【作者】 刘爱霞；

【导师】 王长耀；

【作者基本信息】 中国科学院研究生院（遥感应用研究所） ， 地图学与地理信息系统， 2004， 博士

【摘要】 荒漠化是世界瞩目的严重生态问题之一，它直接影响到人类经济的发展和社会的稳定。研究荒漠化的分布现状和动态变化趋势，将为荒漠化的治理提供重要的科学依据，而荒漠化监测指标的研究是其关键问题之一。 本文在综合分析荒漠化的研究现状和趋势的基础上，在荒漠化指标选取原则指导下，基于前人的研究成果，建立了一套适合于大尺度荒漠化监测的遥感指标体系，使用NOAA数据和MODIS数据定量反演了5个荒漠化遥感监测指标，分别对中国及中亚地区1995年和2001年的荒漠化分布状况进行了监测评价，并对7年以来的荒漠化动态变化进行了分析；同时利用1982-2000年NOAA—AVHRR NDVI时间序列数据，分析了中亚及中国干旱半干旱区的荒漠化多年变化。 主要研究成果与创新点如下： 1)提出了荒漠化遥感监测综合指标，并建立了监测指标体系 在前人研究荒漠化指标的基础上，结合荒漠化指标的选取原则和本文的研究目的，提出了适合于应用遥感进行大尺度荒漠化监测的荒漠化监测指标，并通过对监测指标和指标组合进行荒漠化监测精度的分析得出，改进型土壤调节植被指数、植被覆盖度、反照率、陆面温度和TVDI(土壤湿度指标)的组合在荒漠化遥感监测中的分类精度最高。根据荒漠化气候类型的不同，把中国和中亚地区划分为亚湿润干旱区、半干旱区、干旱区和高寒区四个区域，并对每一区域分别建立了不同的荒漠化程度遥感监测指标体系。 2)选择了适用的荒漠化遥感反演方法实现了研究区荒漠化空间分布特征的分析 利用NOAA和MODIS数据，采用适合于进行大尺度遥感定量反演的方法，对5个荒漠化遥感监测指标进行反演，得到1995年每旬和2001年每16天的各个监测指标数据，建立了荒漠化遥感监测指标数据库。并对反演出的各个荒漠化监测指标在中国以及中亚地区的分布特征进行了分析。 3)确定了荒漠化最佳分类方法，实现了中国荒漠化现状评价和动态监测 在荒漠化监测试验区科尔沁沙地，通过对非监督分类、最大似然法(MLC)、和决策树三种分类器的精度比较，得出决策树的分类精度最高。在荒漠化遥感监测指标体系的建立和各个监测指标反演结果的基础上，利用生长季MSAVI累积值、生长季平均反照率、生长季平均陆面温度、年最大植被覆盖度和生长季平均TVDI值共5个指标，用决策树分类方法对中国1995年和2001年的荒漠化现状进行监测分析，并对1 995一2001年的荒漠化的动态变化情况进行了分析。结果表明，中国荒漠化呈整体扩展，局部改善的趋势;荒漠化发展的速度大于逆转的速度。 4)实现了中亚地区荒漠化现状评价和动态监测 中亚地区对中国干早区半干旱区的生态环境有重要影响。通过使用基于中国荒漠化土地样本建立的荒漠化遥感监测指标体系，对中亚地区相对应的四个分区1995年和2001年的荒漠化状况进行了监测评价，并分析了1995一2001年的荒漠化动态变化情况。按照国家进行荒漠化土地统计表明，中亚6国的荒漠化发展速度大于逆转速度，呈整体扩展，局部改善的趋势。 5)采用植被萌芽事件分析植被与非植被区界线多年变化 使用1982一2000年skm分辨率的NOAA-AVHRR NDvilo日合成时间序列数据集，根据区域内是否发生过植被萌芽事件，估计了每年中亚及中国干早半干旱区的植被与非植被区的界线，并对18年的植被区界线进行GIS的叠加，分析了多年的界线变化情况。干旱半干旱区的沙漠界线具有很强的变化性，在沙漠的中心地带从来没有监测到有植被的分布，然后从沙漠经沙漠草原，然后过渡到典型草原带，有植被萌芽事件发生的频率逐渐增大。 6)利用NDVI变异系数(CoV)分析了中亚及中国干旱半干旱区荒漠化动态变化规律 根据1982一2000年NOAA-AVHRRNDvllo日合成时间序列数据，分别计算了每年的NDvi变异系数(CoV)，然后用最小二乘法求得18年的NDvi CoV坡度，即CoV多年变化趋势。通过分析COV坡度的变化，对中亚及中国干旱半干旱区的多年荒漠化状况进行了评价分析。并通过分析中国干旱半干旱区多个气象站的降雨和温度数据表明，荒漠化发展趋势不仅与降水有一定关系，而且与近年来温度的升高密切相关;而在有人类活动干扰的区域，每个像元的COV坡度变化揭示出了土地荒漠化趋势同时受人为因素的严重影响。更多还原

