ISWC OpenIR  > 水保所知识产出(1956---)
陕甘宁黄土高原区 NDVI 变化及其在土地利用 变化检测中的适宜性
张文帅
Subtype硕士
Thesis Advisor王飞
2014-05
Degree Grantor中国科学院研究生院
Place of Conferral北京
Keyword归一化植被指数 陕甘宁地区 土地利用变化检测 直接比较法 交叉相关光 谱匹配
Abstract

土地利用/覆盖变化(LUCC)被认为是全球环境变化与可持续发展研究的重要内容。
目前,地表土地利用/覆盖信息难以实现现势性,因此实时自动地进行变化检测是一个十
分紧迫的问题。近十几年,陕甘宁地区在一系列植被建设工程实施背景下,植被覆盖得
到明显改善,生态效应正在显现。这些植被建设工程可以归结为土地利用/覆盖类型的变
化,因此实时检测陕甘宁地区地类的变化对区域水土流失治理和生态环境建设具有重要
的理论和现实意义。本文利用 1999~2012 年 SPOT VGT NDVI 1km/10d 分辨率数据,采
用差值法、一元线性回归法和 Hurst 指数 R/S 分析法等方法,系统分析了近 14 年陕甘宁
地区植被覆盖时空变化特征及未来变化趋势;采用偏相关分析法,计算了 NDVI 与同期
降水、气温数据的关系,并结合土地利用数据和造林面积数据,分析研究区 NDVI 变化
和人类活动的关系,从而分析了 NDVI 变化与气候因子和人为因子的关系;在前两部分
的基础上,选择人为影响较大的区域作为变化检测研究区,采用直接比较法,分析了基
于粗分辨率 NDVI 数据土地利用/覆盖变化检测在陕甘宁地区的适宜性。研究结果表明:
(1)近 14 年来,陕甘宁地区年均 NDVI 整体呈极显著(P<0.001)上升趋势,生
长季、四季 NDVI 也呈极显著(P<0.01)上升趋势,增速由快到慢依次为夏季 >生长季 >
秋季 > 春季 > 全年 > 冬季。在空间上,NDVI 以轻微改善和中度改善为主,二者比
重之和为 98%以上,总体上研究区东南部改善大于西北部,其中延安市东北部地区、彭
阳固原地区以及沿河(秃尾河、无定河和北洛河)两岸地区改善最为显著(P<0.001);
呈现退化趋势区域比重仅为 0.05%,主要原因为城市化。
(2)陕甘宁地区的植被覆盖 NDVI 年内变化范围为 0.149~0.456,其中 2 月最低,8
月最高。年内变化曲线基本呈现开口向下的抛物线变化特征,经历缓慢增加(2~4 月)、
快速增加(4~5 月)、缓慢增加(5~8 月)、快速下降(8~11 月)、缓慢下降(11~次年 1
月)、平稳过渡(次年 1~2 月)6 个阶段。在空间上,从春季到夏季,“绿色植被带”在
研究区东南向西北方向上随时间变化逐渐加宽。
(3)近 14 年间,陕甘宁地区气候呈现不明显的“冷湿化”特征,气温降低和降水
量增加趋势均不明显,年际间波动较大。研究区 NDVI 变化由气候因素和人为因素共同
作用驱动,其中人为因素占主导地位,主要驱动因素为近十几年的退耕还林草工程、植
被建设以及农业生产水平的提高,其生态环境效应正在显现;降水对 NDVI 变化为正效应,植被覆盖越少的区域 NDVI 对降水敏感性越强,但人类活动可以掩盖这种正效应,
气温与 NDVI 没有明显的相关关系,主要表现其在对 NDVI 季节韵律的控制上。
(4)陕甘宁地区的主体土地利用类型为耕地和草地,分别占研究区总面积 35%和
47%以上,其次为林地,占 10%以上,其他地类占 8%以下。2000~2005 年 5 年内,耕地
大幅度减少,主要转化为林地和草地,使得林草地面积大量增加,同时存在垦荒现象;
草地总体覆盖度明显增加,由低覆盖度逐渐向高覆盖度转化。退耕还林草和草地自然恢
复为研究区 NDVI 增加提供了“绿色基础”。
(5)在榆林延安接壤区域,变化检测结果总体精度较差,Kappa 系数仅为 0.214,
因此应用这种方法在陕甘宁地区进行变化检测需要慎重;但在部分植被覆盖增加极显著
且 NDVI 与降水相关性较弱的区域有较高的精度,总体精度达 73.24%,Kappa 系数为
0.455。降水对变化检测结果的精度影响很大,有效剔除降水的影响是提高变化检测精度
的关键。
(6)变化检测结果误差来源主要包括三方面:a)混合像元问题,一个 1km×1km
的像元平均包含 3 种地类和 7 个斑块,一定包含丘陵旱地和草地两种地类,因此一个变
化像元中可能会包含未发生地类转变的部分,导致错分误差和漏分误差增大;b)同物
异谱和异物同谱问题,容易把“草地—草地”和“耕地—耕地”错判为变化类型,而把
“草地—耕地”和“耕地—草地”漏判为不变类型;c)变化检测指标和阈值确定方法
不合适,有待设计更加合理的变化检测指标和阈值确定方法,有效抑制各种外源“噪声”,
以达到准确地进行变化检测。
关键词:归一化植被指数;陕甘宁地区;土地利用变化检测;直接比较法;交叉相关光
谱匹配

