High-resolution numerical climate modelling
On the path of ``2020 Global Change Challenge'', we propose to set up a scientifically validated high resolution configuration of our ARPEGE-NEMO climate model to simultaneously improve its computing performances on present best machines and explore its scientific potentialities.
This task is the most significant action which aims at adapting climate models to fully take advantage of the new generation of high performance computers. In parallel of an enhancement of model algorithms (strong scaling), the resolution increase of each coupled system module (weak scaling) is another solution to continue benefiting from the growing computing resources. It is obvious that present low resolution models (ARPEGE for IPCC AR5, 25,000 horizontal grid points) are undersized to be parallelized on more than a few hundreds of cores.
Scientifically speaking, higher resolution coupled models are mandatory to simulate flows explicitly down to smaller scales and to capture potential nonlinear interactions between a wider range of spatial and temporal scales of the different components of the Earth system. These configurations will contribute to enhance performances of different ARPEGE-NEMO based prediction systems (from seasonal to decadal time scales).
Coupled models used by the community are constantly in evolution. Any modification, like physical parametrization enhancements, but also resolution increase and even porting on a new machine, could radically change the main characteristics of the coupled system like heat budget, atmospheric or oceanic circulation, mean state or variability ...
Each new configuration requires a complete check of the parametrization on both atmosphere and ocean components. Increase of resolution could even implies a complete rewriting of some parametrizations.
However the ARPEGE-NEMIX model (the NEMO ocean model is transformed to a mixed layer model called NEMIX, where only vertical processes are represented) is a particular case for which the tuning simple by design: thanks to the flux correction, biases are strongly reduced and balances at the interface (heat and water) are easy to reach.
It won't be the same with the ARPEGE-NEMO version of the system, which implies that the newly created ocean-atmosphere interactions and their impact on the whole system at high resolution will have to be investigated.
This task is the most significant action which aims at adapting climate models to fully take advantage of the new generation of high performance computers. In parallel of an enhancement of model algorithms (strong scaling), the resolution increase of each coupled system module (weak scaling) is another solution to continue benefiting from the growing computing resources. It is obvious that present low resolution models (ARPEGE for IPCC AR5, 25,000 horizontal grid points) are undersized to be parallelized on more than a few hundreds of cores.
Scientifically speaking, higher resolution coupled models are mandatory to simulate flows explicitly down to smaller scales and to capture potential nonlinear interactions between a wider range of spatial and temporal scales of the different components of the Earth system. These configurations will contribute to enhance performances of different ARPEGE-NEMO based prediction systems (from seasonal to decadal time scales).
Coupled models used by the community are constantly in evolution. Any modification, like physical parametrization enhancements, but also resolution increase and even porting on a new machine, could radically change the main characteristics of the coupled system like heat budget, atmospheric or oceanic circulation, mean state or variability ...
Each new configuration requires a complete check of the parametrization on both atmosphere and ocean components. Increase of resolution could even implies a complete rewriting of some parametrizations.
However the ARPEGE-NEMIX model (the NEMO ocean model is transformed to a mixed layer model called NEMIX, where only vertical processes are represented) is a particular case for which the tuning simple by design: thanks to the flux correction, biases are strongly reduced and balances at the interface (heat and water) are easy to reach.
It won't be the same with the ARPEGE-NEMO version of the system, which implies that the newly created ocean-atmosphere interactions and their impact on the whole system at high resolution will have to be investigated.
Back
