EN中文

Journal of Geo-information Science >

2017 , Vol. 19 >Issue 10: 1355 - 1363

DOI: https://doi.org/10.3724/SP.J.1047.2017.01355

The Breakpoints Detection Method Using Time Series of Vegetation Fractional Coverage

  • WANG Enlu ,
  • WANG Xiaoqin , * ,
  • CHEN Yunzhi
Expand
  • Key Laboratory of Spatial Data Mining & Information Sharing of Ministry of Education, National Engineering Research Centre of Geo-spatial Information Technology, Fuzhou University, Fuzhou 350116, China
*Corresponding author: WANG Xiaoqin, E-mail:

Received date: 2017-01-19

  Request revised date: 2017-06-14

  Online published: 2017-10-20

Copyright

《地球信息科学学报》编辑部 所有
Fold

Abstract

Detecting breakpoints plays an important role in plotting and analyzing time series of the changing characteristics such as firing, logging, diseases and insect pests in vegetation. It is a useful technique of extracting the significant information in time series data. We focused on the method of Detecting Breakpoints and Estimating Segments in Trend (DBEST). We studied the detection of vegetation breakpoints by using vegetation fractional coverage (VFC) data which is derived from MODIS NDVI remote sensing images ( 250 m) from 2000 to 2015 in Changting County of Fujian Province. In order to determine if the results of breakpoints detection are reasonable, the primary experiment is to test the applicability of DBEST method by using the VFC data of various changing types in time series. We select several samples of time series data which covered the key water and soil erosion conservation area. The vegetation changes more frequently in this area for conducting the break-points detection experiments. We make an accuracy evaluation of changing time and changing types by using the temporal trajectories and Landsat remote sensing images of every point. We find that DBEST is suitable for VFC time series data of Changting, by using the default first and second level-shift-thresholds (θ1 = 0.1, θ2 = 0.2) which indicated that DBEST could define the changing level of VFC, but the duration-thresholdφ should be adjusted according to the study area and the type of time series data (we setφ=3). Those parameters have weak influences on the accuracy of breakpoints positions, but have more effects on the changing types of breakpoints. On the whole, the excessive intervention is not necessary for detecting vegetation in DBEST. However, through a lot of experiments we believe that the threshold of the changing magnitude can be modified by our own need to gain a satisfying results. Finally, we set β = 0.2 to fit our own research targets. The precision of the changing time is 92%, greater than the changing types (80%), indicating that DBEST method works well in extracting the important changing information for VFC time series. Meanwhile, the experimental results are broadly consistent with the varying conditions of the local vegetation.

Key words： breakpoints detection; time series; MODIS NDVI; vegetation fractional coverage; DBEST

Cite this article

WANG Enlu , WANG Xiaoqin , CHEN Yunzhi . The Breakpoints Detection Method Using Time Series of Vegetation Fractional Coverage[J]. Journal of Geo-information Science, 2017 , 19(10) : 1355 -1363 . DOI: 10.3724/SP.J.1047.2017.01355

1 引言

时间序列断点检测通常被认为是一种时间序列分割和分段趋势估计的过程,其意义在于可以更细致地对数据进行研究,如时序数据的简化与存储以及分段线性表示等[1-2]。通过遥感技术获取的植被指数（Vegetation Index, VI）时间序列数据为陆地植被覆盖变化研究提供了极大的便利,在处理和研究时间序列遥感VI数据集的过程中,时序总体趋势估计能够判断一段时间内植被的变化趋势,如最常用的普通最小二乘方法（Ordinary Least Squares, OLS）将时间序列进行一元线性拟合,用线性方程的斜率描述变化趋势[3-5]。但往往时序VI数据的变化趋势并不总是平缓地、循序渐进地上升或下降,对植被而言,当受到病虫害、火烧、砍伐或雨雪冰冻等事件影响时[6-7],其VI值会产生较大波动,而断点检测方法能够提取这些重要特征为后续的研究和分析提供帮助,因此人们一直致力于研究能够有效提取时序遥感植被数据集中变化信息的方法[8-9]。
早期的变化信息提取方法大多聚焦于双时相或多时相的变化检测,这类方法在时间序列数据上的应用具有很大的局限性,而断点检测技术能够完成连续变化信息的提取,逐步成为了研究时序遥感数据处理方法的热门方向之一[10]。目前的断点检测多是基于时序轨迹的分析,通常用于调查大区域范围内的植被变化和土地覆盖变化等[11]。曾用于气候、气象等研究的经典断点检测方法如滑动t检验法[12]、Mann-Kendall检验法[13]等,多用于检测数据序列中迅速而显著的突变情况;有学者运用这些方法检测时序NDVI数据中的均值突变,进而发掘植被的变化特征[14],但它们有时会得到不一致的检测结果[15],对植被变化规律的把握和分析造成了一定程度的困扰。还有一些广泛用于生物、水文和气象等领域的断点检测方法[16-19],这些方法通常对数据服从的分布和序列长度等方面有特殊的要求,因此在遥感数据上的应用有一定的局限性。
针对以上断点检测方法存在的问题,有学者提出了一些专用于处理时序遥感数据的断点检测方法,如Kenndy等[20-22]提出的LandTrendr（Landsat-based detection of Trends in Disturbance and Recovery）方法和Verbesselt等[23-24]提出的BFAST（Breaks For Additive Seasonal and Trend）方法是目前时间序列遥感数据断点检测研究中较为常见的方法。LandTrendr方法是专为Landsat数据设计的森林干扰检测算法,可以较为准确地捕捉植被的突发性变化和逐步的缓慢退化,但需要调控的参数较多;BFAST是用于分析时序遥感植被专题产品数据的断点检测算法,同样可以检测植被的突变和渐变,并能够提供断点位置的估计区间,但一般适用于密集型长时间序列植被参数数据。由此可知,这些方法受限于卫星传感器、数据类型和数据长度等方面的因素,其通用性有待完善。Jamali等[25]提出了一种不限卫星传感器类型和数据长度的、用于时间序列遥感植被产品数据断点检测和分段趋势估计（Detecting Breakpoints and Estimating Segments in Trend, DBEST）的方法,可以处理周期性和非周期性时序VI数据,使用较为简便且具有较好的通用 性。目前与DBEST方法相关的应用较少,本文以福建省长汀县为例开展DBEST断点检测方法 研究。

2 实验区概况及数据

2.1 实验区概况

实验区福建省长汀县地处福建省西部,地形以丘陵为主,处于中亚热带海洋性季风气候区,雨量充沛且降雨多集中在春夏两季。由于长汀县地形较为破碎,高强度的降水易造成以花岗岩风化物为主的土壤受到侵蚀。历年来各级政府对长汀县水土流失治理十分重视,采取了一系列的治理措施大力开展水土保持工作,目前已初见成效。自然因子和人为因素的共同作用,使不同时期长汀县的植被变化呈现多样性,另自2000年以来的水土流失治理资料可作为宝贵的验证数据,因此本文选取长汀县作为实验区,其区位示意图如图1所示。图1中浅黄色的区域即为长汀县水土流失重点治理范围,红色的点为断点检测试验所选取的若干样点,大部分均匀分布在治理区域内用于评价检测效果。

