EN中文

Journal of Geo-information Science >

2016 , Vol. 18 >Issue 8: 1110 - 1122

DOI: https://doi.org/10.3724/SP.J.1047.2016.01110

Analysis of NDVI Time Series in Grassland Open-cast Coal Mines Based on SSA-Mann Kendall

  • JIA Duo ,
  • MU Shouguo , * ,
  • ZHAO Hua
Expand
  • 1. School of Environment Science and Spatial Informatics, China University of Mining and Technology, Xuzhou 221116, China;2. Institute of Land Resources, China University of Mining and Technology, Xuzhou 221116, China
*Corresponding author: MU Shouguo, E-mail:

Received date: 2016-02-28

  Request revised date: 2016-05-08

  Online published: 2016-08-10

Copyright

《地球信息科学学报》编辑部 所有
Fold

Abstract

Vegetation index time series have difficulty to depict the detailed response of vegetation dynamics to coal mining and the signals of change trend and periodic oscillation at an inter-annual scale. While at the monthly scale, the signals are so weak that they are hard to be extracted due to the disturbance of vegetation phenology. In addition, the physical significance of the transformation detection algorithm is still unclear. In order to solve these problems and to reveal the change trend and periodic oscillation of vegetation growth in the disturbance area of grassland open-cast mine, this paper selected Shengli open-cast mine area as an study area to extract, amplify and quantitatively compute the signals of the monthly change trend and periodic oscillation based on the MODIS NDVI time series from January 2001 to December 2013 in the mining disturbance area and the unchanged feature region. SSA-Man Kendall was adopted to extract the change trend and periodic oscillation. Moreover, we quantitatively analyzed the significant degree and jump time. For the change trend, the Sen slope of the change components were calculated, which could indicates the change direction. Then the trend significance was measured based on the results of Mann Kendall trend analysis. Combined with the moving-t test for some fuzzy catastrophe points, the time points of abrupt changes were also detected. For the periodic oscillation, it was estimated based on the utilization of power spectrum analysis. The evolution characteristics of NDVI time series′ periodic oscillation in different feature regions were also studied. The results show that SSA-Mann Kendall can effectively extract the signals of change trend and be competent in depicting the periodic oscillation at different time scales, as well as quantitatively express the change trend signals. A downward trend of NDVI time series in the stopes is significant, and it is more significant than the waste dumps within the same mine area, while the trend in the unchanged feature region is relative stable. In addition, damage of vegetation is a sudden event in the grassland opencast mine area, and the catastrophe points of NDVI time series usually occur at the beginning of the mine construction. In the partial open-cast mines, NDVI time series′ periodic oscillation in the stopes and waste dumps are different, which are related to the different disturbance forms within these areas. In particular, the vegetation almost vanishes in the stopes with coal mining, however, the dynamics of vegetation growth are more complex in the waste dumps due to the effective vegetation restoration.

Key words： NDVI time series; singular spectrum analysis; Mann Kendall; trend of change; periodic oscillation

Cite this article

JIA Duo , MU Shouguo , ZHAO Hua . Analysis of NDVI Time Series in Grassland Open-cast Coal Mines Based on SSA-Mann Kendall[J]. Journal of Geo-information Science, 2016 , 18(8) : 1110 -1122 . DOI: 10.3724/SP.J.1047.2016.01110

1 引言

植被指数是利用多光谱遥感数据经过分析运算产生对植被长势、生物量等具有一定指数意义的数值,其与叶面积指数、叶绿素含量、植被覆盖度、生物量等植被生物物理参数密切相关,同时也是气候参数、植物蒸散、土壤水分等地表生态环境参数的指标之一[1]。用于反演植被生长状况的遥感植被指数以NDVI最为常见[2]。基于NOAA/AVHRR、SPOT/VEGETATION及MODIS等高时间分辨率卫星传感器的NDVI时间序列已广泛用于植被监测[3]、植被覆盖分类[4]、土地覆盖类型解译[5]、植被物候参数提取[6]及变化检测[7]的研究中。露天煤矿区生产中的表土剥离、煤炭开采、表土回填、植被恢复过程将不可避免地对植被生长产生重要影响[8],但露天煤矿区采矿扰动下植被生长的变化规律尚不明确。
植被生长因外界环境要素复杂变化而具有动态特征,NDVI时间序列有效地记录了其动态过程[9]。对其进行趋势、突变和周期性分析是认识植被系统动态演化特征的基本方法[10]。NDVI变化趋势被视为植被覆盖、物候和生物量的动态。回归分析法和相关系数法是NDVI时间序列变化趋势测度的主流方法[11],但该方法规避误差能力较弱,易受NDVI误差出现时间位置的影响,也无法对变化趋势显著程度进行度量[12]。Sen斜率估计对误差有较好的规避能力,结合Mann-Kendall趋势检验度量NDVI时间序列变化趋势的显著程度,可更合理地衡量NDVI时间序列的变化趋势[11]。周期振荡是NDVI时间序列的另一个重要动态特征。郭笑怡等[10]指出经去季节性循环处理的NDVI异常值序列存在多尺度的振荡特征,不同时间尺度的振荡特征能反映不同尺度间的标度行为特点。刘亚龙等[13]指出植被演变存在内在的周期循环长度,该特征是植被非线性和复杂性研究的重要组成。然而,现有NDVI时间序列周期振荡的研究较少,且多集中于年际时间尺度,难以细致刻画植被生长对开采活动的响应特征。而月际尺度NDVI时间序列的物候性周期波动,会导致人类活动引起的时间序列变化趋势和周期振荡信号微弱,而常规手段难以度量。奇异谱分析无需正弦性假定,对功率谱信号具有明显的放大作用,且波形由实际序列确定[14],在非线性地球物理数据的动态分析中具有较好的适用性,现已广泛用于非参数时间序列分解、滤波、预测及数据挖掘[15],然而此类基于变换的方法虽有效地突出植被动态信息,但其物理意义尚不明确[16]。
鉴此,为细致刻画草原露天矿区植被生长对采矿扰动的响应特征,以胜利露天煤矿区为例,以奇异谱分析在月际尺度放大并重建草原露天矿区主要特征区域NDVI时间序列趋势成分和周期成分,通过Mann Kendall检测对趋势成分的显著程度和突变状况进行定量化,并分析各特征区周期振荡特征及异常振荡的原因,以期揭示采矿扰动下草原露天煤矿区植被生长的变化规律。

2 研究区概况与数据源

2.1 研究区概况

胜利矿区为中国14个大型煤炭基地之一,位于内蒙古自治区锡林郭勒盟锡林浩特市北2~5 km,胜利苏木和伊利勒特苏木境内,锡林河以西的缓坡丘陵干草原与河谷冲积、湖积平原过渡带（图1）。地理坐标为115°24′26″~116°26′30″E,43°54′15″~44°13′52″N,海拔970~1202 m。矿区呈北东、南西向条带状,走向长45 km,南北宽平均7.6 km,总面积342 km2。处温带丛生禾草典型草原腹地,除河滩、丘间洼地和盐化湖盆低地外均为典型草原。矿区共划分10个井（矿）田,包括6个露天煤矿田、1个露天锗矿田和3个井工矿田。其中,西一号露天矿、西三号露天矿、露天锗矿及东三号露天矿已投入生产。
Fig.1 Location map of Shengli open-cast mine area

