Variable zoom on the climate

The facts and the project

Climatology is probably the most demanding scientific discipline in terms of computing time. With so many parameters, on a range of scales, that need to be included, modelling climate change is not a simple matter but it is one of crucial importance for the future of the planet.

All climatology models, even the most advanced, have shortcomings, well-known to the scientists, as a result of the huge complexity of the phenomena to be represented… The more detailed the resolution, for example, the better the representation of both their heterogeneity and their interactions on the very smallest scale.

Can simply increasing the resolution of existing climate models help to correct the observed shortcoming? Finding an answer to this question was the aim of the work carried out by a team from the Institut Pierre Simon Laplace (IPSL) using the Curie supercomputer. The aim: To gradually increase the resolution of a reference climate model from 100 km (the current standard) to 10 km, and then to check to what extent the shortcomings have been overcome.

The results

This was a real technical challenge and the resolution was increased across the model as a whole and for the local areas (zones restreintes), major sources of errors. This approach made it possible to quantify exactly how any local increase in the resolution in the critical regions resolved the bias in the model. This is a world first, both in terms of the resolutions achieved and in the ability to “zoom in” on certain regions of the world.

The scientific data produced by this initial simulation now needs to be put to use but it has already confirmed the place of French climatology among the “leaders” in terms of research.


PULSATION: Petascale mULti-gridS ocean-ATmosphere coupled simulatIONs 22,5 million core hours on Curie (TGCC)

Principal Investigator: Sébastien Masson - UPMC