2.2 研究数据

研究选用2000-2015年空间分辨率为250 m、时间分辨率为16 d的MODIS NDVI产品,经过S-G滤波（Savitzky-Golay）后采取像元二分法进行植被覆盖度（VFC）的估算[26],获得值域在0到1之间的植被覆盖度时序数据,其计算如式（1）所示。
VFC = NDVI - NDV I min NDV I max - NDV I min （1）
式中： V I max 代表完全被植被覆盖的像元值; V I min 代表完全没有植被覆盖的像元值;VFC为所求植被覆盖度。计算完成后再进行年际时间序列影像最大值合成（Maximum Value Composite, MVC）处理,可以较好地避免噪声和植被生长季相等因素产生的干扰,从而准确地反映地表植被覆盖变化的情况[27-28]。
Fig. 1 Location map of Changting County, Fujian Province

图1 福建省长汀县示意图

3 研究方法

3.1 断点检测方法

DBEST模型[14]由趋势估计（提取）和趋势分割2部分组成,趋势提取是针对时间序列VI数据的预处理过程,而断点检测主要通过趋势分割来实现,技术流程如图2所示。
Fig. 2 Flowchart of DBEST

图2 DBEST模型技术流程

3.1.1 趋势估计
趋势估计过程的核心思想是首先根据第一、第二水平变化阈值和持续时长等参数（表1）估计时间序列VI数据的水平变化点（level-shift-point）,所得水平变化点代表时间序列中最明显的变化特征,是估计子序列趋势的重要参考依据;然后,利用基于Loess的周期-趋势分解（Seasonal-Trend decomposition procedure based on Loess, STL）方法将原始时间序列分解为周期项、趋势项和余项,并用趋势项参与趋势分割计算。需要注意的是,如果原始数据是非周期性时间序列,则不需要STL分解过程,可直接进行水平变化点的提取和趋势分割。
DBEST算法设置的全部调控参数如表1所示。第一、第二水平变化阈值、距离阈值和持续时长等参数可以根据数据本身的特性选取,也可以使用DBEST算法推荐的阈值。通常情况下用户仅需要指定序列中断点的数目或者压缩率,就能得到相应的断点检测结果。
Tab. 1 Thresholds used in DBEST

表1 DBEST参数

阈值 含义
第一水平变化阈值 序列中水平变化点和下一点间最小的差值绝对值
持续时长 相邻水平变化点之间的最小时间步长
第二水平变化阈值 水平变化点前后子序列均值最小的差值绝对值
距离阈值 相邻波峰、波谷之间的连线和相距最远数据点间的最小垂直距离（注：DBEST可估计的默认值）
断点数目 最主要的或最感兴趣的断点数目
变化级别 子序列的最大简化程度或认为发生变化的最小级别
压缩率 对原数据序列进行最大化压缩的比率
显著性水平 用于检验变化的显著性
3.1.2 趋势分割
DBEST要求数据序列长度N至少大于2,趋势分割的过程主要分5步：
（1）先利用f函数获取波峰、波谷点
f ( i ) = 1 , sign ΔV I ( i - 1 , i ) = - sign Δ V I ( i , i + 1 ) 0 , 其他 （2）
其中， ΔV I （ i - 1 , i ) = V I ( i ) - V I ( i - 1 （ 3 ） ΔV I （ i , i + 1 ) = V I ( i + 1 ) - V I ( i ) （ 4 ）
式中：sign为符号函数,所有f=1的点是时间序列中的波峰、波谷点,而f=0的点包括无明显变化的点和在可能存在大幅变动的转点。
（2）使用g函数获取转点
g ( i ) = 1 , sign ΔV I ( i - 1 , i ) = - s ign ΔV I ( i , i + 1 ) 1 , f ( i ) = 0 且 d ( i ) > ε 0 , f ( i ) = 0 且 d ( i ) < ε （5）
式中：g函数的作用是标记f=0的点中大于距离阈值（表1）条件的转点,DBEST给出了默认的距离阈值,也可以根据实际需要自定义。
（3）使用h函数获取局部趋势变化（Trend Local Change,TLC）
h （ i ） = 0 , g ( i ) = 0 V I ( z ) - V I ( i ) , g ( i ) = 1 （6）
式中：TLC是相邻2个转点i和z（即g=1的点）之间的差值,非转点的TLC值被赋为0,全部计算完成后将TLC序列按降序排列,TLC值与变化幅度成正比。
（4）根据贝叶斯信息准则（Bayesian Information Criterion, BIC）推测断点数量
BIC准则能够推测合适的断点数量,可显著降低时间序列最小二乘拟合的误差,而又不会造成过拟合现象。
（5）分段最小二乘线性拟合
利用最小二乘法求解由各个转点之间的斜率构成的线性方程组,得到分段趋势估计结果。其中,线性拟合用到的断点数量可以由BIC准则得到,也可根据需要由用户自定义。

3.2 断点检测实验方案设计

3.2.1 植被变化类型归类
植被的变化主要源于自然因素和人为因素的共同干扰,本文将其归为突变、渐变和混合变化3种类型（表2）。突变的含义是植被在短时间内发生的较为剧烈、显著的变化,如火烧、砍伐、收割和栽种等,则对应遥感影像上的VFC数值也会表现出较大的上升或下降幅度;渐变往往是指植被的持续变化过程,其VFC数值不会产生瞬间、大幅的振荡。然而,实际情况往往不会只出现一种变化类型,且时间序列越长,植被覆盖度的变化过程就可能越复杂。
Tab. 2 Main changing types of vegetation in Changting County, Fujian Province

表2 福建省长汀县植被变化类型

植被变化类型 突变类型 渐变类型 混合类型
举例 森林火烧、砍伐或人工种植等 自然恢复或
土地沙化等		 火烧后自然恢复、经砍伐后人工造林或反复火烧等
3.2.2 样点的选取
长汀县的地形和气候特征加之人为因素造成的植被退化、土地沙化情况较为严重,自2000年以来各级政府和水保部门不断开展长汀县的水土流失治理工作,其中植树造林是水土保持工作的重要手段之一。考虑到重点治理区域内的植被变化类型较为丰富,因此侧重选取位于治理区域内的样点利用DBEST算法进行断点检测试验,样点分布如图1所示。
3.2.3 断点检测分析与评估方式
DBEST断点检测结果的分析和评估可由2部分内容组成：① 算法试验阶段,首先利用选取的样点开展断点检测实验,以时序轨迹可视化视图、长汀县水土流失重点治理区域矢量图（包含16年来的治理时间和治理措施）和Landsat影像等进行检测结果的精度评估和分析;② 在处理单个时间序列数据的基础上,逐像元获取断点的空间分布特征,同样结合辅助数据分析所得结果的准确性和植被变化的原因。

