Reviewed by
Andreas Koch
University of Salzburg
According to the editors, this book "was conceived as a beginning point to explore the current state of spatial/temporal simulations that integrate human behavior and environmental factors as applied to decision-making in spatially referenced dynamic environments" (p. vii). Despite this remark, the volume indeed has a pioneering work, published in 2002 and also edited by Randy Gimblett "Integrating Geographic Systems and Agent-based Modeling Techniques for Simulating Social and Ecological Processes" (Gimblett 2002). In this context, the volume has at least two implications. From a theoretical perspective the challenge is to create geosimulation models which take the mutual relations of social and spatial systems - each as independent entities - into consideration. This leads to increasing methodological efforts to link Geographical Information Systems and Simulation Software Tools. Contemporary results of this concern are presented in this volume in a great variety - and very convincingly. From an applied perspective this book is primarily addressed to spatial planners, environmental managers and geographers; those who are interested in programming details or related technological issues may be somewhat disappointed. The book is structured in four thematic paragraphs with 23 chapters in all; for this review I have chosen a selection of representative articles from each paragraph. In the first (three papers), entitled "Wayfinding and Prediction of Behavior in Visitor Landscapes", Elands and Van Marwijk present an approach which tries to operationalize environmental perception of individuals via environmental values. The expression of "nature experience" is translated into "use value", "perception value", "narrative value", and "appropriation value". Use and perception values are represented as "attraction parameter" and "crowding parameter", respectively, and both parameters are created by empirical data (questionnaires and GPS). They represent, based on a simulation study and an empirical recreation study, visitor behavior in the Dutch Dwingelderveld National Park. The technique applied in this case is multi-agent simulation with autonomous, cognitive agents, using a software tool called MASOOR (Multi Agent Simulation Of Outdoor Recreation), which is used by several other contributors to this volume. Their intention is to elaborate multi-agent simulation (MAS) as a means of bringing together different experts in recreational science and practice. Geographical Information System (GIS) is used in order to visualize the environment that consists of a network of qualitatively different paths in this study.
In the next paragraph, which is dedicated to "Monitoring Systems for Capturing Visitor Patterns of Use" (four papers), both Xia/Arrowsmith and Loiterton/Bishop deal with techniques and technologies for counting and tracking movements of visitors in recreational areas in space and time. While the first duo present a comprehensive overview of different techniques like infrared sensors, GPS and monitoring systems to collect and analyze geo-referenced data, the second pair focus on the necessity to calibrate and validate agent behavior by presenting a case study of spatial patterns of visitor behavior in the Royal Botanic Garden in Melbourne. One interesting issue here is that they try to implement agent feeling algorithms like "exhaustion" or "boredom". Here too the idea is to integrate practitioners of spatial planning and protection management into the process of model creation and model specification. One methodological claim in this regard is to reduce psychological factors of human-environment interactions.
The problem of operationalizing spatial visibility is addressed by Skow-Petersen and Snizek. They criticize the common dichotomous differentiation of how agents perceive their spatial environment and hence propose the extension of the Boolean visibility approach by introducing a so-called Probabilistic Visibility approach, i.e. a fuzzy parameter depending on landscape settings like topography and vegetation height has been developed. The benefit is that it equips agents with a more realistic capability of perceiving environmental features while interacting with them and with each other.
Simulating spatial behavior of individual recreationists in the Austrian Danube Floodplains National Park using empirical and GPS data to feed the MASOOR simulation platform, in order to develop a decision support system for park managers, is presented by Taczanowska, Muhar and Arnberger. In so doing, they define three types of agents that are derived from differences in trail preferences. They too aim at representing the empirical reality in a model to evaluate potential management decision scenarios.
