CERFACS CERFACS Scientific Activity Report Jan. 2010 – Dec. 2011

4 HPC Climate modelling
4.1 General context
Based on recent findings and thanks to new observational products (mostly satellite) traditional spatial grid
scales used for climate (on the order of 100-200 km) appear to be insufficient to adequately resolve physical
processes such as those involved in air-sea interaction which are critical if one wants to assess the true
predictability and associated mechanisms of the climate system. Higher resolution is also necessary (but
not sufficient) to generate more reliable extreme event probabilities that are critical for impact assessment.
The current and future high-end computational power will likely enable soon climate prediction at the
regional (10-25km) and later local (less than 10 km) scales. These scales are required to make further
progress : (1) for evaluation of processes at the cloud and eddy scale (air-sea coupling, aerosol activation,
cloud microphysical and dynamical interactions) that affect regional and global climate sensitivity and drive
climate biases and (2) for estimation and assessment of impacts using self-consistent models to produce
more accurate statistics of climate variability (e.g., intensity of precipitation) and extreme events (e.g.,
hurricanes) to drive impact assessment models (e.g., of the hydrological cycle).
This scientific context has led us in 2011 to a project of a Grand Challenge for the next decade. Our 2020
HPC Grand Challenge is to develop, in close relation with our national, (Météo France, IPSL, DRAKKAR
group) and international partners, a global climate model system at very high atmospheric and oceanic
resolution (horizontal 10-25 km, vertical 70-90 levels) on various massively parallel computers (with O(10 4 -
10 5 ) processors). The considered climate variability and predictability temporal scales are from interannual
to decadal.
Many challenges, both scientific and technical, need to be adressed along the path of our 2020 Grand
Challenge. Among them, our initial focus in 2011 was on the development and evaluation of an intermediate
high resolution configuration of our ARPEGE-NEMO climate model (atmospheric resolution of 50 km or
spectral truncature T359, oceanic resolution of 0.25 o ). This intermediate configuration is necessary to
prepare and ease the on going and widespread vector to scalar computing migration and the optimal use of
ultra-high resolution climate models on the new generation of high performance computers.
4.2 Development and implementation of the ARPEGE-NEMIX
model
In 2011, a simplified version of this intermediate ARPEGE-NEMO coupled model (called ARPEGE-
NEMIX) has been developed and evaluated in the framework of the ANR funded project SPADES (08-
ANR-SEGI-025). This simplification deals with disabling of any ocean horizontal motion and the coupling
of the atmospheric model with a mixed layer ocean. The mean effect of the tracer advective transport is then
represented by a flux correction methodology (NEMIX, [CC17]). This configuration has been widely used
within the international community for a wide range of studies going from climate sensitivity to greenhouse
forcing to influence of air-sea coupling upon tropical to midlatitude intraseasonal variability, including
the representation of tropical cyclones (for our own recent science applications, see 2.1.2 in paragraphe
“Climate Variability and Global Change” and [CC21]).
108 Jan. 2010 – Dec. 2011