IEECAS OpenIR  > 博士研究生论文
基于遥感的青海湖流域植被分类、空间分布与影响因素研究
王瑾
学位类型博士
导师安芷生 ; 蔡演军
2013-05
学位授予单位中国科学院研究生院
学位授予地点北京
学位专业第四纪地质学
关键词植被分类 遥感 植被空间分布 高寒灌丛 气候因子 地形 土壤 退化草
其他摘要         自然植被的分布一方面受到区域物理环境和气候的影响,另一方面自然植被
的分布与变化还在一定程度影响和作用于生态和环境。因此,自然植被的分布不
仅是生态和环境优劣的重要指示,更是生态和环境建设的核心内容。每一气候类
型或环境分区都有与之相适应的植被类型,因而,植被的空间分布是探讨植被-
气候相互关系最基础的环节。青海湖流域是青藏高原东北部重要的生态脆弱带,
其丰茂的植被是控制西部荒漠化向东部蔓延的天然屏障。查明青海湖流域植被的
空间分布特征,研究植被分布的影响因素,探讨植被动态变化的驱动因子,对于
认识青海湖流域生态和环境的现状、厘定青海湖流域植被与自然气候变化和人类
活动的影响具有重要的意义。本文以TM 遥感影像为主要数据源,以高分辨率遥
感影像为辅助数据,通过室内分析和野外实地验证,建立了TM 遥感影像植被分
类体系和决策树分类方法,编制了青海湖流域的植被类型分区图,分析了影响区
域植被空间分布的不同因子,基于具有代表性的峻河流域多时相遥感影像的解译
分析,探讨了流域高寒灌丛分布变化的驱动因子,得到以下主要结论:
1、对典型区域峻河流域TM 影像运用不同方法进行植被分类,对比结果表明,
在高寒半干旱山区,使用融入地形、植被指数等多源辅助信息的决策树分类方法
能够有效地解决光谱混淆问题,可将分类精度由非监督分类法的66.25%提高到
90.62%。以此为基础,在整个青海湖流域使用该方法,将地形特征、植被指数、
植被覆盖度和光谱特征融合到一起,并将决策树分类和非监督分类相结合、分层
提取与分区分类相结合,使青海湖流域TM 影像植被分类的总体精度达到
87.20%,Kappa 系数0.86,达到“非常好”,表明在峻河流域获取的分类方法可以
应用到整个青海湖流域,并能够获得良好的分类结果。
2、通过总结前人的研究成果,建立青海湖流域遥感影像植被分类体系,以2009
年TM 影像为基础,运用融入多源辅助信息的决策树分类方法,对青海湖流域遥
感影像分层分区进行解译,编制完成青海湖植被类型的分区图。分析不同植被类
型的分布特点,发现青海湖流域植被分布与海拔高度、坡度、坡向具有相关性,
每种植被的生长和发育都有一个最优地形组合:温性草原发育于海拔高度在
3300m 以下的平坡地区,高寒草原发育于3300-3600m 的平坡和缓坡地区,高寒
草甸发育于3600-4100m 的平坡和缓坡地区,高寒沼泽草甸发育于3800m 以上的
平坡和缓坡地区,高寒稀疏植被发育于大于4100m 的平坡和缓坡地区,河谷灌
丛发育于3600m 以下平缓的河谷地带,高寒灌丛发育于3600-4100m 陡斜的阴坡
和半阴坡地区。同时,研究表明不同的植被分布与不同的气候特征相联系,例如
在天峻地区气候干冷,植被组合更趋于高寒化,刚察地区相对暖湿,植被类型则
由温性植被逐渐向高寒过渡。
3、与土壤分布图进行对比分析,确认了土壤发育与植被类型分布之间的密切联
系,也就是:温性草原和高寒草原植被生长于栗钙土之上,高寒草甸分布在高山
草甸土区域,高寒稀疏植被则分布于海拔较高的高山草甸土和高山草原土之上,
高寒沼泽草甸和河谷灌丛下覆土壤类型为沼泽土,而高寒灌丛则与高山草甸土和
山地草甸土相对应。
4、选取月份相同的刚察县2000 年和2009 年的TM 影像进行NDVI 分析,并计
算植被覆盖度,结果表明,该区域退化草地主要分布在湖周,与超载过牧、盲目
垦荒和人为滥伐等人类活动因素密切联系,而且2000 年到2009 年来,发生轻度
退化的草地占刚察县陆地面积的2.39%,中度退化草地占2.51%,重度退化草地
占0.69%,草地退化较为显著,确认了人类活动是环湖地区草地退化的主要影响
因素。
5、通过对峻河流域1977 年、1990 年、2000 年、2009 年高寒灌丛面积提取和分
析,得到该地区高寒灌丛面积的变化与蒸发量和降水量比值呈反相关;同时与整
个青海湖流域的干旱强度具有反相关关系,揭示了自然气候变化对灌丛分布的可
能影响。
6、对青海湖流域部分区域的IKONOS-2 和GeoEye-1 高分辨率数据进行预处理
和解译,发现融入纹理信息和地形特征能够将高分辨率遥感数据分类精度由监督
分类的71.84%提高到92.63%,Kappa 系数由0.63 提高到0.89。应用高分辨率遥
感数据植被分类结果对同一地区TM 影像分类结果进行对比,两种分类结果的一
致性超过80%,进一步说明青海湖流域2009 年TM 植被分类结果是可信的。;           Vegetation distribution is largely affected by regional environmental changes.
