【作者】 徐昔保；

【导师】 李吉均；

【作者基本信息】 兰州大学 ， 人文地理学， 2007， 博士

【摘要】 元胞自动机(Cellular Automata，CA)是一种时间、空间、状态都离散，空间上的相互作用和时间上的因果关系皆局部的格网动力学模型，它“自下而上”的研究思路，强大的复杂计算功能、固有的并行计算能力、高度动态特征及具有空间概念等特征，使得它在模拟空间复杂系统的时空演化方面具有很强的能力，充分体现了“复杂结构来自于简单子系统的相互作用”这一复杂性科学的精髓，非常适用于具有复杂时空特征的地理系统模拟，现已成为城市土地利用空间演化模拟的重要工具与研究热点。本文在前人研究基础上较全面地综述了城市元胞自动机模型应用研究的最新进展，系统分析了基于CA的城市土地利用动态演化模拟与优化的理论基础，讨论了CA与其相关研究理论方法的异同点；侧重于从城市元胞自动机应用与扩展研究角度出发，首先详细介绍了SLEUTH和和DUEM模型的原理及实现过程，对其进行适当改进与应用；其次，综合集成GIS、神经网络(Artificial Neural Network，ANN)、遗传算法(Genetic Algorithm，GA)和元胞自动机(CA)分别构建了城市土地利用演化ULEM(Urban Land-Use Evolution Model)模型和城市土地利用优化ULOM(Urban Land-Use Optimization Model)模型；最后，以兰州市主城区为例，分别对这四个模型进行实证研究。本文的主要研究内容和结论如下：1兰州市城市土地利用动态演化特征及驱动机制分析结合GIS和遥感等技术手段，建立了兰州市1949-2005年8期中等分辨率的城市空间扩展GIS空间数据库和1976-2005年4期高分辨率的城市土地利用演化GIS空间数据库。宏观尺度上，1949-2005年城市空间扩展呈现出明显的阶段性特征：1949-1959年为有机快速扩展阶段；1959-1976年为无序高速扩展阶段；1976-1980年为城市建设恢复与调整期；1980-1993年为低速扩展更新阶段；1993-2001年为城市发展加速期阶段；2001-2005年为新一轮快速扩展阶段。微观尺度上，1976-2005年城市土地利用内部演化呈现出以下特征：(1)1976-1995年以政府社团用地和公共绿地的快速增长为主；(2)1995-2001年以商业用地增长最快，其次为居住、工业和政府社团用地；(3)2001-1005年城市用地年均增长速度比前两个时期都低，主要以居住、商业和工业用地的增长为主。兰州市城市土地利用动态演化主要受自然环境、经济与人口、交通、政策和历史因素等影响。2基于SLEUTH模型的城市土地利用空间扩展模拟侧重于对SLEUTH模型的蒙特卡罗迭代数、繁荣系数、萧条系数、坡度阈值及校准指标的选择等进行了适当改进，应用中充分考虑了城市规划及政策等对城市空间扩展的影响。结合SLEUTH模型和景观指数重建了兰州市1949-2005年城市空间扩展过程，表现为1949-1978年以离心式发展为主、1978-1993年以向心式填充发展为主和1993-2005年再次以离心式发展为主。两种不同情景的校准显示了两种不同的城市增长形式，与建国后社会制度的变迁一致。按两种不同情景预测2006-2050年城市扩展趋势表明：(1)情景2的城市空间扩展趋势明显快于情景1；(2)情景1预测的城市增长方式以向心式内部填充与更新改造为主，情景2预测的城市增长方式以离心式外延扩展为主；(3)从兰州市土地资源紧缺性角度看，情景1预测的城市增长方式更趋向于可持续发展，而情景2预测的城市增长方式更接近于未来城市发展趋势；(4)兰州市城市发展近期内应加强城市内部土地利用的空间优化，控制城市快速扩展趋势；(5)SLEUTH模型在受地貌作用限制显著的河谷型城市应用中具有通用性。3基于DUEM模型的城市土地利用动态演化模拟结合GIS和DUEM模型，重建兰州市1995-2005年城市土地利用演化过程，按两种不同情景预测2006-2020年城市土地利用动态演化趋势表明：(1)兰州市城市土地利用演化主要集中在城关区雁滩、七里河区马滩和崔家大滩、安宁区安宁堡、沙井驿、崔家庄和迎门滩等；(2)情景2预测的居住用地和商业政府社团用地的增长速度明显快于情景1，工业用地的增幅也略高；(3)情景2预测的演化趋势将更接近未来真实演化趋势；(4)DUEM模型在受地貌作用限制显著的河谷型城市应用中具有通用性。4耦合ANN与CA的城市土地利用演化ULEM模型侧重于在空间研究尺度、土地分类标准、空间变量、模型开发方式、神经网络训练随机采样方法和算法、校准方法和模型开发平台等方面进行扩展，基于Matlab7.2平台耦合GIS、ANN和CA开发了城市土地利用演化ULEM模型。该模型在时间步长、模型循环控制及校准方法等方面存在一些不足。以兰州市主城区为例进行实证研究表明：(1)两种不同仿真方案五种不同参数组合都获得较高的模型精度；(2)仿真方案1中随机变量v的增大可以明显提高模型精度，α参数的增大使得模拟的土地利用空间分布更趋向于分散；而对仿真方案2几乎不起任何作用：(3)2006-2015年城市土地利用演化预测中，α=1.5和T=0.8的组合方案将最接近未来城市土地利用演化趋势；(4)城市规划对兰州市城市土地利用演化影响深远。5耦合GA与CA的城市土地利用优化ULOM模型基于土地利用的适宜性和约束性，综合考虑土地利用的效益优化、数量优化和空间优化，采用较精细的城市土地利用分类标准和微观研究尺度，基于Matlab7.2平台耦合GIS、GA与CA开发了城市土地利用优化ULOM模型。以兰州市主城区为例进行实证研究表明：(1)兰州市城关区内的工业应尽快迁出市区；(2)继续加强张掖路、白银路和雁滩路的商业开发；(3)政府社团用地的增长将主要集中在安宁区迎门滩和城关区雁滩；(4)居住用地除了城市内部用地的置换增加外，非城市用地区域的居住用地开发将是重点；(5)优化结果基本上反映了兰州市未来城市土地利用演化的趋势，与城市发展政策、城市规划和政府意愿基本保持一致；(6)ULOM模型对空间尺度具有一定的敏感性，邻居构型将影响模拟的城市发展形态或增长形式。本文中SLEUTH和DUEM模型的应用研究、ULEM和ULOM模型的开发扩展研究，将有助于发展和丰富地理系统时空模拟的理论和方法，丰富和促进城市元胞自动机的扩展应用研究，支持与辅助兰州市城市规划和土地利用规划，具有一定的理论方法创新和实践应用价值。同时，城市CA模型研究在国内尚处在起步阶段，其理论研究基础、校准方法、模型的尺度、精度和真实性等都有待进一步探讨与扩展。更多还原

【Abstract】 Cellular Automata (CA) is a kind of discrete grid dynamic model, whose time, space and state are all discrete, and the spatial interaction and causality on time are completely specified in terms of a local relation. The traits of CA such as "from bottom to top" approach, strong complicated computing capability, inherent parallel computing capability, highly dynamic characteristic and spatial concept, etc., have made it very strong on spatial-temporal evolution modelling of the complex system. CA has shown thoroughly the essence of complexity science that "the complicated structure comes from the interaction of the simple subsystem". So CA is suitable to study complex spatial-temporal geographic system, especial for urban land use, and it has been an important tool and research focus for urban land use modeling.This thesis reviewed the literatures of CA model and progress on urban land use modeling thoroughly based on the previous researches. The theoretic framework about cellular automata on urban land use evolution simulation and optimization were analyzed, and the similarities and differences between the interrelated