Outline Cartography: Cartographic Visualization: Background It is the science or art of mapmaking With the introduction of GIS, computerized Cartographic Visualization Cartography & Cartoviz/Geoviz It is a practice that has a long history tools became available making it easier to Recent work in Cartoviz/Geoviz Until the last couple of decades, its primary generate maps and do spatial analysis April Webster An introduction to Geoviz methods Cartographic (or geographic) visualization Animation for spatiotemporal data exploration purpose has been that of communication November 21, 2008 is a relatively new development in the field Conditioned choropleth maps and the storage of information via static Cartograms paper maps of cartography Summary It is the marriage of cartography, Future of Cartoviz/Geoviz information visualization and exploratory data analysis
Cartographic Visualization: Papers presented Goal of paper: Its purpose is to support Geovisualization illustrated Menno-Jan Kraak, ISPRS Journal of Photogrammetry & Remote Data exploration Sensing 57(2003), 390-399. Geovisualization Illustrated To increase awareness within the Data analysis Geographic visualization: designing manipulable maps for geographic community of the Hypothesis generation exploring temporally varying georeferenced statistics A. M. MacEachren, F. P. Boscoe, D. Haug, and L. W. Pickle. Proc. geovisualization approach and its Knowledge acquisition InfoVis '98, 87-94 Conditioned Choropleth Maps and Hypothesis Generation. benefits. Why? – the “magnitude and complexity of the Carr, D.B., White, D., and MacEachren, A.M., Annals of the Association available geospatial data pose a challenge as to of American Geographers, 95(1), 2005, pp. 32-53 CartoDraw: A Fast Algorithm for Generating Contiguous how the data can be transformed into information Cartograms. and ultimately knowledge.” Keim, D.A, North, S.C., Panse, C., IEEE Transactions on Visualization and Computer Graphics (TVCG), Vol. 10, No. 1, 2004, pp. 95-110
Author’s approach : Napoleon’s March on Moscow (Minard): Minard’s Map: Small multiples for time series: Show how alternative graphic How can we take an alternative look representations can stimulate & support at this map and its data to improve visual thinking about spatial patterns, our understanding about this event? relationships, & trends Applies the geoviz approach to one of the most well-known maps in the history of cartography “Napoleon’s March on Moscow” (Minard)
Napoleon’s 1812 Russian campaign: Change perceived by succession of maps depicting the successive events. Troop movement shown in a single flow; size of army encoded in width of bands. B=position of troops Time inherently illustrated. C=adds overview of campaign up to date Temperature diagram linked to the retreat path.
Animation to represent time: 2-d chart: 3D view of the size of Napoleon’s army: Space-time cube of Napoleon’s march: Variations introduced to represent an event are deduced from real movement on the map Animated map provided on author’s website http://www.itc.nl/personal/kraak/1812
Example of visualization not influenced by traditional cartographic rules: Reveals info not shown in original map: (1) 2 battles took at Pollock (2) Napoleon stayed in Moscow for a month before returning west Column height = # of troops (colour could be added to represent temperature as well) Alternative use of space-time cube: temperature vs troops vs time Interactivity necessary to look at 3-d map from different views (to deal with occlusion) Could benefit from interactive options – sliders on each axis to highlight time period or location Critique: The Project: Goal of paper: Geographic visualization: designing manipulable maps for exploring temporally varying Presents the 2nd part of a 2-part project Light intro to the geoviz – basic georeferenced statistics (1) Overall project goal: to understand the To present a geoviz interface techniques prototype to support domain experts Stresses that new & different views can cognitive aspects of map use & to develop reveal new insights appropriate geoviz tools with an emphasis in the exploration of time series No in depth description of techniques on data exploration. multivariate spatial health data No discussion of pros/cons Project: (2) To understand the cognitive aspects Could have made better use of small Part 1: spatial pattern analysis of map use multiples Part 2: spatiotemporal analysis
HealthVisB Interface: HealthVisB Interface Prototype: HealthVisB Interface Prototype: HealthVisB : Provides spatiotemporal analysis methods Authors use a single integrated display with Single integrated manipulable map animation or discrete time steps 1 or 2 variables: Mortality cause Acknowledge that small multiples could also (Risk factor) be used but they don’t use it as: Data classification schemes: Disaggregation of info & small map size needed to 5-class diverging scheme fit many views on a page may make comparison of 7-class diverging scheme variables difficult 2-variable binary scheme – with focusing Can step through time or can use animation (VCR controls) 2-class binary scheme (crossmap) 7-class diverging scheme Mortality rates: blue=high, grey=low Risk: dark shades=high, light shades=low
