Past Awardee

Exploring the Frontier of Multivariate Geovisualization: Virtual Reality and Parallelization in GeoDa

Luc Anselin

College: Liberal Arts and Sciences
Award year: 2004-2005

This project is to extend the current capability of a software tool for the exploration of geospatial data, GeoDa, developed under the direction of the PI. The extension is intended to deal more effectively with 3-dimensional visualization, the exploration of multiple linked views of the data, and the analysis of large (> 100K observations) space-time data sets. This will be approached by investigating the use of NCSA technologies for the interactive exploration of 3-dimensional scatter plots, such as the CAVE virtual reality environment, and the application of a high resolution display wall in the visualization of multiple linked windows, beyond what is possible in a standard desktop. In addition, research will focus on the introduction and implementation of parallel algorithms in the analysis of spatial and space-time autocorrelation, to leverage the computational power in the wall. The extended data exploration environment will be applied in empirical investigations of space-time patterns in several examples, including the association between air quality and hospital admissions in the Los Angeles basin, clusters and spatial outliers in house sales prices, and patterns in cancer incidence and mortality. The methodological exploration, software development, and empirical applications leverage ongoing funded research projects under the direction of the PI, and extend the frontier of what is currently possible in terms of user interaction, high resolution visualization and size of data sets.

GeoDa was developed as part of an NSF funded research infrastructure project and is designed to be an "introduction to spatial data analysis." It is highly interactive and extends the principle of dynamically linked windows (observations highlighted in one graph or map being simultaneously highlighted in all) to incorporate the notions of global and local spatial and space-time autocorrelation from spatial statistics. This allows the investigation of clusters and outliers in geospatial data sets, a powerful collection of methods with applications ranging from criminology, epidemiology, to public health and environmental analysis, with an important outreach component. The software includes these techniques in addition to standard exploratory data analysis tools, such as scatter plots, parallel coordinate plots, and 3-dimensional scatter plots.

While powerful, the current implementation of the software and associated methods is somewhat limited in terms of the number of analyses that can reasonably be shown on a standard desktop (limited desktop real estate), the counterintuitive nature of brushing and linking three dimensional graphs in two dimensions (projected onto a two dimensional plane), and the size of the data sets that can reasonably be explored. A crucial aspect of such exploration is immediate feedback to the user. The three core aspects of the project are intended to address these limitations. The use of virtual reality in the CAVE environment provides new opportunities for interactive visualization of three dimensional patterns (including space-time patterns), and the wall provides not only extensive screen real estate for multiple linked windows but also extensive algorithms for parallelization. The application of parallel algorithms in spatial statistics is still in its infancy and its further development is an important methodological aspect of the project.