4 结果与分析

4.1 断点检测实验

DBEST算法有2种输出模式,即Generalisation和Change Detection。二者的区别在于,Generalisation模式可以按用户指定的压缩率或断点数量进行时间序列分割,并对不同分割条件下的子序列进行拟合,得到一种时序简化结果;Change Detection模式是将序列中波动较大的断点全部输出,以及关于每个断点的详细信息,如起始时刻、持续时长、终止时刻、变化类型、显著性等。但2种输出模式使用的核心算法基本一致,用户可以根据不同的需要选择合适的输出模式。本文将其输出表达方式进行改进,结合了Generalisation和Change Detection模式对检测结果进行表示和分析。
DBEST原文中利用以月为时间间隔的VI测试数据进行初始实验时,将第一水平变化阈值θ1设为0.1,第二水平变化阈值θ2设为0.2,变化持续时长 φ为24（即两年）,变化级别阈值β设为0.1,收到了较好的断点检测效果,可以运用在气候干旱、植被稀少的区域（如伊拉克）,对植被的任何微小变化都较为敏感。但本文的实验区长汀县植被覆盖率很高,且发现VFC在0-0.2之间的变化在原始影像上的反映不够明显,如果仍以β=0.1作为TLC的筛选条件,则会提取一些变化幅度处于0.1-0.2之间的、非感兴趣的变化信息,不利于有效检测植被发生重大变化的事件。因此,合理地使用DBEST模型开展植被断点检测,需综合考虑时序植被参数数据自身的特点和感兴趣的变化级别等,来适当地调整模型参数。
根据经验,长汀县植被遭受砍伐、火烧等干扰后,需要3-4年时间恢复到与发生干扰前相近的植被覆盖水平。本文使用的是最大值合成的以年为时间间隔的VFC数据,故将φ调整为3,同时θ1和θ2分别为0.1、0.2保持不变可有效判别断点类型（大于0.2为均值突变）;再将β阈值条件设为0.2（为获取更明显的变化）即可获取感兴趣的断点位置信息。长汀县某样点时间序列VFC数据断点检测实验如图3所示。
图3分为2部分,图3（a）是某单点时序数据的可视化表示结果,包括时序轨迹、断点位置、断点类型和变化幅度等信息;图3（b）-（d）是对应空间位置上的TM影像,通过观察比空间分辨率为250 m的MODIS影像更清晰的TM原始影像,有助于校验断点检测结果的准确性。此外,由于MODIS影像在长汀县范围内的地形表征不明显,因此本文不考虑MODIS VFC数据在250 m空间尺度下带来的地形效应。
图3（a）中,黑色轨迹为原始时间序列,蓝色轨迹为经过断点分割后的简化时间序列。从图中可以看到有2个变化非常明显的点,位于2003年和2004年,其中2003前、后3年的2个时间段的植被覆盖度均值发生了剧烈变化,属于突变类型,用红色实心点表示;从2004年后植被逐渐开始恢复,属于渐变类型,用红色空心点表示。
图3（a）中时间序列轨迹下方是局部趋势变化（TLC）柱图,红色柱子代表每个潜在断点的变化程度,与输入参数中的第一、第二水平变化阈值相对应,柱子越高表示相对应的位置变化越剧烈,根据TLC的值来划分断点级别。选取第二水平变化阈值为0.2（变化点之间最小差值为0.2）来检测发生剧烈变化的时间点,可以看到2个变化最为明显的点TLC绝对值均超过了0.2,印证了断点检测结果的合理性。选择断点变化前后的Landsat影像验证了所检测断点的准确性（图3（b）-（d））。
Fig. 3 Examples and verification of breakpoint detection

图3 断点检测结果示例与验证

4.2 结果与分析

为了进一步测试DBEST断点检测结果的准确性与合理性,将50个样点结合水土流失治理专题图和Landsat影像分别进行断点检测结果的主客观分析,主要从变化的时间（包括开始、结束时间和持续时长）与变化类型2个方面对所提取的变化信息进行评估,统计数据如表3所示。
表3反映的统计精度是对每个样点时间序列断点检测结果结合时序轨迹和Landsat影像进行主客观分析得到的,变化时间误检的情况多体现在时序曲线的细节处,如连续的波动干扰等。变化类型主观判定不合理的因素主要是与Landsat影像目视解译结果不符,如大量采伐或人工种植等短期内发生的事件使植被覆盖度突然降低或升高的情况,应表现为突变类型,但由于初始最小步长的设置不具有完全适用性或下一时段的数值恢复到原来水平的速率较快导致该点两侧VFC的均值水平变化没有达到0.2,造成检测结果有偏差（部分偏差仅处于毫厘之间）;同时,变化类型的区分非常严格,误检的结果都是突变类型被认为是渐变类型,没有出现同一时间序列反复、多次误检的情况。
Tab. 3 Accuracy of breakpoints detection

表3 断点检测精度评估

变化时间（起、止和持续时长） 变化类型
合理 46 40
一处不合理 3 10
两处及以上不合理 1 0
精度/% 92 80
从表3可知,DBEST检测变化时间（断点位置）的准确性要略好于变化类型,分析原因主要有如下2点：① 第一、第二水平变化阈值和持续时长等参数的初始化和植被变化幅度之间的对应关系会存在差别,如森林火烧事件造成植被覆盖度的突然降低应属于突变类型,但相同的变化类型在影像不同区域上反映出的VFC数值下降幅度和后续植被恢复速率可能存在差别,这种差别会对检测结果造成直接的影响;② 由于DBEST算法能够很好地捕捉时间序列曲线的波峰、波谷点和转折点,这种特性对于检测发生重要变化的时间点具有较大优势,但同时由于局部的变化级别由TLC决定,如果时间序列存在频繁的随机波动则会增加断点位置误检的可能性。

4.3 实验区植被变化时空表达与分析

DBEST断点检测的验证和分析结果表明,该方法对于处理年际时间序列植被覆盖度的影像数据具有很好的适用性,能够有效检测出时间序列中存在的变化时间、变化类型等信息。图4是利用DBEST断点检测结果合成的福建省长汀县植被时空变化情况。
Fig. 4 Spatial-temporal visualization of the vegetation in Changting County, Fujian Province

图4 福建省长汀县植被变化时空表达

图4（a）表示包含突变、渐变特征（单一、混合并存）、无明显变化和非植被区域的空间分布; 图4（b）表示图4（a）中存在变化的时间序列中最大的变化幅度,按设定的第二水平变化阈值（0.2）对断点的TLC值进行划分得到4个变化级别,其中TLC值处在0~0.2和-0.2~0之间的像元被归到无明显变化的类别中;图4（c）表示图4（b）中对应像元变化开始的时间;图4（d）表示图4（c）中对应像元变化持续的时长;无明显变化和非植被区域作为掩膜在图4（a）-（d）的空间范围保持一致。
图4以时间序列VFC数据中变化幅度最大的断点属性为基础,提供了包括断点类型、变化幅度和变化时间等信息的时空一体化表达和分析方式,能够直观地反映一些显著的植被变化特征,如 图4（a）中长汀县西北至东南方向的红色轨迹表示其植被覆盖度最主要的变化存在突然下降的情况,再观察图4（b）-（d）可看到下降幅度超过了0.4,且变化开始时间处于2004-2005年,持续时长为1-2年,对比查看2004-2007年的Landsat影像可知植被覆盖度下降原因应是2004-2005年开山修路等事件所致。