Usability Study : Usability Study Results: Critique: Purpose: to investigate how spatial data exploration is facilitated (if at all) by the In-depth usability study of spatial data Conditioned Choropleth Maps geoviz tools (animation, time stepping, exploration by domain experts focusing). Good description of usability study protocol Method: observation with think-aloud Not very flexible in data classification (2, 5 or 7 protocol & system-generated logs classes only) Participants: 9 domain experts (doing Analysis would have been stronger if had also research on analysis of health-related Task: Time trend in heart disease Task: Comparison of time trend compared animation to small multiples data) between 2 diseases Only those users who used animation were able to identify Users preferred to use primarily No support for standardizing data Tasks: users were asked to look for spatial the subtle spatiotemporal animation or primarily time- trends over time pattern. stepping. Somewhat unclear in their discussion of results
Conditioned Choropleth Maps (CCMaps): Purpose of CCMaps: Motivation for conditioning: CCMaps: What is a choropleth map?: To support the generation of hypotheses to In many cases, we already know some of the Dependent variable: coarsely partitioned factors contributing to the spatial variability of a A map in which data collection units are shaded with an explain the spatial variation of a variable; a into 3 classes intensity proportional to the data values associated with variable. those units. common and important task for Region’s class indicated by colour: red=high, What we really want to figure out is what else gray=medium, blue=low What is conditioning? geographers. might be contributing to its spatial pattern? It partitions data for a variable of interest into subsets to 2 conditioning variables: also coarsely E.g., What causes the spatial variation of lung To get at this underlying spatial pattern, we need control the spatial variation of this variable that can be a way to remove or “control for” known variation. partitioned into 3 classes (low, med, high) attributed to explanatory (or conditioning) variables cancer mortality across the United States? Region’s class indicated by its location in a 3-by-3 Why? - our visual-cognitive system cannot do it matrix of panels What is a CCMap? for us. We can’t mentally remove known A choropleth map with conditioning components of variation and envision the Initial classification done using equal intervals (33% underlying patterns. of regions included in each class) Cognostics?: CCMaps Interface - Demo: Critique: Cognostics mentioned briefly as a possible Helps support the explanation of variability in CartoDraw way for choosing where to locate spatial patterns partitioning sliders Dynamic and interactive But they don’t really describe how their Good default colour scheme cognostic ranks different possible Also provides traditional statistical views partitions Limited to two explanatory variables They indicate that this is an area for future No flexibility in the classification of variables No guidance for choosing conditioning work variables No support for dynamic data standardization
What is a cartogram?: Key property of a cartogram: Cartogram Example – World Population: Cartogram Example – US Election results: Conventional maps only show data in relation to land area, not population or some other variable of interest. To be effective, a cartogram must be By intentionally distorting individual map regions so that recognizable. That is, a viewer must be able to their areas are proportional to some other input quickly determine the geographic area that is parameter this alternative information can be being presented! communicated more effectively.
Maps transformed in this way are called CARTOGRAMS.
Typical applications: social, political, & epidemiological.
An example of an effective cartogram. An example of a less effective cartogram. Still somewhat recognizable as the USA.
Motivation behind CartoDraw: Goal of CartoDraw: CartoDraw Method: Scanline placement: Drawing cartograms by hand is a laborious task Step 1: intelligent decimation of vertices Automatic versus Interactive The goal of the authors is to produce a Vertices with no noticeable effect on polygon shape & that Automatic placement of scanlines: Interactive placement of scanlines: Previous computer-aided techniques: dynamic “on the fly” cartogram drawing don’t belong to multiple polygons removed Are prohibitively slow method that preserves global shape to Step 2: heuristic, scanline-based incremental E.g., Kocmoud & House report a time of 18 hours for a st medium-sized map (744 vertices) create a recognizable cartogram. repositioning of vertices (global 1 , then interior) Produce significant deformation of global shape Heuristic: area error function and shape error function For each scanline, the repositioned vertices produced are General technique for “trading off shape and only accepted if the area error and shape error are below the area adjustments” – wider applicability than just specified thresholds geoviz. Step 3: fitting undecimated polygons to the decimated mesh to get the output cartogram
Comparison: efficiency and area error Comparison: population cartograms Critique: Future of Geovisualization: Dynamic generation of cartogram The demand for geovisualization 2 seconds for US population data techniques will continue to increase as the 19 sec for 90,000 polygons sheer volume of geospatial data continues Minimizes error in global shape to to grow exponentially and the popularity of promote recognition spatial information increases due to tools Total area error Polygon error – sorted Tobler’s population cartogram Kocmoud & House’s Highly sensitive to scanline placement – such as Google Maps. user defined scanlines better Not much guidance provided to the user for placing scanlines appropriately
Efficiency comparison Scanline-based method