Development of Next-Generation Computer Model for Space-Weather Specification and Forecasting, 15-9026Printer Friendly Version
Inclusive Dates: 04/01/97 - 04/01/00
Background - The term 'space-weather' refers to conditions on the sun, in the solar wind, magnetosphere, ionosphere, thermosphere, and mesosphere, that can influence the performance and reliability of space-borne and ground-based technological systems and can endanger human life or health. Adverse conditions in the space environment can cause disruption of communications, navigation, electric power distribution grids, and satellite operations, leading to a broad range of socio-economic losses. The National Space Weather Program (NSWP) is a new initiative designed to address many of the unresolved aspects of space weather, including theory, modeling and measurements, in a unified manner. The NSWP initiative is jointly funded by the U.S. Air Force, Navy, National Science Foundation, and NASA. One goal of the NSWP is to produce weather forecasts for the various regions of space ranging from the sun to the Earth's middle atmosphere.
Approach - The goal of this research project is to develop a space weather model spanning the mesosphere, ionosphere, and thermosphere. The new model is based on an existing computer code that runs on CRAY supercomputers at the National Center for Atmospheric Research (NCAR). This code, called the Thermosphere-Ionosphere-Mesosphere-Electrodynamics General Circulation Model (TIME-GCM), is widely acknowledged as the premier space weather code in existence. SwRI modified the code to run in a distributed parallel computing environment and to use a variable grid size. The Research Initiative Program in Advanced Modeling and Simulation (RIP-AMS) was an interdivisional collaboration that resulted in the enhancement and expansion of SwRI capabilities in high-performance parallel computing. The RIP-AMS program resulted in parallel computing techniques that permit significant improvements in the runtime of computer codes. Specifically, algorithms based on domain decomposition strategies have been developed, providing a framework that will be applied to the TIME-GCM code, allowing a natural method for parallelization and incorporation of variable grid-size regions.
Accomplishments - The existing serial code that runs on CRAY computers was ported to SwRI workstations, and the specialized CRAY commands currently in the code were replaced by system-independent commands. Within machine precision, the results are the same for the different machines. The code was parallelized using tools and techniques developed under the Institute's RIP-AMS program. Domain decomposition was performed by latitudes. The code was modified to run on the SwRI Distributed Computing Facility using PVM. The research team has tested the parallelized code and demonstrated significant speedup factors using up to 10 nodes. The model runs in real time with 4 processors.
The TIME-GCM code had significant limitations. It was possible to increase the latitude resolution, but not the longitude resolution. The team also discovered several scalability issues, which were described in the published manuscripts. The most important issue is the requirement to broadcast a large (24 MB) file to each slave processor in turn. This procedure dominates the model run-time and limits the scalability to larger numbers of processors. The team has taken this model as far as reasonably possible and has learned important lessons. New models without these limitations should be developed. SwRI has begun to develop a new high-resolution overset grid model, which uses the same basic chemistry and physics as the TIME-GCM, but which is being developed specifically for distributed parallel computing environments.
The research team studied plasma motions in the high-latitude F-region ionosphere. Although the team had planned to simulate the interval using the new parallelized high-resolution model, the limitations of the model resulted in revised plans, as mentioned previously. The team performed a data-analysis study that was published in a technical journal. This study will form the basis of a future simulation with a truly high-resolution model.