Optimising MITgcm on ARCHER2: efficient numerical simulation and data assimilation tools for studying the ocean, atmosphere, and cryosphere

ARCHER2-eCSE02-06


image

Motivation

MITgcm is a state-of-the-art general circulation model used to enhance understanding of the atmosphere and ocean and their interactions with land- and sea-ice. We sought with this project to minimise the hours of compute time taken when running MITgcm on ARCHER2, which in turn reduces the energy expended by this research and increases its range of usefulness as more compute-intensive problems can be studied.

image

The above figure is from Morlighem, Goldberg et al (2021) and shows how and where the sea level contribution of the glaciers of the Amundsen Sea Embayment are sensitive to surface mass balance and ice-shelf melting. The simulations with STREAMICE are only possible because of eCSE-cryo.

Summary

We ported three models that are built on the MITgcm core: two address the Amundsen Sea in the Antarctic (one of which incorporates the flow of ice from Antarctica) and one global ocean state estimate.

The three exemplar models were optimised and achieved significant speed up of the order of 2-fold in a widely used setup. This is already helping researchers address more complex problems with their models, as well as allowing for more efficient use of valuable computer resources.

There are approximately 100 MITgcm users in the UK research community, who will also benefit from specific files developed for compiling MITgcm efficiently on ARCHER2.

In order to disseminate the lessons learned from this project to the wider community, an ARCHER2 webinar was presented. This webinar, the MITgcm user’s guide to ARCHER2, aimed to significantly increase the impact of the project. A recording of the webinar, along with the slides, can be found on the event page.

The work done in this eCSE also contributed to the following publication: Morlighem, M., Goldberg, D., Dias dos Santos, T., Lee, J., & Sagebaum, M. (2021). Mapping the sensitivity of the Amundsen Sea Embayment to changes in external forcings using automatic differentiation. Geophysical Research Letters, 48, e2021GL095440.

Information about the code

The entry point for documentation is the Archer2 MITgcm web page.

All code related to the project is available from repositories:

Webinar

Webinar presenting this work

Technical Report

Download as PDF