Other Abstract

Land use and land cover change (LUCC) has been considered to be an important content
of global environment change and sustainable development research. At present, we find land
use and land cover information difficult to achieve present situation, so the real-time
automatic change detection is an extremely urgent problem. In recent years, at the background
of implementation in a series of vegetation construction projects in Shaan-Gan-Ning Region,
vegetation coverage has been improved significantly and the ecological effects become
obvious. These projects can be regarded as LUCC; therefore, the detection of LUCC in
Shaan-Gan-Ning Region means great important theoretical and practical significance to
regional soil and water conservation and ecological environmental construction. In this paper,
we compared the temporal and spatial change of vegetation in Shaan-Gan-Ning Region in
recent 14 years, covering the evolution and future tendency based on the data of SPOT VGT
NDVI 1km/10d resolution in 1999~2012, with difference method including linear regression
method and Hurst index R/S analysis method, etc. We calculated the relationship between
NDVI, precipitation and temperature within the same period by using partial correlation
analysis. Furthermore, we combined land use data and afforestation area data to analyze the
relationship between NDVI change together with climatic factors and anthropogenic factors.
On the basis of the two parts, we selected the artificial factors dominated area as the study
area to detect change of vegetation. We discussed the applicability of LUCC detection in
Shaan-Gan-Ning Region based on coarse resolution NDVI data by the direct comparison
method. The results showed that:
(1) The yearly average NDVI showed a very significantly increased trend (P<0.001) in
Shaan-Gan-Ning Region in recent 14 years. The growing season and seasonal NDVI was also
very significantly increased (P<0.01). The fast growth speed was summer, followed by
growth season, autumn, spring, annual and winter. NDVI changed slightly and moderately in
most area, which accounting for more than 98%. In general, the improvement in southeast
part is larger than northwest part of the study area, among which the northeast part of Yan’an
city, Pengyang-Guyuan area as well as the river area (Tuwei River, Wuding River and Beiluo
River) had a most significantly increased trend with a 0.001 significance. Only 0.05% area
had a degenerate tendency because of the urbanization.
(2) The annual average NDVI change from 0.149 to 0.456 within one year, which in
February was the smallest and the biggest in August in Shaan-Gan-Ning Region. The NDVI
variation curve within one year approximately existing as an down open side parabola, as
slowly increase (February to April), rapid increase (April to May), slow increase (May to
August), rapid decline (August to November), slowly decrease (November to next January),
smooth transition (next January to February)six stages. The "green vegetation zone" gradually
widened with the time from spring to summer and from southeast to northwest direction in
space in the study area.
(3) In recent 14 years, the climate became "cold and wet" non-significantly, the
temperature decreased and precipitation increased not obviously, it had an inter-annual
fluctuation in Shaan-Gan-Ning Region. The combination of climate factors and human factors
caused the NDVI change which was predominated by the human factors in study area. The
main driving factors were the recent years of conversion of cropland to forest and grassland
project, the vegetation construction and raising the level of agricultural production, as a result,