5 结论与讨论

本研究以福建省长汀县2000-2015年植被覆盖度数据作为实验数据,并结合长汀县2000-2010年水土流失治理矢量数据和时序Landsat影像为验证数据,开展了DBEST断点检测方法研究。主要结论如下：① DBEST算法对于长汀县年际时间序列VFC数据具有较好的适用性,变化时间和变化类型的检测精度分别达到了92%和80%,总体上效果较好;② DBEST参数设置无需过多人为干预,调节θ1、θ2和φ等参数会决定断点的变化类型（突变或渐变）,但不会影响断点位置;③ 变化级别阈值β应依据感兴趣的变化幅度进行调整,本文将植被覆盖度的β值设为0.2能够得到更为合理的检测结果;④ 为降低误检机率、获取良好的实验结果,不建议使用波动过于频繁（例如无去噪处理）的年际时间序列植被参数数据。
由于DBEST模型不限制数据类型和数据长度,本文选用了序列长度较短的、非周期性的、基于最大值合成的年际VFC数据参与实验,如果同时采用周期性密集长时间序列数据进行断点检测结果的辅助验证,并在此基础上选择更多的样点序列开展进一步的精度评定,则对于DBEST算法性能的考察应会更为全面,相应得到的植被变化结论也会更具说服力。此外,图4所示的可视化方法能够较好地反映实验区植被最主要变化的信息,但针对时间序列中多个断点属性的表达和分析方式仍有待进一步的思考和研究。
虽然DBEST模型具有较好的通用性和可移植性,但若要尽可能高精度地获取植被变化信息。首先,需要确保实验资料的准确性,如果研究区地形复杂,而所用的遥感专题数据空间分辨率又较高,则不可避免地要讨论地形等因素对断点检测结果的影响;其次,要根据实验数据的类型调整模型参数,如持续时长φ应根据时序数据的时间间隔和当地植被覆盖特点做出适当的改动。总体来看,DBEST断点检测算法能够较好地提取植被变化信息,具有较广阔的应用前景。

The authors have declared that no competing interests exist.

References

Publishing order | Descend order by publishing year | Descend order by cited within
[1]
Pratt K B, Fink E.Search for Patterns in compressed time series[J]. International Journal of Image and Graphics, 2011,2(1):89-106.We describe a technique for fast compression of time series, indexingof compressed series, and retrieval of series similar to a givenpattern. The compression procedure identifies "important" points of a series and discards the other points. We use the importantpoints not only for compression, but also for indexing a database oftime series. Experiments show the effectiveness of this technique forindexing of stock prices, weather data and electroencephalograms.

DOI

[2]
廖俊,周中良,寇英信,等.一种基于重要点的时间序列分割方法[J].计算机工程与应用,2011,47(24):166-170.分段线性表示是时间序列降维的有效方法，其关键在于分割点的确定。在时间序列分段线性表示的基础上，提出一种新的基于重要点的时间序列分割方法。与一般方法比较相邻三点关系不同的是，将时间窗扩展为前一重要点、待考察点和一个指定时间窗组成的区间，再通过比较数据点前后模式变化来确定重要点。通过与其他7种分割方法进行实验比较，证明该方法适应能力强，不但分割结果总体质量高，在压缩率相同时具有更小的拟合误差，而且能够有效滤除噪声，发现时间序列的模式特征。 <BR>

[ Liao J, Zhou Z L, Kou Y X, et al.Method for time series segment based on important point[J]. Computer Engineering and Applications, 2011,47(24):166-170. ]

[3]
梁爽,彭书时,林鑫,等. 1982-2010年全国草地生长时空变化[J].北京大学学报(自然科学版),2013,49(2):311-319.基于遥感技术,利用1982—2010年间NOAA-AVHRR卫星获得的归一化植被指数(normalizeddifference vegetation index,NDVI)数据对全国29年来的草地生长动态变化及其与温度、降水的相关关系进行分析。结果表明:29年来,全国草地生长季NDVI总体上呈显著增加趋势(R2=0.25,p0.01),尤其是在1982—1999年间(R2=0.60,p0.01),而自20世纪90年代末开始失速。值得注意的是,最近十几年(1999—2010年)草地生长季NDVI变化趋势的大小和方向基本上取决于起始年份的选择,因而这一阶段草地生长的总体趋势不显著。不同地区、不同时段生长季NDVI变化趋势的空间分布存在较大差异。草地生长在1982—1999年间呈广泛(约占85%)增加趋势,在1999—2010年间出现较大范围(约占50%)的逆转,尤以干旱、半干旱地区的草地植被生长下降趋势最明显。水热组合状况是影响这一变化的主要原因。生长季平均NDVI与气候因子之间的相关关系分析表明,全国大部分地区草地生长变化主要受降水驱动,对于部分高寒和湿润、半湿润地区的草地,生长季NDVI与温度关系密切,温度上升对该地区草地生长有利。

[ Liang S, Peng S S, Lin X, et al.NDVI-Based spatial-temporal change in grassland growth of China from 1982 to 2010[J]. Acta Scientiarum Naturalium Universitatis Pekimensis, 2013,49(2):311-319. ]

[4]
丁明军,张镱锂,刘林山. 1982-2009年青藏高原草地覆盖度时空变化特征[J].自然资源学报,2010,25(12):2114-2120.利用GIMMS和SPOT VGT两种归一化植被指数(NDVI)数据对青藏高原地区1982&mdash;2009年期间草地覆盖的时空变化进行研究,结果如下:①青藏高原草地植被覆盖的年际变化存在着显著的空间差异。趋于升高的区域主要分布在西藏的北部和新疆的南部;趋于下降的地区主要分布在青海的柴达木盆地、祁连山、共和盆地、江河源地区及川西地区。②青藏高原草地覆盖度年际变化趋势分析表明,在90%的显著性检验水平上,降低和增加面积的比率为0.31,草地植被覆盖水平总体趋于升高态势。③以10 a为步长的分析表明:草地盖度呈现持续增加的区域主要分布在西藏北部;阿里地区草地盖度表现为先减少后增加;雅鲁藏布江流域草地盖度呈现先增加而后减少;而持续减少的区域主要分布在青海省以及川西地区,其中青海省分布最广;统计结果显示,高原大部分地区草地盖度具有升高的态势。

DOI

[ Ding M J, Zhang Y L, Liu L S.Temporal and spatial distribution of grassland coverage change in Tibetan plateau since 1982[J]. Journal of Natural Resources, 2010,25(12):2114-2120. ]

[5]
孙庆龄,李宝林,许丽丽,等. 2000-2013年三江源植被NDVI变化趋势及影响因素分析[J].地球信息科学学报, 2016,18(12):1707-1716.基于2000-2013年三江源MODIS NDVI数据,本文系统地分析了三江源植被生长季累计NDVI的时空变化特征,并结合三江源生态保护与建设工程实施的相关统计数据,探讨了人类活动对三江源植被变化的影响,最后通过气候因子与生长季累计NDVI的相关性分析,揭示了影响三江源不同地区植被变化的主要气候限制因素.结果表明,2000-2013年三江源植被NDVI整体上呈增加趋势,NDVI明显增加的区域面积比例达17.84%,主要分布于研究区的西部和北部;明显减少的区域仅占0.78%,多零星分布于研究区中部;NDVI变化稳定或没有显著变化趋势的区域面积比例为59.64%,主要位于研究区东部和南部.三江源生态保护与建设工程的实施虽然促进了植被恢复,但对区域植被整体变化的影响有限,研究时段内区域植被整体好转主要受气候因素控制.西部长江源区的植被生长主要受气温影响,东北部黄河源区主要受降水制约,南部澜沧江源区降水和气温的限制性均不明显.

