Simulation in spatial analysis and modeling

GeoComputation is an evolving research field with a primary focus on exploring new modeling paradigms and techniques derived from advances in computation for the goal of enriching geographic analysis of highly complex, and often non-deterministic problems (Longley, 1998). Agent-based models, cellular automata, fuzzy sets, genetic algorithms, and neural networks have attracted attention as useful research tools. In recent years, computational geometry, interactive exploratory data analysis and mining, numerical modeling, and many other research themes have been entering the scope of GeoComputation (Couclelis, 1998; Xie & Ye, 2006). Simulation in spatial analysis and modeling has been one of the key approaches of many researchers of GeoComputation. A dynamic geographic simulation represents the spatiotemporal dynamics of a physical system, a human system, or a coupled human–physical system that incorporates multiple stochastic and dynamic processes with the aim of solving or understanding problems or systems with multiple and often conflicting objectives (Maxwell & Costanza, 1997; Batty, 2005). Solving these multi-objective problems presents a significant challenge and may have to rely on alternative evaluations of optimal solutions (Bennett, Xiao, & Armstrong, 2004). As Page (2003) so cogently argues: ‘‘. . . our models become better, more accurate, if they make assumptions that more closely match the behavior of real people . . . ’’ In addition to the model inputs, algorithms, assumptions, and outputs, the human dimension (modeler) is the most important fifth element to keep honesty toward the sensitivity test and to establish credibility in complex modeling and simulation (Clarke, 2005). The diversifications and complexities embedded in geo-simulations become the main theme of this special issue. This special issue draws together seven papers that were presented at the GeoComputation 2005 conference, held August 1–3, 2005 in Ann Arbor, Michigan and jointly hosted by the School of Natural Resources and Environment at the University of Michigan and the Institute for Geospatial Research and Education at Eastern Michigan University. Xiao et al. reviews recent development in evolutionary algorithms and presents a conceptual framework for multi-objective spatial decision making. This framework supports the generation of alternatives for solving spatial optimization problems using evolutionary algorithms and provides visual tools to evaluate multi-objective spatial decisions. Duh and Brown tackle multi-objective spatial allocation problems by integrating several techniques in an innovative method called Knowledge-Informed Pareto Simulated Annealing. Duh