Increased demand by the public for diverse and quality recreation opportunities has placed considerable pressure on the natural resource and it's management. This problem is compounded by the general lack of understanding of interactions between people and forest recreation environments that result in wide variations in perceptions, expectations, patterns of choice and use. Emerging technologies, such as distributed artificial intelligence provide a mechanism to integrate advances in recreation research with a Geographic Information Systems (GIS) based environment. Distributed artificial intelligence provides the foundation for a modeling system to simulate the interactions between recreators and their environment. Despite the work done by many researchers in the development of object-oriented modeling and simulation languages, GIS, non-human agent design and simulations, no single system has been constructed to handle the complexity of goal-oriented autonomous human agents seeking recreational opportunities in natural environments. This paper describes a theoretical framework and a model for simulating hiker behavior in a natural environment using intelligent agents, discrete event simulation (DEVS) and GIS data. The results of hiker interactions are summarized to provide feedback on the implications for alternative recreation management planning.
Increased demand by the public for diverse and quality recreation opportunities has placed considerable pressure on the natural resource and it's management. This problem is compounded by the general lack of understanding of interactions between people and forest recreation environments that result in wide variations in perceptions, expectations, patterns of choice and use. In national forests of fixed size, this has led to unhealthy competition and conflict among recreation uses and between recreation and other resource commodities.
In order to meet the demands of the public, forest recreation managers must take a more proactive role in utilizing computer- based models that explain how physical, biological, social, and managerial attributes of forest environments effect recreational perceptions, uses and choices. Such models will allow managers to predict changes that will occur from management options, to improve resource inventory systems at a variety of scales, to achieve a fuller integration of recreation objectives with other components and products of the forest.
Development and implementation of models to aid recreation management has been hampered by the lack of critical information in three interrelated areas: an understanding of how changes in the physical, biological, social, and managerial attributes of forest recreation environments influence people's perceptions and enjoyment and visa versa; how people's recreation choice behavior interacts with and consequently impacts the attributes of forest recreation environments; how models of human perception and behavior in recreation environments integrate with concepts, methods, or results of research on biological and ecological systems. The integration of these concepts and methods with research on biological and ecological systems is the most perplexing and yet perhaps the most challenging.
As a means to gaining a better understanding the complexity of forest systems, researchers have sought the assistance of advanced computer-based technologies in the development of integrated modeling and simulation systems. GIS technology has led these developments because it has provided powerful databases for storing and retrieving spatially referenced data. Geographic Information Systems (GIS) technology provides powerful data bases for storing and retrieving spatially referenced data. Spatial information is stored in many different themes representing quantitative, qualitative, or logical information. GIS operators provide the means for manipulating and analyzing layers of spatial information and for generating new layers. Since it allows distributed parameterization, a GIS is useful for ecological models that need to explicitly incorporate the spatial structure and the variability of system behavior (Band and Wood 1988; Running et al. 1989; Band et al. 1991). A raster-based GIS represents spatial information as a grid of cells, and each cell corresponds to a uniform parcel of the landscape. Cells are spatially located by row and column and the cell size depends on the resolution required. GIS provides an excellent means of capturing real world data in multiple layers (3-dimensional) and resolutions (spatial scales) over time.
Due to the complexity of ecosystem management, interest has increased in using GIS for simulation of spatial dynamic processes (Berry 1987; Itami 1988; Gimblett, 1989; Deadman , Brown and Gimblett, 1993; Ball, 1993; Green et al. 1989; Costanza et al., 1990; Baumans and Sklar 1990; Sklar and Costanza 1991; Vasconcelos and Guertin 1992; Ball and Guertin 1992). However, GIS systems do not include procedures for handling time, they are designed to process entire arrays of data, and cannot easily address varying localized operations across the spatial grid (Berry 1987; Running et al. 1989; Band et al. 1991; Vasconcelos and Guertin 1992). Maidment (1991) summarized the major deficiency of GIS as the need to incorporate some form of time dependent data structure if it is to be capable of tracking the evolution of spatial phenomena.