DOI

[ Sun Q L, Li B L, Xu L L, et al.Analysis of NDVI change trend and its impact factors in the three-river headwater region from 2000 to 2013[J]. Journal of Geo-information Science, 2016,18(12):1707-1716. ]

[6]
Abbes A B, Essid H, Farah I R, et al.Rare events detection in NDVI time-Series using Jarque-Bera Test[J]. Geoscience & Remote Sensing Symposium, 2015:338-341.Nowadays, Normalized Difference Vegetation Index (NDVI) time-series has been successfully used in research regarding global environmental change. NDVI time series have proven to be a useful means of indicating drought-related vegetation conditions, due to their real-time coverage across the globe at relatively high spatial resolution. In this paper, we propose a method for detecting rare events in NDVI series. These events are particularly rare and infrequent which increases their complexity of detecting and analyzing them. The proposed method is based on the analysis of the random component by Jarque-Bera test to verify the presence of rare events and obtain their features (time and amplitude). For validation, we have used a database for regions in Northwestern of Tunisia. These data come from MODIS for a period from 18 February 2000 to 17 November 2013 at a spatial resolution of 250 m by 250m.

DOI

[7]
杨伟,张树文,姜晓丽,等.基于MODIS时序数据的黑龙江流域火烧迹地提取研究[J].生态学报,2015,35(17):1-11.

[ Yang W, Zhang S W, Jiang X L, et al.The burned area mapping for Heilongjiang basin based on MODIS time series data[J]. Acta Ecologica Sinica, 2015,35(17):1-11. ]

[8]
Cai S, Liu D.Detecting change dates from dense satellite time series using a Sub-Annual change detection algorithm[J]. Remote Sensing, 2015,7:8705-8727.Multi-temporal satellite images are available at very high revisit frequency, allowing the characterization of land-cover transitions and trajectories in greater detail. However, most change detection methods aim to capture a snapshot of land cover, for instance on an annual scale, which do not describe changes that occurred between the annual time points. In this study, we present a sub-annual change detection (SCD) approach to detect change dates from dense satellite time series. SCD estimates change dates by analyzing differences between two consecutive annual segments in two steps. To validate the proposed method, SCD was applied to real and simulated time series of MODIS 16-day NDVI from 2000 to 2012 for MODIS tile h11v05 and a sub-area in southeast Ohio, USA. The results show that SCD can successfully detect from dense time series the dates of changes representing land-cover transitions among various vegetated land covers. In addition, SCD can achieve comparable accuracy on the date of abrupt land-cover changes for the real time series tested in the study area when compared with the trend and seasonal changes identified by a Breaks for Additive Seasonal and Trend (BFAST) method. Future effort will be given to apply the proposed approach to various remotely sensed time series for other areas with extreme climates.

DOI

[9]
Yuan Y, Meng Y, Lin L, et al.Continuous change detection and classification using hidden Markov model: A case study for monitoring urban encroachment onto farmland in Beijing[J]. Remote Sensing, 2015,7:15318-15339.In this paper, we propose a novel method to continuously monitor land cover change using satellite image time series, which can extract comprehensive change information including change time, location, and from-to information. This method is based on a hidden Markov model (HMM) trained for each land cover class. Assuming a pixel's initial class has been obtained, likelihoods of the corresponding model are calculated on incoming time series extracted with a temporal sliding window. By observing the likelihood change over the windows, land cover change can be precisely detected from the dramatic drop of likelihood. The established HMMs are then used for identifying the land cover class after the change. As a case study, the proposed method is applied to monitoring urban encroachment onto farmland in Beijing using 10-year MODIS time series from 2001 to 2010. The performance is evaluated on a validation set for different model structures and thresholds. Compared with other change detection methods, the proposed method shows superior change detection accuracy. In addition, it is also more computationally efficient.

DOI

[10]
Goodwin N R, Coops N C, Wulder M A, et al.Estimation of insect infestation dynamics using a temporal sequence of Landsat data[J]. Remote Sensing of Environment, 2008,112(9):3680-3689.The current outbreak of mountain pine beetle ( Dendroctonus ponderosae ) in western Canada has been increasing over the past decade and is currently estimated to be impacting 9.2million hectares, with varying levels of severity. Large area insect monitoring is typically undertaken using manual aerial overview sketch mapping, whereby an interpreter depicts areas of homogenous insect attack conditions onto 1:250,000 or 1:100,000 scale maps. These surveys provide valuable strategic data for management at the provincial scale. The coarse spatial and attribute resolution of these data however, make them inappropriate for fine-scale monitoring and operational planning. For instance, it is not possible to estimate the initial timing of attack and year of stand death. In this study, we utilise eight Landsat scenes collected over a 14year period in north-central British Columbia, Canada, where the infestation has gradually developed both spatially and temporally. After pre-processing and normalising the eight scenes using a relative normalisation procedure, decision tree analysis was applied to classify spectral trajectories of the Normalised Difference Moisture Index (NDMI). From the classified temporal sequence of images, key parameters were extracted including the presence of beetle disturbance and timing of stand decline. The accuracy of discriminating beetle attack from healthy forest stands was assessed both spatially and temporally using three years of aerial survey data (1996, 2003, and 2004) with results indicating overall classification accuracies varying between 71 and 86%. As expected, the earliest and least severe attack year (1996), recorded the lowest overall accuracy. The relationship between the timing of stand attack ( i.e. moderate to severe beetle infestation) and NDMI (initial year of detected disturbance) was also explored. The results suggest that there is potential for deriving regional estimates of the year of stand death using Landsat data and decision tree analysis however, a higher temporal frequency of images is required to quantify the timing of mountain pine beetle attack.

