Topic 2: New Directions in Distributed Geovisualization Alan M. MacEachren GeoVISTA Center Department of Geography, Penn State & International Cartographic Association Commission on Visualization & Virtual Environments maceachren@psu.edu www.geovista.psu.edu Overview A. Brief background 1. cartography & geovisualization 2. geovisualization, scientific visualization, and information visualization B. Distributed geovisualization defined 1. drawing upon distributed resources for geovisualization 2. geovisualization distributed among individuals 3. geovisualization distributed in the real world C. Use & Usability D. Looking ahead 1
A: Background traditional foci in cartography: capturing information about and representing the physical world developing innovative strategies for visual analysis and information abstraction that take advantage of the power of human vision both are inputs to geovisualization Geographic Visualization (geovisualization) the use of visual representations of geospatial information to facilitate thinking, understanding, and knowledge construction about aspects of the geographic scale human and physical environment the creation of those visual representations 2
functions of geovisualization explore analyze synthesize present dimensions of use tasks users interaction MacEachren, et al 2004 links to other developments in visualization traditional foci in cartography: capturing information about and representing the physical world (e.g., visualizing landscapes, geology, climate processes) SciVis developing innovative strategies for visual analysis and information abstraction that take advantage of the power of human vision (e.g., visualizing health statistics, demographic trends, multivariate inputs to landcover classification) InfoVis 3
lines = relationships between grants color = semantic regions e.g., InfoVis and Spatialization scale 1:3 physical human tech./eng. relationships weak medium strong NSF IGERT awards 1998-2003 figure courtesy of Sara Fabrikant e.g., InfoVis and Spatialization a news landscape peaks = dense document clusters fog hides all news below a query threshold figures courtesy of Sara Fabrikant 4
B: Distributed Geovisualization drawing upon distributed resources for geovisualization geovisualization use distributed among individuals multiuser geovisualization geovisualization use distributed in the real world multisite, synchronous geovisualization B1: Distributed resources for geovisualization traditional modular visualization environments: e.g., AVS, IRIS Explorer, IBM DX open, component-based visualization environments GeoVISTA Studio 5
Open, component-based visualization environments Snap-together-visualization: Chris North Center for Human-Computer Interaction, and Department of Computer Science, Virginia Tech infovis.cs.vt.edu/snap Geovisualization: GeoTools: started in 1996, University of Leeds. GT1 Java Applet API, now GeoTools-Lite. GT2 seeks to implement OpenGIS Consortium (OGC) specifications www.geotools.org GeoVISTA Studio: Mark Gahegan, GeoVISTA Center, Penn State University geovistastudio.sourceforge.net/ Maps in Snap: with ESRI MapObjects DataMaps: To be distributed on the upcoming Counties USA cdrom by US Census Bureau Includes 8000 census variables, supports multiple Dynamic Queries, (developed with participation of HCIL, Univ. of MD, College Park) figure courtesy of Chris North 6
http://docs.codehaus.org/display/geotools/map+and+style+tutorial Geotools: OpenGIS compliant components map structure in GeoTools can think of a map as an ordered list of layers, where each layer is rendered on top of the previous ones according to the data it contains (the feature source) following certain painting rules (the style). figure courtesy of James Macgill http://docs.codehaus.org/display/geotools/map+and+style+tutorial Geotools: OpenGIS compliant components GeoTools2 maps Geotools2 provides map symbolizing by implementing the Styled Layer Descriptor OpenGIS standard. Map shown with magnifier on. figure courtesy of James Macgill 7
Maps in Snap: with Geotools figure courtesy of Chris North GeoVISTA Studio Conceptual background and architecture Multivariate visualization Color selection tools Geovisualization + computational analysis adding visual-computational knowledge representation tools Project Director: Dr. Mark Gahegan system architect now: Dr. James Macgill geovistastudio.sourceforge.net/ 8