图1 胜利矿区位置图

注：A为露天锗矿;B为西三号露天矿;C为西一号露天矿;D为东三号露天矿

2.2 数据源与数据预处理

2.2.1 数据源
本文选取MOD13Q1的16天合成数据产品构建NDVI时间序列（数据来源：http://ladsweb.nascom.nasa.gov/）。其空间分辨率为250 m,全年23景,2001-2013年共299景影像。以Landsat8 OLI影像提取胜利矿区各特征区（数据来源：http: //glovis.usgs.gov/）,成像时间为2013-06-01,数据经辐射定标、大气校正、裁剪及融合处理并与MOD13Q1数据精确配准。
2.2.2 时间序列数据处理
（1）NDVI时间序列数据重建
利用MRT工具将MOD13Q1级产品的NDVI数据由Sinusoidal投影转换至UTM投影,WGS84-50N坐标,利用IDL 8.0以研究区边界进行批量裁剪。以自适应S-G滤波实现NDVI时间序列重建。具体过程为如式（1）所示。
R i = ∑ i = - m m C i × I i + j 2 m + 1 （1）
式中： R i 表示处理后的NDVI值; C i 为由多项式 f ( t ) 最小二乘拟合给定的系数; i 为滑动窗口中心横坐标; I i + j 为滑动窗口中第 j 个原始NDVI; 2 m + 1 为滑动窗口宽度,其大小决定了时间序列重建的拟合与去噪效果。 m 值越小拟合效果越好,但去噪效果不佳; m 值过大会导致NDVI时序数据失去更多的细节信息。经多次实验设定 m 值为4。多项式 f ( t ) 表示为式（2）。
f ( t ) = C 1 φ 1 ( t ) + C 2 φ 2 ( t ) + … + C n φ n ( t ) （2）
多项式系数C通过最小二乘求解法方程组求得（式（3））。
A T AC = A T b （3）
其中：
A ij = w i φ j t i , b i = w i y i （4）
式中： w i 指第 i 个数据的权重,权重值越高数据拟合时产生的影响越明显。以MOD13Q1产品的VI质量（VI quality）数据提供的像元尺度NDVI数据综合质量评价指数确定权重。VI质量数据域将16 bit的数据位分配给9个字段,分别表示不同的质量属性和等级[17]（表1）。将0~1 bit数据转换为10进制得到VI质量总评数据,分级打分赋权重值（表2）。
Tab.1 MODIS VI quality data field

表1 MODIS VI质量数据域

比特位 含义 比特位 含义 比特位 含义
0-1 VI质量总评 8 云边界 11-13 陆地/水域标记
2-5 VI可用性 9 大气二向性校正 14 雪/冰
6-7 气溶胶含量 10 混合云 15 云阴影
Tab.2 VI quality overall assessment data

表2 VI质量总评数据

VI质量总评指数 质量描述 权重
0 产生质量较高的植被指数数据 1
1 产生植被指数但需参考其他QA数据 0.6
2 产生植被指数但大部分像元受云影响 0.2
3 因其他原因无植被指数 0
设置迭代次数为3,即以3次迭代后的NDVI值构建时间序列,上述过程借助MATLAB平台的Timesat实现[18]。然后,利用最大值合成将NDVI时间序列转换为月最大值序列,并进行尺度转换获得长度为156的时间序列。
（2）时间序列数据标准化
重建的NDVI时间序列为相对光滑的波动曲线,具有明显的物候特征。为提取由采矿等人为因素引起的露天矿区不同特征区的NDVI变化趋势及周期振荡特征,需去除由季节变化导致的物侯性周期波动信号将原始时间序列转化为Z标准值[19],如式（5）所示。
Z i , j = NDV I i , j - μ i σ i , j （5）
式中： Z i , j 表示第 j 年第 i 月的 Z 标准值; NDV I i , j 指第 j 年第 i 月的区域NDVI原始值的平均值; μ i 为所有年份 i 月NDVI原始平均值的均值; σ i , j 为第 i 月 j 年的NDVI空间标准差。标准NDVI值的正负分别表示NDVI高于或低于当月NDVI平均水平。

2.3 矿区不同特征区域划分与时间序列提取

2.3.1 不同特征区域划分
露天矿区生产主要表现为采剥-排弃-造地过程,该过程的影响区主要分布于矿区露天采场和排土场,则设定采矿扰动区为各矿井露天采场和排土场。分别获取图1中A、B、C、D 4个矿井的采矿扰动区NDVI时间序列,进行变化趋势和周期振荡分析。同时,在矿区未受扰动区域设定伪不变特征区,以该区域NDVI时间序列作为对照组与上述采矿扰动区的NDVI动态特征进行对比。
2.3.2 各特征区识别与NDVI时间序列提取
（1）采矿扰动区
矿井各特征区范围以2013年Landsat 8 OLI数据为基准,以突出由于采矿范围不断扩大导致的NDVI时间序列的动态过程。基于决策树分类手段[20],并对照收集的矿区规划图及Google Earth历史数据精确识别各特征区范围。为有效规避基准年份各特征区面积差异导致的误差,借助ArcGIS的Creat Random Points在各特征区生成20个随机点,以随机点提取时间序列曲线,并计算均值作为该区域时间序列。此外,为减小混合像元（MOD13Q1空间分辨率为250 m）导致的误差,对特征区边界附近的随机点进行替换。
（2）伪不变特征区
为保证伪不变特征区不受采矿活动影响,假设以工业场地、排土场、露天采场等采矿活动区域及居民点为中心5 km范围内受到了人为活动的影响,则伪不变特征区应位于该影响范围外,采用同样方法提取伪不变特征区NDVI均值序列。

3 研究方法

经数据重建和标准化的NDVI时间序列杂乱无章,难以识别变化趋势和周期振荡特征。采用SSA-Mann Kendall分别提取各特征区标准化NDVI时间序列的趋势和周期振荡信号,并对趋势信息进行定量化,结合周期振荡特征,揭示草原露天矿区植被动态特点。SSA-Mann Kendall即奇异谱分析与Mann Kendall结合,以奇异谱分析提取一维时间序列的微弱趋势和周期振荡信号,并以Mann Kendall检测对趋势信号进行定量化,其计算流程如图2所示。首先,通过奇异谱分析重建矿区各特征区NDVI时间序列的趋势和周期振荡成分。对提取的趋势成分,计算其Sen斜率,以计算结果的正负分别表示NDVI上升和下降趋势[21];然后,以Mann Kendall的计算结果度量变化趋势的显著程度,利用Mann Kendall突变检测计算趋势成分的突变时间,对存在模糊突变点的时间序列进行滑动 t 检验最终确定突变点;最后,对得到的周期振荡信息,通过绘制傅里叶功率谱图计算显著周期[22],并分析各特征区NDVI时间序列周期振荡的演变特征,探究异常周期振荡现象的原因。奇异谱分析与Mann-Kendall检验的具体实现过程见下文,功率谱的计算过程见文献[22]。
Fig.2 Flow chart of SSA-Mann Kendall based NDVI time series in the grassland open-cast mine area

