多尺度土地资源配置研究

【作者】 尹珂；

【导师】 李航； 魏朝富；

【作者基本信息】 西南大学 ， 农业资源利用， 2010， 博士

【摘要】 随着我国经济高速发展,城市化已经成为一个不可逆转的社会发展趋势。在此过程中人们对土地利用空间和土地产品的需求也在不断增长,人地矛盾的日益激化引发了很多社会、经济和环境问题,并有可能造成生物栖息地消失或破碎化、生物多样性降低以及生态系统服务功能的减弱。如何应对这些生态环境问题,并将其纳入到政策、规划和项目的决策过程中已成为研究的热点。本研究以三种尺度：渝北区(县域尺度)、龙兴镇(镇域尺度)、河堰等三个行政村(村域尺度)为例,综合考虑不同尺度上影响土地资源配置的因素,采用不同的方法对多尺度土地资源的优化配置进行研究,本研究完善了城乡统筹背景下的土地利用优化配置体系,为实现土地资源的可持续利用提供理论依据。1县域尺度的土地资源配置研究本研究以重庆市渝北区各乡镇作为功能分区的基本区域单元,以渝北区自然生态、社会经济发展以及土地资源现状和特点为基础,建立统筹发展的土地利用主体功能分区。首先根据一定的原则构建指标体系,土地利用指标是以重庆市渝北区1989、2000和2004年3个时段的TM卫星遥感影像数据为数据源,根据3个年份对应期间的土地利用现状图、相关统计资料,结合研究区实际情况,采用遥感图像处理软件ERDAS 8.4、地理信息系统软件ARCGIS 9.2,通过非监督分类与目视解译相结合的方法,对3个时段的卫星遥感影像进行土地利用解译得到,而社会经济指标的数据源自2008年《重庆市统计年鉴》、2008年《渝北年鉴》以及2008年《重庆市国土资源统计年鉴》；然后采用主成分分析法对指标进行处理及筛选；最后使用聚类分析方法对渝北区进行主体功能分区。(1)主成分分析方法能够科学合理地提炼出分区指标。本研究从经济发展、社会进步和环境保护三个方面选出了15个初选指标,运用主成分分析法进行降维处理,以期用较少的几个综合指标来代替原来的变量指标,于是从15个初选指标中遴选出财政收入、粮食单产、林地面积等9个指标组成县域主体功能分区的综合指标评价体系,这9个指标不仅能够反映土地利用原始信息的代表性和准确性,而且有效地减轻了工作量,提高了分区精度。(2)对比快速聚类法和分层聚类法的分区结果以及单因子ANOVA分析和判别分析的验证,使分区更加合理。在9个遴选指标的基础上构建了19×19的模糊相似矩阵,根据不同准则分别进行快速聚类和分层聚类,对比发现只有木耳镇的划分结果存在分歧。单尾ANOVA分析表中只有“农民人均纯收入”指标的显著性不佳,而判别分析中交叉验证样本的分类正确率为88.9%,说明分类的结果比较理想。基于以上分析,考虑到功能区单元位置的连续性并参考渝北区自然环境条件,将渝北区划分了四个主体功能区,北部新区、双凤桥街道与回兴街道为优先开发区；龙溪街道、龙山街道、龙塔街道与双龙湖街道为重点开发区；兴隆镇、古路镇、大湾镇、龙兴镇、玉峰山镇和统景镇为限制开发区；硌碛镇、石船镇、木耳镇、大盛镇、茨竹镇和悦来街道为禁止开发区。2镇域尺度的土地资源配置在满足界限约束和空间连续性约束的条件下,以重庆市渝北区龙兴镇为例探索量化分区方案。首先针对土地利用分区问题建立数学模型,然后编写C++程序在ArcGIS平台上实现模拟退火算法对优化模型的求解,利用虚拟场景的分区结果评价模型的生态效应,而在现实场景中运用游憩价值法和GeoDA的空间自相关分析功能分析龙兴镇的区域发展特点,并在此基础上进行土地利用优化分区。(1)虚拟场景的测试证明了土地利用优化分区模型的生态保护效应是十分明显的,尤其是对生物多样性的保护。Fragstats的计算结果显示出：取消连接性的限制时,则呈现最多的斑块数55和最小的连接度指数6.12,因此,取消连接性的限制条件后的结果表明连接性限制对分区结果的景观破碎度和异质性都具有相当的影响力；AI值代表了区域之间的聚集度,越强烈的聚集度区域之间进行的物种或者基因的交流越有可能发生,土地兼容性均为1时得到了极端的AI值93.12,而土地兼容性均为-1时出现了许多破碎的斑块,使得边缘总长度(TE)达到最大值1003,说明非对称的土地兼容性系数设置具有重要生态意义；在优化模型中使用了欧氏平方距离因而得到与土地适宜性相吻合的3个清晰的核心区。(2)在区域发展特点的分析基础上,结合基础图库建立土地适宜性数据库,可以实现既适应区域社会经济发展又符合自然生态保护要求的合理分区。游憩价值分析得出龙兴镇自然环境的经济价值为5696.72万元,肯定了古镇旅游产业的发展潜力；全局Moran’s I指数0.5138和局部自相关分析图反映出龙兴镇的经济发展已表现出较强的空间分异格局,说明受“鱼嘴组团”的影响龙兴镇非农产业已产生集聚效应。利用Access数据库软件包编制栅格属性数据,导入ArcGIS后形成一对多和多对多的数据属性关联关系,从而实现将土地适宜性量化地引入分区模型,量化分区结果与龙兴镇镇域分区规划相比我们发现：定量模型的分区体现了相似的土地利用形式拥有更好的聚集效应,而相斥的土地利用形式则呈现远离的态势,因而生态核心区十分明显。3村域尺度的土地资源配置村域尺度的研究以龙兴镇河堰村、粉壁村及石溪村为例,运用三种不同的多目标线性规划方法构建土地利用情景决策系统,探索村级土地利用结构的优化配置。(1)不同的目标优先等级对帕累托最优解有显著影响。在渝北区龙兴镇的研究中,要获得较高的毛利润,则必须付出较高的生产成本和较多的农业化学品用量；如果既要实现毛利润和工时需求最大化,又要满足生产成本和农业化学品用量最小化,那么所得的帕累托最优解中高产量水果和蔬菜作物以及针叶树和水杉的种植面积必然会增加,在此基础上所有目标完成度才能超过40%从而实现均衡状态；如果生态目标的优先级设置越高,则林地面积增加的越多；如果在实现工时需求最大化和农业用地面积最大化的同时,满足毛利润最大化这个经济目标,则水果和蔬菜的用地面积会相应的有所增加。(2)目标优先级设置应该有客观标准。在运用土地利用情景决策模型时,应满足以下几个标准：a.研究区必须接近城市,如城乡结合部；b.要有较为明显的旅游特征或潜质；c.研究区是以农业生产为主的区域；d.研究区有较多样的农业活动。当研究区有较为明显的旅游特征或潜质的时候,则适宜将环境目标作为优先考虑,从而促进林业物种的增加和其他非密集性的土地利用形式。而对于以农业生产为主的区域,则应该优先考虑经济目标,这样可以同时解决局部的社会问题,因为受惠的直接群体就是当地的农民。至于多种农业活动的区域,潜在的利益冲突就可能存在,因此优先考虑社会问题是必然的。4多尺度土地利用优化系统土地利用优化是一个繁复的运作体系,涉及到多个尺度下不同的因素,因此不同尺度研究体系的融合问题成为土地利用优化研究的热点和难点。本研究结合MAS技术提出基于利益相关者agent的多尺度土地利用优化系统框架。(1)县域尺度下的研究单元是以行政区划图为基础的各个乡镇,镇域尺度下是利用土地利用现状图与行政区划图相叠加后利用栅格划分获得的,而村域尺度下则是以农户行为作为土地利用结构的主导因素,因此分析单元为农户,这些研究单元是多尺度土地利用优化系统运行的基础。由于研究的主要目的是为了探索土地利用的优化配置格局,因此在不同的尺度下选择了不同的土地优化配置的方法：县域尺度下考虑的是土地利用功能的引导性,所以选取了模糊数学方法进行主体功能分区；镇域尺度既要有一定的政策指导性,还必须拥有一定的实际操作性,因而基于模拟退火算法建立了土地利用优化分区模型,以优化算法求解定量模型的形式进行土地利用分区；而村域尺度需要更强的可操作性,以便指导农户行为,因此结合了三种多目标线性规划方法创建了土地利用情景决策系统,运用目标优先级的差异性体现决策者偏好对土地利用配置结果的影响,从而辅助决策者制定土地利用结构优化方案。(2)多尺度土地利用优化系统的组织结构是Agent个体之间交互的框架,由Agent个体、Agent组及Agent域三级组织构成。在本研究中,Agent个体就是政府、农户、科研工作者等,他们共同组成了利益相关者Agent组,而本系统最核心的部分就是三个Agent域一理论概念域、实证分析域以及模型方法域。Agent个体之上有“组”和“域”两级管理协作机构,可以有效减少系统的内部冲突,协调问题求解,具有更强的问题求解能力,提高交流的效率。而“组”和“域”是开放、动态的概念,其中的成员数目是动态变化的,成员可以动态地加入或退出。这个动态演化的过程可视为Agent随环境及目标需求的不断变化,连续进行协作关系的适应性调整过程。在这样的系统中探寻土地利用优化方案,可以有效的调动利益相关者,使各尺度的土地利用优化方案都是经过利益相关者反复交流的结果,从制定的过程上一改往日“由上而下”的规划体系,实现了真正意义上的参与式规划。综上所述,本文系统研究不同尺度下的土地资源优化配置,不再局限于单一尺度或者采用某一数学方法得出优化配置方案的传统研究思维,构建了适合多尺度研究的土地资源优化配置体系框架,为城乡统筹背景下的土地利用研究提供了新的思路和方法,弥补国内在土地利用研究的不足。但是由于某些尺度的研究只能在假设成立的条件下进行,所以还需要不断完善。此外三个尺度研究的耦合还需要具体的技术手段来加以实现,因此未来应继续加大时间段上的研究和尺度耦合的研究,以便应用到实践中。更多还原

