The Potential of Spaceborne High Performance Computers for Onboard
Space Physics Data Analysis

Abstract submitted to 1999 Fall Meeting of the American Geophysical Union.

M. Rilee (Raytheon ITSS)
S. Curtis
A. Figueroa-Vinas 
B. Farrell, 
J. Houser
M. Kaiser
M. Reiner (Raytheon ITSS)
NASA/Goddard Space Flight Center

Future Space Science missions will involve multiple spacecraft seeking
to sense phenomena that occur across a wide range of spatial and
temporal scales. Some missions will be required to provide real-time
feedback for space weather studies or monitoring.  The operational and
communication needs of constellations of spacecraft with highly
capable sensors are immense and for some missions are currently beyond
our technical ability.

To examine the problem of science data flow at its source, the
Solar-Terrestrial (ST) Probe Science Application Team of NASA's Remote
Exploration and Experimentation (REE) project has developed two data
analysis applications for a spaceworthy scalable parallel computer
being developed by NASA's High Performance Computing and Communication
Program.  To investigate the possible benefits of spaceborne high
performance computing for Solar-Terrestrial missions, the team chose
to model data analyses that arise in radio frequency interferometry
and plasma particle spectrometry.  Radio frequency interferometry was
chosen as a fundamental technique for imaging and plasma wave
spectrometry. Plasma particle spectrometry was chosen because of its
central importance in ST physics. Both kinds of science instruments
produce large amounts of high-dimensional data that places severe
demands on mission communications infrastructure. The science
communication burden may be eased if a way could be found to extract
scientifically meaningful information from the raw data in place
onboard the spacecraft. Our work shows that the REE computer
architecture is well suited to the data analysis problems we have
studied.

We present results of our modelling efforts and discuss how well our
data analyses perform on the {\it REE Flight Processor} testbed. In
particular, new results from our plasma spectrometry model will be
presented: a highly parallel, efficient, robust, and flexible approach
to reducing plasma spectrometer data will be presented.