土地利用覆盖

中文题名土地利用/土地覆盖变化空间信息的遥感和地理信息系统方法研究 副题名 外文题名 论文作者庄大方 导师刘纪远研究员 学科专业地图学与地理信息系统 研究领域\研究方向 学位级别博士 学位授予单位中国科学院遥感应用研究所 学位授予日期2001 论文页码总数148页 关键词地理信息系统 环境遥感 土地利用 土地覆盖 馆藏号BSLW /2003 /P208 /19 【中文摘要】 在全球气候变化日益影响着人类生产生活诸多方面的今天，土地利用/土地覆盖变化也日益成为当今地球系统、人类生存环境和区域可持续发展研究的热点。而在土地利用/土地覆盖变化的研究领域里，遥感和地理信息系统技术起着举足轻重的作用，这是不争的事实。本文在遥感和地理信息系统技术的支持下，对土地利用和土地覆盖时空信息的获取、处理与分析利用作初步探讨。
文章首先综述了当前国内外土地利用/土地覆盖变化研究的现状，以及国内外在这一领域的研究理论、方法、最新进展及主要成果。在此基础上，论述了国内外土地利用/土地覆盖信息的基本特点及几种主流的土地利用分类指标体系。进而探讨了土地利用/土地覆盖信息的遥感获取方法、土地利用/土地覆盖数据库建立、土地利用/土地覆盖变化信息系统集成技术、土地资源遥感调查中的静态信息采样方法、耕地城镇动态变化空间采样技术以及基于我国资源环境数据库的中国人口空间化模型。文章前一部分重在土地利用/土地覆盖变化研究的理论方法上的探讨，而后一部分则把重点放在基于现有土地利用/土地覆盖变化数据库的应用研究，以下是由这些研究工作得出的初步的结论：
(1)在GIS系统下，采用多时相遥感影像与数字地学影像复合的综合分类方法，进行大、中尺度植被分类时，地理数据与遥感光谱数据产生的新数据集，改变了遥感数据的单一光谱信息结构，丰富了图像处理的信息源。通过地理数据与遥感光谱数据的复合，可以以影像的方式，将地表植被覆盖状况的内在成因和外在表现有机地结合起来，改善植被分类的精度。基于遥感技术的土地利用与土地覆盖变化调查，其关键技术是图像分类处理获取信息的方法。目前，遥感图像的分类技术远远跟不上遥感技术本身的发展，时至今日，成功的能支持实用系统的分类方法仍是目视解译。
(2)土地资源信息在计算机辅助下的人工解译判读目前主要依赖于判读专家的长期实践经验，而后期的干扰地物的去除则主要依赖于根据实际情况而制定的抽样方案，TM细小地物抽样与航片细小地物抽样是相辅相成，互为补充的抽样方案的两个

部分。实践证明，TM细小地物抽样在平原地区实施效果较好，但在丘陵山区则难度较大，而航片细小地物抽样对对除去TM影像无法侦测到的细小地物具有明显的作用。
(3)经典采样调查技术与空间采样方法的发展为稳定区域的空间分析提供了有力工具。但经典统计采样方法用于空间对象的动态变化的监测研究，尤其是针对全部国土面积的监测，尚存在着采样样本太大和精度无法满足需求的难题。传统采样方法需要大量的资料、人力、财力，更为重要的是需要花费大量的时间，从而降低资源调查的时间分辩精度。多重空间采样框架为实现特定采样对象从全国范围内的原始随机小样本采样到特定区域的随机大样本采样提供了依据及实现手段，使得针对空间信息的采样样点分布更为符合空间对象的分布特征，从而达到提高采样精度的目的。多重采样框架为正态分布理论应用于空间对象采样数据分析提供了依据。随机采样的空间差异性不仅体现在样本统计参数的变化，也体现在分布方式随空间范围的大小而不同。基于多重空间采样框架的样本估计能有效地提高采样调查精度，节省采样样点，保证资源调查的时间分辨率。
(4)遥感与空间分析技术支持下的耕地城镇动态采样方法的最大优势在于对遥感和采样方法进行了比较好的结合，从示例研究来看，和采用全覆盖的遥感监测相比，监测使用的遥感数据减少了3/4，监测的工作量相应减少3/4。同时，由于采用遥感和空间采样相结合的方法，使得针对整个国土面积的大样本随机采样得以实现，克服了传统采样方法针对整个国土面积无法保证随机抽样的缺点，并保证了相应的采样精度。本方法不仅适用于耕地城镇的年际动态监测，同时也适用于特定空间对象的年际动态监测。
(5)离散点数据的连续空间化问题一直是将某些社会经济离散点数据与其他地理空间数据相结合，应用于各项实际研究的瓶颈问题。本文通过全国县级行政单位人口数据的空间化过程，一方面可以重新得到各种类型区人口居住密度的预设值或经验值；另一方面，可以将空间化后的人口数据与其他自然资源、环境数据和社会经济等数据进行有效的融合，将能得到1km格网的人均自然资源占有量、人口密度以及其他经济、环境指标的派生数据信息，这对于我们国家的自然资源管理和宏观调控、人口的动态监测等，均具有重要的实际意义。另一方面，本研究提出的人口点数据的连续空间化方法是基于多层次空间数据及社会经济数据的综合研究的有益的尝试，为此类工作积累了可资借鉴的经验。

