$proj_team_count = count($proj_team); ?>

ABC-MUSE

Robust high-resolution multi-physics simulations

The generalization and optimization of the multi-purpose software environment

escience methodology

ABC-MUSE

ABC-MUSE

Project Highlights

Transferring recent advances from the domain of Computational Astrophysics to the domain of Climate Research

Performing high resolution global climate simulations to assess the impact of changes in ocean currents on local sea level height

The generalization and optimization of the multi-purpose software environment

The Astrophysical Multipurpose Software Environment (AMUSE) provides a homogeneous interface to a wide variety of packages enabling the study of astrophysical phenomena where complex interactions occur between different physical domains, such as stellar evolution and dynamics, (magneto-)hydrodynamics, radiative transfer, and astrochemistry. Applications are numerous. For example, AMSUSE enables studying the co-evolution of planetary systems within cluster environments, the formation and evolution of black holes in galaxies, or the interplay of gas, radiation, and chemistry in star formation process.

A unique opportunity

The ABC-MUSE project will broaden the scope of AMUSE and transform it into a public, robust facility with a much wider functionality than the current implementation aimed at astrophysics. Utilizing the current AMUSE team in collaboration with the Netherlands eScience Center provides a unique opportunity to transform AMUSE to a general-purpose facility of national and international significance and kick-start long-term development. Main issues are algorithmic generalization, diversification of the method, high performance computing, data mining, and visualization.

Applying AMUSE in the domain of climate research

The ABC-MUSE system will be applied to problems in the domain of climate research. Today’s Global Climate Models (GCMs) are constructed in a very similar way as models in the domain of astrophysics: GCMs consist of a variety of compute kernels that are combined to allow for complex interactions between, among others, separate simulations of atmosphere, sea, ice, and landmass. Although much work on GCMs is being performed, the domain of climate research is in need of a simple homogeneous interface to enhance the creation and expansion of GCMs.

 Today’s Global Climate Models are constructed in a very similar way as models in the domain of astrophysics

As an important application we consider the problem of the strong weakening (collapse) of the Meridional Overturning Circulation (MOC) in the Atlantic Ocean. State-of-the-art GCMs have indicated that a weakening of the MOC would lead to a strong cooling of the North Atlantic region. Constrained by compute power, most GCMs operate at a horizontal resolution of about one degree for ocean and sea-ice, and about two degrees for atmosphere and land. As a result, the ocean component does not resolve many important oceanic features, including so-called eddies (vortices with scales ranging from 10-100 km). Recent research has indicated that high spatial resolution in both ocean and atmospheric model components does improve the simulation of many features in the climate system. In particular, the impact of changes of the Atlantic MOC are significantly different in eddy-resolving ocean models compared to models in which eddies are parameterized .

Achieving high resolution climate simulations

This project aims to achieve high resolution simulations that are at the moment not easily realized. To assess the impact of a severe reduction of the strength of the MOC, we aim to run coupled simulations at such high resolution that ocean eddies are fully resolved. Although steps in this direction are currently being made in the NLeSC eSALSA project, part of the simulation development remains complex and ad hoc, given the fact that the applied community-developed software is not based on design principles similar to that of AMUSE. For example, it is currently very hard to apply fast, specialized GPU kernels. The use of GPU kernels has been shown to be highly effective, providing significant speed improvements for full simulations or – alternatively – allowing for a significant increase in model resolution. An implementation based on the design principles of a generalized ABC-MUSE would provide the necessary flexibility. visualization, and data mining capabilities.

Image: 'White Marble' Arctic View by NASA Goddard Space Flight Center (CC License) 

Sign up for our newsletter