Landscape Models and Explanation in Landscape Ecology— A Space for Generative Landscape Science? Daniel G. Brown University of Michigan Richard Aspinall Macaulay Institute David A. Bennett University of Iowa Further development of process-based spatial models is needed to facilitate explanation in landscape ecology. We discuss the dual modeling goals of prediction and explanation and identify challenges faced in explaining land- scape patterns. These challenges are especially acute in attempts to explain patterns that result from complex adaptive systems. We compare examples of two process models used to describe landscape changes in Yellowstone National Park as a consequence of predator-prey interactions. Generative landscape science is offered as a complementary approach to explanation, combining models of candidate processes that are believed to give rise to observed patterns with empirical observations. Key Words: complex systems, spatial modeling, spatial pattern. A central theoretical concern in landscape ecology is understanding the interaction between observed landscape patterns and a di- verse set of social and environmental processes. The domain of landscape ecology in the United States has been defined, primarily, around the causes and consequences of spatial pattern, ex- pressed primarily in terms of biotic and abiotic processes (Naveh 1982; Risser, Karr, and For- man 1984; Turner 1989; Nassauer 1995; Mlad- enoff and Baker 1999; Turner, Gardner, and O’Neill 2001). Although Turner, Gardner, and O’Neill (2001, 7) ‘‘do not think it necessary to include a human component explicitly in the definition of landscape ecology,’’ we take a more inclusive view, derived from the European or- igins of landscape ecology (Naveh 1982), that concerns itself not just with the interaction of landscape pattern with biophysical processes, but also with human actions. This view would seem essential to any attempts to make claims about the implications and/or appropriateness of management interventions, and to contribute to the emerging ‘‘integrated land science’’ de- scribed by Klepeis and Turner (2001). Models of landscape change have been used extensively to study the effects of both natural and human processes on landscape patterns since the emer- gence of landscape ecology in the United States during the 1980s (Baker 1989). An important thread of work in geographic information science (GIScience) deals with spa- tial models that can formally represent patterns and processes (e.g., Goodchild, Parks, and Steyaert 1993). Pattern-based models focus on describing spatial distributions and identifying correlates of those distributions (e.g., Guisan and Zimmermann 2000), whereas process- based models describe process using a number of different representations of mechanisms. Among the variety of purposes for which land- scape models are built, two are most important in driving the nature and structure of models: (1) to make inferences about how and why land- scapes change, sometimes (not always) with the intent to produce more favorable outcomes, and (2) to predict future landscape states and pat- terns. These two goals are difficult to separate from one another. On the one hand, we can hardly expect to make reasonable predictions about future landscape patterns if we do not have reasonable explanations for how they The Professional Geographer, 58(4) 2006, pages 369–382 r Copyright 2006 by Association of American Geographers. Initial submission, March 2005; revised submissions, January 2006; final acceptance, April 2006. Published by Blackwell Publishing, 350 Main Street, Malden, MA 02148, and 9600 Garsington Road, Oxford OX4 2DQ, U.K.