3rd HyMeX workshop 1-4 June 2009 Heraklion (Gournes), Crete-Greece
Distributes computing experiment: Building a grand ensemble of simulations of present day and projected Mediterranean and North African climateLaura Mariotti (International Centre for Theoretical Physics)
Extreme events are intrinsically local and intermittent in nature making long simulations and large ensembles a necessity in order to produce meaningful statistics. Even so climate models have difficulty in reproducing the statistics of extreme events correctly, let alone extreme events themselves. Moreover the discrepancy between model simulations and observations is highly dependent on parameter settings, while the spread in simulated projections of climate are highly dependent on both parameter settings and forcing uncertainty. Existing distributed computing experiments have conclusively shown that members of the public can be engaged to donate the dormant resources in their idle home computers to carry out vast numbers of climate simulations with state-of-the art models overshadowing the most powerful conventional supercomputers available to the climate community.
Distributed computing resources would allow us to construct a large ensemble of simulations where not only parameter values are perturbed, but uncertainties are explored as related to varying forcings and driving global models. We refer to this as a grand ensemble.
We plan to build an experiment framework in which the RegCM would be driven by boundary conditions from the next generation of IPCC GCM simulations for all greenhouse gas reference concentration pathways (RCP).
Individual simulations, i.e. the unit computational effort requested of each participant, will consist of a 20 year time slices, with choice of different domains and resolutions. At present we envision 4 domains, covering the Eastern, Central, Western and Whole Mediterranean Basin and three resolutions, approximately 50 km, 25 km and 10-15 km. The selected time intervals will investigate both near climate change (1980 to 2000, 2000 to 2020 and 2020 to 2040) and century scale climate change (2080 to 2100, 2100 to 2120, 2180 to 2200, 2200 to 2220, 2280 to 2300).
Results from the grand ensemble of regional Mediterranean and North African climate simulations would build on the experience from established distributed computing experiments and would be used to investigate the dependence of the statistics of extreme events on model structural properties, GCM-derived boundary conditions and radiative forcing projections. Focus will be not only on constraining the changes in the statistics of extreme events but also on quantifying the uncertainties associated with statistical values, and the transfer of uncertainty from model parameters, boundary conditions and forcing scenarios.