【Abstract】 With the rapid economic development in China, Urbanization has become an irreversible trend of social development. In this process, people’s requirement of land use space and land products is also growing, the increasingly intensified contradiction between people and land led to many social, economic and environmental issues, and may result in the disappearance or fragmentation of habitat, the reduction in biodiversity as well as the weakening of ecosystem services. How to respond to these ecological problems and integrate them into policy, planning and project decision-making process has become a hot spot of research. Therefore, this paper dealt with multi-scale optimal allocation of land resources under consideration of influencing factors with various scale land use pattern, taking representative blocks of Yubei district, Longxing town and Heyan village etc. for example. This study would enrich and improve land quality assessment system by anatomizing similarities and differences of multi-scale land quality in hilly-land regions. The research will improve the system of land use optimization in the context of rural and urban development.1. Land resources disposition at county scaleThis research makes the township of Yubei district in Chongqing as basic zoning units to establish co-ordinate developing land use main function zoning. First of all, we construct the index system according to certain principles, and the index data derive from "Chongqing Statistical Yearbook", "Yubei Statistical Yearbook" and "Chongqing Statistical Yearbook of Land and Resources". Then, we handle and select the indicator by principal component analysis. Finally we use cluster analysis for land use main function zoning in Yubei district.(1) The principal component analysis can scientifically and reasonably refine zoning index. This study selected 15 primary indicators from economic development, social progress and environmental Protection which is to deal with dimension-reduction treatment by principal component analysis, So we can get less comprehensive index to replace the original variable indicator which is a few more. Therefore, we chose 9 indicators such as "the Financial revenue", "the per unit area yield of grain" and "forest area" to constitute the comprehensive evaluation system of main function zoning at county scale. These indicators not only reflect the representativeness and accuracy of the land use original information, but also effectively reduce the workload and improve the zoning precision.(2) The comparison of results from K-means cluster and hierarchical cluster as well as the verification of Single factor ANOVA analysis and discriminant analysis make the zoning results more accurate. We constitute 19×19 fuzzy similar matrix based 9 selected indicators to use K-means cluster and hierarchical cluster by different criterion. We found the zoning results from each method were similar except the Muer town. In the one-tailed ANOVA analysis table only the index "rural per capita net income" had the highest F value and the classification accuracy of Cross-validation samples is 88.9% in discriminant analysis which show the classification results is ideal. Based on the above analysis, we divided Yubei district into four main functional areas which resulted from considering continuity of position of functional units and consulting natural conditions in Yubei district.2. Land resources disposition at town scaleWe quantify the zoning plan to explore Longxing town in satisfying the boundary constraint and spatial continuity constraint conditions. First of all, we set up a mathematical model for the problem of land use zoning. Then we wrote C++ program to solve the optimization model by implementing SA on the platform of ArcGIS. On the one hand we evaluated the model’s ecological effects from the results of virtual scene; On the other hand we carried out optimization of land use zoning on the basis of analyzing the characteristics of regional development by CVM method and spatial autocorrelation analysis.