【作者】 刘荣；

【导师】 高敏华；

【作者基本信息】 新疆大学 ， 地图学与地理信息系统， 2010， 硕士

【摘要】 土地利用/土地覆被变化(LUCC)是全球环境变化的主要原因和重要组成部分,也是调控人类行为的科学决策依据。它广泛涉及到各个地区资源的有效开发与合理利用、耕地保护与粮食安全、社会经济的可持续发展及生态环境的治理与保护等一系列重大问题。土地利用变化研究一直以来是地理学及其相关学科研究的前沿与热点。数据与尺度问题是纵贯土地利用/土地覆被变化研究的综合性研究内容。因此,研究多尺度土地利用变化和驱动力之间的关系,揭示不同尺度下土地利用变化的驱动机制是当前土地利用/土地覆被变化研究的必然要求,并将对今后土地利用/土地覆被变化研究的发展产生重要影响。哈巴河县位于新疆额尔齐斯河流域中下游地区,是一个生态环境脆弱、自然资源比较丰富的贫困落后地区。随着经济的快速发展,人口数量的不断增加,土地利用的数量、质量、类型、结构也在不断的发生着变化。土地资源的开发利用与生态环境保护、农牧林协调发展等之间的矛盾日益突出,脆弱的干旱半干旱荒漠生态系统受到严重的干扰和破坏。寻求土地资源合理开发利用的途径,对该区域土地的可持续利用起着至关重要的作用。本研究选取哈巴河县为研究区,开展了土地利用格局的多尺度空间自相关分析和模拟。通过多尺度的空间自相关分析,分析研究区各土地利用类型及土地利用格局驱动因子在不同的尺度上所具有的空间自相关性。实现了用多尺度空间自相关统计方法在土地利用分布规律研究中的应用。然后选取主要的驱动因子模拟研究区土地利用格局,在不同尺度下定量分析土地利用变化的驱动因子对土地利用格局的作用程度,建立了在明确空间定位基础上的、综合集成的多尺度Logistic回归模型。最后在确定研究区域的最佳模拟尺度的基础上,构建AutoLogistic回归模型,在考虑空间自相关因素的情况下,对研究区土地利用格局进行模拟。本文的主要结论如下:(1)各土地利用类型及其土地利用驱动因子的空间自相关性随着权重距离的增加,其空间自相关性减弱。从小尺度至大尺度的聚合规模上,哈巴河县各种地类的Moran’s I值呈现出随研究尺度的增大空间自相关性增强的趋势。(2)通过多尺度Logistic回归分析显示,影响土地利用格局的空间因素表现出尺度效应。不同的模拟尺度下,入选的变量有所不同,各驱动因子对土地利用格局的作用程度也存在着一定的差异。结果表明,当模拟尺度在500m时,各地类模拟模型的拟合优度均达到了最高。因此,可以确定500m×500m是构建研究区土地利用格局模拟模型的最佳尺度。(3) AutoLogistic回归模型由于考虑了土地利用类型空间格局的空间自相关性特征,对土地利用格局具有较强的解释能力。在500m×500m尺度上运用AutoLogistic回归模型对土地利用格局进行模拟,哈巴河县各地类的ROC值分别为:耕地0.914、林地0.736、草地0.742、其他用地0.855、水利用地0.931、建设用地0.924,拟和度都大于0.7,得到了较好的拟合效果。认为纳入了空间自相关性后的AutoLogistic回归模型具有较强的解释能力和适用性,在模拟区域土地利用格局过程时是相对合理的,是值得应用与推广的一种进行土地利用格局模拟模型。更多还原

【Abstract】 As the major course and the principal constitution of the global environment change,changes in Land Use and Land Cover Change is also a factor from which we regulate our behavior and make proper decision. These involve a series of important issues.Including efficient exploitation of resource, protection of environment, maintenance of farmland, safety of food and sustainability of our social and economic development. So, in the fields of geography and others related, Land Use and Land Cover Change has been a hot spot recent years.Also,the problems of data and scale are very import in many all-around researches at all times. For now, it is critical to future development for us to research the relation between multi-scale of land use and their motivation and to unveil the mechanism.Habahe country which lays in the middle and lower reaches of the Ertysh River is a backward area with rich deposits but a fragile ecosystem. The sustained increase of economics there leads to a expansion in population, which cause several changes of amount, quality, type and structure in land use and have arouse more contradiction among land, stock and environment. Great devastation has been made to the brittle arid and semi-arid desert ecosystem .So, it very important to manage land use in a rational way. Still, only when we have a complete view of the process of land use change and thus expose the true driving forces can the dynamic model of vast spatial scale of land use been understood driving the research ahead.We choose Habahe country as the study area to analysis and simulate multi-scale spatial autocorrelation of land use patterns. We used the multi-scale spatial autocorrelation to analysis the space autocorrelation between land use types and land use driving factors at different scales in the study area. And also ,we achieved the study of using multi-scale spatial autocorrelation of statistical methods in land-use distribution.Secondly,Selected a certain simulation study of driving factors of land-use patterns, and to consider different scales of land-use change driving factors acting on different levels of land-use pattern, established a multi-scale Logistic regression model on the basis of a clear spatial location, comprehensive integrated multi-scale. At last, in determining the best analog scale in study area, we used the AutoLogistic regression model which is considering spatial autocorrelation to simulate land-use patternsThe main conclusions of this research are as follows:(1) Spatial autocorrelation degree of various land use types and land use driving factors reduced as the increasing of the weight; from the small scale to the large scale, space autocorrelation of Moran’s I of different land use types in Habahe County increased by the enhancing of the study scale.(2) Through a multi-scale Logistic regression analysis showed that the impact of land use pattern-scale spatial factors showed effects of different analog scale, the selected variables are different. And variables for describing the land-use type and driving factors of theβcoefficient of the variable relations also have some differences. Found that when analog scale is in the 500m ,the goodness of fit of the simulation model are up to the highest. Therefore, we can determine the 500m×500m is the best-scale to build simulation model in the study area land use pattern.(3) In the 500m×500m scale,local class ROC values are: cultivated land is 0.914, forestland is 0.736, grassland is 0.742, other land is 0.855, waterbody is 0.931, construction land is 0.924, all the goodness of fit are greater than 0.7, showed that we have a better fitting results. We thought that the AutoLogistic regression model which is considering spatial autocorrelation has strong applicability and explanatory power, it is relatively reasonable to simulated pattern of regional land-use change, and it is also a simulation model of land use which is worthy to be applied and promoted.更多还原