LAGOON
Large scAle Global storm surge simulatiOn of OceaNs
Visit the website of the project here
Coastal areas host around 10% of the world’s population and a huge amount of economic activities. Climate change is expected to increase coastal flooding hazard in years to come. In this project, we propose to develop a numerical tool for the stormsurges predictions. For four years, a joint effort between the partners of this project among others has been done for the development of a numerical tool able to tackle planetary computations with high resolution at the coast.
The scope of this project is to increase the computational performance of our modelling platform, in order to upgrade it as an efficient and accurate tool for storm-surge predictions in different future climate scenarios. To achieve this goal and producing results which go beyond the state-of-the-art, our efforts will be focused on the following numerical and informatics developments, devoted to decrease the run time of the model in operational conditions.
(1) In deep ocean, tidal waves and atmospheric surges are very linear processes under the influence of earth rotation. In the low Froude regime, standard numerical schemes may suffer from excessive numerical diffusion on the coarser elements (typically located offshore). This is expected to damp the water level. Also, the very short time step needed to ensure a stable computation of barotropic waves induces long runs with explicit time stepping. In this project the aim is therefore to develop a robust Implicit-Explicit (ImEx) strategy for the high order discontinuous Galerkin scheme, to drastically increase the time step.
(2) The good parallel efficiency of the computational core of the library was obtained with explicit time stepping. We will therefore need to address the efficiency of the inversion of the (non)linear systems induced by implicit time stepping. As the stiffness of the linear system at low Mach number is similar to the Laplace equation, an efficient strategy consists in developing an aggregated multigrid preconditioned Krylov method. This method is attractive, because it is based on the computation of the residual (which is already well parallelized), but on aggregated grids. Input-Output (IO) is another aspect that shall also be addressed. With data size and output frequency needed for the global reanalysis application, IO becomes an important bottleneck of the code. Two stages IO, in-situ and in-transit data post-processing are strategies that will be evaluated with existing technologies and will be implemented to improve the performance of the production code when it will come to realize the numerical experiments.
(3) We will exploit the numerical and HPC developments proposed in the above paragraphs to assess the impacts of climate change on flood risk at the global scale in future projections. To do so, we will perform sea-level future projections under different climate scenarios following a multi-model approach, so far out of reach. The pursued high resolution (10 km offshore and below 1 km at the coast) also goes beyond the state of art for global sea level predictions, and will only be attainable thanks to our high performance numerical model. The numerical tool will be validated on 1979-2014 sea level reanalysis, and be used for the generation of a database of sea level projections on future climate CMIP6 projections.
The code developed within this project will be freely distributed, with a strong effort put on reproducibility of results. Data generated for both the sea level reanalysis and the database of sea level projection for future climate projections will be distributed towards the community.
