16 August 2010
Geographic information is everywhere these days. It comes from GPS-enabled mobile devices, economic, demographic, health and other government statistics, satellites and other sensors, RFID tags on packages, and many other sources. Place names in text are also being extracted and translated into coordinates automatically. Maps that depict this bewildering array of topics pop up on the web every day. But, how much of the geographic information is used effectively, or at all, and how many of the maps are useful, or even usable?
We at Penn State’s GeoVISTA Center pursue a range of geographic information science (GIScience) that helps people figure out how to leverage geographic information efficiently and effectively – and interactive web mapping is a big part of our work. Our research involves fundamental information science focused on processing, analyzing, displaying, and enabling exploration of large volumes of spatial and spatiotemporal data. It also involves research to understand human acquisition of spatial information from maps, mobile devices , and the environment, plus research to enable human use of associated geotechnologies. In teaching, our goal is to help students develop knowledge and skills to successfully build and apply geographically-aware analytical and presentation tools.
A key focus of our research and education activities is geovisualization method and tool development. We approach geovisualization as much more than helping users “see” geographic information. The goal is to actively explore, hypothesize, and construct knowledge. Our faculty and students have used Adobe Flex and Flash to quickly implement creative geovisualization ideas, to support experiments that explore the perceptual and cognitive issues in dynamic tool design, and to enable larger tool development projects. One example of the latter is our Flash-based MapStats for Kids learning site developed for the U.S. FedStats program (live since spring 2005). The Network Challenge (one of five educational games in the site) is illustrated below.
Over the past few years, we have taken advantage of Adobe Flash and Flex to create map-oriented, cross-platform rich Internet applications. Below, we highlight one Flash-based student project, the development of our FlexLayers mapping tools, and a new prototype, geographically-enabled document query interface that incorporates FlexLayers.
DC CrimeViz is a Flash-based, data-driven web map application for exploring geo-located crime data (data that is published online by the District of Columbia). Three grad students—Kevin Ross, Rob Roth, and Craig McCabe — produced the initial application as part of course work in a Dynamic Cartographic Representation course last spring. The current version includes a suite of space-time analysis tools that extend the value of the data by supporting quick exploration of crime patterns in place and time and drill-down to access details. The figure illustrates a composite time option, which shows daily totals aggregated for the whole time period, a composite week that exhibits a distinct mid-week, Wednesday peak of arsons, along with drill-down to one incident report.
A key component of our web-based geovisualization activities that ActionScript supports effectively is the generation of custom, interactive geographic features on maps in the client user interface. We developed FlexLayers specifically to support direct interaction; Flex provided a great platform for this database-driven, on-the-fly, interactive mapping.
Health GeoJunction is one example of the geovisualization tools we are developing in Flex using FlexLayers. It is designed to support quick, place-time-concept filtering of health science-related documents. Health GeoJunction is implemented with a client server architecture that uses the open source PostgresSQL database with its PostGIS extension for handling spatial data, Geoserver as a spatially-aware connection between database and client, and the client built completely in Flex. The application supports Open Geospatial Consortium Standards (OGC) web map and web feature service specifications. Behind the scenes on the server, text-processing algorithms (developed by Senior Research Associate Dr. Ian Turton) identify and disambiguate geographic names and provide the geographic coordinates for each place mentioned. Additional tools identify keywords, authors and dates for cross-referencing documents.
Health GeoJunction’s goal is to “find the document needle in the haystack.” Our initial prototype uses subsets of abstracts from the National Library of Medicine PubMed repository, a citation database of biomedical articles. For example, a PubMed search returned approximately 6,000 abstracts related to avian influenza. With Health GeoJunction, a user can quickly answer questions such as, “Which papers published in 2006 are about ‘avian influenza’ + ‘disease outbreaks’ and about Thailand and neighboring countries?” By manipulating the timeline sliders and making three clicks (one on the map and two in the tag cloud), a user can drill down to find the six publications meeting the specification. Results below include the “geographic footprint” and abstract for one of the six abstracts.
Above, the tag cloud is sorted by frequency of two-word phrases for the entire abstract set for avian influenza. The lower tag cloud shows the place-time-concept filtered results. The abstract view lists date, title, and authors and provides a link to the PubMed entry through the Document Details tab.