图2 基于SSA-Mann Kendall草原露天矿区NDVI时间序列分析流程

3.1 奇异谱分析

奇异谱分析将一维时间序列嵌入至高维的欧式空间,通过选取能够代表感兴趣区的具有现实意义的子空间构造新的时间序列。经奇异值分解得到的前几个主分量包含了时间序列的大部分趋势和周期信息,经重建可有效将其提取。具体过程为：设长度 N 为156的NDVI时间序列, X N = ( x 1 , ⋯ , x N ) ,按照嵌入空间维数 M ,建立相空间矩阵[23]（式（6））。 M 取值越大,可提取原始时间序列的细节信息越多,同时也会包含更多的噪声信息。假设经时间序列重建,噪声已被最大程度削弱,为突出时间序列的细节信息,根据限制条件 M ≤ N 2 ,设定嵌入空间维数为 N 2 （即78）。
X = x 1 x 2 ⋯ x N - M + 1 x 2 x 3 ⋯ x N - M + 2 ⋮ ⋮ ⋮ x M x M + 1 ⋯ x N = X 1 , X 2 , ⋯ , X N - M + 1 （6）
构造矩阵 S = X X T ,其特征值为式（7）。
λ 1 ≥ λ 2 ≥ ⋯ ≥ λ m ≥ 0 （7）
将上述特征值开方得到奇异值,按照奇异值大小依次排序（式8）。
λ 1 ≥ λ 2 ≥ ⋯ ≥ λ m ≥ 0 （8）
定义最大特征值对应的特征向量为第一阶模式,代表信号的最大变化趋势,第二大特征值对应的特征向量为第二阶模式,代表与第一阶模式无关的剩余信号的最大变化趋势,依此类推[24]。按照上述矩阵特征值大小,依次进行奇异值分解（式（9））。
x i + j = ∑ k = 1 M a ik E kj （9）
式中：原始时间序列的第 k 个时间主分量 a ik （式（10））,记为TPC[24],为时间序列在特征值 λ k 对应的特征向量 E k 上的投影,可用于识别周期信号。
a ik = X i · E k = ∑ j = 1 M x i + j E kj （10）
定义特征向量 E k 为时间经验正交函数,其为滞后时间步长的函数,表示时间序列振荡的主要周期模态。以第 k 个TEOF和TPC重建第 k 个 x i 的成分 x i k ,以描述NDVI时间序列中线性或非线性振荡信号（式（11））,由于嵌入窗口较大导致重建分量较多,仅选取其中有意义的分量进行重建,按照其方差贡献率大小由高至低排序依次记作R1、R2等。其余分量视作噪声不参与趋势成分和周期成分重建。
x i k = 1 i ∑ j = 1 i a ijk E kj , 1 ≤ i ≤ M - 1 1 M ∑ j = 1 M a ijk E kj , M ≤ i ≤ N - M + 1 1 N - i + 1 ∑ j = i - N + M M a ijk E kj , N - M + 2 ≤ i ≤ N （11）
基于各重建分量（R）的演变特点,对各重建分量进行组合,构建NDVI时间序列的趋势成分和周期振荡成分。该过程表示为式（12）。
x i = ∑ k = 1 k x k , i = 1,2 , … , N （12）
式中： k 表示前 k 个有意义的重建分量,本文 k 值取9。构建的趋势成分或周期成分以参与构建的重建分量相加表示,如R1+R2表示该成分由重建分量R1与R2的组合构建。

3.2 Mann-Kendall检测

3.2.1 Sen+Mann-Kendall趋势检验
Sen趋势度与Mann-Kendall趋势检验结合对长时序植被数据趋势检测具有较好的适用性[11]。其中,Sen趋势度用于判别时间序列的变化趋势。
β = mean x j - x i j - i ， ∀ j > i （13）
式中： β 正负分别表示上升和下降趋势[21]。变化趋势的显著性度量采用Mann-Kendall方法,其计算过程为式（14）。
S = ∑ i = 2 n ∑ j = 1 i - 1 Sign ( X i - X j ) （14）
式中： Sign 为符号函数,当 X i - X j ＜ 0 时, Sign = - 1 ;当 X i - X j = 0 时, Sign = 0 ;当 X i - X j ＞ 0 时, Sign = 1 。M-K统计量计算公式如式（15）所示。
Z = S - 1 / n n - 1 2 n + 5 / 18 , S < 0 Z = 0 , S = 0 Z = S + 1 / n n - 1 2 n + 5 / 18 , S > 0 （15）
式中： n 为156。按照置信水平 α = 0.01 ,若 Z > Z 1 - α / 2 ,则认为存在明显变化趋势。
3.2.2 Mann-Kendall突变检测
Mann-Kendall突变检测的计算过程为：假设趋势成分序列 x 1 , x 2 , … , x n 平稳且随机, n 为156, r i 表示序列的第 i 个样本 x i 大于 x j （ 1 ≤ j ≤ i ） 的累计数,则统计量 S k 如式（16）所示。
S k = ∑ i = 1 k r i ， k = 2,3 , … , n （16）
在原序列平稳且随机独立的假设下, S k 的均值、方差如式（17）-（18）所示。
E ( S k ) = n ( n - 1 ) / 4 （17）
VAR ( S k ) = n ( n - 1 ) ( 2 n + 5 ) / 72 （18）
S k 标准化如式（19）所示。
U F k = S k - E S k / VAR ( S k ) （19）
若 U F k > Uα ,本文取显著水平 α = 0.05 。则NDVI的变化趋势显著。将NDVI时间序列的逆序列重复上述过程,再令 U B k = - U F k ,其中 k = n , n - 1 , … ; U B 1 = 0 。当 UF 超过置信线时,表明NDVI时间序列呈现显著上升或下降趋势;若 U F k 和 U B k 相交于置信线之间,则交点处对应的时点即为突变开始的时间[25];若交点位于置信线外,或存在多个明显交点,以滑动 T 检验确定是否为突变点[26],逐一记录各特征区域时间序列趋势成分的突变时间。

4 结果与分析

4.1 各特征区原始与标准化NDVI时间序列

图3（a）显示了各特征区NDVI时间序列表现出明显的物候性周期波动。不同区域时间序列平均振幅有所差异,其中伪不变特征区平均振幅最大,该区域植被生长状况最佳。因植被周期性物候波动的干扰,难以判别原始NDVI时间序列的变化特点。标准化NDVI时间序列（图3（b））的物候性周期变化信号被减弱,但采矿扰动区的线性变化趋势仍然较微弱,直观上难以识别。
Fig.3 The changing condition of standardized NDVI time series in different feature regions of the open-cast mine area

图3 矿区各特征区域NDVI变化状况

4.2 标准化NDVI时间序列趋势成分变化特征

露天矿区各特征区NDVI时间序列变化趋势及突变时间计算结果如表3所示。在总体变化趋势上,采矿扰动区Sen斜率均为负值,M-K值均达到0.01置信水平,即各区域NDVI均显著下降。同一矿井露天采场下降趋势较排土场更为显著,而不同矿井采矿扰动区的变化趋势存在明显差异。露天采场中,西三号露天矿的下降趋势最显著。在排土场,露天锗矿的下降趋势最显著,东三号、西三号露天矿次之,西一号露天矿NDVI下降趋势相对平缓。
Tab.3 Changing trend and jump time of NDVI time series in the open-cast mine area′s

表3 露天矿区各特征区NDVI时间序列变化趋势及突变时间[27]

特征区 地点 方差贡献率/（%） β M-K 变化趋势 突变时间/月
露天采场 东三号露天矿 30.43 -0.00829 -8.9110 下降 89
西三号露天矿 20.67 -0.00094 -11.3752 下降 124
西一号露天矿 19.88 -0.00175 -3.7253 下降 124
露天锗矿 24.38 -0.00170 -5.7451 下降 122
排土场 东三号露天矿 17.84 -0.00148 -8.6812 下降 114
西三号露天矿 19.93 -0.00336 -6.3153 下降 101
西一号露天矿 17.89 -0.00066 -3.3851 下降 71
露天锗矿 59.59 -0.00392 -9.3554 下降 90
伪不变特征区 植被覆盖区域 14.57 -0.00027 -1.4573 - -

注：突变时间以2001年1月为1,依此类推2013年12月为156;“-”指无显著变化趋势

趋势成分突变点的计算结果显示,同一矿井的露天采场突变时间略滞后于排土场,但突变时间相差不大。不同矿井突变时间各异,在露天采场,东三号露天矿最早发生突变,83-84月间NDVI时间序列明显下降,露天锗矿、西一号露天矿及西三号露天矿突变时间大致相同。在排土场,西一号露天矿于70-71月时最先发生突变,西三号露天矿最晚突变,其突变时间为100-101月。经现场调查,4个矿井中,露天锗矿和西一号露天矿的开发时间较早,东三号露天矿开发相对较晚,各矿井排土场区域的突变先后恰好与各矿井的开发时序相差不大,因此上述各矿井突变时间的差异可能与矿井开发建设和开采时序密切相关。
露天采场和排土场趋势成分演变特征（图4）均表现为先平稳随即在某一时刻（突变点）迅速下降。该现象与露天矿区生产过程中的采剥-排弃-造地过程对植被生长产生的影响一致。在挖损作用下,地表植被荡然无存,因此趋势成分从某一时刻开始迅速下降。对比伪不变特征区与采矿扰动区的趋势成分（图5）,2个特征区趋势成分演变特征差异明显。Sen斜率结合M-K的计算结果显示（表3）,伪不变特征区趋势成分较平缓,该区域植被生长状况虽有所下降,但趋势并不明显。露天采场和排土场的趋势成分具有明显的人为干扰特征,2个区域趋势成分均在某一时点（突变点）迅速下降。
Fig.4 Trend component of NDVI time series in the mining disturbance area

