Seamless Modelling of the Southern European Seas with an optimized MPI-based Unstructured-Grid Model

This interdisciplinary work combines computational science and physical oceanography. The parallelization of the SHYFEM circulation model and its application for simulating the system of interconnected basins in the Southern European Seas (SES) are the main areas of focus. The SES system is made up of different water bodies that can generally be regarded as separate and having independent dynamics. Most modeling studies treat these water bodies as standalone. The recent studies that model the system as a whole typically focus only on a particular process or a short time-scale. Representing the dynamics generated by narrow straits is difficult in regular grid models, and so those based on an unstructured grid represent a natural solution. Modelling aimed at a complete description of such complex physical systems incurs a considerable computational cost, which is directly affected by the size of the physical domain, the model resolution and the length of the simulation. Modern simulations involve numerical codes of circulation models that exploit the ongoing increases in computer power. The code of state-of-the-art models can be executed in parallel through high-performance computing facilities, enabling the simulations to be completed quickly. In this context, the parallelization of SHYFEM model is the first task, which has been successfully applied in coastal and regional configurations. The task of parallelization of SHYFEM includes an assessment of its reproducibility and parallel performances on a realistic test case. In the second task, we address the applicability of this parallel unstructured model for the more demanding SES system, demonstrating that this parallel code can effectively simulate the SES system from a physical point of view, seamlessly representing the physical domain in the long term. The SES system is simulated for one year with a fully baroclinic run. The results obtained are assessed qualitatively and by comparison with observations where possible.