methods and theories with CA were discussed. This thesis mainly focused on the applications and extensions of CA in urban land use modeling and optimization. Firstly, two foreign classical CA models-SLEUTH and DUEM, were introduced with the principles and implementations, and applied with some proper revisions with the parameters. Secondly, tow extended CA models were designed and developed at the platform of MATLAB 7.2. The first CA model is named ULEM model (Urban Land-Use Evolution Model), integrating GIS, Artificial Neural Network (ANN) and CA to model urban land use dynamic evolution; the second CA model is named ULOM model (Urban Land-Use Optimization Model), integrating GIS, Genetic Algorithm (GA) and CA to optimize urban land use spatial structure. Lastly, all the four CA models were applied in the built-up of Lanzhou, China to test the respective suitability and capability. The main contents and conclusions of this thesis were illustrated as following:1 Analysis on characteristics and drive forces of land use dynamic evolution in LanzhouThe medium-resolution GIS database for urban expansion of 1949-2005 with eight different periods and high-resolution GIS database for land use evolution of 1976-2005 with four different periods were established with GIS and remote sensing. At the macro-level, the expansion of 1949-2005 presented obvious phased characteristics: the organic fast expansion period of 1949-1959; unordered high-speed expansion period of 1959-1976; recovery and adjustment period of 1976-1980; accelerated development period of 1993-2001 and new round fast expansion period of 2001-2005. At the micro-level, the characteristics of urban land use dynamic evolution of 1976-2005 in Lanzhou were illustrated as following: (1) The main fast growth of urban land use in 1976-1995 was Institutional and Green land use; (2) The fastest growth of urban land use in 1995-2001 was Commercial land use, then Residential, Industrial and Institutional land use; (3) The growth rate of 2001-2005 was both lower than those of the former two periods, with the main growth of Residential, Commercial and Industrial land use. Urban land use dynamic evolution in Lanzhou was mainly influenced by physical, economic, population, transportation, policy and historical factors, etc.2 Modeling urban land use spatial expansion in Lanzhou based on SLEUTH modelTo SLEUTH model, Monte Carlo iteration number, boom and bust parameters, threshold value of slope, calibration metrics were revised properly; and urban planning and policies influencing urban expansion were all considered synthetically. Urban expansion of 1949-2005 in Lanzhou was simulated with SLEUTH model and landscape spatial metrics, characterizing with diffusive growth in 1949-1978, coalescence growth in 1978-1993 and diffusive growth in 1993-2005. The calibration results under two different scenarios reflected two different urban growth forms, corresponding to the transformation of social-economic institution after the foundation of PRC. Predictions of 2006-2050 under two different scenarios were shown as following: (1) Urban spatial expansion tendency of the second scenario was faster than the first; (2) Coalescence infilling development, renewal and reconstruction were the main growth form for the first scenario, while the diffusive expansion was the main growth form for the second; (3) In the regard of land use restriction, the growth of the first scenario was more sustainable development, while that of the second was closer to the future urban growth; (4) Future urban development in Lanzhou should strengthen land use spatial optimization in the inner-city and control the fast expansion tendency; (5) SLEUTH model has high suitability to river-valley city.3 Modeling urban land use dynamic evolution in Lanzhou based on DUEM modelUrban land use dynamic evolution of 1995-2005 in Lanzhou was simulated with GISand DUEM model. Future evolution