Climate and Weather Modelling

The consequences of climate change are difficult to predict, with extreme weather phenomena hitting certain locations that have escaped forecasts and prevented people from taking precautions. Many of these predictions and climate research are based on numerical simulations using models developed over decades.

Recent advances in supercomputers have made it possible to overcome existing limitations, with scientists facing the possibility of simulating phenomena such as the formation of individual clouds or storms at a global level.
Faced with this progress, one of the challenges is the different time scales of technological development, on the one hand, and model development cycles, on the other. Importing extremely heavy software systems, the result of decades of work, into extremely complex hardware - which may be replaced by a new architecture in the next few years - is becoming increasingly complicated.

Within WP4, Project 1 — Replicability and skill of Earth system models - Workflow and Interface will develop a workflow for testing three aspects of Earth System Models (ESM):

1

Replicability

2

Projection Skill

3

Performance

Most model developers have their own kit of diagnostics and tests that are applied to gather information on those three aspects. The knowledge of how to obtain sensible results from specific tests often lies at individual sites and is rarely shared. However, there are robust methodologies available and widely accepted by the community. Researchers aim to combine those methodologies inside a single workflow with a generic interface applicable to any weather and climate model and accessible on any cluster.

Still on WP4, Project 2 will focus on optimising, porting and testing European and Japanese weather and climate codes.

The High-Performance Climate and Weather benchmark (HPCW) is a domain-specific benchmarking framework which has been developed In the H2020 projects ESCAPE-2 and ESiWACE2 jointly by ATOS and the German Climate Computing Center, DKRZ, with the support of ECMWF. It contains a suite of benchmark codes intended to capture the state-of-the-art in high-performance weather and climate simulation. HPCW can be used as a tool in lieu of more traditional and limited benchmarks, such as High-Performance Linpack and High-Performance Conjugate Gradients, and provides a more pragmatic assessment of a computing system.

The suite currently includes entirely European community climate and weather codes, but through this project, researchers can expand the scope of HPCW to include Japanese codes, too.

Meet the Team

Keep exploring
our project!

Results