DOI

[11]
Gong J, Ma G, Zhou Q.A review of multi-temporal remote sensing data change detection algorithms[J]. Protein Expression & Purification, 2008,82(2):308-316.ABSTRACT Change information of the earth's surface is becoming more and more important in monitoring the local, regional and global resources and environment. The large collection of past and present remote sensing imagery makes it possible to analyze spatio- temporal pattern of environmental elements and impact of human activities in past decades. Research has been widely reported on methodology of remote sensing change detection and monitoring. The present reviews have assorted the detection approaches and drawn many useful conclusions. Based on the former classification methods, this article classifies change detection methods from its essence into seven groups, including direct comparison, classification, object-oriented method, model method, time-series analysis, visual analysis and hybrid methods. At the same time, in pre-processing, the effect and methods of geometric correction and radiometric correction is discussed, and in accuracy assessment, this paper summarize the present methods into exterior check and interior check and emphasis on that how to get the ground truth. The challenges that the change detection is currently facing and possible counter measures are also discussed.

DOI

[12]
周顺武. 近50年拉萨夏季降水趋势和突变分析[J].西藏科技,1999(4):35-39.

[ Zhou S W.Analysis of climatic trend and jump of summer rainfall in Lhasa during recent 50 years[J]. Tibet Science & Technology, 1999,4:35-39. ]

[13]
Goossens C H, Berger A.Annual and seasonal climatic variations over the northern hemisphere and europe during the last century[J]. Annales Geophysicae, 1986,4(4):385-400.Provides a method which allows the detection of an abrupt climatic change and which localizes approximately the date of its beginning. The identification of such a change is made by means of a non-parametric test, the Mann-Kendall rank statistic. The efficiency of this test has been compared to the efficiency of other usual techniques used for detecting a change and the reliability of the localization of the change has been analysed using a Monte-Carlo technique. The data used are the long-term series of average northern hemisphere temperature, the instrumental records of temperature and precipitation for western european countries and the long-term european reconstructed series. Most of the temperature series exhibit a general long-term warming which supports the reality of the early 20th century northern hemisphere warming, as noted by many authors. However, this study shows that an abrupt change towards a warming occurred generally around 1900+/-25 years and that we are still experiencing this longterm warm climate.

DOI

[14]
孟猛,宗美娟.基于NDVI的气候突变对植被影响研究[J].安徽农业科学,2011,39(24):14942-14944.

[ Meng M, Zong M J.Study on the influence of climate abrupt variation on the vegetation based on NDVI[J]. Journal of Anhui Agricultural Sciences, 2011,39(24):14942-14944. ]

[15]
王园香,唐世浩,郑照军. 1982-2006年中国5-9月的NDVI变化与人类活动影响分析[J].地球信息科学学报, 2015,17(11):1333-1340.植被是联系陆地、大气和生态系统的自然纽带,其随着气候和人类活动而发生变化,因此,研究它的突变和变化趋势具有重要意义。本文利用3年滑动t检验、Mann-Kendall检验（MK检验）和距平分析法,研究了中国1982-2006年5-9月平均的NDVI（NOAA/AVHRR GIMMS）突变和变化趋势及其主要原因。3年滑动t检验和MK检验表明,1998年华东地区的NDVI出现了突变,而东北地区和青藏高原NDVI没有出现突变。NDVI变化趋势的分析表明,1982-1998年华东地区NDVI为较平稳的趋势,1998年出现突变后,1998-2006年转为明显下降的趋势（以1998年为转折点）。因此,华东地区NDVI存在明显的变化趋势。NDVI突变和变化趋势的成因分析表明,20世纪90年代后期至21世纪初,随着华东地区大规模的城市化建设和房地产的过度开发,导致耕地面积减少,华东地区植被1998年出现了突变,并从偏多转为明显偏少的趋势,而卫星仪器和气候因子并非是导致该地植被出现突变和变化趋势的主要原因。

DOI

[ Wang Y X, Tang S H, Zheng Z J.Analysis of NDVI and the impact of human activity in China from May to September during 1982 to 2006[J]. Journal of Geo-information Science, 2015,17(11):1333-1340. ]

[16]
Bleakley K, Vert J P. The group fused Lasso for multiple change-point detection[EB/OL]. (2011-06-21) [2017-06-14]. https://arxiv.org/pdf/1106.4199v1.pdf

[17]
汪丽娜,陈晓宏,李粤安,等.水文时间序列突变点分析的启发式分割方法[J].人民长江,2009,40(9):15-17.水文时间序列突变点的检测是水文统计分析中的一项重要内容,但由于水文时间序列的非线性、非平稳性,导致对检测水文时间序列突变点的技术要求较高.建立了时间序列变点分析的启发式分割数学模型,以此来分析水文时间序列均值的突变.以理想的水文时间序列为仿真基础,结合合水水库的月径流序列,采用启发式分割算法得出去除周期成分后的月径流序列的突变点,以此证明了启发式分割算法检测水文时间序列突变的可行性和优越性.结果表明:4个突变点中,3个突变点集中于水库建成后不久,说明人类活动的剧烈影响为主要驱动力,加剧了对入库月径流量的影响,造成月径流的频繁波动和数次突变点的产生.

DOI

[ Wang L N, Chen X H, Li Y A, et al.Heuristic segmentation method for change-point analysis of hydrological time series[J]. Yangtze River, 2009,40(9):15-17. ]

[18]
封国林,龚志强,董文杰,等.基于启发式分割算法的气候突变检测研究[J].物理学报, 2005,54(11):5494-5499. ]气候系统的非线性、多层次性和非平稳性对气候突变的检测方法提出了较高的要求.基于t检 验将非平稳序列分割为多个不同尺度的自平稳子序列，Bernaola Galvan提出的启发式分割 算法(BG算法)，对非平稳时间序列的突变检测效果较好.在BG算法的基础上，通过理想时间 序列验证BG算法处理非平稳时间序列的有效性，并对近2000a北半球树木年轮距平宽度序列 基于不同层次的思想，检测和分析其中包含的各种尺度的气候突变事件，成功地区分不同尺 度的突变.定义的新物理量——突变密度的分析表明，自然因素作用的基础上，人为因素影 响的加剧可能导致近1000a来突变密集段和稀疏段分布失衡，这可能是全球变化的重要表现 之一.

DOI

[ Feng G L, Gong Z Q, Dong W J, et al. Abrupt climate change detection based on heuristic segmentation algorithm[J]. Acta Physica Sinica, 2005,54(11):5494-5499. ]

[19]
Zhao X, Chu P S.Bayesian change point analysis for extreme events (typhoons, heavy rainfall, and heat waves): an RJMCMC approach.[J]. Journal of Climate, 2010,23(5):1034.A hierarchical Bayesian framework is developed to identify multiple abrupt regime shifts in an extreme event series. Specifically, extreme events are modeled as a Poisson process with a gamma-distributed rate. Multiple candidate hypotheses are considered, under each of which there presumably exist a certain number of abrupt shifts of the rate. A Bayesian network involving three layers09”data, parameter, and hypothesis09”is formulated. A reversible jump Markov chain Monte Carlo (RJMCMC) algorithm is developed to calculate posterior probability for each hypothesis as well its associated within-hypothesis parameters. Based on the proposed RJMCMC algorithm, a simulated example is designed to illustrate the effectiveness of the method. Subsequently, the algorithm is applied to three real, rare event time series: the annual typhoon counts over the western North Pacific (WNP), the annual extreme heavy rainfall event counts at the Honolulu airport, and the annual heat wave frequency in the Chicago area. Results indicate that the typhoon activity over the WNP is very likely to have undergone a decadal variation, with two change points occurring around 1972 and 1989 characterized by the active 196009“71 epoch, the inactive 197209“88 epoch, and the moderately active 198909“2006 epoch. For the extreme rainfall case, only one shift around 1970 is found and heavy rainfall frequency has remained stationary since then. There is no evidence that the rate of the annual heat wave counts in the Chicago area has had any abrupt change during the past 50 years.

