The Visualisation of Uncertainty in Spatially-Referenced Attribute Data using TRUSTworthy Data Structures

This paper presents the use of hierarchical spatial data structures to visualise attribute and spatial uncertainty when using spatial information systems. This is demonstrated using spatially-referenced data from the New Zealand 2001 Census. Firstly, selected current spatial visualisations created to show uncertainty were assessed in an Internet survey, revealing that overall there is not many usable techniques and also users want to see the original information display as well as a display of uncertainty. This forms a background to the other purpose of this paper, to discuss hierarchical tree structures as a potential visualisation-of-uncertainty technique for socioeconomic data. The major question is: can an uncertainty be effectively communicated using data structures whilst simultaneously displaying the attribute information? Two such structures were compared and assessed in another survey: the region quadtree and the Hexagonal or Rhombus (HoR) quadtree, both variable resolution structures. These structures work in the following way: an area where attribute data is uncertain will show less resolution through the data structure than an area that is not, exemplifying a level of detail metaphor. Data structures can disseminate aggregated census data and standardise spatial units, thus reducing subjectiveness in boundary definitions for socioeconomic data. The principles behind possible implementations are presented through a demonstration program, TRUST (The Representation of Uncertainty using Scale-unspecific Tessellations). The results of the second survey revealed that the region quadtree and the HoR quadtree can display uncertainty information to a user, and also that the region quadtree is more visually complicated. The HoR tessellation ranked similar to common visualisation-of-uncertainty techniques like overlay and blinking pixels.