Cloud-based data-proximate computation provides a framework to analyze tera- and peta-bytes of data as the resolution and complexity of ocean and climate simulations increases and as SWOT data will become available. The success of the framework, which allows for the community to fully harness and systematically analyze the outputs from submesoscale-permitting ocean models, which are associated with enormous carbon emissions, will depend on the ability of scientists to convince funding organizations to recognize its potential and switch fundings from local computational resources and storage to cloud-based computation and storage.
Ocean and climate scientists have used numerical simulations as a tool to examine the ocean and climate system ever since the 1970s. Since then, owing to the continuous increase in computational power and advances in numerical methods, they have been able to simulate increasing complex phenomena. The fidelity of the simulations in representing the phenomena remains a core issue in the ocean science community.
Ahead of the Surface Water and Ocean Topography (SWOT) satellite launch, expected in November 2022, many simulations were developed in order to allow for the instrumental calibration of SWOT and to disentangle the internal wave signals from (sub)mesoscale flows, as SWOT is expected to observe the superposed field of the two dynamics.
The paper Cloud-based framework for inter-comparing submesoscale permitting realistic ocean models proposes a cloud-based framework to inter-compare and assess such simulations. The authors carried out example analyses with the output of eight submesoscale-permitting ocean models for one of the SWOT Crossover (Xover) regions around the Gulf Stream separation. Models were compared based on surface diagnostics of the temporal mean and variability and three-dimensional diagnostics on physical processes.
Despite the similar horizontal resolution amongst many models in the study, the spatial scales represented vary widely as consequence of model parameters and numerical schemes, thus highlighting the need for collaborative work to inter-compare realistic simulations, both from a scientific point of view, so to assure the fidelity of submesoscale-permitting ocean models in representing the underlying physics and tracer transport, and from an engineering perspective on the numerics of ocean models.
The cloud-based framework proposed in the paper would also allow for a systematic inter-model comparison necessary for compiling best practices for future model runs, which in turn could reduce the enormous computational cost and carbon emission associated with the tuning of submesoscale-permitting models.
The authors conclude that the strength of cloud storage and computing comes from it being decentralized from any specific institution, but this also leaves open the question about who pays for the operational cost. The success of the framework will depend on the scientific community to convince its peers and funding organizations to recognize its benefit and switch fundings from local computational resources and storage to cloud-based computation and storage.
FOR MORE INFORMATION
Uchida et al. (2022). Cloud-based framework for inter-comparing submesoscale permitting realistic ocean models. Geoscientific Model Development. https://doi.org/10.5194/gmd-2022-27