimensional, time-evolving ocean circulation of the Subpolar North Atlantic Ocean. With the goal of building a collaborative sharing environment, our model outputs are publicly available on the Johns Hopkins SciServer System. Because of the large size of these datasets (from TB to PB), severe barriers exist to using the data in practice. To address this problem we are developing OceanSpy, an open-source and user-friendly Python package suited to enable scientists and interested amateurs to use oceanographic datasets as virtual sandboxes. For example, OceanSpy can help observational oceanographers to optimize the design of field campaigns, or to study the dynamics underlying features found in observations. OceanSpy has the potential to enable interactive analyses on petascale datasets because it builds on software packages that support lazy evaluation and parallelism (Pangeo NSF EarthCube project). Our package already has Eulerian capabilities, and we are currently implementing a Lagrangian particle code (see 3D movie, Gelderloos et al. 2016).