Printversion of this page
PDF-Version of this page

 

Scientific Computing

The scientific computing group is concerned mainly with the growing need of the users at AWI for computer resources. The user support, covering a broad variety of topics and work in close connection with several research groups at AWI.

 

Climate modells are parallelized for the use on massively parallel computers at AWI, at DKRZ or at HLRN. Among these are the modular ocean model (MOM), the Bremerhaven Ice-Ocean-System (BRIOS), a parallel version of the regional atmosphere model HIRHAM (from DNMI) and an unstructured grid ocean/sea-ice model COM (community ocean model).

 

Research on sequential data assimilation algorithms for parallel computers is another field of work. Here different algorithms based on the Kalman filter are compared. An efficient local variant of the ensemble-based Singular "Evolutive" Interpolated Kalman (SEIK) filter was developed. This algorithm, together with other fully-implemented algorithms based on the Kalman filter, are part of the assimilation framework PDAF, which was developed in the Scientific Computing group. PDAF supports the implementation of parallel data assimilation systems based on existing numerical models. Within the project Community Ocean Model (COM), finite element models for assimilation of oceanigraphic data are parallelized and optimized.

 

Also algorithmic support is provided, especially when code parallelization by domain decomposition and message passing techniques is performed.

 

Members of the group participate also in national an international projects like the Tsunami early warning system GITEWS (German contribution to Indonesean Tsunami Early Warning System) and C3Grid (Collaborative Climate Community Data and Processing Grid).

 

When computing resources at AWI are insufficient for certain grand challenge projects, the group manages project initialization at national supercomputing centers like at FZ-Juelich, ECMWF, or HLRN. The latter is done within the joint project BremHLR.

 

New parallelization techniques are explored with MPI and OpenMP for mutiprocessor workstations and high performance computers such as the IBM Blade CenterSGI UV100, and NEC SX-8R. Users of the compute server at AWI are also supported concerning optimization questions and numerical libraries.

 

The group maintains contacts to other scientific institutes and therefore is capable to import and mediate certain knowledge for scientific projects at AWI from different sites.

 

In 2004 in cooperation with Hochschule Bremerhaven a research group in bioinformatics has been established to support several research groups from molecular biology.

 


 
Printversion of this page
PDF-Version of this page
 
 

Related Pages

Publications
Open Positions