CyberGIS: What Still Needs to Be Done? Michael F. Goodchild University of California Santa Barbara
Progress to date Interoperable location referencing coordinate transformations geocoding addresses point-of-interest databases 34 deg 24 min 42.7 seconds north, 119 deg 52 min 14.4 sec west 236150m easting, 3811560m northing, UTM Zone 11 Northern Hemisphere National Grid reference 11SKU36151156 909 West Campus Lane, Goleta, CA 93117, USA Mike Goodchild s house
Standards Live access: WMS, WFS, WCS Server GIS service-oriented architectures Metadata OGC, ISO Semantic interoperability INSPIRE
Engagement Citizens as both producers and consumers enabled by standards, GPS, cartographic software neogeography OpenStreetMap and Haiti
So why the fuss? Why cyber geographic information system? why not cyber geriatric information system? Four points all represent impediments all call for fundamental research
Location as common key The stack of layers
But in reality Spatial databases are organized as layers horizontal integration not vertical property z about all places rather than all properties about location x tell me everything about location x overlay must be invoked explicitly graphical overlay or topological overlay many mashups are merely graphical overlay
The spatial join Using location as a common key to link tables All location references are subject to uncertainty measurement error vagueness in feature identification indeterminate limits The probabilistic join
Multiple attribution Shapes ESRI ~~~ USGS Names D aowaga Lake Tahoe Sierra Lake Types +Water Body Plate carre - Lake - Reservoir
The true spatial join is still elusive Much better techniques needed especially to deal with vague, vernacular references in text, speech, human discourse generally beyond formally defined coordinates well-defined metrics of confidence We are a long way from realizing the fully interoperable vision
The functionality of cybergis What is the appropriate level of granularity of cybergis functions? How many functions are there? 542 in the ArcGIS 9.3.1 toolbox How to navigate among them? 18 top-level categories vaguely defined, overlapping Analysis, Spatial Analyst, Spatial Statistics, Geostatistical Analyst
ArcGIS ModelBuilder
Requirements A standard set of functions interoperable across all servers defined granularity an atomic level in reality functionality is determined in part by legacy and non-interoperable hidden from the user where appropriate
The spatial join What is the correlation between two variables? 1. identical support, e.g. congruent rasters 2. different support, e.g. non-congruent polygons Analysis requires that both data sets share the same support Case (2) requires a GIS operation resampling, overlay, point-in-polygon, etc a spatial join but the operation can be invoked automatically
What is needed? Basic research on the functionality of GIS and consequent standards requires an exceptionally comprehensive knowledge of applications across all domains of science Search and discovery for services supported by appropriate metadata as we have done for data
Selling the vision Why is cybergis important? because it enables new applications, new discoveries possibilities that were not realized before Has the case been made? or is this a matter of faith? We need a set of compelling examples of what could not be done without cybergis
The GIS analog Where are the examples of discoveries that could not have been made without GIS? a FAQ of the early 1990s today compelling examples exist in virtually every application domain even in history Anne Knowles and Gettysburg
What is cybergis? GIS on steroids high-performance computing distributed computing the geospatial Web Google Earth in 2005 a primitive GIS needing better analysis? or something entirely new answering new questions implementing new use cases
The cyberinfrastructure vision Atkins: A third kind of science beyond theory and empiricism intensely computational studying complex systems emphasis on simulation agent-based modeling cellular automata
Double-negative science Confirming a hypothesis by rejecting a null hypothesis an effect exists because the data are inconsistent with a non-existent effect Complete spatial randomness (CSR) events are equally likely everywhere one event does not make other events more or less likely in its vicinity
Getis A, Franklin J 1987 Second-order neighborhood analysis of mapped point patterns. Ecology 68(3): 473-477
What makes spatial special? Anselin s two properties spatial heterogeneity spatial dependence the antithesis of CSR CSR can always be rejected out of hand acceptance of CSR is always a Type II error acceptance of the null when in fact it is false Why was CSR ever adopted? because it was mathematically tractable in computational systems that requirement goes away
Towards a positive spatial science A suite of reasonable hypotheses about spatially distributed phenomena pattern templates Part of the functionality of a system for computational spatial science a new paradigm for spatial analysis enabled by cybergis not simply GIS on steroids Not a simple task equifinality
Four points and a summary Enabling location as a common key Defining GIS functionality Finding the breakthroughs A new kind of science CyberGIS is a powerful vision there are many research questions that still need to be addressed