Application of Cellular Automata in Landscape Pattern Optimization
元胞自动机在景观格局优化中的应用

General treatment of landscape pattern spatial optimizations is in its infancy.Among various modes of landscape pattern optimization,the most advanced and sophisticated one is a mode that searches optimal solutions in state space of simulation results of landscape changes.It has exhibited many advantages such as relatively high objectivity and automaticity of optimization.However,at present,there are no enough quantitative theories of interactions between pattern and process in landscape for extensive use of this optimization mode.Fortunately,due to inherent advantages of cellular automata(CA),the spatial explicit model based on CA is likely to bring the optimization mode into operation,on condition that a series of technical problems are conquered.In in paper,principles of cellular automata and its applications in landscape ecology were introduced.A CA has three essential characteristics: spatial discreteness of expressions of states,temporal discreteness of changes of states,spatial correlation of rules of state transformation.Therefore,it has not only general advantages for analysis and simulation of spatial-temporal development,but also relative advantages for spatial simulation of landscape changes.The reasons for both aspects of advantages in the applicability of CA into landscape pattern optimization were elucidated in the paper.The former is that a CA model is constructed "from bottom to top",based on interactions of microscopic individual units,so that it can simulate landscape changes directly at a relatively small scale,regardless of quantitative laws at landscape scale.The latter is that its definition of neighborhood and associated rules of transformation can naturally meet the requirement of landscape ecology that attach much importance to horizontal processes.The challenges of application of CA into landscape pattern optimization were analyzed,among which the two most distinguished are as fellows: one is contradictions between its simplicities of construction and complexities of landscape change.The other is how to define rules of transformation that can reflect natural and human factors during landscape change.Other problems in application of CA into landscape pattern optimization include definition of scales,calibration of temporal paces,and computational complexity.Some partial and tentative solutions to these problems were presented.Firstly,the conception of CA model was expanded with most generalized expressions so as to enable CA to simulate complex landscape changes.Secondly,models of dominant ecological process were integrated into rules of state transformation.Last but not least,in order to reduce computational complexity and ensure the mode to be practically operated,some computer algorithms and techniques of software engineering,such as search strategies and calculation multiplexing,were utilized.An architecture and flow chart for the landscape pattern optimization mode,centering on generalized-CA based spatial explicit