图4 采矿扰动区NDVI时间序列趋势成分

Fig.5 Comparison of the trend component of NDVI time series between the mining disturbance area and the unchanged feature region

图5 采矿扰动区与伪不变特征区NDVI时间序列趋势成分对比

4.3 标准化NDVI时间序列周期振荡特征

4.3.1 采矿扰动区周期振荡特征
图6为奇异谱分析重建的胜利矿区主要矿井采矿扰动区NDVI时间序列周期成分,各周期成分组合及其贡献率与显著周期如表4所示。各区域周期成分的贡献率较低,振荡信号较微弱。除东三号排土场及露天锗矿采场的贡献率较接近外,各周期成分贡献率表现出了一定的变化梯度,不同周期成分对应的不同振荡周期直观刻画了不同时间尺度的细节信息,多尺度振荡间的不同标度行为说明了采矿扰动区外部环境的复杂性。此外,方差贡献率的高低在一定程度上反映了该特征区域周期振荡的主要形式。多数矿井的周期成分中,短期振荡成分的贡献率最高,说明矿区以9-13个月的短期振荡为主。在振荡强度方面,贡献率较高周期成分的振荡能量高,振荡现象明显,且振荡能量多集中于前2个分量。
Tab.4 Information of period component in the mining disturbance area

表4 各矿井采矿扰动区NDVI时间序列周期成分

特征区 重建成分组合 贡献率/(%) 周期/月 重建成分组合 贡献率/(%) 周期/月 重建成分组合 贡献率/(%) 周期/月
东三号露天矿采场 R4+R5+R6 13.30 11.1 R2+R3 12.33 31.2 R7+R8 6.16 13
东三号露天矿排土场 R2+R3 13.16 10.4 R4+R5+R9 13.00 6 R6+R7+R8 10.48 14
西一号露天矿采场 R2+R3+R4 19.47 13 R5+R6 9.61 11.1 R7+R8 8.80 7.8
西一号露天矿排土场 R2+R3+R4 28.01 12 R5+R6+R7 13.93 6 R8+R9 8.01 -
西二号露天矿采场 R2+R3 16.43 10.4 R4+R5+R6 15.67 5.5 R7+R8 6.29 -
西二号露天矿排土场 R2+R3+R4 21.01 10.4 R7+R8+R9 9.54 - R5+R6 8.43 39
露天锗矿采场 R2+R3 11.78 9 R5+R6+R7 10.45 12 R4+R8+R9 10.37 -
露天锗矿排土场 R3+R4 11.18 13 R5+R6 10.44 11.1 R7+R8+R9 5.20 -

注：-指无显著周期

Fig.6 Evolution characteristics of period component of NDVI time series in the mining disturbance area

图6 采矿扰动区NDVI时间序列周期成分演变特征

注：a为东三号露天矿;b为西一号露天矿;c为西三号露天矿;d为露天锗矿;1表示露天采场;2表示排土场

周期成分对区域异常变化的响应较为敏感,各矿井周期振荡强度均出现异常。对比趋势成分突变时间与上述周期振荡强度异常时间,东三号露天矿采场突变时间为88-89月,该区域第一周期成分振荡强度自72月开始发生异常,东三号露天矿排土场NDVI于113-114月时突然下降,该时间恰与该区域周期成分振荡强度的异常现象出现时间相差不大。西一号、西二号露天矿采场和排土场周期振荡强度的异常时间与趋势成分突变时间也大致相同,但异常振荡的开始时间略滞后于突变时间。采矿作用导致趋势成分的突变,而采矿扰动区异常周期振荡现象的发生与该区域NDVI异常下降有关。
各矿井采场和排土场前2个周期成分蕴含周期振荡成分的主要能量。经对比发现,各矿井采矿扰动区周期振荡的演变特征存在较大差异（图7）,不同矿井异常振荡现象出现时间不一致,且上述及异常振荡的发生是由该区域NDVI异常下降导致的,不同矿井异常振荡时间的差异则可能与建设和开采时序密切相关。
Fig.7 Comparison of the component for the first two periods in the mining disturbance area

图7 各矿井采矿扰动区NDVI时间序列前两周期成分演变特征对比

注：m1为露天采场第一周期成分;m2为露天采场第二周期成分;n1为排土场第一周期成分;n2为排土场第二周期成分

4.3.2 露天采场与排土场周期振荡演变特征对比
西一、西三号露天矿采场、排土场均为短期振荡且振荡周期较为接近,周期振荡演变特征及振荡强度也较为相似（图8）。而东三号露天矿和露天锗矿出现较大差异,前者振荡演变特征与排土场较接近,但露天采场振荡强度明显高于排土场;后者采场和排土场的周期振荡演变特征近似对称分布。推测露天锗矿采场区域植被在早期就因采矿等外界因素干扰而发生明显变动,采矿活动导致该区域植被消失殆尽,因此周期振荡逐渐恢复为稳定状态。排土场植被生长动态较为复杂,一方面,周期振荡强度的逐渐增强可能与排土场植被受损面积不断增加有关;另一方面,实地调查发现,排土场的植被恢复措施使该区域仍然存在植被生长而非完全消失。此外,振荡强度的逐渐增强可能与排土场植被生长对外界因素干扰的响应具有累计效应有关。
Fig.8 Comparison of period component for the stopes and waste dump

图8 露天采场和排土场NDVI时间序列周期成分对比

4.3.3 伪不变特征区与采矿扰动区周期振荡演变特征对比
伪不变特征区与采矿扰动区第一周期振荡演变特征差异明显（图9）。前者在1-24月内振荡强烈,此后振荡强度逐渐减小并趋于平稳,从第84月开始振荡再次加强。后者的振荡强度从第96月开始逐渐加强,该特点与伪不变特征区较类似,但2个区域异常振荡的原因有所差异。结合2区域趋势成分演变特征（图5）,伪不变特征区的异常振荡与该区域NDVI在该时间内上下往复有关,而采矿扰动区的异常振荡现象是由采矿活动造成的植被损伤导致的。
Fig.9 Comparison of period component of NDVI time series between the unchanged feature region and the mining disturbance area

图9 伪不变特征区与采矿扰动区NDVI时间序列周期成分对比

5 结论

为充分反映植被生长对采矿扰动的响应特征,在月际尺度以MODIS-NDVI时间序列为数据源,通过奇异谱分析结合Mann-Kendall检测,放大并重建露天矿区采矿扰动区和伪不变特征区的趋势成分和周期成分,并对上述区域的变化趋势和周期振荡特征进行分析,探究SSA-Mann Kendall用于矿区时间序列变化检测的适用性以及采矿活动影响下露天矿区植被生长状况。得到以下主要结论：
（1）SSA-Mann Kendall能将NDVI时间序列中的微弱信号充分放大、提取,并重建为具有物理意义的趋势与周期振荡成分,并对趋势成分的显著程度和突变时间进行定量表达,其在一定程度上克服了现有基于变换方法的时间序列分析意义不够明确的问题。SSA-Mann Kendall重建的不同频率周期振荡成分能直观刻画不同时间尺度的细节信息,结合周期振荡与趋势成分的演变规律有助于识别各区域植被生长的动态特点。
（2）在变化趋势方面,露天采场的下降趋势显著,各矿井均达到0.01置信水平,同一矿井露天采场下降趋势较排土场更为显著,伪不变特征区植被虽处于下降趋势,但M-K结果尚未达到显著水平。地表植被损伤具有突发性,各特征区突变时间与矿井开发时间大致吻合,从突变点开始趋势成分迅速下降,即矿井开发建设初期地表植被即遭严重破坏,随着开采范围不断扩大,受其影响地表植被损伤程度逐渐增加。同一矿井,露天采场突变时间滞后于排土场,即矿井建设初期排土场植被受损较露天采场更为严重。在周期振荡特征方面,露天采场和排土场的不同扰动形式导致矿井露天采场和排土场周期振荡演变特征存在差异,采矿扰动下露天采场植被消失殆尽,排土场因植被恢复措施,其时间序列动态更为复杂。
（3）对各特征区异常周期振荡现象发生的原因进行了解释,其中,周期振荡强度异常增大的时间对应着该点露天矿井开发建设的开始,即在尺度较小的露天煤矿区,NDVI时间序列依然存在微弱的周期振荡信号,且不同于大区域尺度下,气候变化导致的周期振荡,矿区周期振荡现象与采矿活动密切相关。但对各特征区不同的振荡周期的原因尚缺乏合理的解释,同时也缺乏周期振荡强度的有效度量。

The authors have declared that no competing interests exist.

