【作者】 柳晓华；

【导师】 周勇；

【作者基本信息】 华中农业大学 ， 土壤学， 2004， 硕士

【摘要】 传统的土地利用变化模型在处理空间格局时往往以国家或行政区作为基本单元，从而在很大程度上难以详细地反映土地利用斑块的动态变化规律。另外研究的结果往往存储在封闭独立的数据系统内，研究成果信息难以共享。本研究在对潜江市土地利用／覆盖变化进行分析过程中，引入了景观格局分析方法开展土地利用变化过程的研究；使用了通用的关系型数据库SQL Server来存储大量的土地利用信息的空间数据与属性数据，对数据的共享技术做了一些尝试；同时将遥感与地理信息系统技术有机地结合起来，综合运用计算机自动分类技术与人工目视解译，大大提高了土地利用分类的精度与效率。主要研究结果如下： 1、选取了1986年和2000年的两幅TM遥感影像作为提取两期土地利用景观类型的数据源。在没有高分辨率SPOT全色影像的情况下，尝试使用Landsat 7的第8波段(全色波段)代替SPOT全色影像，与分辨率为30米的多光谱图像进行了基于主成分变换的影像融合(分辨率为15米)，用来获得较准确的2000年土地利用景观分布图，并进行了精度评估，分类精度达到了90.64％。 2、采用了阈值法来提取实验区两期TM影像的水体界限；对于建设用地(城镇居民地)则采用基于光谱知识的城镇提取模型来提取，以修正监督分类的结果。对于水体的提取主要采用遥感图像处理软件ERDAS 8.5提供的空间模型模块，通过条件函数(Criteria Function)来反复实验提取水体的阈值(TM5＜45)；然后将重新提取的水体和建设用地导入ARC／INFO，使用UPDATE命令来更新监督分类的结果。从而找出了一条利用ERDAS与ARC／INFO的优势功能与人工目视解译相结合的方法来进行遥感图像的自动分类的途径，提高了分类精度。 3、应用样区遥感图像解译的结果，将ARC／INFO下的INFO数据库导入到其他数据库软件中进行了面积统计和相关景观指数的计算。最终获得了土地利用转换矩阵，从而可以分析各种土地利用景观的变化，结合景观格局分析来找出影响潜江市土地利用景观变化的主要因素。为实时动态监测土地利用变化找出了一条有效的途径。为农业产业结构的调整以及城市化进程的合理调控等提供依据。 4、通过土地利用转换矩阵和景观指数的计算，获得了如下有意义的结果：①潜江市境内有大量的农场，地块平整而规则，耕地的地块反而不那么破碎，相比而言水面和林地则比较破碎，水面由原来的规则的湖面转化成大量零星的鱼塘与不规则的水产养殖水面，水面面积大幅度增加，增幅达12.08％；②林地主要分布在潜江市的北部和主要支流的周围，一般由与大的支流和防洪大堤相伴的人工林和天然林场构成。由于各建制镇和潜江市区的城市化进程在过去的14年中不断加快，某些城市面积扩张了近1／3；③在研究中引入了景观生态学的景观格局分析方法，交叉使用斑块和景观两个尺度来分析潜江市土地利用景观类型整体的空间结构特征和变化过程。结果表明在斑块尺度上两期土地利用景观都以农田为主，占整体面积的66%，其他景观镶嵌其中。两期TM影像对比，农田、林地、未利用地均有不同程度减少，而水面和建设用地(人文景观)则有所增加，特别是林地减少特别快，两期对比林地减少9.6%，而且斑块面积和数量变小，而人文景观增加明显，增幅达到3.06%。在景观尺度上多样性指数和均匀度指数都有所下降，人工干扰的结果是造成景观多样性下降的主要原因。分析潜江市土地利用景观动态变化可以看出，城市化进程和农业产业结构的调整和政策导向是改变景观格局的根本原因，相反自然因素作用甚小。更多还原

【Abstract】 The processing of spatial pattern of traditional land use change model is always constrained in taking the canton as a basic unit and is difficult to reflect the dynamic change principle of land use change in detail. In the process of studding land use change of Qianjiang city in Hubei Province, a landscape pattern analysis is introduced into this study for exploring of the land use change in Qianjiang city. Additionally, the result of traditional land use change study is often stored into an independent and isolated database system so as to hardly be reused and produce a little benefit as well as share the information in public. This study tries to change this embarrass situation and firstly use the fashionable RDBMS to store the land use information that contains large number of spatial and nonspatial information. Meanwhile, synthetically using the automatically classification by computer and interpreting by expert experience and integrating the remote sensing and GIS technologies, the accuracy and efficiency of land use classification is improved greatly. The main research results are shown as followings:1. For this study, the TM images of the year 1986 and 2000 are selected as the data source of the extracting the landscape patches. Without the high resolution images of SPOT, this study tries to use a resolution merge with the pan band (the eighth band of TM 2000) instead of the pan band of SPOT imagines that are commonly used in other studies. Based on this, the image resolution is conducted by means of principal component, and implemented with a supervised classification to acquire the land use map of the year 2000 later. The research results indicate that the accuracy of land use classification reaches 90.64%, which is met with the requirement of land use classification2. In the process of studying the whole land use and landscape of the Qianjiang City, a threshold method is adopted with a limit value of 45 for image translation of the water area, while for image translation of the residential area a model based on the spectral knowledge is selected. All the results will update the classification by overlapping and intersecting based on the spatial analysis of topology. Especially, the remote sensing imagery processing software ERDAS 8.5 is used with its spatial analysis modular. Through Criteria Function of ERDAS Imagine, this research teststhe threshold value repeatedly to gain a fine result with an accuracy coming to 91.5%, and it is imported to the ARC/INFO database to modify the classification by the command of UPDATE. Thus, it is proved that a method taking advantage of ERDAS and ARC/INFO to integrate the classification automatically by computer and other good ways as well as interpreting by expert experience can overcome the low accuracy of classification by computer.3. Using above-gained results of image translation, the INFO tables in ArcGIS are exported to database software and an area statistics is conducted by grouping land use code and other criteria. After that, the land use change matrix and relevant landscape patter index are obtained to analyze the factor that result in the land use change of specific types. In this way an effective approach for the real-time and dynamic monitoring of the land use change and controlling the land use change is formed.4. With the land use transformation matrix and some landscape pattern index selected in this research, some significant results are shown as followers: (1)This study sample plot is The Qianjiang City with large number of farms, and the patches is regular other than fragmented in the county. Comparing with the water and forestland, the water land is changing with trivial stews and irregular aquatic water land, and the water area is increased largely with a value of 12.08% and the water area is fragmented. (2)The forestland mainly is distributed at the northern and ambient river branches. With the faster and faster urbanization in past fourteen years some the construction land is increased 33% in the area. (3)With the two period ima更多还原