Development of a 3-D Visualization Environment for Integration and Analysis of Computer Simulations and Satellites and Ground-based Observations Based on Grid Environment

Kazunori Yamamoto

Authors: Kazunori YAMAMOTO, Graduate School of Science and Engineering, Ehime University
Ken T. MURATA, Center for Information Technology, Ehime University
Eizen KIMURA, Graduate School of Medicine, Ehime University
Rie HONDA, Department of Information Science, Kochi University


In the Solar-Terrestrial Physics (STP) field, there are two major research methods for geo-space; one is computer simulation and the other is satellite and/or ground-based observation. Combination of these two methods is crucial to develop our understanding of the structure and dynamics of the Earth’s geo-space. Especially, 3-D visualizations would be effective and plausible for the combinations of both data. We have constructed a 3-D visualization environment for integration of simulation and satellites and ground-based observations data analysis: the Virtual Earth’s Magnetosphere System (VEMS). On the VEMS, hetero data are integrated in the following manner: (1) Search and download of observation data files, (2) Convert of data format, (3) Preprocess of data (resampling), (4) Visualize in 4-D (3-D in space and 1-D in time), and
(5) Interactive analysis.

(1)Search and download of observation data files: Using the STARS (Solar-Terrestrial data Analysis and Reference System) meta-database that provides meta-information of observation data files managed at distributed observation data sites over the Internet, users download data files without knowing where the data files are managed.

(2)Convert of data format: We developed an original data class to support our reading of data files and conversion them into a common data format. The data class defines schemata for several types of data. Since this class encapsulates data files, users easily read any data files without paying attention to their data formats.

(3)Preprocess of data (resampling): Resampling time series with linear interpolations, time scale is regularized over data files.

(4)Visualize in 4-D: A smaller scale coordinate system is laid on a larger one. Due to this hierarchical structure of coordinate systems, any observation data given in the smaller scale coordinate system is visualized in the 3-D space in the larger coordinate system without any geometrical conversion.

(5)Interactive analysis: In addition to the observation data, global MHD simulation data are also visualized in the 4-D format as long as the simulation is performed with real parameters. We can directly compare observation data (scalar or vector) with global MHD simulation results. Through our research uses of the VEMS for scientific data analyses, we found several practical problems of the system, especially in case of long durational data analyses. Herein, let us consider one year data analysis. A data file usually consists of one day observation data, thus there are 365 files for one year. On the process (1) in our system, we need to download 365 files for that analysis per data. It usually takes long time, and makes our real time data analysis hard. We attempt to overcome this problem with introducing a download and cache agent server into our system.

Another problem is the computational load to make re-sampling of data on the process (3). Assume that the sampling time of a data is 10 seconds. The total number of time in one year data is 3,153,600. Since it is unreasonable number, we usually resample them. This process requires CPU power and memory size to the computer. To overcome this problem, we introduce a GRID system to obtain load balance environment. Data files and data processes are parallelized on the GRID, and we will achieve high performance statistical data analyses.

Keywords: Computer Simulations, Satellites and Ground-based Observations, Data Integration, 3-D Visualization, Load Balancing