Abstract
(Englisch)
|
METROMED aimed at studying and modelling the key processes of matter transfer (exchange and storage) and the biogeochemical cycles in the coastal zone area, as a follow up of the previous MAST II program. During the early phase of the METROMED program, Limnoceane collected and assembled all the hydrographic data previously gathered in the Gulf of Lion and in the Thermaikos area.. These have been validated and assembled in a CD-ROM, which also contained a new, user's friendly, visualisation software providing a fast access to data sets of the numerous cruises. The CD ROM was distributed among the other project partners, in order to be included in their national database. Part of these data have also been subsequently used as initial/boundary conditions to run hydrodynamic models and to evaluate the quality of simulations. One key topic of the Limnoceane's contribution was indeed to study the influence of the hydrographic conditions and of the meteorological forcing on the particle's transport and dispersion under various hydrographic conditions and meteorological forcing. The simulations of the circulation in the Gulf of Lion were based on climatological data sets for the summer period, while in the Gulf of Thermaikos, the results of one of our previous intensive cruise provided realistic and reliable initial conditions. The Prosper General Circulation Model (PGCM), previously developed by our group, was a basic tool for these studies. Running the model in a standalone mode demonstrated to be time consuming and not appropriated for sensitivity analysis of the dispersion processes. The strategy for such analysis has been greatly improved by a storage of the velocity fields calculated with the PGCM into a distributed database, and by the off-line coupling of the PGCM results with a random walk model. That approach favoured a close and fruitful collaboration with the Institute d'Intelligence Artificielle at University of Neuchâtel as a practical application of their interactive visualisation platform Zoomin, developed during the time of the project. This allowed rapid comparisons between the distribution pattern of a tracer obtained by using classical eulerian simulations, with the domain of occupation of a large population of individual particles produced by the random walk model. The eulerian dispersion appears thus to be strongly influenced by the diffusion coefficients selected for the tracer, and also by the grid size of the model, a choice which is usually limited by the computing resources. The numerical diffusion in such models increases the dispersion, especially in the surrounding of land based sources where strong gradients occur. This may induce misleading conclusions on the potential impact of tracers or pollutants released by such sources. The main drawback when running the random walk method offline is the storage capacity needed to memorise the velocity fields, although this problem becomes less critical with the increasing capacity of the storage mediums.
|
