Combining exploratory data analysis and scientific visualization in the study of very large, space-time data sets.

摘要

Understanding the contents of very large spatio-temporal datasets provides a starting point for hypothesis generation and testing. Geographic Information Systems or GIS are used to examine spatial datasets but currently are unable to manage or display temporal components of spatial data. This research develops a methodological framework for examining the contents of very large space-time datasets through a combination of exploratory data analysis and scientific visualization techniques within a supercomputer environment. This combination is achieved by implementing a browse function within a software system using PHIGS+, a 3-dimensional computer graphic language. The software provides animation to show the temporal components of spatial data on a 3-D globe. In addition, a number of other methods, including the use of time as an elevation value to show the temporal component of the data in a static environment, are provided. The software also provides 3-D transformations including the ability to rotate the globe on any axis and the ability to zoom in and out (change scale) are available in real-time. The software demonstrates the usefulness of this methodology with the use of a space-time dataset of global seismic activity. The research shows that current supercomputer environments, which are targeted for numerical analysis and viewing results of simulations, can provide a limited amount of real-time interaction but still lag in performance for accessing and displaying large quantities of space-time data.