DOI

[20]
Kennedy R E, Cohen W B, Schroeder T A.Trajectory-based change detection for automated characterization of forest disturbance dynamics[J]. Remote Sensing of Environment, 2007,110:370-386.Satellite sensors are well suited to monitoring changes on the Earth's surface through provision of consistent and repeatable measurements at a spatial scale appropriate for many processes causing change on the land surface. Here, we describe and test a new conceptual approach to change detection of forests using a dense temporal stack of Landsat Thematic Mapper (TM) imagery. The central premise of the method is the recognition that many phenomena associated with changes in land cover have distinctive temporal progressions both before and after the change event, and that these lead to characteristic temporal signatures in spectral space. Rather than search for single change events between two dates of imagery, we instead search for these idealized signatures in the entire temporal trajectory of spectral values. This trajectory-based change detection is automated, requires no screening of non-forest area, and requires no metric-specific threshold development. Moreover, the method simultaneously provides estimates of discontinuous phenomena (disturbance date and intensity) as well as continuous phenomena (post-disturbance regeneration). We applied the method to a stack of 18 Landsat TM images for the 20-year period from 1984 to 2004. When compared with direct interpreter delineation of disturbance events, the automated method accurately labeled year of disturbance with 90% overall accuracy in clear-cuts and with 77% accuracy in partial-cuts (thinnings). The primary source of error in the method was misregistration of images in the stack, suggesting that higher accuracies are possible with better registration.

DOI

[21]
Kennedy R E, Yang Z Q, Cohen W B.Detecting trends in forest disturbance and recovery using yearly Landsat time series:1. LandTrendr-Temporal segmentation algorithms[J]. Remote Sensing of Environment, 2010,114:2897-2910.We introduce and test LandTrendr (Landsat-based detection of Trends in Disturbance and Recovery), a new approach to extract spectral trajectories of land surface change from yearly Landsat time-series stacks (LTS). The method brings together two themes in time-series analysis of LTS: capture of short-duration events and smoothing of long-term trends. Our strategy is founded on the recognition that change is not simply a contrast between conditions at two points in time, but rather a continual process operating at both fast and slow rates on landscapes. This concept requires both new algorithms to extract change and new interpretation tools to validate those algorithms. The challenge is to resolve salient features of the time series while eliminating noise introduced by ephemeral changes in illumination, phenology, atmospheric condition, and geometric registration. In the LandTrendr approach, we use relative radiometric normalization and simple cloud screening rules to create on-the-fly mosaics of multiple images per year, and extract temporal trajectories of spectral data on a pixel-by-pixel basis. We then apply temporal segmentation strategies with both regression-based and point-to-point fitting of spectral indices as a function of time, allowing capture of both slowly-evolving processes, such as regrowth, and abrupt events, such as forest harvest. Because any temporal trajectory pattern is allowable, we use control parameters and threshold-based filtering to reduce the role of false positive detections. No suitable reference data are available to assess the role of these control parameters or to test overall algorithm performance. Therefore, we also developed a companion interpretation approach founded on the same conceptual framework of capturing both long and short-duration processes, and developed a software tool to apply this concept to expert interpretation and segmentation of spectral trajectories (TimeSync, described in a companion paper by Cohen et al., 2010). These data were used as a truth set against which to evaluate the behavior of the LandTrendr algorithms applied to three spectral indices. We applied the LandTrendr algorithms to several hundred points across western Oregon and Washington (U.S.A.). Because of the diversity of potential outputs from the LTS data, we evaluated algorithm performance against summary metrics for disturbance, recovery, and stability, both for capture of events and longer-duration processes. Despite the apparent complexity of parameters, our results suggest a simple grouping of parameters along a single axis that balances the detection of abrupt events with capture of long-duration trends. Overall algorithm performance was good, capturing a wide range of disturbance and recovery phenomena, even when evaluated against a truth set that contained new targets (recovery and stability) with much subtler thresholds of change than available from prior validation datasets. Temporal segmentation of the archive appears to be a feasible and robust means of increasing information extraction from the Landsat archive.

DOI

[22]
Cohen W B, Yang Z Q, Kennedy R.Detecting trends in forest disturbance and recovery using yearly Landsat time series:2. TimeSync - Tools for calibration and validation[J]. Remote Sensing of Environment, 2010,114:2911-2924.Availability of free, high quality Landsat data portends a new era in remote sensing change detection. Using dense (~annual) Landsat time series (LTS), we can now characterize vegetation change over large areas at an annual time step and at the spatial grain of anthropogenic disturbance. Additionally, we expect more accurate detection of subtle disturbances and improved characterization in terms of both timing and intensity. For Landsat change detection in this new era of dense LTS, new detection algorithms are required, and new approaches are needed to calibrate those algorithms and to examine the veracity of their output. This paper addresses that need by presenting a new tool called TimeSync for syncing algorithm and human interpretations of LTS. The tool consists of four components: (1) a chip window within which an area of user-defined size around an area of interest (i.e., plot) is displayed as a time series of image chips which are viewed simultaneously, (2) a trajectory window within which the plot spectral properties are displayed as a trajectory of Landsat band reflectance or index through time in any band or index desired, (3) a Google Earth window where a recent high-resolution image of the plot and its neighborhood can be viewed for context, and (4) an Access database where observations about the LTS for the plot of interest are entered. In this paper, we describe how to use TimeSync to collect data over forested plots in Oregon and Washington, USA, examine the data collected with it, and then compare those data with the output from a new LTS algorithm, LandTrendr, described in a companion paper (Kennedy et al., 2010). For any given plot, both TimeSync and LandTrendr partitioned its spectral trajectory into linear sequential segments. Depending on the direction of spectral change associated with any given segment in a trajectory, the segment was assigned a label of disturbance, recovery, or stable. Each segment was associated with a start and end vertex which describe its duration. We explore a variety of ways to summarize the trajectory data and compare those summaries derived from both TimeSync and LandTrendr. One comparison, involving start vertex date and segment label, provides a direct linkage to existing change detection validation approaches that rely on contingency (error) matrices and kappa statistics. All other comparisons are unique to this study, and provide a rich set of means by which to examine algorithm veracity. One of the strengths of TimeSync is its flexibility with respect to sample design, particularly the ability to sample an area of interest with statistical validity through space and time. This is in comparison to the use of existing reference data (e.g., field or airphoto data), which, at best, exist for only parts of the area of interest, for only specific time periods, or are restricted thematically. The extant data, even though biased in their representation, can be used to ascertain the veracity of TimeSync interpretation of change. We demonstrate that process here, learning that what we cannot see with TimeSync are those changes that are not expressed in the forest canopy (e.g., pre-commercial harvest or understory burning) and that these extant reference datasets have numerous omissions that render them less than desirable for representing truth.