【外文摘要】 Abstract
Today, when global

climate change plays an increasingly influential role in all aspects of human's life, Land-Use/Land-Cover Change (LUCC) gradually turns into the highlight in the research field of earth system, environment system and sustainable development. It is self-evident that remote sensing and Geographical Information System (GIS) techniques function as the main force in LUCC studies. This paper, with the support of remote sensing and GIS, carries out a preliminary study on the acquisition, processing and application of geographic temporal-spatial data. Based on a review of the LUCC study's present situation, theory, methodology and major advancement, the paper first comes to the characteristics of LUCC and some major land use/cover classification systems, hence the remote sensing acquisition of LUCC information, the development of LUCC database, the static/dynamic sampling methods for land resource survey and the population spatialization of China. The first Part of the paper is mainly addressing the theoretical and methodological aspects of LUCC. For the second part, major attention is laid on the LUCC database and its application. The following conclusions have been reached:
(1) In the study of large/middle-scale land-cover classification, the integrated method of composing multi-temporal remote sensed imagery with digital geographic images was adopted to generate new data sets, thus the single spectrum information structure of remote sensing was changed, which enriched the information source and made it possible that the classification accuracy would be improved. Among the remote sensing LUCC survey techniques, the key is to acquire information by image processing. However, the techniques of remote sensed imagery classification are far behind the remote sensing techniques. Still, the efficient classification method is visual interpretation.
(2) The computer-assisted visual interpretation of land-cover information mainly relies on interpreter's experiences, but the finalization of statistical measurements of area depends on the sampling schema based on real situation. On one hand, the sampling of small features on TM imagery was efficient in the plain area, but the sampling of small features on aerial imagery is efficient in the hill area, and they are complementary. On the other hand, the sampling of small features on TM imagery and the sampling of small features on aerial imagery are different in the small feature's size. How to apply a unique sampling method to a unique area to achieve satisfactory results deserves further investigation.
(3) The multiple sampling frame makes it possible to sample from small random samples in the country scale to large random samples in a specific area. The multiple sampling frame also makes the spatial information sampling meet the distribution attributes of spatial features, which improves the sampling accuracy. Also, based on the multiple sampling frame, the normal distribution theory can be applied to

analyze the data resulted from the spatial feature sampling. The spatial discrepancy of random sampling is embodied not only in the variations of statistical sample parameters, but also in the variations of the distribution attributes related to different spatial scales. The sample estimation based on multiple sampling frame can efficiently improve the sampling accuracy, decrease the sampling points and guarantee the temporal resolution of land resource survey.
(4) The dynamic farmland and urban area sampling method based on remote sensing and GIS has a great advantage, which combines the remote sensing and spatial sampling technology. From the case study, compared with the full area remote sensing monitoring method, this method reduced the remote sensing data used to one fourth, so is the survey workload. At the same time, the combination of remote sensing and spatial sampling method actualized the random sampling on large samples at the country level and secured the sampling accuracy. This method is not only suitable for monitoring annual changes of farmland/urban area, but also suitable for monitoring the annual variation of some specific spatial features. As for the instable features, some spatial analyzing method are to be studied, such as spatial interpolation, the control scope of sample points, the size of the sample area, error conductivity model and accuracy analysis.
(5) Discrete point data can be applied with other GIS data to address some specific questions in our researches. But how to spatialize the discrete point data has been a question itself. In this paper, the case study- the spatialization of China's population at the county level in the country scale can, on one side, obtain the population density of all kinds of terrain features across China. On the other side, the population density data at the county level can be integrated with environment data and social-economic data to generate 1-km grid data sets of average population density, average capacity of natural resources and others derived from environment data. And these data could be useful to monitor China's resource management, population dynamics and so on. The example study has demonstrated a plausible method of wisely using discrete point data to carry out some practical studies. The study of population spatialization also urges that some GIS software containing population data input, analyzing and updating modules be developed, thus bridges the geography and dermograhy, and actualizes the studies of population digitalization and spatialization.