References

Publishing order | Descend order by publishing year | Descend order by cited within
[1]
赵英时. 遥感应用分析原理与方法[M].北京:科学出版社,2003:387-396.

[ Zhao Y S.Theory and method of remote sensing application and analysis[M]. Beijing: Science Press, 2003:387-396. ]

[2]
侯学会,牛铮,高帅.近十年中国东北森林植被物候遥感监测[J].光谱学与光谱分析,2014,34(2):515-519.基于归一化差值植被指数 (normalized difference vegetation index,NDVI),提出了一种新的物候遥感监测方法,基于森林植被的年NDVI时间累积曲线,利用Logistic模型对NDVI累积曲线进行拟 合,依据曲率极值方法提取森林植被物候期的关键参数(生长季开始日期,SOS;生长季结束日期,EOS),并对森林植被的生长季长度(length of season,LOS)进行分析,探讨近10年东北森林物候的时空变化。主要结论为:(1)2001年~2010年间,东北森林生长季开始日期集中在 110~140天但在10年间没有明显变化;(2)第260~290天,森林逐渐停止生长,生长季结束日期从北向南逐渐推迟,但在十年间几乎没变化; (3)与生长季开始和结束日期相对应,东北森林生长季长度集中在120~160天之间,但存在空间差异,大兴安岭地区森林生长季长度较短,集中在 120~140天之间,小兴安岭、长白山、辽东半岛地区的森林生长季长度可达到160天,对整个研究区来讲,近10年间变化的区域仅占研究区的 14.9%,变化趋势集中在1d/10年。研究结果与物候观测数据及已有的研究具有较好的一致性,说明利用遥感数据动态监测东北森林植被物候期具有一定的 可靠性。

DOI

[ Hou X H, Niu Z, Gao S.Phenology of forest vegetation in northeast of China in ten years using remote sensing[J]. Spectroscopy and Spectral Analysis, 2014,34(2):515-519. ]

[3]
Chen B Z, Xu G, Coops N C, et al.Changes in vegetation photosynthetic activity trends across the Asia-Pacific region over the last three decades[J]. Remote Sensing of Environment, 2014,144:82-41.The updated Global Inventory Modeling and Mapping Studies (GIMMS) third generation global satellite Advanced Very High Resolution Radiometer (AVHRR) Normalized Difference Vegetation Index (NDVI) dataset provides very detailed global information on the state of vegetation from 1982 to 2011. Using these data we investigated the changes in the vegetation photosynthetic activity in the Asia–Pacific (AP) (including Australia, South East Asia, China, and the Pacific Coast of North America) region, by discerning gradual changes into two key metrics: 1) the cumulative annual NDVI in each year and 2) the seasonality or variance in that index. We then assessed changes using break and turning points using three statistical models (least-square linear, expanded paired-consecutive linear and piecewise regression models). We found that the AP region overall experienced increasing NDVI from 1982 through 2011 with an average rate of 5.3002×0210 61024 02NDVI02yr 61021 (0.13%02yr 61021 ). The annual NDVI increased from 1982 at a faster rate of 26.1402×0210 61024 02NDVI02yr 61021 (0.65%02yr 61021 ) until a break in the trend after 1991 (after that the trend reduced to 5.7802×0210 61024 02NDVI02yr 61021 ). In the Asia–Australia (AA) subarea of the AP, vegetation greening slowly increased at 8.7102×0210 61024 02NDVI02yr 61021 before 2003 and then increased to 28.3002×0210 61024 02NDVI02yr 61021 after 2003. In contrast, in the North America (NA) subarea NDVI rapidly increased initially at 18.7202×0210 61024 02NDVI02yr 61021 before 1992 and then marginally increased (3.9602×0210 61024 02NDVI02yr 61021 ) after 1992. The turning points were found to be 2008 and 1987 for the AA and NA subareas respectively. Analysis of monthly NDVI data showed that the trends were positive for most of the months of the study period, particularly during the growing season. Geospatial analyses demonstrated that cumulative annual NDVI and the variance or seasonality across the large AP region varied across the different subareas. As well, we found evidence for different spatial patterns of the NDVI changes with strong spatial heterogeneity in the patterns of the break and turning points. This suggests complex and nonlinear responses of vegetation photosynthetic activity to regional climatic changes and other drivers.

DOI

[4]
贾明明,任春颖,刘殿伟,等.基于环境星与MODIS时序数据的面向对象森林植被分类[J].生态学报,2014,34(24):7167-7174.林区地形复杂、植被分布无序,且森林植被光谱信息相近,因而森林二级类型边界的确定成为土地覆盖遥感分类的难点.选择吉林省东部山区为研究区,以环境星影像(HJ-1 CCD)和中等分辨率成像光谱仪(MODIS)时序数据为基础,采用面向对象的分类方法进行森林植被类型的提取.分类特征参数主要选取了HJ-1 CCD的光谱和纹理特征,以及MODIS时序数据的物候特征.研究区总体分类精度为91.5%,Kappa系数为0.88,森林二级类型的分类精度均较高,其中落叶阔叶林的制图精度达到了97.1%.所用的面向对象分类方法与未加入物候特征的面向对象分类方法相比,森林二级类型的分类精度得到大幅度提高.

DOI

[ Jia M M, Ren C Y, Liu D W, et al.Object-oriented forest classification based on combination of HJ-1 CCD and MOD1S-NDVI data[J]. Acta Ecologica Sinica, 2014,34(24):7167-7174. ]

[5]
Brown J, Kastens H J, Coutinho A C, et al.Classifying multiyear agricultural land use data from Mato Grosso using time-series MODIS vegetation index data[J]. Remote Sensing of Environment, 2013,130:39-50.MODIS 250-m NDVI and EVI datasets are now regularly used to classify regional-scale agricultural land-use practices in many different regions of the globe, especially in the state of Mato Grosso, Brazil, where rapid land-use change due to agricultural development has attracted considerable interest from researchers and policy makers. Variation exists in which MODIS datasets are used, how they are processed for analysis, and what ground reference data are used. Moreover, various land-use/land-cover classes are ultimately resolved, and as yet, crop-specific classifications (e.g. soy-corn vs. soy-cotton double crop) have not been reported in the literature, favoring instead generalized classes such as single vs. double crop. The objective of this study is to present a rigorous multiyear evaluation of the applicability of time-series MODIS 250-m VI data for crop classification in Mato Grosso, Brazil. This study shows progress toward more refined crop-specific classification, but some grouping of crop classes remains necessary. It employs a farm field polygon-based ground reference dataset that is unprecedented in spatial and temporal coverage for the state, consisting of 2003 annual field site samples representing 415 unique field sites and five crop years (2005-2009). This allows for creation of a dataset containing "best-case" or "pure" pixels, which we used to test class separability in a multiyear cross validation framework applied to boosted decision tree classifiers trained on MODIS data subjected to different pre-processing treatments. Reflecting the agricultural landscape of Mato Grosso as a whole, cropping practices represented in the ground reference dataset largely involved soybeans, and soy-based classes (primarily double crop 'soy-commercial' and single crop 'soy-cover') dominated the analysis along with cotton and pasture. With respect to the MODIS data treatments, the best results were obtained using date-of-acquisition interpolation of the 16-day composite VI time series and outlier point screening, for which five-year out-of-sample accuracies were consistently near or above 80% and Kappa values were above 0.60. It is evident that while much additional research is required to fully and reliably differentiate more specific crop classes, particular groupings of cropping strategies are separable and useful for a number of applications, including studies of agricultural intensification and extensification in this region of the world. (C) 2012 Elsevier Inc. All rights reserved.