tendencies of 2006-2020 were predicted under two different scenarios and illustrated as following: (1) Future urban land use evolution will be mainly centralized in Yantan of Chengguan District, Matan and Cuijiadatan of Qilihe District, Anningbao, Shajingyi, Cuijiazhuang and Yingmentan of Anning District; (2) The growth of Residential and Commercial-Institutional land use for the second scenario will be obviously faster than the first, and the growth of Industrial land use for the second will be a little faster than the first; (3) Evolution tendency for the second will be closer to the future evolution tendency than the first; (4) DUUM model has high suitability to river-valley city.4 Urban land use evolution model (ULEM) integrating ANN and CAAn extended CA model integrating GIS, ANN and CA, named ULEM, was developed based on MATLAB7.2 for urban land use evolution modeling. To ULEM model, the spatial scale, land use criterion, spatial variables, development mode, random sample method of training and algorithm for ANN, calibration method and model developing platform were extended. And there were also some disadvantages for the model, such as time step, the control of model circle and calibration method, etc. Its application in Lanzhou was illustrated as following: (1) The accuracies under two different scenarios with five different parameter combinations were all very high; (1) The increase of random variable v could improve the accuracy of the model obviously, and the increase of the parameterαwould make the simulated spatial patterns of urban land use more scattering for the first scenario; while the changes of v andαhave almost no any influence on the accuracy and pattern for the second; (3) The prediction withα= 1.5 and T = 0.8 may be the best parameters’scenario to model future urban land use dynamic evolution tendency; (4) Urban planning influenced land use evolution in Lanzhou deeply.5 Urban land use optimization model (ULOM) integrating GA and CAAn extended CA model integrating GIS, GA and CA, named ULOM, was developed based on MATLAB7.2 for urban land use spatial optimization. This model was based on land use suitability and restriction, and considered with land use benefit optimization, land use amount optimization and land use spatial optimization with refined land use criterion at the micro-scale. Its application in Lanzhou was shown as following: (1) The industry in Chengguan District should be transferred out of the built-up area as soon as possible; (2) Commercial land use along Zhangye Road, Baiyin Road and Yantan Road, should be further to develop; (3) The growth of Institutional land use will be mainly centralized in Yingmentan of Anning District and Yantan of Chengguan District; (4) The growth of Residential land use will play the emphasis on developing non-urban land use, expect with the replacement from other land use in the inner-city; (5) Urban land use evolution tendency optimized by ULOM model is lined with urban planning, policy and the will of the government; (6) ULOM model has some sensitivity to the spatial scale, and its neighborhood extent will influence urban development pattern or growth form.The applications of SLEUTH and DUEM models, developments of ULEM and ULOM models in this thesis will be meaningful to explore and enrich the theories and methods for geographic system modeling; to enrich and promote the extended applied research of CA in urban modeling; to support and assist urban planning and land use planning in Lanzhou and supply a sound background for sustainable development in Lanzhou. In a world, there are certain innovations in theory and methodology, and applied value for real urban land use system modeling of this thesis. Simultaneously, the research of urban CA model in China is still at the rising stage, the theory, accuracy, spatial scale, facticity and calibration methods for CA model should be further explored and extended.更多还原