【Abstract】 Desertification is one of the most serious ecological and environmental problems in the world. It directly influences regional economic development and social stability. Understanding the distribution and development trend of desertification provides us important scientific basis for desertification control and rehabilitation, in which desertification index system is a key indicator of the level of desertification dynamic change.Based on investigation of current research and previous efforts on desertification and guided by the design principle, this dissertation proposes a desertification index system suitable for large-scale desertification monitoring using remote sensing techniques. First, five desertification indices were retrieved from NOAA and MODIS satellite data. Then, the desertification status and the dynamic changes from 1995 to 2001 in central Asia and China were analyzed. In addition, we delineated the annual boundary between vegetated and non-vegetated areas and assessed the desertification status in arid and semiarid region of central Asia and China based on the NOAA-AVHRR NDVI time-series dataset from 1982 to 2000.Main research results and initiatives in this thesis include as following:1. An integrated desertification indexes are proposed and the construction of a desertification monitoring index system using remote sensing techniques is recommended.According to the desertification index design principle and the research aim of this dissertation, we selected five desertification indexes (MSAVI, FVC, Albedo, LST and TVDI) suitable for large-scale desertification monitoring using remote sensing technique. After applying different index and index combinations on desertification monitoring and its precision evaluation in test area, the result shows that the precision of index combination of MSAVI, FVC, Albedo, LST and TVDI is superior than others. In term of the desertification climate types, the potential extent of desertification in China and central Asia was respectively divided into four categories: dry sub-humid area, semi-arid area, arid area, high and cold area. Different desertification index system was built for each area.2. Selection of a suitable method on remote sensing retrieval of desertification indexes andthe spatial distribution of desertification was analyzed in study area.Based on analysis and comparison of current retrieval algorithms, we utilized a suitable algorithm on large scale to retrieve five desertification indexes with ten-day NOAA AVHRR data set in 1995 and 16-day MODIS data set in 2001, and built the database of desertification monitoring indexes in China and central Asia. At the same time, we analyzed the spatial distribution characteristic of the indexes in China and central Asia.3. An optimal classifier for desertification monitoring was determined and desertification status and dynamic change analysis in China were evaluated and analyzed.In test area Kerqin desert, by assessing the classification accuracies of three types of classifiers(unsupervised classifier, maximum likelihood classifier and decision tree classifier), we select decision tree classifier for desertification monitoring. Supported by desertification index system and the database of desertification indexes, utilizing the cumulated MSAVI values, averaged albedo, averaged land surface temperature and averaged TVDI in growing season, and the annual maximum fraction of vegetation cover in China, the desertification status in 1995 and 2001 was classified by decision tree classifier, and analysis of desertification changes from 1995 to 2001 was also completed in study area. The result shows that there is a trend of development as a whole accompanied with occasionally local improvement in desertificated areas in China, the desertification extension is faster than rehabilitating.4. Evaluation of desertification status and desertification dynamic monitoring in central Asia. Central Asia has great influence on the ecological environment of China, especially in thear更多还原