DOI

[6]
Ma X L, Huete A, Yu Q, et al.Spatial patterns and temporal dynamics in savanna vegetation phenology across the North Australian Tropical Transect[J]. Remote Sensing of Environment, 2013,139:97-115.The phenology of a landscape is a key parameter in climate and biogeochemical cycle models and its correct representation is central to the accurate simulation of carbon, water and energy exchange between the land surface and the atmosphere. Whereas biogeographic phenological patterns and shifts have received much attention in temperate ecosystems, much less is known about the phenology of savannas, despite their sensitivity to climate change and their coverage of approximately one eighth of the global land surface. Savannas are complex assemblages of multiple tree, shrub, and grass vegetation strata, each with variable phenological responses to seasonal climate and environmental variables. The objectives of this study were to investigate biogeographical and inter-annual patterns in savanna phenology along a 1100 km ecological rainfall gradient, known as North Australian Tropical Transect (NATT), encompassing humid coastal Eucalyptus forests and woodlands to xeric inland Acacia woodlands and shrublands. Key phenology transition dates (start, peak, end, and length of seasonal greening periods) were extracted from 13 years (2000-2012) of Moderate Resolution Imaging Spectroradiometer (MODIS) Enhanced Vegetation Index (EVI) data using Singular Spectrum Analysis (SSA).<br/>Two distinct biogeographical patterns in phenology were observed, controlled by different climate systems. The northern (mesic) portion of the transect, from 12 S, to around 17.7 degrees S, was influenced by the Inter-Tropical Convergence Zone (IT) seasonal monsoon climate system, resulting in strong latitudinal shifts in phenology patterns, primarily associated with the functional response of the C4 grass layer. Both the start and end of the greening ( enhanced vegetation activity) season occurred earlier in the northern tropical savannas and were progressively delayed towards the southern limit of the Eucalyptus-dominated savannas resulting in relatively stable length of greening periods. In contrast, the southern xeric portion of the study area was largely decoupled from monsoonal influences and exhibited highly variable phenology that was largely rainfall pulse driven. The seasonal greening periods were generally shorter but fluctuated widely from no detectable greening during extended drought periods to length of greening seasons that exceeded those in the more mesic northern savannas in some wet years. This was in part due to more extreme rainfall variability, as well as a C3/C4 grass-forb understory that provided the potential for extended greening periods. Phenology of Acacia dominated savannas displayed a much greater overall responsiveness to hydroclimatic variability. The variance in annual precipitation alone could explain 80% of the variances in the length of greening season across the major vegetation groups. We also found that increased variation in the timing of phenology was coupled with a decreasing tree-grass ratio. We further compared the satellite-based phenology results with tower-derived measures of Gross Ecosystem Production (GEP) fluxes at three sites over two contrasting savanna classes. We found good convergence between MODIS EVI and tower GEP, thereby confirming the potential to link these two independent data sources to better understand savanna ecosystem functioning. (C) 2013 Elsevier Inc. All rights reserved.

DOI

[7]
Jamali S, Jönsson P, Eklundh L, et al.Detecting changes in vegetation trends using time series segmentation[J]. Remote Sensing of Environment, 2015,156:182-195.DBEST was also tested using data from the Global Inventory Modeling and Mapping Studies (GIMMS) NDVI image time series for Iraq for the period 1982–2006, and was able to detect and quantify major change over the area. This showed that DBEST is able to detect and characterize changes over large areas. We conclude that DBEST is a fast, accurate and flexible tool for trend detection, and is applicable to global change studies using time series of remotely sensed data sets.

DOI

[8]
李晶,Carl E Z,李松,等.基于时序NDVI的露天煤矿区土地损毁与复垦过程特征分析[J].农业工程学报,2015,31(16):251-257.露天煤矿区是人类活动强扰动地区之一。该文以阿巴拉契亚煤田区韦恩县为研究区域，应用遥感时序分析法分析了像元尺度的土地损毁和复垦过程特征。得出结论：1984－2010年间，韦兹县露天开采扰动区域占采矿权范围的45.80%，其中植被恢复区域占开采范围的66.45%，开采时间越早，植被恢复像元比例越高；开采造成的地表无植被覆盖期时长中位数为6 a，均值为7 a；已充分复垦的区域，NDVI值恢复至采前水平的加权平均时长为12 a。基于像元变化轨迹的研究，除揭示土地损毁-复垦过程特征外，能较好地反映空间异质性，可以为土地复垦管理和相关政策决策提供科学依据。

[ Li J, Carl E Z, Li S, et al.Character analysis of mining disturbance and reclamation trajectory in surface coal-mine area by time-series NDVI[J]. Transactions of the Chinese Society of Agricultural Engineering, 2015,31(16):251-257. ]

[9]
Djebou D C, Singh P V, Frauenfeld W O.Vegetation response to precipitation across the aridity gradient of the southwestern United States[J]. Journal of Arid Environments, 2015,15:35-43.Atmospheric water demand affects a variety of factors, including primary production and the terrestrial water balance. Precipitation gradients from arid to humid regions also impact the water balance and play a large role in vegetation dynamics. Focusing on a 23-year period (1989–2011), we examine precipitation during the growing season in conjunction with the Normalized Difference Vegetation Index (NDVI) series for 21 satellite scenes spanning across the southwestern United States. We classify the satellite scenes into three different groups, based on the United Nations Aridity Index (AI). Group 1 is categorized as relatively humid with AI02≥020.65, group 2 is intermediate with 0.50≤AI023mm and >13mm) changed during this time. We also use cross-correlation analyses to establish the lagged behavior of the three types of vegetation in relation to precipitation amount and number of events. The vegetation response is similar between precipitation amount and number of precipitation events. However, in the arid region, we find distinct responses to precipitation depending on the vegetation type. The magnitude and significance of the vegetation response to precipitation patterns increase with environmental aridity. There is thus a meaningful disparity of vegetation behavior in time and space.

DOI

[10]
郭笑怡,刘德赢,张洪岩.大兴安岭NDVI时间序列的长程相关性特征分析[J].地球信息科学学报,2013,15(1):152-158.运用消除趋势波动分析方法和GIMMS数据集,研究大兴安岭1982-2006年NDVI长程相关性,揭示了大兴安岭NDVI动态的时间尺度特征,为气候-植被综合模型中时间尺度特性分析提供依据。主要结论有:(1)NDVI是一个复杂的动力系统,其动态过程受到多个因素和过程影响,NDVI可能存在多种不同的时空尺度。(2)一阶消除趋势波动分析可以避免忽略较小时间尺度特征,适合应用于15d分辨率的时间序列;大兴安岭NDVI在2-15个月内有较强的长程相关性;表明NDVI动态过程不是随机序列,是一个由内在自相似机制决定的长程相关过程;(3)随着时间的推移,NDVI动态时间特征发生变化,大兴安岭各生态地理区标度指数都存在&quot;拐点&quot;其出现时间在6.5-8.5个月之间,北段西侧与南段&quot;拐点&quot;出现时间比北段和中段早1-2个月。在较小时间尺度内,NDVI呈现出接近1/f噪声特点,在较大时间尺度上,NDVI具有长程幂律相关性,不同生态地理区NDVI持续性强度有差异。

DOI

[ Guo X Y, Liu D Y, Zhang H Y.Detecting long-range correlations in NDVI over Greater Khingan Mountains[J]. Journal of Geo-information Science, 2013,15(1):152-158. ]