Concepts for integrated modeling and simulation support systems have been around for some time (Oren and Zeigler, 1979), it is only recently that serious efforts have been undertaken to implement such concepts. Hendriksen (1983) recognized that there must be support for the model development process in addition to the runtime execution facilities provided by most simulation languages, at the time. Balmer (1986) and Marray et al. (1987) proposed employing the newly emerging tools of artificial intelligence and knowledge-based systems to support aspects of the simulation enterprise. Reddy et al. (1986) departed from traditional procedure-based simulation languages to incorporate declarative knowledge representation and reasoning facilities. Ruiz-Mier and Talavage (1989) addressed the requirement to integrate both declarative and procedural knowledge-representation schemes. There is now good evidence that knowledge-based simulation can serve as a strong foundation for complex systems design (Sevinc and Zeigler 1988; Rozenblit et al., 1990).
Zeigler (1976, 1984) has developed a uniform model description language based on the DEVS (Discrete Event System Specification) formalism. This formalism has been implemented in a number of environments and shown to be a workable code-independent model description language. One of the current implementations of the DEVS formalism is seen in DEVS-Scheme (Zeigler, 1990), an object oriented (OO), LISP-based simulation environment for hierarchical, modular models. The strength of this approach has been demonstrated with models describing landscape succession and fire growth (Vasconcelos and Zeigler, 1993; Vasconcelos, et al., 1993; Vasconcelos, Zeigler & Graham, 1993 ). Here the DEVS-Scheme Modelling and Simulation Environment is interfaced to the widely-used Geographic Resource and Analysis Simulation System (GRASS). Although LISP/OO implementations appear the most attractive candidates from a pure capability standpoint, other considerations, impact on implementation. Although many newer versions of LISP based languages have made significant improvements, there is the question of space/time efficiency of LISP execution.
Work by Marti (1990) and Catsimpoolas (1992) using case based reasoning in a DEVS like environment simulates intelligent behavior of military units across GIS linked terrain data. Their work provides good evidence that object oriented simulations can provide a strong foundation for complex systems design. Their work demonstrates that the discrete event, object simulations are "particularly suited to situations involving movement and interactions over terrain." The Rand Integrated Simulation Environment (RISE) capitalizes on merging goal-directed behavior with geographic information into the planning mechanism.
Other attempts have been made to incorporate information about landscape directly through the use of object oriented programming approaches. Models developed by Saarenmaa et al. (1988,1993); Stone (1990); Folse et al. (1989); Berry et al. (1993) and others were designed to describe the behavior of non-human agents (animal or insect predator-prey relationships) in natural habitats. Many of these approaches model individuals directly as autonomous entities. To simulate large populations, the individuals are replicated and simulated, including their interactions with one another and with the environment. While this work and others modeling individuals populations as autonomous agents is fruitful, to be useful for resource managers whose simulation requires realism, attention must be focused on a thorough examination of the behavior mechanisms of individuals and to more accurate spatial representations of the environment.
Distributed artificial intelligence (DAI) has been used to capture behavioral conditions and sets of intercommunicating among agents coexisting in a common environment. Research by Cohen et al., (1989) and most recently by Anderson (1994), Anderson and Evans (1991;1994), Saarenmaa et al.(1993) have contributed extensively to this area of single and multiple intelligent agent requirements . Most recent work by Anderson and his colleague have resulted in the development of an intelligent agent simulation test bed referred to as Gensim. Gensim was originally intended for testing and examining intelligent agent designs. They are currently working with a number of simple Gensim domains for the purpose of testing intelligent agents. To demonstrate Gensim's potential to natural resource management they have developed a simple domain that illustrates many of the needs of ecological modeling. The domain consists of a theoretical grid-based ecosystem that simulates predator-prey relationships. While most intelligent agent development in natural resources is still utilizing theoretical environments, the next step is to make direct linkages to real-world environments.
Despite the work done by many researchers in the development of object-oriented modeling and simulation languages, GIS, non-human agent design and simulations, no single system has been constructed to handle the complexity of goal-oriented autonomous human agents seeking recreational opportunities in natural environments. While GIS alone cannot solve this problem, it can provide two and three dimensional data of the real world, location and positional information on the earth's surface, and the capability to store semantic knowledge about qualities of landscape, crucial for human behavioral modeling. AI on the other hand can provide the added benefit of capturing and storing knowledge about sociological functioning of humans such as interactions traits with one another, preferred movement patterns and rates etc, and biological requirements such as rest, sleep, eating, drinking etc., as decision rules. These together incorporated into an integrated modeling and simulation system can provide a tool to aid the manager in making more informed decisions about forest recreation environments.