Conceptually our focus is on understanding the process of geospatial information use model driven data driven Exploratory data analysis and hypothesis generation Confirmatory analysis and knowledge construction Evidence integration and decisionmaking Current GIS Presentation and results assessment Time adapted from figure by Mark Gahegan GeoVISTA Studio: typical application A Java, component-based, visual programming environment for development of applications and applets that integrate visual, statistical, & computational methods for (geospatial) data exploration, analysis & knowledge construction 9
Studio Architecture: Application Builder Studio employs JavaBean technology to construct tools. The JavaBean specification defines a set of standardized APIs for the Java platform. From this, the builder automatically constructs a syntactic description of the functionalities and i/o methods of any bean. GeoVISTA Studio: Types of Users Component Developers Build JavaBeans components to produce new tools. These new tools are imported into Studio and tested. Application Developers Construct and disseminate data analysis/visualization applications to address specific problem domains Application Users Do not use Studio directly. Instead, use standalone applications or applets produced by Studio. 10
Studio functionality a sampling data transformation and statistics (conventional, spatial) a visual classifier including color selection (ColorBrewer) multivariate classification: self-organising map, learning vector quantisation, k-means, maximum likelihood a multiform matrix (scatterplots, maps, ) interactive parallel coordinate plot (PCP) 2D dynamic map 3D rendering, including dynamics spatial & multivariate clustering minimum spanning tree time series plots Multivariate visualization methods Dynamic Parallel Coordinate Plot (PCP) Multiform Matrices 11
Parallel Coordinate Plot (PCP) breast cancer mortality rate MD/ 100,000 cervical cancer mortality rate 35 175 35 McKean 30 150 30 Lycoming 25 125 25 Mifflin 20 15 100 75 20 15 10 50 10 5 25 5 0 0 0 NCI: Cancer incidence and risk factors NSF # 9983451 white female lung cancer mortality rates purple = very high, green = very low health service areas per capita income 93 79-81 82-84 85-87 88-90 91-93 12
ESTAT: Exploratory Spatial-Temporal Analaysis Tool cervical breast cer-% mort. mort. local rate rate stage br-% local stage MD ratio mamm per mamm screen capita test facility inc pap smoke no test ever insur 13
cervical breast cer-% mort. mort. local rate rate stage no br-% MD per mamm insur local ratio capita test stage inc pap mamm smoke test screen ever facility Multiform Displays Bivariate Matrix matrices extending the scatterplot metaphor: UniForm Bivariate Matrix matrix with one representation form (e.g., all scatterplots); 14
Census: DC race & housing Multiform Displays Bivariate Matrix matrices extending the scatterplot metaphor: UniForm Bivariate Matrix matrix with one representation form (e.g., all scatterplots); MultiForm Bivariate Matrix matrix with two (or more) bivariate representation forms (e.g., maps + scatterplots). 15
Multi-BiMatrix: bivariate map + scatterplot local (early) stage cervical/breast cancer diagnosis % breast % cervical % all high % for both low % for both breast one cervical all breast + cervical breast + all cervical + all % breast % cervical % all computational grouping and sorting Correlation selecting subspaces selecting relationship measure and sorting Conditional Entropy 16
design incorporating several multiform displays and computational sorting Coordination manager computational sorting multiform displays Typology of interaction, an (Inter)action ontology (with Mark Gahegan, Frank Hardisty & Junyan Lou) Session events: start, end Selection events: brush, focus, sample, indication Data events: data, metadata Visual events: visual classification, visual mapping How should different components respond to these events? Which events should they respond to? Do all users and applications share this list? How can we be more flexible? 17
Color schemes ColorBrewer online univariate color scheme selectin tool Studio ColorBrewer extended ColorBrewer promoting logical & effective use of color for data visualization Cynthia Brewer & Mark Harrower NSF (EIA-9983459) 18
ColorBrewer: sequential 5-step ColorBrewer: sequential 9-step 19
ColorBrewer: diverging NCHS Atlas colors ColorBrewer: with roads, gray background 20