in press (following revisions) for Conference on Visualization and Data Analysis 2005 (part of the IS&T/SPIE Symposium on Electronic Imaging 2005), 16-20 January 2005, San Jose, CA USA A Typology for Visualizing Uncertainty Judi Thomson a and Beth Hetzler a and Alan MacEachren b and Mark Gahegan b and Misha Pavel c* a Battelle Memorial Institute, PNWD, Richland, WA, USA; b Penn State University, State College, PA, USA; c Oregon Health and Sciences University, Portland, OR, USA; ABSTRACT Information analysts, especially those working in the field of intelligence analysis, must assess not only the information presented to them, but the confidence they have in that information. Visual representations of information are challenged to incorporate a notion of confidence or certainty because the factors that influence the certainty or uncertainty of information vary with the type of information and the type of decisions being made. Visualization researchers have no model or framework for describing uncertainty as it relates to intelligence analysis, thus no consistent basis for constructing visualizations of uncertainty. This paper presents a typology describing the aspects of uncertainty related to intelligence analysis, drawing on frameworks for uncertainty representation in scientific computing. Keywords: uncertainty, framework, geospatial information 1. INTRODUCTION Analysts make allowances for uncertainty in data on a daily basis. They understand that the information upon which decisions are based is rarely absolutely true, and that the resulting decisions are based on best guesses and assumptions. In most situations, uncertainty is not a life threatening issue- people learn to cope with potential variants in the information they are given and make their decisions accordingly. However, where the information is used in high risk decisions affecting large numbers of people, a comprehensive understanding of the uncertainty in the data becomes far more important. The thesis of this paper is that those analysts who have consistent, comprehensive representations for the multiple uncertainties associated with data, and understanding of the impact of those uncertainties on decisions, make better decisions. Implied in this thesis is the notion that the relevance of particular types of uncertainty changes depending on the task or decision facing the analyst. For instance, suppose an analyst is trying to determine the immediate impacts of a chemical explosion in a building. The analyst would be quite concerned about accurately knowing the number of people in the building, but possibly less concerned with accurate information about the wind speed and weather patterns for the area around the building. On the other hand, an analyst tasked with determining longer term impacts of the same explosion might be more concerned with accurate information about wind and weather patterns. Uncertainty in the same data affects the different analysts in different ways. Mechanisms for flexible representations of the uncertainties important to the analytical process are needed and to provide those representations, an understanding of the different facets of uncertainty and how they affect analysis is required. Significant work has been done to understand and approach inaccuracies in measures such as location, time of day, etc., which allows rudimentary understandings of uncertainties, but are insufficient for representing the myriad uncertainties encountered by analysts. In addition to uncertain measures, analysts are concerned with abstract * Further author information: (Send correspondence to J.T.) J.T.: E-mail: judi@jthomson.net, Telephone: 1 509 375 4438 B.H.: E-mail:beth.hetzler@pnl.gov, Telephone: 1 509 375 6690