While it impossible in this paper to develop a system to examine how physical, biological, social, and managerial attributes of forest environments effect recreational perceptions, uses and choices, as a first step of our research we will describe the cognitive framework used in the design of the agents, requirements for these agents and the testing of a simple hiker agent in an environment whose landscape cells are populated with GIS data.
Several attempts in the last decade have been made to develop a common language with respect to terminology, theoretical and applied procedures for landscape assessment. Zube, Sell and Taylor (1982) strongly conclude that perceptual theories developed within the framework of the cognitive paradigm, provide a useful generalized basis for improving scenic assessment and recreation behavior models. They describe the contribution of the cognitive paradigm as an exploration into human meaning associated with landscapes or landscape properties where information is received by the human observer and in conjunction with past experience, future expectation, and sociocultural conditioning, lends meaning to landscape. Cognitive assessment models are extremely important in improving landscape assessment and recreation behavior models, as well as provide a framework for intelligent agent design because they evaluate landscape quality using an explicit theoretical structure describing underlying cognitive processes influencing human perception, preference for landscape and perceived opportunities to engage recreation activities.
One such model considered within the cognitive paradigm that has received the most attention for application in natural resource assessment is the Kaplan's 'Informational Processing Model'. Stephen and Rachel Kaplan (1982, 1989), in both their books 'Environment and Cognition', and 'The Experience of Nature', present a cohesive construct of human preference for landscape which is based on the view that humans have evolved with mental and perceptual capacities for processing visual information which is important to survival. A comprehensive body of work has been undertaken by the Kaplans' and their research associates, testing the components of the model as they relate to preference for landscape scenery (Kaplan & Kaplan, 1989). The components of the model, (ie. complexity, coherence, mystery and legibility), provide a conceptual framework for the assessment of the perceived value of landscape scenery. Studies undertaken by (Itami, 1979; Herbert, 1981; Brown & Itami, 1982; Wright et al. (1984); Gimblett, Itami & Fitzgibbon, 1985; Smith, 1986; Gimblett & Fitzgibbon, 1987, Gimblett, 1990; Lynch & Gimblett, 1992; Kroh & Gimblett, 1993) and many others have clearly demonstrated the usefulness and potential of the conceptual framework.
While, from a theoretical point of view this conceptual framework does provide a solid foundation from which a basic understanding of the process of human perception and movement through landscape can be acquired, there has been minimal work undertaken to apply this theoretical understanding to real world environments. Only recently, with technological advances in spatial mapping systems such as geographic information systems (GIS) and design of intelligent agents have new opportunities arisen to assess human perception and behavior issues and apply them to assess opportunities for acquiring a high quality recreation experience.
It has been shown that humans can perceive the landscape according to the theory proposed by the Kaplans' and there is now a general understanding of the biophysical variables that affect their perception and experience of the landscape. However, the question still arises as to how experience of landscape will change between or within individual landscape settings as one travels (hikes) through the landscape. This understanding is crucial if models are to be constructed for and application in natural resource planning and management situations and where intelligent agents can be of value.
To develop intelligent agents for natural resource applications under a cognitive paradigm as outlined above, it is essential that the requirements for these agents are first examined. Agents must; have knowledge of cartographic location and position with respect to the environment it is functioning in; have knowledge of it's own state i.e.. mind. physical well being, and be capable of obtaining states of all other agents in the simulator; have an internal and external clock with ability to keep track of how long it has been wondering around, how far it has travelled and must comprehend and adapt changes to the physical world such as sun changes, seasonal variations; be goal-oriented, opportunity and/or benefit driven to accomplish one's goals such as a researcher seeking more information on habitat ranges, hiker seeking solitude at a specific location, runner seeker physical challenge; have autonomous ability to move through a GIS-based environment, internalize necessary information and make decision as to where it should move next; have a record keeping ability to know where it has been and where it is going; have vision or perceptual ability to see the world around them, internalize what has been seen, represent the knowledge in some meaningful way and translate into some potential action. The ability to internalize biophysical and psychophysical information surrounding the agent is crucial to intelligent behavior.
