Mitigating seismic and tsunami risk :  the contribution of numerical simulation
René CRUSEM, CEA/DAM, DIF/DASE/LDG

At the time scale of our human societies, major earthquakes and the large tsunamis which are sometimes associated, are the natural phenomena which are most costly in terms of damage and human lives.
Precise deterministic prevision of an earthquake (time, location, magnitude) is currently – and probably still for a long time -  an illusion; however the evaluation of the possible effects, at a given geographic location, of a given earthquake, should be possible if the geological media and the physics encountered are sufficiently known, and if the software used for that purpose are sophisticated enough.
This is true also concerning tsunamis and their destructive effects; in addition, in that case, since tsunami propagation is relatively slow in deep ocean, it can be expected – if the earthquake generating the tsunami is detected and diagnosed early enough, if the impacted coasts are far enough from the source, and if the software computing the propagation of the tsunami waves are fast enough – that the height of the waves can be predicted well before their arrival on the shore, so that the exposed population can be evacuated in time.
The numerical tools required to simulated the detailed effects of earthquakes and tsunamis share some common features: simulations need to cover a large time (from a few seconds to several hours) and space scale (from a few tens of meters to several thousands of kilometres), they have to address several domains of physics (solid mechanics, fluid mechanics, elasto-plasticity, …), they need to incorporate models that are a least partially non-linear (fracture, wave run-up, ..), and, finally, the modelled geometry should in general be 3D. Therefore detailed and/or fast simulations request huge computer resources.
René CRUSEM, Hélène HEBERT,  Philippe HEINRICH, Françoise LE PIVER,  Christian MARIOTTI, Marc NICOLAS

Weather prediction uncertainty estimation with ensemble prediction systems
Philippe ARBOGAST, Météo France

Risk, energy and High Performance Computing : recent demonstrations
Etienne DE ROCQUIGNY, EDF

Power production involves the management of a diverse risk portfolio related to guaranteeing the adequate safety level, controlling the ageing of the power plant fleet as well as the supply-demand balance. Numerical modelling of increasing complexity are developing in order to better characterize the underlying factors: multi-physics, multi-scale or complex portfolios all imply increasing computing power.
Probabilistic quantification of the associated risks amounts to an additional technological challenge as one needs to multiply at a large scale these already-costly unit simulations in a framework that becomes stochastic.
This also alters the way the computer power is invested in the sense that massive distribution becomes necessary;  to best value decision-support computing power, one needs to re-work the compromise between the sophistication of best-estimate models and meshes and the stochastic exploration. Examples of Blue Gen-based recent demonstrators for the simulation of nuclear reactor or power production portfolio will be introduced.

Realistic simulation through HPC computing: Application to Risk at Sea
Argiris KAMOULAKOS, ESI Group

HPC and risk
Pierre BENJAMIN, EADS