ABSTRACT
The study and understanding of complex geographic phenomena often depends on the analysis of multivariate spatial data to discover complex structures and gain new knowledge. Figure 12.1 shows a conceptual representation of a typical data set that contains multiple variables and geographic information, which can be viewed as a spatial data matrix (Haining 2003), lattice data (Cressie 1991), or a “map cube” [a simple case of the map cube model introduced in Shekhar et al. (2001) and Chapter
12.1 Introduction ................................................................................................. 325 12.2 Related Work ............................................................................................... 327
12.2.1 Cluster Analysis ............................................................................. 327 12.2.2 Regionalization .............................................................................. 329 12.2.3 Geovisualization ............................................................................ 330
12.3 An Integrated Approach to Multivariate Clustering and Geovisualization ................................................................................... 331 12.3.1 A Theoretical Framework ............................................................. 331 12.3.2 Multivariate Clustering and Pattern Encoding .............................. 334 12.3.3 Multivariate Visualization and Mapping ....................................... 335 12.3.4 Regionalization with Spatially Constrained
Hierarchical Clustering ................................................................. 336 12.4 Application in Global Climate Change Analysis ........................................ 337
12.4.1 Data Source and Preprocessing ..................................................... 337 12.4.2 Discovering Climate Change Patterns .......................................... 337 12.4.3 Comparing Regionalization and SOM Results ............................. 339
12.5 Conclusion and Discussion.......................................................................... 341 References .............................................................................................................. 342
4 of this book]. Such data sets are commonly encountered in various spatial research elds such as socioeconomic analysis, public health, climatology, and environmental studies, among others. For example, to study global climate patterns and their change over time, we not only examine temporal trends or patterns of climate variables (e.g., temperature) at a specic location, but we are also interested in the geographic variation of such trends or patterns. It is a challenging task to explore large multivariate spatial data sets and tease out complex (and often unexpected) patterns, which may take various forms (linear or nonlinear) and involve multiple spaces (e.g., multivariate space and geographic space) (National Research Council 2003).