While it is important to understand from an information seeking perspective how people react to these environments, likewise a manager must understand how hikers perceive the activities of others who share that environment. One focus of academic explorations of recreation behavior has been directed toward conflict and crowding in wilderness settings and the concept of carrying capacity (Graefe et al 1984). The concept of crowding in wilderness areas has been a topic of discussion for some time (Shelby and Neilson 1975). Many people who travel to wilderness areas do so out of a desire to experience an environment that appears unaffected by the activities of man. However, the activities of people in these environments, and their interactions with one another, may degrade the quality of their own experience.
A number of theories have been outlined to describe a recreators dynamic response to the environment. The expectancy theory (Graefe et al 1984) proposes that people develop expectations for what they will encounter, and then gauge what they actually experience against those expectations. In a backcountry hiking context, hikers may expect to experience solitude during their trip. Encounters with other hikers during that trip may run contrary to there expectations, leading to a feeling of dissatisfaction. The fundamental notion here is that as more people enter the wilderness and encounters between groups rise, the satisfaction of the individual recreators will be adversely affected.
Discrete event, object-oriented simulation can prove a useful tool in this investigation. If the movements and interactions of hikers can be modelled and related to surveys of hikers expectations in these environments, it should be possible to develop a tool to model the activities of hikers in a particular environment Models capable of modelling hikers movements and interactions, giving an indication of the interaction between the hikers and the environment, could prove to be a valuable tool for managers.
The discrete event simulation specification system, first introduced by Zeigler (1976) provides the means to specify a modelling system. The system contains a time base, inputs, outputs, state variables, and functions for determining new states and outputs given current states and inputs (Zeigler 1990). DEVS focuses on the changes in variable values at time segments determined by the rules built into the individual components of the system.
Discrete event simulation formalisms and Artificial Intelligence knowledge representation schemes form a powerful combination known as knowledge-based simulation (Zeigler 1990). In contrast to other knowledge-based simulation systems, DEVS-Scheme is based on the Discrete Event Simulation (DEVS) formalism, a theoretically well grounded means of expressing hierarchical, modular discrete event simulation models (Zeigler 1990).
Figure 1: System Entity Structure for the Model EF-LAND
The building blocks of the overall model EF-LAND are the atomic models shown in the system entity structure above, LAND, HIKER, GENERATOR, and TRANSDUCER.
Each atomic model contains the following components:
At least two state variables are present in an atomic model: phase and sigma. Phase represents the current state within which the model is operating. Sigma represents the time remaining until the next internal transition. Without an external event, the model remains in the current phase for the time specified by sigma. When sigma elapses, the model executes its output function, to send any required messages on specified output ports, followed by its internal transition function, to change the model to a new phase. If the model receives a message on an input port before sigma has elapsed, it executes its external transition function to change the phase of the model and reset sigma accordingly.
The atomic model of the hiker is designed to represent the actions of a hiker moving through a landscape and interpreting its experiences of that landscape (See Figure #2). The actions of the model are recursive over a five time unit cycle. At the beginning of the cycle, the model resting in phase 'passive' at a specific location on the landscape grid, represented by the corresponding state variable location. Upon receiving a go message from the generator, the model executes its external transition function to select a new adjacent location on the landscape grid. This is followed immediately by an output to the corresponding newly occupied land cell upon which it has arrived, informing the land cell that it is present, The hiker model waits for a reply from the land cell informing it of the local conditions. It receives a reply from the land cell telling it of other hikers within the eight cell neighborhood. If other hikers are present, the satisfaction level of the hiker, represented by the state variable satisfaction, is decremented. If other hikers are not present, the satisfaction level is incremented. At the end of the 5 unit time cycle, the hiker model makes an output to the land cell that it is occupying, telling it that it is leaving that location and transitions into state passive where it waits until the beginning of the next cycle when it receives another go message from the generator and moves again. As the simulation cycles, hikers move across the grid of land models interpreting the information that they obtain and incrementing or decrementing their satisfaction level. At the completion of the simulation, the hiker model receives an end message from the generator. Upon the receipt of this message, the hiker reports its name, its final satisfaction level, and a record of where it has traveled over the landscape grid to the transducer.
Figure 2: Pseudo-code for HIKER model.