(1) Performance of virtual scene proved that the ecological protection effects of land use optimal zoning model is very clear, especially the protection of biodiversity. The calculated results of Fragstats show that:When we canceled connectivity constraints the result presented the most patch number 55 and the lowest connectance index 6.12, this result indicated connectivity constraints have considerable influence to landscape fragmentation and heterogeneity; Aggregation index value represents the aggregation between regions, the more intense aggregation the more perhaps exchange occurred between species or genetic in different areas, when land compatibility value is 1 there is the most AI value 93.12, but when land compatibility value is-1 there are many broken patches which lead the total edge achieve max value 1003, these result indicate asymmetric land compatibility had important ecological significance; When the model use squared Euclidean distance there would be 3 clear core areas which consistent with the land suitable.(2) Based on Characteristics analysis of regional development we establish a database of land suitability through combination of basis maps, there could be achieved reasonable zoning which is not only to adapt to the regional socio-economic development but also to consistent with natural ecological protection requirements. CVM analysis indicates the economic value of the natural environment in Longxing town is 56967200 yuan, it confirms the development potential of town tourism industry. Global Moran’s I index 0.5138 and the map of partial autocorrelation analysis all reflect Longxing town’s economic development has shown a strong pattern of spatial variation, this indicate that the combined effect of non-agricultural industries has generated. We compiled raster attribute data by using access database package which formed incidence relationship of one-to-many or many-to-many between the data attributes, in order to achieve quantitatively introduce land suitability into zoning model. We compared quantitatively zoning result with zoning planning in Longxing town, found that:Zoning result of quantitative model reveal similar forms of land use have a better mass effect, and repulsive form of land use present removed state, so ecological core area is very obvious.3. Land resources disposition at village scaleThe research at village scale study Heyan village, Fenbi village and Shixi village by Land use scenarios decision-making system which is constituted with three different multi-objective linear programming methods, the goal is exploring the optimization of village land use structure. All land use technical data are come from 347 valid household surveys.(1) Different priority objectives on the Pareto optimal solution have a significant effect. In this study if you want to get a higher gross profit you must pay higher production costs and greater usage of agricultural chemicals; the more higher priority of ecological objectives set, the more forest areas increase; if we want to maximize work requirements and agricultural area while maximize gross margin, so the fruits and vegetables will be a corresponding increase in area.(2) Goal priority setting should be objective criteria. The use of land-use scenario decision-making system should meet the following criteria:a. areas close to urban areas; b. areas of pronounced attractiveness for tourism; c. predominantly agricultural areas; d. areas with diversified agricultural activities. In areas of pronounced attractiveness for tourism, it seems logical to apply an environmentally oriented philosophy (Env> Soc> Eco) leading to promotion of native forestry species and other non-intensive land uses. In predominantly agricultural areas, an economically oriented philosophy (Eco> Soc> Env) would