Geography and Computational Science

n many U.S. campuses, a rising tide of scholarly activity, called computational science , is being recognized by academic administrators through the establishment of research centers and degree programs (Rice 1994). This nascent discipline has emerged from several traditional fields, including for example, cosmology, geography, and pharmacology, that contain subareas in which computing is the primary tool used to pursue research questions. The common thread that ties together such seemingly disparate activities is a shared focus on the application of advanced computing to problems that heretofore were either intractable, or, in some cases, unimagined. The purpose of this paper is to sketch out the opportunities for geography that lie at the intersection of computational science and geographical modeling. At the outset, it is important to draw a distinction between computational science and computer science. Though they are related, computational science is concerned with the application of computer technology to create knowledge in particular problem domains. Sameh (1995:1), for example, envisions computational science as a multifaceted discipline that can be conceptually represented as a pyramid with an application field (e.g., geography) at the apex. At the four base corners are: (1) algorithms, (2) architectures, (3) system software, and (4) performance evaluation and analysis. O’Leary (1997: 13) articulates a different but related view in which an interdisciplinary and team-oriented computational science rests on the foundational elements of a particular science or engineering discipline, together with mathematics, computer science, and numerical analysis. Despite some apparent variability in these (and other) views of computational science (cf., Stevenson 1994, 1997), they share a consistent unifying principle: the use of models to gain understanding. While most traditional views of science hold sacred the dyadic link between theory and experimentation, computational scientists have expanded this view to include a separate but equal role for simulation and modeling (Figure 1). Geographers have been specifying and testing models for decades (Hägerstrand 1967; Chorley and Haggett 1969) and are well positioned to make significant contributions to interdisciplinary computational-science teaching and research initiatives. Despite substantial progress (Longley et al. 1998), however, in many cases, the use of models to support scientific investigations and decisionmaking has been hampered by computational complexity and poor performance. In the next section of this paper, I describe the underlying causes of this computational complexity. Then I focus discussion on the development of synergistic interactions between geography and computational science, placing a particular emphasis on the use of new computer architectures to improve the performance of models and foster their application in an enlarged set of scientific and policy contexts. I next describe how emerging computational technologies will begin to alter approaches to the development of geographical models. Because visualization is an important element of computational science, enabling researchers to gain insights from the results of their numerical computations, in the final section of the paper, I initiate a discussion about how a form of advanced visualization, immersion, is creating a need to rethink aspects of cartography.