Physically based hydrological modelling and High Performance Computing: why and how? by Dr. Daniel Caviedes-Voullième (Agrosphere, Forschungszentrum Juelich GmbH, Germany)

Hydrological modelling has experienced a transition from phenomenological and observation-matching modelling approaches into process and physics based formulations. The are many reasons for this, including the need of specific spatially-explicit quantitative information, dealing with a changing environment, and the interest of having virtual laboratories for hypothesis testing. Such physically based formulations are non-trivial to design, from many points-of-view: mathematical formulation, numerical discretisation, computational implementation, cost and efficiency. Design considerations go back and forth among these aspects, and are currently undergoing a revolution with dramatic increase of High Performance Computing capabilities. This talk will present some of the current established approaches and some ongoing developments in surface and subsurface hydrological modelling, against the backdrop of changing computational technology which is accelerating the evolution of such models. We will explore the pros and cons of these approaches, and argue how HPC-enabled physically based models open new avenues for hydrological research and management.