The atomic model LAND functions in this exercise as raster cell in a GIS, containing information about the study area and relaying that information to the hiker agent as requested (See Figure #3). The land model is designed to provide the hiker who occupies it with a report of the local conditions, including those in its neighborhood. When it receives a message from a hiker that it has been occupied, the land cell queries the status of the cells in the neighborhood comprising the eight cells immediately adjacent to it to determine if any hikers are in the vicinity. If the model receives a message from a cell in its neighborhood telling it that it is occupied, then the state variable status is incremented by one. If another hiker has occupied the same central land cell, the 'status variable is incremented by two. The land model then makes an output to the hiker(s) that are occupying that cell, reporting to them its status. If the land cell has been occupied by two or more hikers during the cycle, it reports this encounter to the transducer, stating the its location of and the names of the hikers involved.
state-variables:
Figure 3: Pseudo-code for the LAND model.
The transducer serves to terminate the simulation at the preselected time. The transducer receives encounter reports from the land cells during the simulation run and final reports from the hikers at the end of the run. The transducer then creates a report of the results of the simulation, including; the number of encounters, where they were and which hikers were involved, as well as the locations covered by each hiker and the final satisfaction level of each hiker.
The generator sends a go message at the beginning of each five time unit movement cycle instructing the hiker cells to relocate. When the generator receives a stop message from the transducer at the end of the simulation run, it makes an output to the hikers telling them to submit their final reports to the transducer, and then passivates.
A landscape grid of 9 by 9 cells was selected for this simple simulation. This appeared sufficiently large to allow hikers to move around and interact while remaining small enough to be manageable. The effects of different levels of user activity could be simulated by pruning the hikers broadcast model to different sizes. Prunings of three and five hikers were selected. Two simulation runs were chosen to facilitate discussion of the behavior of the model. The simulation was allowed to run for twenty movement cycles, or a total of 100 time units. Discussion of the results will focus on the behavior of the models and on recommendations for future developments of the model.
An examination of the movement patterns of the hiker agents in this simplistic environment revealed a number of interesting observations. Agents seemed to move freely from cell to cell based on simple rules of behavior. Hiker agents, in some cases, displayed frequent backtracking or random movement between adjacent cells resulting in oscillating behavioral patterns. This occurred because of the lack of structured goal-oriented behavior being established for each agent at this time. This situation can be remedied by establishing this behavior beforehand or allowing each agent to have a different set of objectives such as would occur in multiple recreation forest environments.
Preliminary simulations of the hiker models have indicated a number of areas in which future investigations should be pursued. These areas include the development of new linkages between modelling systems and Geographic Information Systems, the development of a greater understanding of the relationship between management practices and natural processes, and the better understanding of experiences of recreators moving through a landscape.
Any dynamic modelling of the ecological systems or the management of those systems requires that connections be made between the dynamic modelling system and the spatial information contained in a GIS. For this simulation, spatial information is embodied in the land models to be queried by the hiker models. Future developments of this modelling system will endeavour to forge direct links between the modelling system and the GIS data base.
Although a system of controlled random movement of the hikers was utilized in this model to facilitate early testing, ultimately this system must be refined to reflect the behavior patterns of a real hikers in a real landscape. Future efforts are to be directed towards this goal. A number of avenues of investigation are being pursued towards the goal of simulation hiker movement.
Any hiker model will have to be able to react to the conditions in its immediate surroundings, while taking steps to achieve an overall goal. Such a model will have to be able to incorporate both short term and long term planning. Anderson and Evans (1991) outline an architecture for an intelligent agent which incorporates this short term/long term planning approach. They further stress the importance of the agents senses, the agents connection with its environment. Developing in an agent the capability to receive and interpret GIS data within the viewshed of its current location, and interpret that data in light of an overall plan to decide where to move next is one of the goals of this investigation.
The development of such short term and long term planning mechanisms will require the development by researchers of a comprehensive theory as to how such factors as the immediate environment and the predetermined goals of the an individual affect human movement in a natural landscape. A simplified model could be developed in which hikers follow a predetermined trail through the landscape.
Another area to be addressed is the nature of the experiences that a hiker has during a trip and how these affect satisfaction with that trip. A simple approach would be to gauge the satisfaction of the hiking experience on the basis of the number of encounters with others during that hike. There have been studies to measure the relationship between reported contacts and satisfaction (Hammitt et al 1984). Research in this field would have to be directed at a specific recreator group, such as backcountry hikers. Even within such a group perceptions of the impact of encounters with others, and the nature of that encounter, on the satisfaction of the hiker would vary. However it seems possible, with additional data on encounter frequency preference, to develop a model for a specific area that would allow managers to determine the relationship between level of hiker use in that area and the number of daily encounters. A carrying capacity for the area could conceivably be developed by comparing the results of the model with data from surveys on ideal numbers of encounters per hike.