[11]
蔡博峰,于嵘.基于遥感的植被长时序趋势特征研究进展及评价[J].遥感学报,2009,13(6):1170-1186.基于遥感的植被长时序变化特征是植被生态学研究的核心领域,也是全球变化研究的重点方向.AVHRR、SPOT VGT和MODIS是当前研究植被长时序趋势变化的主要数据源.海量数据不断积累的同时,植被长时序趋势特征研究方法却缺乏对比评价和分析.当前常用的方法有代数运算法、傅里叶变换、主成分分析、小波变换法、回归分析法和相关系数分析法等.在对各种方法评述和分析的基础上.重点讨论和对比了主流方法中的回归分析法和相关系数分析与新兴方法Sen+Mann-Kendall法.结果表明,Sen+Mann-Kendall能克服主流方法的不足,不需要数据服从某一特定分布,并且对数据的误差具有较强的抵抗能力.<dt><strong><t>Abstract：</t></strong></dt><dd>The long time series vegetation trends (LTSVT) research based on remote sensing in large area is the core field of vegetation ecology and an important direction in the global change study. AVHRR, SPOT VGT and MODIS are currently the main data resources of LTSVT research. With volumes of remote sensing data, the analysis and evaluation methods for LTSVT study emerged as an urgent issue. Algebra calculation, Fourier transformation, PCA analysis, wavelet transform, linear trend analysis (LTA), correlation analysis (CA), etc., are the main methods. After the assessing and grouping of the methods, we focused on comparing the LTA and CA, which were well accepted methods, with the newly introduced Sen + Mann-Kendall method. Our review showed Sen + Mann-Kendall had a strong strength of errors resistance and was not constrained by the data statistical distribution.

[ Cai B F, Yu R.Advance and evaluation in the long time series vegetation trends research based on remote sensing[J]. Journal of Remote Sensing, 2009,13(6):1170-1186. ]

[12]
Wessels K J, Prince S D, Malherbe J, et al.Can human-induced land degradation be distinguished from the effects of rainfall variability? a case study in South Africa[J]. Journal of Arid Environments, 2007,68(2):271-297.<h2 class="secHeading" id="section_abstract">Abstract</h2><p id="">Advanced Very High Resolution Radiometer (AVHRR), Normalized Difference Vegetation Index data (NDVI, 1&#xA0;km<sup>2</sup> 1985&ndash;2003) and modeled net primary production (NPP, 8&#xA0;km<sup>2</sup> 1981&ndash;2000) data were used to estimate vegetation production in South Africa (SA). The linear relationships of Log<em><sub>e</sub></em>Rainfall with NPP and ΣNDVI were calculated for every pixel. Vegetation production generally had a strong relationship with rainfall over most of SA. Therefore, human-induced land degradation can only be detected if its impacts on vegetation production can be distinguished from the effects of rainfall. Two methods were tested (i) Rain-Use Efficiency (RUE=NPP/Rainfall or ΣNDVI/Rainfall) and (ii) Residual Trends (RESTREND), i.e. negative trends in the differences between the observed ΣNDVI and the ΣNDVI predicted by the rainfall. Degraded areas mapped by the National Land Cover in north-eastern SA had reduced RUE; however, annual RUE had a very strong negative correlation with rainfall and varied greatly between years. Therefore, RUE was not a reliable indicator of degradation. The RESTREND method showed promising results at a national scale and in the Limpopo Province, where negative trends were often associated with degraded areas in communal lands. Both positive and negative residual trends can, however, result from natural ecological processes, e.g. the carryover effects of rainfall in previous years. Thus, the RESTREND method can only identify potential problem areas at a regional scale, while the cause of negative trends has to be determined by local investigations.</p>

DOI

[13]
刘亚龙,王庆,张明明,等.山东地区NDVI与气象因子持续性分析[J].资源科学,2010,32(9):1777-1782.探究植被的长期波动规律，对其持续性做出准确判断，是当前LUCC研究的重要课题。本文运用R/S(重标极差分析)分析了1998年到2008年旬值NDVI序列和与之对应的旬值气温、降水、日照数等气象因子序列的长程相关性，结果显示：NDVI序列和各个气象因子序列均存在长程相关性，并且Hurst指数存在突变点，气温、降水、NDVI序列的突变点分布大致相同，分布约在55旬左右。在1~55旬时间尺度上，气温、NDVI序列的Hurst指数大小很相近，呈很强烈的长程相关性。分析得出气温、降水序列的长程相关性影响NDVI的长程相关性，日照时间对NDVI其持续性影响甚小。NDVI时间序列在时间分布上具有分形特征,植被演变存在状态持续性及其内在的周期循环长度,从而为植被的非线性和复杂性研究提供了新的研究视角和实证结论。

[ Liu Y L, Wang Q, Zhang M M, et al.Analysis of the persistence of NDVI and climatic factors in the Shandong Peninsular[J]. Resources Science, 2010,32(9):1777-1782. ]

[14]
王澄海,崔洋.西北地区近50年降水周期的稳定性分析[J].地球科学进展,2006,21(6):576-584.<p>用小波分析和奇异谱分析两种方法相结合对西北地区26个站1951&mdash;1999共49年的降水周期随时间的变化进行了分析。结果表明： 西北地区降水周期随时间变化具有较强的区域性；宁夏陕北地区和河西走廊地区近50年的降水变化趋势呈反位相关系；根据近30年西北地区降水变化形式，西北地区大致可分为3种类型的降水变化：①不稳定变化型（新疆地区）；②过渡型（河西走廊地区）；③稳定型（宁夏陕北地区）。近50年西北地区降水普遍存在的准3年左右的周期在20世纪70~80年代显著性下降，5～7年左右的周期和9～14年左右的长周期也随时间有着不同的变化；即各地区所有周期成分的稳定性都显著性地随时间发生变化，各地降水普遍存在一个相对比较稳定的但不太显著的10年左右的平衡态。表明西北地区的降水随时间变化具有多平衡态和非周期性振荡的特点。但西北大部分地区都存在某一特定的周期成分，在该固定周期成分出现前后，其振幅变化与该地区降水趋势变化之间存在着密切关系。</p>

[ Wang C H, Cui Y.A study of the stability of the precipitation cycle over northwest China in the past 50 years[J]. Advances in Earth Science, 2006,21(6):576-584. ]

[15]
Golyandina N, Korobeynikov A.Basic singular spectrum analysis and forecasting with R[J]. Computational Statistics and Data Analysis, 2014,71:934-954.Singular Spectrum Analysis (SSA) is a powerful tool of analysis and forecasting of time series. The main features of the RSSA package, which efficiently implements the SSA algorithms and methodology in R, are described. Analysis, forecasting and parameter estimation are demonstrated using case studies. These studies are supplemented with accompanying code fragments. (C) 2013 Elsevier B.V. All rights reserved.

DOI

[16]
殷守敬,吴传庆,王桥,等.多时相遥感影像变化检测方法研究进展综述[J].光谱学与光谱分析,2013,33(12):3339-3342.近年来，随着遥感平台和传感器的发展，已经实现了对地球表面大部分区域的连续重复遥感观测，积累了海量的多源、多尺度、多分辨率遥感数据。这些数据详细记录了地表上各种地物的变化过程，使得基于遥感影像的中长期变化检测等全球变化研究成为可能，并极大地推动了遥感影像处理方法和应用的研究。但是，尽管许多学者已经开展了大量相关的研究工作，目前基于多时相遥感影像的变化检测仍然面临许多挑战，还没有形成相对完整、成熟的理论体系，对相关研究进展的系统性总结工作仍然相对缺乏。回顾了多时相遥感变化检测方法的发展现状，并根据输入数据类型和数量的不同将这些方法分成单时相分类比较法、双时相比较法和时序分析法三类，对其进展情况和特点分别进行总结分析，然后就多时相遥感影像变化检测方法研究中现存的问题加以分析，并尝试探讨了其发展趋势。

DOI

[ Yin S J, Wu C, Wang Q, et al.Review of change detection methods using multi-temporal remotely sensed images[J]. Spectroscopy and Spectral Analysis, 2013,33(12):3339-3342. ]

[17]
LPDAAC. MODIS user guide[EB/OL]. .

