� Amap Interface for Exploring Multivariate Paleoclimate Data

We begin with an abbreviated review of recent approaches to the display of three or more geographically-referenced data variables from the literatures of statistics, computer graphics and cartography. In comparison we describe a method we have developed for exploring relationships among multivariate paleoclimate data pro­ duced by a global circulation model at Penn State's Earth System Science Center. Key features of the interface include 1) display of multivariate data in two-dimen­ sional, planimetrically-correct map formats ; 2) optional display of a single map on which up to three variables are superimposed or four maps among which up to four variables are juxtaposed ; and 3) the ability to quickly revise data variable selections, to focus on subsets of the data and to experiment with different combi­ nations of point, line and area symbols . CONTEXT This paper discusses methods for displaying three or more geographically-refer­ enced data variables . In particular we are concerned with methods appropriate for exploratory data analysis in Earth system science. Earth system science is an inte­ grative, multidisciplinary approach to understanding "the entire Earth system on a global scale by describing how its component parts and interactions have evolved, how they function, and how they may be expected to continue to evolve on all time scales" (Earth System Sciences Committee 1988) . Numerical modeling is an important method of formalizing knowledge of the behavior of the Earth system and for predicting natural and human-induced changes in its behavior . The volume of quantitative data produced by model simulations and earth observing satellites imparts a high value on effective graphical techniques for identifying potentially meaningful patterns and anomalies . Because the Earth system comprises many interrelated phenomena, the ability to display multiple data variables in a format that fosters comparison is particularly desirable . In the following we will first discuss several recent approaches to the display of three or more variables from the literatures of statistics, computer graphics and cartography. Then we describe an interface we have developed for displaying up to four paleoclimate data sets produced by a global climate model at Penn State's Earth System Science Center . Key features of the interface include 1) display of multivariate data in two-dimensional, planimetrically-correct map formats ; 2) op­