As a result of this initial investigation into the use of intelligent agents in natural resource management, the following benefits are identified;
A more detailed model, in which the interaction with the environment as well as with other hikers affected satisfaction, would again require from researchers a consolidated theory as to the nature of a hikers interaction with, and perception of, back country environments. This model will also require further development to correlate time units in the simulation environment with time in the real system.
Clearly, this work is still in the preliminary stages. However, the process of the development of a functioning agent that ties social research results to current knowledge of natural systems, will create an important link in the bridge between the social and ecological sciences within a managerial framework that has been needed for a long time.
Anderson, J. & Evans, M.1991. An Architecture for Reactive and Strategic Planning. Technical Report 91-04. Department of Computer Science. University of Manitoba. Winnipeg, Manitoba.
Anderson, J. 1994. Constraint-Directed Improvisation for Everyday Activities, forthcoming Ph.D. Dissertation, Department of Computer Science, University of Manitoba.
Anderson, J. & Evans, M. 1994. Intelligent Agent Modelling for Natural Resource Management. (In Press).
Ball, G. 1990. A Spatial Dynamic Approach To Ecological Modeling: Simulating Fire Spread. Ph.D. Dissertation, School of Renewable Natural Resources, University of Arizona,Tucson.
Ball, G. 1993. Ecosystem Modeling with GIS. Environmental Management (In Press).
Ball, G. & Guertin, P. 1992. Improved fire growth modeling. International Journal of Wildland Fire, 2 (2) pp. 47-54.
Balmer, D.W. & Paul, P.J. 1986. CASM - The Right Environment for Simulation. J. Operational Research Society, No. 37, pp. 443-452.
Band, L.E., Peterson, D.L., Running, S.W.,Coughlan, J., Lammers, R., Dungan, J. & Nemani, R. 1991. Forest ecosystem processes at the watershed scale: basis for distributed simulation. Ecological Modelling, 56 pp.171-196.
Band, L. E. & Wood, E. F. 1988. Strategies for large scale distributed hydrologic simulation. Applied Mathematics and Computation, 27 (1) pp.23-38.
Baumans, R. & Sklar, F.E. 1990. A polygon-based spatial (PBS) model for simulating landscape change. Landscape Ecology, 4 (2/3) pp.83-97.
Berry, J. 1987. A mathematical structure of analysing maps. Environmental Management, 11 (3) pp.317-325.
Berry, J.S., Belovsky, G., Joern, A. Kemp, W.P., Onsager, J. Object- oriented simulation Model of Rangeland Grasshoper Population Dynamics. In Proceeding of Fourth Annual Conference on AI, Simulation, and Planning in High Autonomy Systems. September 20- 22, 1993. Tucson, Arizona. Pgs. 102-108
Brown,T.J. & Itami, R.M., 1982. Landscape Principles Study:Procedure for Assessment and Management- Australia. In: Landscape Journal, 1:Fall pp. 113-121.
Brown, T.J. and Itami, R.M., 1979. Landscape Principles Study: Upper Yarra Valley & Dandenong Ranges. Vol.2 Procedures for Landscape Assessment and Management. Report to Upper Yarra Valley & Dandenong Ranges Authority, Victoria, Australia.
Catsimpoolas, N. & Marti, J. Scripting Highly Autonomous Simulation Behavior Using Case-Based Reasoning. IEEE, April 1992.
Cohen, P.R., Greenberg, M.L., Hart, D.M. & Howe, A.E. 1989. Trial by fire: Understanding the Design Requirements for Agent in Complex Environments. AI Magazine 10(3):32-48.
Costanza, R., Sklar, F.H. & White, M.L. (1990). Modeling coastal landscape dynamics. BioScience, 40 (2) pp.91-107.
Deadman, P., Brown, R.D. Gimblett, H.R. "Modelling Rural Residential Settlement Patterns with Cellular Automata". Journal of Environmental Management. (1993) 37, 147-160. 1993
Folse, L.J., J.M. Packard & Grant, W.E. 1989. AI Modelling of animal movements in heterogeneous habitat. Ecological Modelling, 46 pp.57-72.