DOI

[23]
Verbesselt J, Hyndman R, Newnham G, et al.Detecting trend and seasonal changes in satellite image time series[J]. Remote Sensing of Environment, 2010,114:106-115.A wealth of remotely sensed image time series covering large areas is now available to the earth science community. Change detection methods are often not capable of detecting land cover changes within time series that are heavily influenced by seasonal climatic variations. Detecting change within the trend and seasonal components of time series enables the classification of different types of changes. Changes occurring in the trend component often indicate disturbances (e.g. fires, insect attacks), while changes occurring in the seasonal component indicate phenological changes (e.g. change in land cover type). A generic change detection approach is proposed for time series by detecting and characterizing Breaks For Additive Seasonal and Trend (BFAST). BFAST integrates the decomposition of time series into trend, seasonal, and remainder components with methods for detecting change within time series. BFAST iteratively estimates the time and number of changes, and characterizes change by its magnitude and direction. We tested BFAST by simulating 16-day Normalized Difference Vegetation Index (NDVI) time series with varying amounts of seasonality and noise, and by adding abrupt changes at different times and magnitudes. This revealed that BFAST can robustly detect change with different magnitudes (>0.1 NDVI) within time series with different noise levels (0.01鈥0.07 蟽 ) and seasonal amplitudes (0.1鈥0.5 NDVI). Additionally, BFAST was applied to 16-day NDVI Moderate Resolution Imaging Spectroradiometer (MODIS) composites for a forested study area in south eastern Australia. This showed that BFAST is able to detect and characterize spatial and temporal changes in a forested landscape. BFAST is not specific to a particular data type and can be applied to time series without the need to normalize for land cover types, select a reference period, or change trajectory. The method can be integrated within monitoring frameworks and used as an alarm system to flag when and where changes occur.

DOI

[24]
Verbesselt J, Hyndman R, Zeilies A, et al.Phenological change detection while accounting for abrupt and gradual trends in satellite image time series[J]. Remote Sensing of Environment, 2010,114:2970-2980.A challenge in phenology studies is understanding what constitutes phenological change amidst background variation. The majority of phenological studies have focused on extracting critical points in the seasonal growth cycle, without exploiting the full temporal detail. The high degree of phenological variability between years demonstrates the necessity of distinguishing long-term phenological change from temporal variability. Here, we demonstrate the phenological change detection ability of a method for detecting change within time series. BFAST, Breaks For Additive Seasonal and Trend, integrates the decomposition of time series into trend, seasonal, and remainder components with methods for detecting change. We tested BFAST by simulating 16-day NDVI time series with varying amounts of seasonal amplitude and noise, containing abrupt disturbances (e.g. fires) and long-term phenological changes. This revealed that the method is able to detect the timing of phenological changes within time series while accounting for abrupt disturbances and noise. Results showed that the phenological change detection is influenced by the signal-to-noise ratio of the time series. Between different land cover types the seasonal amplitude varies and determines the signal-to-noise ratio, and as such the capacity to differentiate phenological changes from noise. Application of the method on 16-day NDVI MODIS images from 2000 until 2009 for a forested study area in south eastern Australia confirmed these results. It was shown that a minimum seasonal amplitude of 0.1 NDVI is required to detect phenological change within cleaned MODIS NDVI time series using the quality flags. BFAST identifies phenological change independent of phenological metrics by exploiting the full time series. The method is globally applicable since it analyzes each pixel individually without the setting of thresholds to detect change within a time series. Long-term phenological changes can be detected within NDVI time series of a large range of land cover types (e.g. grassland, woodlands and deciduous forests) having a seasonal amplitude larger than the noise level. The method can be applied to any time series data and it is not necessarily limited to NDVI.

DOI

[25]
Jamali S, Jonsson P, Eklundh L, et al.Detecting changes in vegetation trends using time series segmentation[J]. Remote Sensing of Environment, 2015,156:182-195.61DBEST, a user-friendly program for analysing vegetation time series is introduced.61It enables users to capture trend features, from details to main changes.61It detects abrupt and gradual changes, and estimates change time and magnitude.61It can be applied to vegetation time series at different spatio-temporal scales.61DBEST was tested using simulated data and real NDVI data series over Iraq.

DOI

[26]
佟斯琴,包玉海,张巧凤,等.基于像元二分法和强度分析方法的内蒙古植被覆盖度时空变化规律分析[J].生态环境学报,2016,25(5):737-743.

[ Tong S Q, Bao Y H, Zhang Q F, et al.Spatiotemporal changes of vegetation coverage in inner Mongolia based on the dimidiate pixel model and intensity analysis[J]. Ecology and Environmental Sciences, 2016,25(5):737-743. ]

[27]
D. Stow, S. Daeschner, A. Hope, et al.Variability of the seasonally integrated normalized difference vegetation index across the north slope of Alaska in the 1990s[J]. International Journal of Remote Sensing, 2003,24(5):1111-1117.ABSTRACT The interannual variability and trend of above-ground,photosynthetic activity of Arctic tundra,vegetation in the 1990s is examined,for the north slope region of Alaska, based on the seasonally integrated normalized difference vegetation index (SINDVI) derived from local area coverage,(LAC) National Oceanic and Atmospheric Administration (NOAA) Advanced,Very High Resolution Radiometer,(AVHRR) data. Smaller SINDVI values occurred,during,the three years (1992鈥1994) following the volcanic eruption,of Mt Pinatubo. Even after implementing,corrections for this stratospheric aerosol effect and,adjusting for changes in radiometric calibration coefficients, an apparent increasing trend of SINDVI in the 1990s is evident for the entire north slope. The most,pronounced increase was observed,for the foothills physiographical,province.

DOI

[28]
宋怡,马明国.基于Spot Vegetation数据的中国西北植被覆盖变化分析[J].中国沙漠, 2007,27(1):90-92.基于遥感和地理信息系统的技术,利用SPOT-VEGETATION NDVI(Normalized Difference VegetationIndex)数据对我国西部地区植被覆盖的情况进行了动态监测。采用MVC(Maximum Value Composites)、一元线性回归趋势分析和变化幅度百分比等方法分析西部地区植被变化特征,并结合西北五省土地利用类型图,分析不同植被类型的年最大化NDVI(MNDVI)变化趋势及特点。其结果是:近7 a来植被覆盖存在普遍退化的趋势,且2000-2001与2001-2002年度的变化幅度较大。在局部区域植被有改善的趋势,但总的改善幅度小于退化幅度。分析结果表

DOI

[ Song Y, Ma M G.Study on vegetation cover change in northwest China based on SPOT vegetation data[J]. Journal of Desert Research, 2007,27(1):90-92. ]

Options
Outlines

/

〈 〉