Earth System Models: Challenges in a changing HPC environment

Abstract

Climate models are continuously further developed into comprehensive Earth System Models (ESMs) representing more and more detailed process descriptions at increasingly higher model resolution. This transition poses several challenges, which are exacerbated by the concurrently changing HPC environment. The steadily increasing complexity of the model systems requires new approaches for the overall model design to keep them manageable and extendable, their long life cycles require the application of software engineering techniques in HPC and rely on standardisations. One of the fundamental issues is the methodology to combine ("couple") efficiently different domain and/or process models into a consistent ESM. Often, historically grown legacy codes, or parts of them, and model components based on a variety of numerical methods and parallelisation strategies need to be connected. Several approaches to tackle these issues exist. The Modular Earth Submodel System combines different coupling approaches, such as the fine granular modularisation for atmospheric chemistry processes, the indirect and direct coupling of an atmosphere - ocean system, and the on-line nesting of a high resolution limited area atmospheric model into a global model.