The models and methods of land use trend prediction mainly include regression prediction method, Markov method, artificial neural network, grey model, cellular automata model and so on. Regression prediction method is a prediction method that deduces the numerical value of predicted variables from the known values of other variables by using mathematical equations representing the relationship between variables. This method is suitable for standard variable groups with strong correlation between variables. Because different land use types influence and restrict each other, this method is often used to study the relationship between land use change and human factors. The disadvantages of this model are that it is not suitable for large-scale forecasting, and the prediction error caused by economic factors is difficult to quantify.
Markov process is a process without aftereffect. To simulate the dynamic change of land use with Markov process, we must first determine the initial state matrix and transfer probability matrix of land use types. The advantage of this prediction model is that it is simple to calculate and realize, and it can reveal the quantitative transformation law and overall change trend between different land use types. The operation of the model only needs to consider the current information of land use, but does not need to consider the internal mechanism of land use change. The deficiency is that the model lacks the ability to reveal the driving mechanism of land dynamic change and the ability to express it in space. This model is suitable for the case that the driving mechanism of land use change is unclear and the short-term land use change is predictable.
Artificial neural network can vividly simulate part of people's thinking ability, and the process of prediction is to use the information sources obtained in different periods to find out the areas and types of land change on the basis of comprehensive analysis and comparison. The advantage of this model is that it can analyze dynamic data and summarize laws according to historical data. However, due to the limitation of forecast area size and time, the characteristics of some main factors are difficult to determine, and the forecast results are not very accurate.
The dynamic model GM(n, h) of grey system is a modeling method based on the theory of grey system, which uses the data collected discretely by the system to establish its dynamic differential equation, takes the grey module as the basis and takes micro-fitting analysis as the core. Land use system is essentially a grey system. Under the condition of incomplete land data, the grey model can be used to analyze the process of medium and long-term land use structure.
Cellular automata model is the concept and application modeling method of discrete dynamic system. Its framework is simple and open, suitable for simulating complex systems with self-organizing structure, and has strong vitality. The "bottom-up" research idea, powerful complex computing function, inherent parallel computing ability and spatio-temporal dynamic characteristics of the model make it natural, reasonable and feasible to simulate the spatio-temporal dynamic evolution of this spatial complex system of land use change. The combination of CA model and GIS software provides a new idea and modeling method for dynamic modeling of land use. Compared with other models, its advantages are: ① simpler and more natural; (2) The spatial interaction, rather than the interaction between socio-economic indicators, can better reflect the change of spatial pattern and its further feedback.
The complexity of land use change determines that land use research must adopt the theoretical method of complex system, especially the establishment of mathematical model based on complex system thought, which is one of the important fields of land use process research. Therefore, it is the key to establish a scientific land use change model by combining the law of land use change and adopting complex system research methods. At present, under the background of complex system theory, using cellular automata model to study the complex behavior of geographical processes is the frontier in the field of geographical modeling and an interdisciplinary frontier technology. It is of great theoretical value and practical significance to apply it to the prediction and simulation of land use change.
(1) cellular automata model can describe the spatio-temporal dynamic evolution of complex system structure. Using cellular automata to dynamically simulate land use cover change (LUCC) and quantitatively discuss and predict the process of land use change is of great significance to regional sustainable development, land use planning and land management decision-making. This model starts with the study of the current situation of land use in a region, analyzes the dynamic change process of land use with the help of GIS software, studies the transfer law of different land use types, explores the internal mechanism of land use change, and provides theoretical basis for land use regulation with different development goals.
(2) Combining GIS and Geo-CA to analyze land use change can not only improve the efficiency of the model operation, but also simulate the whole process of the model by computer, and can directly display the change of land use nature and the results of land prediction. By adjusting the model parameters, we can get the results of land use in different forecast years in the future, and provide a basis for regional decision makers to evaluate land use.
(3) Studying the land use change in the river basin can not only provide an important decision-making basis for the rational use of land and the rational allocation of river water resources, but also provide an important guarantee for the coordinated development of production and life, ecological environment and economy of river basin residents. The application of Geo-CA model in the simulation and prediction of land use change in river basins is still in the exploratory research stage. By establishing a more suitable geo ca- land use model and applying it to Tarim River Basin, it not only broadens the application field of GeoCA model, but also plays a good demonstration role for land use prediction in other basins.
In 1980s and 1990s, Batty and Xie combined fractal theory with cellular automata to study the formation and expansion of cities in detail. The diffusion-limited aggregation (DLA) model they designed can be regarded as a generalized CA model. 1994, Me proposed the Dynamic Urban Evolution Model (DUEM), which described the city with self-similarity and fractal dimension and its development process with CA theory. Clark (1998) set appropriate model parameters according to the historical data of urban development, traffic and terrain conditions, established the CA model of urban growth, and loosely coupled the model with the GIS platform, successfully simulated and predicted the urban areas of San Francisco and Baltimore in the United States. Since 1990s, Wu F. (1998) has organically combined cellular automata model with multi-factor evaluation model, applied AML and C language in ArcInfo, and realized the integration of GIS, CA model and MCE model on a unified interface. On this platform, the simulation of urban expansion in Guangzhou, Guangdong Province is realized. White of Canada and Engelen of the Netherlands (1994) used the constrained CA model to simulate the dynamic changes of land use, such as the urban growth of Cincinnati in the United States and the changes in land use composition of Caribbean islands affected by global warming.
Wu Xiaobo and Zhao Jian (2002) used cellular automata model based on remote sensing and GIS to simulate the urban development process of Haikou from 65438 to 0987 to 2000. Heping (2004) put forward a prediction model of desertification evolution based on GIS software and CA model, and took Beijing and its surrounding areas as an example, and achieved good simulation results. Chen Longquan (2004) based on the characteristics of Markov and CA models, discussed the feasibility of dynamic simulation and prediction of land use/land cover change with Markov-CA model. Markov-CA absorbs the advantages of Markov and CA theory in time series simulation and prediction. Experiments using two TM data show that Markov-CA model can simulate and predict land use/land cover change well. Zhang Xianfeng (2000,2001) proposed a new method to simulate and predict the geographical spatio-temporal process by integrating GIS and CA model, that is, firstly, the 4-tuple of the standard CA model was extended to meet the requirements of spatio-temporal dynamic simulation in GIS environment, and then the dynamic simulation and prediction model (LESP) of urban land use evolution was established by taking the dynamic process as an example. Finally, the model is used to successfully simulate the urban expansion and the evolution of sustainable land use in Baotou City. Zhou Chenghu et al. (200 1) constructed a practical and operable spatial dynamics model (GeoCA-urban) by using geoca. The model is suitable for simulating the development and evolution of cities, and the simulation effect is good.
With the support of GIS software and Geo-CA theoretical model, the ecological professional analysis subsystem of the system takes Sanyuan River Basin and Alar Artificial Oasis Area as research sample areas, and starts with the superposition analysis and transfer analysis of two thematic maps of land use in different periods, and combines the econometric model of land use change to analyze the dynamic change characteristics of land use and quantitatively study the relationship between land use change and water system, traffic, soil and topography. On this basis, the control factor layer is constructed, and the model parameters are adjusted by combining social and economic data to determine reasonable neighbors and transformation rules. After repeated debugging and modification, a reasonable land use dynamic evolution model (GeoCA-Landuse model) and its software system are finally constructed.