| dc.description.abstract | Numerical ensemble models are commonly used in the analysis and forecasting of a wide range of environmental processes. Use cases include assessing the consequences of nuclear accidents, pollution releases into the ocean or atmosphere, forest fires, volcanic eruptions, or identifying areas at risk from such hazards. In addition to the increased use of scenario analyses and model forecasts, the availability of supplementary data describing errors and model uncertainties is increasingly commonplace. Unfortunately most current visualization routines are not capable of properly representing uncertain information. As a result, uncertainty information is not provided at all, not readily accessible, or it is not communicated effectively to model users such as domain experts, decision makers, policy makers, or even novice users.
In this research the state of the art of uncertainty visualization is reviewed. Building upon the literature review, a conceptual framework is developed for visualizing uncertainty information in an interactive web-based mapping environment. Attribute uncertainty in ensemble datasets is quantified using various uncertainty metrics, such as the standard deviation, inter-quartile range, or interval probabilities derived from a cumulative probability density function. The metrics are well defined and mapped to a representation which uses dynamic circular glyphs to represent uncertainty. The application was developed in line with the data state model and makes the uncertainty visualizations available in an online mapping and visualization environment (UVIS). The visualization routines incorporate aggregation (upscaling) techniques to adjust the uncertainty information to the zooming level, resulting in a new and visually pleasing bivariate display in which both attribute value and uncertainty are embedded.
The web-application was presented to groups of test users of varying degrees of expertise. The interface and the visualizations were usable even for non-expert users, and the dynamic circular glyphs were found to be an effective way to identify areas of uncertainty, quantify uncertainties, and to estimate probabilities of attribute value intervals occurring. | |