Therefore, variations of vegetation distribution can serve as an indicator of regional
ecological balance and it provides important guidelines for environmental protection.
As vegetation types has its compatible climatic and environmental preferences, the
spatial distribution of natural vegetation became crucial to investigating the
interaction between climate variation and vegetation changes. Located in the
ecologically fragile zone in the northeastern Tibetan Plateau, the thriving vegetation
coverage of Lake Qinghai Lake catchment acts as a natural barrier for desertification
to the east. Thus, knowledge of its spatial distribution and driving forces will benefit
our understanding of the ecological status and climatic changes of Lake Qinghai basin.
This study established a vegetation classification system and its corresponding
interpretation procedure based primarily on high-resolution TM remote sensing
images and supplementary data. Accompanied by field investigation, we compiled
qualitative data to enable a construction of a vegetation distribution map. Based on the
analysis of the dominate control on it spatial distribution, we reached the following
conclusions:
1. Compared with the conventional approach, the new classification system using
ancillary datum improved the overall accuracy from 66.25% with supervised
classification method to 90.62%. Based on the successful application in mapping
Junhe sub-watershed, we constructed an analogous classification system for the
program of mapping vegetation which combined decision tree and unsupervised
classification technique and added ancillary information such as elevation, slope,
aspect, NDVI, vegetation coverage and spectral information into mapping vegetation
in Qinghai Lake Basin. The result shows that the overall accuracy can be improved to
87.20% and the Kappa coefficient to 0.86, achieved "very good".
2. Spatial distribution of vegetation in Qinghai Lake Basin is correlated to the
elevation gradient, slope and aspect: The most suitable terrain combination for the
growth of the temperate steppe is flat slope land at or below3300m above the sea level,
of the alpine steppe is flat and gentle slope land between 3300-3600m, of alpine
meadow is flat and gentle slope land between 3600-4100m, of alpine swamp meadow
is flat and gentle slope land above 3800m, of alpine sparse vegetation is flat and
gentle slope land above 4100m, of the shrub in plateau valley is flat valley land below
3600m, and of alpine shrub is north-facing or semi-north-facing slope land between
3600-4100m. In Tianjun, vegetation is more fit for alpine climate because of the
arid-cold climate in this region, while the vegetation in Gangcha shows a transition
from steppe in temperate to alpine meadow.
3. Comparison between various soil types confirmed a close relationship between soil
development and vegetation types. Temperate savannah and high latitudinal grassland
usually grow on chestnut soil, alpine meadow grows on alpine meadow soil, alpine
sparse grassland grows on mountain meadow soil, alpine swamp meadow and valley
shrub both grow on swamp soil, and alpine shrub grows on mountain meadow soil
and alpine meadow soil.
4. The NDVI analyses of TM images of Gangcha county in 2000 and 2009 indicate
that retreat of the steppe land mainly distributed along the perimeter of Lake Qinghai,
which is closely related with intensive anthropogenic activities including overgrazing,
deforestation and blindly reclamation. During the period between 2000 and 2009,
grassland shrinking became prominent in Gangcha county. Specifically, the area of
slightly retreated grassland covers 2.39% of that in Gangcha county, intermediate
retreated area is 2.51% and intensively retreated areas is 0.69%.
5. The changes of alpine shrub coverage is inversely correlated with E/P ratio, based
on the interpretation of alpine shrub coverage in this region in 1997, 1990, 2000 and
2009, respectively. The variations also exhibit inverse changes with the aridity of the
watershed, suggesting possible controls of natural climatic fluctuations.
6. Interpretation of high-resolution IKONOS-2 and GeoEye-1 imaging data of
Qinghai catchment suggests that combined texture information and topographical
analysis can increase the resolution from 71.84% to 92.63%, and the Kappa
coefficient from 0.63 to 0.89. Comparison between different interpretations of the
same region result in a consistency as high as 80%, which further validates the results
of the 2009. 
学科领域地质学 ; 环境科学
文献类型学位论文
条目标识符http://ir.ieecas.cn/handle/361006/2599
专题博士研究生论文
推荐引用方式
GB/T 7714
王瑾. 基于遥感的青海湖流域植被分类、空间分布与影响因素研究[D]. 北京. 中国科学院研究生院,2013.
条目包含的文件
条目无相关文件。
个性服务
推荐该条目
保存到收藏夹
查看访问统计
导出为Endnote文件
谷歌学术
谷歌学术中相似的文章
[王瑾]的文章
百度学术
百度学术中相似的文章
[王瑾]的文章
必应学术
必应学术中相似的文章
[王瑾]的文章
相关权益政策
暂无数据
收藏/分享
所有评论 (0)
暂无评论
 

除非特别说明,本系统中所有内容都受版权保护,并保留所有权利。