The dynamic simulation of landscape change is carried out from two aspects: change aggregation degree and mathematical method. Agglomeration can be divided into three landscape change modes, namely, overall landscape change mode, landscape distribution change mode and landscape spatial change mode, as shown in Table 5- 1.
Table 5- 1 landscape change model
(According to PerBrinck et al., 1989)
According to different properties, the dynamic model of landscape pattern can be divided into five types, namely, agent-based landscape change model, empirical statistical model, optimization model, dynamic simulation model and mixed comprehensive model (Fu Bojie, 1995). Landscape change models can be divided into random landscape model, neighbor rule model and landscape process model (including infiltration model, individual behavior model and spatial ecosystem model).
1. Random landscape model
Random landscape model is to study the overall dynamics of landscape pattern and process in time and space (Yu et al., 2006), which does not involve specific ecological processes and is a model that attempts to combine spatial information with probability distribution. This landscape model combines geometric method (description system), statistical method (analysis system) and mechanism method (simulation process), or introduces biofeedback principle into spatial dynamic model or spatial characteristics into traditional ecological model. The most commonly used model is Markov chain model (Jianguo Wu, 2006).
2. Neighborhood rule model
In the process of landscape dynamic change, the change of patches depends not only on the state of the last time point, but also on the properties and changes of adjacent patches, which can be organized into a series of rules that restrict the amplitude and direction of landscape dynamic change. Neighborhood rule model is a kind of landscape dynamic model based on this premise, and it is a discrete dynamic model that can produce complex landscape structure and behavior at the landscape level. At present, the most common and representative neighborhood rule model is the cell self-organization model (CA model).
The components that make up cellular automata are called "cells", and each cell has a state. These cells are regularly arranged on the grid of "cell space", and their respective states change with time and are updated according to a local rule, that is, the state of a cell at a certain moment depends on and only depends on the state of the cell at the previous moment and the state of all neighboring cells of the cell. The cells in the cellular space are updated synchronously according to such local rules, and the whole cellular space presents the change of discrete time dimension.
Mathematically, finite cellular automata is a quadruple:
Study on Forest Landscape Pattern and Ecological Planning —— A Case Study of Baihe Forestry Bureau in Changbai Mountain
Here a stands for cellular automata system; L stands for cellular space, d is a positive integer, representing the dimension of cellular space in cellular automata; S is a finite and discrete state set of cells; N represents the combination of pixels in all neighborhoods (including the center pixel). F is a transformation rule based on proximity function. According to the transformation rules, cells can be transformed from one state to another.
The general characteristics of CA model are: ① spatial dispersion and homogeneity, the change of each cell follows the same law, and the distribution of cells is the same; ② the discreteness of time; (3) the discreteness and finiteness of state; ④ Synchronous calculation: the state change of CA model can be regarded as the calculation or processing of data or information; ⑤ Locality, the current state of each cell may only affect the state of adjacent cells with radius r at the next moment. From the perspective of information transmission, the speed of information transmission in CA model is limited; ⑥ The dimension is high, and the number of variables in a dynamic system is generally called dimension. From this point of view, CA model should belong to an infinite dimensional dynamic system.
CA model is simple, flexible, clear and widely used. Its greatest advantage is that it can combine the observation data on local small scale with the neighborhood transformation rules, and then study the dynamic characteristics of the system on large scale through computer simulation. The model is also good at revealing the increasing and decreasing process of landscape components, the way of biological behavior or the diffusion process of ecological interference under the action of specific constraint systems. From the data structure, because the cell structure in CA model is the same as that in grid-based GIS, the model is easy to integrate with GIS, remote sensing data processing and other systems (Li Habin et al. 1988,1996; Xiao Duning, 1999, Zhang Xiaofeng et al., 2000).
However, the application of CA model in landscape ecology has some limitations, such as: ① overemphasizing the state of adjacent units, only considering local interaction, ignoring the influence of regional and macro factors; ② The unit attributes considered in the model are relatively simple, while the unit attributes in the actual landscape are composed of multi-level and multi-factors, and there are still interactions between units; ③ The transformation rules are predetermined, but the dynamic process of the real landscape usually shows some possibility and tendency, and the state transformation is not completely determined; ④ It is difficult to grasp the temporal and spatial resolution, which will directly affect the accuracy of the simulation results (Jia H et al., 1998).
3. Landscape process model
Landscape process model is to study the occurrence, development and diffusion of an ecological process (such as interference or material diffusion) in landscape space from the mechanism. This method usually has three modeling starting points: ① to simulate the dynamic change process of landscape with a known material movement law, such as infiltration model; (2) Clearly consider the spatial position and behavior of each biological individual in the landscape, and reflect the function and structural dynamics of the landscape through individual behavior and function, such as a process model based on individual behavior; (3) Based on a detailed understanding of the dynamic change mechanism of landscape, the dynamic change process of landscape is truly expressed through simulation, such as the spatial ecosystem model.