Folse, L., Mueller, H. & Whittaker, A. Object-Oriented Simulation and Geographic Information Systems, AI Applications in Natural Resource Management 4(2) 41-47, 1990.
Gimblett H.R. Dynamic Modelling in GIS: A Cellular Automaton Approach to Modelling the Growth Behavior of Urban and Rural Development. GIS National Conference '89. Ottawa, Ontario, Canada. 1989.
Gimblett, H.R.,Itami, R.M. & Fitzgibbon, J.E., 1985. Mystery in an Information Processing Model of Landscape Preference. In: Landscape Journal, Volume 5, Number 2, Fall 1985.
Gimblett, H.R. & Fitzgibbon, J. 1987. A Conceptual Model Mystery: A Component of the Kaplans' Model of Landscape Preference. University of Guelph. Unpublished manuscript.
Gimblett, H.R., 1984. The Perception of Mystery as a Dimension of Rural Landscape Scenery. Unpublished Master's Thesis. University of Guelph, Ontario, Canada.
Graefe A.R., Vaske, J.J., Kuss F.R. 1984. Social Carrying Capacity: An Integration and Synthesis of Twenty Years of Research. Leisure Sciences 6(4) 395-432
Green, D.G., Reichelt, R.E., van der Laan, J., & Macdonald, B.M. (1989) A generic approach to landscape modelling. In Proceedings Eight Biennial Conference, Simulation Society of Australia, pp. 342-347. Canberra, Australia.
Hammitt W.E., McDonald C.D., Noe F.P. 1984. Use Levels and Encounters: Important Variables of Perceived Crowding Among Non-Specialized Recreationists. Journal of Leisure Research. 16(1): 1-8.
Hendrickson, J.O. 1983. The Integrated Simulation Environment , Operations Research, Vol. 31, pp. 1053-1073.
Herbert, Joseph E., 1981. Visual Resource Analysis: Preference and Prediction in Oakland County, Michigan. Unpublished Master's Thesis. University of Michigan. Ann Arbor, Michigan.
Itami, R. M. 1979. Scenic Quality in Australia. Unpublished Master's Thesis. Centre for Environmental Studies, University of Melbourne, Australia.
Itami, Robert M. 1988. Cellular worlds. Models for dynamic conceptions of landscape. Landscape Architecture, 78 (5) pp. 52-57.
Kaplan, S. & Kaplan, R., 1982. Cognition and Environment: Functioning in an Uncertain World. Prager Publishers. New York.
Kaplan, S. and Kaplan, R., 1989. The Experience of Nature: A Psychological Perspective. Cambridge, MA: Cambridge University Press.
Kroh, D. & Gimblett, H.R. Comparing Live Experience with Pictures in Articulating Landscape Preference. Landscape Research 17(2), pp. 58-69.
Lynch, J. A. & Gimblett, H.R. Perceptual Values in the Cultural Landscape: A Computer Model for Assessing and Mapping Perceived Mystery in Rural Environments. Computers, Environment and Urban Systems, Vol. 16, pp. 453-47`, 1992.
Maidment, D. R. 1991. GIS and Hydrologic Modelling. Proceedings First International Conference/Workshop on Integrating Geographic Information Systems and Environmental Modelling, September 15-19, 1991, Boulder Colorado.
Marti, J. "Cooperative Autonomous Behavior of Aggregate Units over Large Scale Terrain." in Proceeding of AI, Simulation and Planning in High Autonomy Systems, IEEE, March 1990.
Marray, K.J & Sheppard, S.V. 1987. Automatic Synthesis Using Automatic Programming and Expert Systems Techniques Toward Simulation Modeling, Proc. Winter Sim. Conf., pp. 534-543.
Oren, T.L. & Zeigler, B.P. 1979. Concepts for Advanced Simulation Methodologies. Simulation, Vol 32., No. 3, pp. 69-82.
Reddy, Y.V. Fox, M.S. Husain, N,. & McRoberts, M. 1986. The Knowledge-Based Simulation System, IEEE Software, March, pp. 26-37.
Rozenblit, J.W., Zeigler, B.P. Kim, T.G. 1990. Concepts for Flexible Testing of Modular, Hierarchically Structured Systems, Trans. of Soc. Comp. Sim., Vol. 7., No. 3, September.
