HR: 0800h
AN: GP31A-0086    [Abstracts]
TI: Scalable Numerical Dynamo Model Developed as a Component of Geomagnetic Data Assimilation Framework
AU: * Jiang, W
EM: jiangw@umbc.edu
AF: University of Maryland at Baltimore County, 1000 Hilltop Circle, Baltimore, MD 21250 United States
AU: Kuang, W
EM: kuang@bowie.gsfc.nasa.gov
AF: NASA, Goddard Space Flight Center, Greenbelt, MD 20771 United States
AU: Sun, Z
EM: sunzhib1@math.umbc.edu
AF: University of Maryland at Baltimore County, 1000 Hilltop Circle, Baltimore, MD 21250 United States
AU: Liu, D
EM: dliu@bowie.gsfc.nasa.gov
AF: University of Maryland at Baltimore County, 1000 Hilltop Circle, Baltimore, MD 21250 United States
AU: Tangborn, A
EM: tangborn@gmao.gsfc.nasa.gov
AF: University of Maryland at Baltimore County, 1000 Hilltop Circle, Baltimore, MD 21250 United States
AB: Geomagnetic data assimilation requires new code architecture, in which numerical dynamo model shall be integrated with geomagnetic field (observation) and data assimilation models. Compared to purely numerical dynamo simulation, the CPU time and storage for geomagnetic data assimilation could increase by several orders of magnitude, primarily due to ensemble approach in data assimilation. Therefore a more efficient and scalable dynamo model is necessary which should be executable on distributive computing systems, including distributive clusters via network. For these purposes, we have modified substantially the original MoSST core dynamics model in NASA GSFC. Two versions of the model are developed: the first is based on the master-slave structure, aiming primarily on examining different parallelization approaches in evaluating nonlinear forces in the model and in solving linear systems for time integration. The slave nodes are responsible for the computing; and the master node is responsible for managing and communicating among the slave nodes. The second version is completely different: no master-slave structure is employed. All computation and communication are coherently managed by and are equally distributed among all nodes. The new version is currently benchmarked with the first version and shall be used as the model component of our geomagnetic data assimilation.
DE: 4255 Numerical modeling (0545, 0560)
SC: Geomagnetism and Paleomagnetism [GP]
MN: Fall Meeting 2005