the ecological environment effect is appearing. Precipitation had a positive effect on NDVI
changes, NDVI respond strongly to rainfall in the less vegetation cover areas. But the positive
effect was masked by the impact of intense human activities. Temperature and NDVI had no
obvious correlation, which mainly performed in the control of the NDVI seasonal rhythm.
(4) The main land use types in Shaan-Gan-Ning Region were cultivated land and
grassland, accounting for more than 35% and 47% respectively. The third largest area in type
was forest which accounted for more than 10%, the proportion of other types were less than
8%. Cultivated land was greatly reduced that mainly converted to forest and grassland from
2000 to 2005. Forest and grassland area increased a lot, and there is the phenomenon of
reclamation. The overall coverage of grassland increased very significantly which
transformed from low to high coverage gradually. Conversion of cropland to forest and
grassland and natural recovery of grassland provided a ―green basis‖ for the NDVI increase in
the study area.
(5) The accuracy of change detection results was low in the border region between
Yan’an city and Yulin city, the Kappa coefficient was only 0.214. Therefore, it needs to be
cautious for us by using the change detection method in Shaan-Gan-Ning Region.
Nevertheless, there were high precision in the part of vegetation dramatically improved and
weak correlation between NDVI and precipitation regions. In this region, the overall accuracy
was 73.24% and the Kappa coefficient was 0.455. Therefore, precipitation has great influence
on the accuracy of change detection results. The key to improve the accuracy of change
detection is eliminated the influence of precipitation effectively.
(6) The main reasons of errors in change detection result in three aspects: a) the mixed
pixel problem. There are 3 types and 7 patches in each 1km × 1km pixel averagely; the land
use types must contain hilly dry land and grassland. Therefore a ―changed pixel‖ may contain
no changed part of land use types which lead to commission errors and omission errors. b)
The existence of same matter with different spectrums and different matters with same
spectrum made it easy to misjudge ―grassland-grassland‖ and ―farmland-farmland‖ as
changed state, and misjudge ―grassland-farmland‖ and ―farmland-grassland‖ as no changed
state. c) The change detection indexes and determining threshold method in this paper were
not suitable for LUCC detection which needs more reasonable design in the future so as to
restrain various kinds of ―noise‖ effectively and conduct change detection accurately.
Keywords: Normalized difference vegetation Index (NDVI); Shaan-Gan-Ning Region; Land
use change detection; Direct comparison method; Cross correlogram spectral matching
(CCSM)

Language中文
Document Type学位论文
Identifierhttp://ir.iswc.ac.cn/handle/361005/8995
Collection水保所知识产出(1956---)
Recommended Citation
GB/T 7714
张文帅. 陕甘宁黄土高原区 NDVI 变化及其在土地利用 变化检测中的适宜性[D]. 北京. 中国科学院研究生院,2014.
Files in This Item:
File Name/Size DocType Version Access License
陕甘宁黄土高原区 NDVI 变化及其在(6461KB)学位论文 开放获取CC BY-NC-SAApplication Full Text
Related Services
Recommend this item
Bookmark
Usage statistics
Export to Endnote
Google Scholar
Similar articles in Google Scholar
[张文帅]'s Articles
Baidu academic
Similar articles in Baidu academic
[张文帅]'s Articles
Bing Scholar
Similar articles in Bing Scholar
[张文帅]'s Articles
Terms of Use
No data!
Social Bookmark/Share
All comments (0)
No comment.
 

Items in the repository are protected by copyright, with all rights reserved, unless otherwise indicated.