Ruiz-Mier, S., & Talavage, J. 1989. A Hybrid Paradigm for Modeling of Complex Systems, In: Artificial Intelligence, Simulation and Modelling (eds. L.A., Widman, K.A. Loparo, and N.Nielsen), J. Wiley, NY. pp. 381-395.
Running, S.W., Nemani, R.R., Peterson, D.L., Band, L.E., Potts, D.F., Pierce, L.L. & Spanner, M.A. 1989. Mapping regional forest evapotranspiration and photosynthesis by coupling satellite data with ecosystem simulation. Ecology, 70 (4) pp. 1090-1101.
Saarenmaa, H., Stone, N.D., Packard, J.M., Grant, W.E., Folse, L.J., Makela, M.E. and Coulson, R.N. 1988. Artificial Intelligence Approaches to Modeling Animal/Habitat Interactions. Ecological Modelling, Vol. 44, pp. 125-141.
Saarenmaa, H., Perttunen, J., Vakeva, J., Power, M. J. 1993. A Framework for Multi-Agent Problem Solving and Construction of Object-Oriented Decision Support Systems for Natural Resource Management. The International Joint Conference on Artificial Intelligence Workshop of AI in Natural Resource Management. Chambery, France, August 29 - September 3, 1993.
Sevinc, S. and Zeigler, B.P. 1988. Entity Structure Based Design Methodology: A LAN Protocol Example, IEEE Transactions on Software Engineering, Vol. 14, No. 3, March, pp. 375-383.
Shelby B., Neilson J.M., 1975. Use Levels and User Satisfaction in the Grand Canyon. Paper presented at the annual meeting of the Rural Sociological Society. San Francisco. CA.
Sklar, F.H., Costanza, R. & Day, J.W. 1985. Dynamic spatial simulation modeling of coastal wetland habitat succession. Ecological Modelling, 29 pp. 261-281.
Smith, P.C., 1986. The Use of the Kaplan Information Processing Model in the Planning and Evaluation of Scenic Roads. Unpublished Master of Science Thesis. University School of Rural Planning and Development, University of Guelph, Guelph, Ontario.
Stankey G.H., Schreyer R. 1984. Attitudes Toward Wilderness and Factors Affecting Visitor Behavior: A State-of-Knowledge Review. Research Paper INT-278. Ogden, UT: U.S. Department of Agriculture, Forest Service, Intermountain Forest and Range Experiment Station 45p.
Stone, N.D. 1990. An Object-Oriented Approach to Modeling Arthropod Predator-Prey Interactions. In proceedings of Resource Technology 90, Washington D.C.
Vasconcelos, M. J. & D.P. Guertin 1992. FIREMAP-simulation of fire growth with a geographic information system. International Journal of Wildland Fire, 2 pp.87-96.
Vasconcelos, M. J. & B. P. Zeigler 1993. Simulation of forest landscape response to fire disturbances. Ecological Modelling, 65 pp:177-198.
Vasconcelos M. J. & Zeigler, B.P. 1993. "Discrete-Event Simulation of Forest . Landscape Response to Fire Disturbance", J. Ecol. Model. (accepted)
Vazconcelos, M.J., Zeigler, B.P. & Graham, L.A. 1993. Modeling Multi-Scale Spatial Ecological Processes Under the Discrete Event Systems Paradigm. Landscape Ecology (In press).
Wright, J.D., Itami, R.M. & Fitzgibbon, J.E., 1984. An Investigation of the Variables Effecting the Perception of Mystery in Rural Landscape Scenery. Unpublished Research Report. University of Guelph. Guelph, Ontario, Canada.
Zeigler, B.P. 1976. Theory of Modeling and simulation. John Wiley, New York.
Zeigler, B.P. 1984. Multifaceted Modeling and Discrete Event Simulation. Academic Press, London and Orlando, FL.
Zeigler, B.P. 1990. Object-Oriented Simulation with Hierarchical, Modular Models: Intelligent Agents and Endomorphic Systems , Academic Press.
Zube, E.H., Sell, J.L. & Taylor, J.G., 1982. Landscape Perception: Research Application and Theory. In: Landscape Planning. Vol 9:1 pp. 1-33.