The most voluminous paragraph - according to contributions 12 papers in all - is related to "Computer Simulation and Management of Visitor Landscapes". Examples of MAS applications about protected natural areas in Australia, Alaska, California, Austria, the Netherlands and other regions, as well as a huge range of methodological approaches like discrete-event simulation (by Lawson, Hallo and Manning), choice models (by Hunt), Bayesian Belief Networks (by Sinay, Smith and Carter) and foremost agent-based modeling techniques, are presented and discussed extensively. While the first two paragraphs are more related to the application side and their potential for managing protected natural areas, this third paragraph is strongly related to issues concerning methodological questions.
Pröbstl, Visschedijk and Skov-Petersen argue on a meta-level when assessing the use of ABMs from a managerial perspective. They identify three significant themes which influence the choice of ABMs: cost-efficiency, efficient management and monitoring, and public awareness and acceptance. A crucial point in the adoption of ABMs depends on the question whether or not this technique is suitable for communicating the management goals in a participatory planning process. The degree of flexibility of the computational tools used to extend the software program gradually and the possibilities of visualizing artifacts are of prime interest.
These instrumental capabilities of ABMs are presupposed in the paper of Jochem, Van Marwijk, Pouwels and Pitt. They first describe the conceptual framework of the MASOOR model very extensively, which makes the contribution an essential starting-point for understanding other presentations which also use MASOOR. The theoretical approach which the authors apply is transactionalism, i.e. besides the interactions between different visitor agents there are different qualitative transactions with a complex environment. The representation of both the environment and agent interactions with this environment is modeled with sophistication in MASOOR. In contrast with this positive assessment it must be said that even in MASOOR (as in all other model approaches in this volume) space is implemented as a static and passive reference. This is partly at odds with the theoretical claim which argues for space being an independent property and an active part in the mutual relationships of agent-space interactions. There are, however, no models presented in this book which conceive visitor landscapes as a changing factor and therefore as a spatial behavior-influencing determinant.
Greater flexibility is given to the activity types (walkers, bikers, joggers, etc.) and behavior types - in this case Jochem and colleagues distinguish between the "optimal" chooser, who maximizes their preferences, and the "fuzzy" chooser, who satisfies their preferences with some variance of attitude. The authors conclude with a critical evaluation of the modeling potentials and future research, among which is the interesting and important issue of "modeling implications of environment on recreational behavior".
In the last paragraph of this volume, the editors draw a balanced conclusion from the collection of results and rationale. Marceau then poses the question: "What can be learned from multi-agent system modeling?" The placing of such a contribution at the end of a book with over 450 pages is very convincing. Her explanations are appropriate for those readers who are not very familiar with ABM techniques and for those who are keen on searching for theoretical explanations that confirm the use of ABMs. Referring to complexity theory and post-normal science and their combined use, respectively, Marceau opts for exploration as a significant research perspective. The role of scientists, she argues, "should shift from attempting to predict and control these systems to understanding their possible outcomes" (p. 418). In order to achieve this goal she claims that a MAS model "should be itself a complex adaptive system and such a purposeful model has not yet been built" (p. 419). In this precise sense an important postulation is explicitly mentioned: "to fully incorporate the environment as an essential component within a MAS", and this, among other things, means to define space "as an active entity with its own processes that can change its own state, independent of the activity of the embedded agents" (p. 419).
I would like to conclude as follows: the book offers great opportunities to applied scientists in the field of planning and managing protected recreational areas like national parks. The idea of popularizing the use of ABMs and similar techniques by offering standardized and flexible simulation platforms seems to be a convincing strategy. Moreover, attempts to link GIS and MAS are a sufficient factor for developing a realistic interplay of social and spatial systems. On the other hand, the lack of sociological theories, social interaction, for instance, along with a strong emphasis on theories of perception, behavior and psychological sensitivities, is possibly too single-edged. This is true of a similar lack of geographical theories of space or spatiality. For instance, "spatial interaction" is taken for granted in several chapters. Because of the great variety of contributions with different intentions, however, this volume can be recommended to JASSS readers.
References
Return to Contents
of this issue
© Copyright Journal of Artificial Societies and Social Simulation, 2008