Eva-Maria Stephan

Visually Exploring Mutual Relationships in Environmental Data

Abstract

Typical visualization methods in GIS are restricted to static 2- and 3-dimensional base maps which may be overlaid by vector graphics; visualization mainly serves descriptive and presentation purposes, where the phenomena represented in the data are already known, and are communicated clearly to others through a visual presentation. Beyond that, GIS do not support the use of graphics for viewing and investigating several continuous and/or contiguous data layers simultaneously. There is a great need to make dynamic graphics part of the (data) analysis process and support Exploratory Visualization (EVIS). EVis in GIS requires to integrate concepts of direct manipulation, data flow and multi-dimensional visualization techniques for spatio-temporal data. Such techniques can be derived from the field of scientific data visualization but must be further developed for environmental data analysis.

The poster (get poster A, poster B) describes graphical techniques for exploratory visualization of environmental data and introduces a prototype system, called DataScaping. DataScaping is an interactive visualization system for viewing and for exploring mutual relationships in environmental data. Surface data is presented graphically, but other than than photorealistic landscapes "all kinds of data-surfaces" can be generated, browsed and manipulated interactively. The introduced graphical toolbox includes data visualization by 3-d surface rendering and animation, as well as statistical plots. Main emphasis is put on the interactive manipulation of views through a graphical user interface (GUI), and the dynamic linking of data in multiple windows. Requirements for interactive manipulation include the change of viewing parameters (perspective, position of object, animation time) and object (exaggeration, color, certainty level, resolution, data values). The dynamic linking, based on dynamic data referencing, provides a link between various representations, e.g., statistical images and pseudo-realistic surface views.

In the presented approach of exploratory visualization for environmental data, graphics are used for directed or indirected search, when the data analyst does not (always) know what he is looking for, and visualization helps him to find patterns and relations in the data and to gain understanding and insight in to the overall nature and relationships of multivariate data. The presentation in gridded or rendered (and animated) surfaces provides an intuitive metaphor for exploring spatio-temporal data. In addition, interaction and selection tools are developed, which are necessary to visually explore data with higher dimensionalities, i.e., multivariate data sets.

Exploratory visualization with the Datascaping prototype system is a promising new spatial data analysis tool. Presented techniques for the investigation of environmental surface data offer potential for better exploring interdependencies and characteristics of multivariate data, by providing the possibility to intuitively browse multiple heterogeneous data layers simultaneously. Various examples, such as the interactive exploration of a temporal glacier data set, or exploring two differently interpolated temperature data sets are given in the poster (get poster A and poster B). The image below shows a screenshot of a typical DataScaping session.

The image shows a screenshot of a typical DataScaping session. Relationsships between a precipitation, a mean temperature and a DEM data set can be explored in multiple linked windows (data resolution: 1 km raster). The geographical location of the selected values in the precipitation histogram plot (lower left) are highlighted in the color-coded temperature surface (DEM with draped temperature map, upper right). The upper left image shows the precipitation data set as color image with the posssibility to interactively query neigbor values and generate cuts and draw profiles. The lower right is a scatterplot of the two attributes.