favour their main social group, farmers. In areas with diversified activities, the presumable conflicts of interest as regards the use of land suggest that a socially oriented philosophy (Soc> Eco> Env) might be most appropriate.4. Multi-scale land use optimization systemLand use optimization is a complex operation system which involves different factors at multiple scales, so the integration problem of systems at different scales has become focus and difficulty in land use optimization research. This study which combined with MAS technology proposed the framework of multi-scale land use optimization System which is based on agent stakeholders.(1) Research units at county scale are towns based on administrative map, research units at town scale are obtained by grid division by superposition of present land-use map and administrative map, and research units at village scale are households because the household behavior is the dominant factor of land use structure. These research units are the basis of running multi-scale land use optimization system. Since the main purpose of the study is to explore the optimal allocation of land use pattern, therefore we chose different method of optimal allocation of land at different scale:the method at county scale is using fuzzy mathematics to functional zoning because of considering function of guiding land use; the method at town scale is land use zoning by quantitative models which is based on simulated annealing algorithm; the method at village scale need more operational in order to guide the household behavior, so we combined three multi-objective linear programming method to create land use scenarios decision-making system which use priority objectives difference to reflect decision-makers preferences on the results of land-use allocation and assist policy makers to formulate land use structure optimization program.(2) Multi-scale land use optimization system structure is a framework for interaction between the individual Agent, which is constitute by three levels of organization of individual Agent, Agent group and Agent domain. In this study individual Agent is government, farmer, researcher and so on, they formed stakeholders Agent group, and the most crucial part of this system is the three Agent domains:theoretical concepts domain, analysis domain and model domain. Above the individual Agent there are two-stage management collaborators of "group" and "domain" which could effectively reduce the system’s internal conflict to coordinate problem and improved the efficiency of communication with more problem-solving ability. The "group" and "domain" is an open, dynamic concept in which the number of members is dynamic and the members can dynamically join or leave. The dynamic evolution process can be regarded as Agent changing with the environment and the target requirement, continuous coordination with the adaptation. In such a system to explore land use optimization program, we can effectively mobilize stakeholders, and land use optimization programs at every scale all result from repeated exchange of stakeholders, finally to achieve a real sense of the participatory planning.In summary, this article systematic study Optimal allocation of land resources under different scale, no longer limited to a single scale or tradition thinking that use certain mathematical method to get optimization program, built a land resource allocation system for multi-scale study, provided new ideas and methods for land use research in context of urban and rural areas, made up for the domestic shortage of land use study. However, the system requires constant improvement because certain scale of research can only carry out on assumption conditions. In addition, the coupling of three scales need the specific techniques to be achieved, so in the future we should continue to increase research and study of coupling in order to apply to practice.更多还原