[18]
Jönsson P, Eklundh L.TIMESAT-a program for analyzing time-series of satellite sensor data[J]. Computers and Geosciences, 2004,30(8):833-845.<h2 class="secHeading" id="section_abstract">Abstract</h2><p id="">Three different least-squares methods for processing time-series of satellite sensor data are presented. The first method uses local polynomial functions and can be classified as an adaptive Savitzky&ndash;Golay filter. The other two methods are more clear cut least-squares methods, where data are fit to a basis of harmonic functions and asymmetric Gaussian functions, respectively. The methods incorporate qualitative information on cloud contamination from ancillary datasets. The resulting smooth curves are used for extracting seasonal parameters related to the growing seasons. The methods are implemented in a computer program, TIMESAT, and applied to NASA/NOAA Pathfinder AVHRR Land Normalized Difference Vegetation Index data over Africa, giving spatially coherent images of seasonal parameters such as beginnings and ends of growing seasons, seasonally integrated NDVI and seasonal amplitudes. Based on general principles, the TIMESAT program can be used also for other types of satellite-derived time-series data.</p>

DOI

[19]
Helldén U, Christian T.Regional desertification: a global synthesis[J]. Global and Planetary Change, 2008,64(3-4):169-176.The paper presents results on the use of NOAA AVHRR data for desertification monitoring on a regional-global level. It is based on processing of the GIMMS 8 km global NDVI data set. Time series of annually integrated and standardized annual NDVI anomalies were generated and compared with a corresponding rainfall data set (1981-2003). The regions studied include the Mediterranean basin, the Sahel from the Atlantic to the Red Sea, major parts of the drylands of Southern Africa, China-Mongolia and the drylands of South America, i.e. important parts of the desertification prone drylands of the world. It is concluded that the suggested methodology is a robust and reliable way to assess and monitor vegetation trends and related desertification on a regional-global scale. A strong general relationship between NDVI and rainfall over time is demonstrated for considerable parts of the drylands. The results of performed trend analysis cannot be used to verify any systematic generic land degradation/desertification trend at the regional-global level. On the contrary, a "greening-up" seems to be evident over large regions.

DOI

[20]
毕如田,白中科.基于遥感影像的露天煤矿区土地特征信息及分类研究[J].农业工程学报,2007,23(2):77-82.该研究利用Landsat TM数据，以安太堡大型露天煤矿为例，在对地物光谱特征深入分析的基础上，设计了大型露天矿区土地剧烈扰动下不同地物特征提取模型，提取了安太堡露天矿区植被高覆盖区、植被低覆盖区、剥离堆垫区、采煤运煤区及边坡区等不同地物信息。采用归一化植被指数<i>(NDVI)</i>提取植被低覆盖区与植被高覆盖区信息，采用(TM4-TM5)>0提取植被高覆盖区信息并与<i>NDVI</i>进行了比较，采用TM4<40提取采煤运煤区信息，采用TM4/TM7在0.99～1.01范围来提取边坡区信息, 并统计计算了各类地物所占面积和分布情况。对研究区TM影像进行主成分分析，剥离堆垫区、采煤运煤区和边坡区等反映矿区扰动特征的信息主要由第1主成分反映，植被低覆盖区和高覆盖区等反映矿区植被覆盖特征的信息主要由第2主成分反映，两个主成分的贡献率达到97.16％，并利用扰动特征和植被特征对研究区地物进行了分类。该技术与方法为露天矿地物变化动态监测以及土地复垦与生态重建均提供了准确数据支持。

[ Bi R T, Bai Z K.Land characteristic information and classification in opencast coal mine based on remote sensing images[J]. Transactions of the Chinese Society of Agricultural Engineering, 2007,23(2):77-82. ]

[21]
Tabari H, Talaee P H.Temporal variability of precipitation over Iran: 1966-2005[J]. Journal of Hydrology, 2011,396(3-4):313-320.Precipitation is a principal element of the hydrological cycle and its temporal variability is important from both the scientific and practical point of view. The annual and seasonal precipitation trends of 41 stations in Iran for the period 1966-2005 have been analyzed using the Mann-Kendall test, the Sen's slope estimator and the linear regression. The effective sample size method was applied to eliminate the effect of serial correlation on the Mann-Kendall test. The results indicated a decreasing trend in annual precipitation at about 60% of the stations. The decreasing trends were significant at seven stations at the 95% and 99% confidence levels. The magnitude of the significant negative trends in annual precipitation varied from (-)1.999 mm/year at Zanjan station to (-)4.261 mm/year at Sanandaj station. The spatial distribution of the annual precipitation trends showed that the significant negative trends occurred mostly in the northwest of Iran. On the seasonal scale, the trends in the spring and winter precipitations time series were mostly negative. The highest numbers of stations with significant trends occurred in winter while no significant positive or negative trends were detected by the trend tests in autumn precipitation. The significant negative trends ranged between (-)0.283 mm/year at Zahedan station and (-)0.807 mm/year at Sanandaj station in winter season. In addition, the highest and lowest significant increases of precipitation values were obtained over Semnan and Mashhad in summer at the rates of (+)0.110 mm/year and (+)0.036 mm/year, respectively. (C) 2010 Elsevier B.V. All rights reserved.

DOI

[22]
魏凤英. 现代气候统计诊断与预测技术[M].北京:气象出版社,2007:71-76.

[ Wei F Y.Modern climatic statistical diagnosis and forecasting technology[M]. Beijing: China Meteorological Press, 2007:71-76. ]

[23]
Vautard R, Yiou P, Ghil M.Singular spectrum analysis: a toolkit for short, noisy chaotic signals[J]. Physica D: Nonlinear Phenomena, 1982,58(1):95-126.The entire toolkit is validated against a set of four prescribed time series generated by known processes, quasi-periodic or chaotic. It is also applied to a time series of global surface air temperatures, 130 years long, which has attracted considerable attention in the context of the global warming issue and provides a severe test for noise reduction and prediction.

DOI

[24]
徐克红,程鹏飞,文汉江.太阳黑子数时间序列的奇异谱分析和小波分析[J].测绘科学,2007,32(6):35-38.本文对小波变换和奇异谱分析方法进行了简要介绍，对离散小波的分解和重构、奇异谱分析的重构进行了详细阐述。结合太阳黑子数1749年至2007年3月期间的月平均值时间序列进行了小波变换的分解和重构及SSA方法的重构，提取了其主要的周期特性，并对两种分析方法进行了比较。

DOI

[ Xu K H, Cheng P F, Wen H J.Singular spectrum analysis and wavelet analysis on time series of Sunspot[J]. Science of Surveying and Mapping, 2007,32(6):35-38. ]

[25]
李常斌,王帅兵,杨林山,等.1951-2010年洮河流域水文气象要素变化的时空特征[J].冰川冻土,2013,35(5):1259-1266.采用M-K检验、小波分析和Sen斜率等方法, 对1951-2010年60 a来洮河流域不同地理-生态区间水文气象要素变化的时空特征进行了综合研究.结果表明: 洮河流域水文气象要素呈现多种周期不同尺度的振荡特性, 气温、降水和径流的年代际变化周期以9~13 a和2~5 a最为常见.气温从1990年代中期开始明显上升, 突变的时间北部略早于南部, 西部明显迟于东部;除上游草原牧区外, 降水总体于1990年代初期开始减少;受降水变化影响, 流域河川径流量1990年代发生明显减少.过去60 a, 洮河流域气温以0.18 ℃&#183;(10a)<sup>-1</sup>的速率增温;降水以0.03 mm&#183;(10a)<sup>-1</sup>的速率减少;河川径流量以11.36 mm&#183;(10a)<sup>-1</sup>的速率减小;近30 a来洮河流域以0.63 ℃&#183;(10a)<sup>-1</sup>的速率增温, 降水以8.86 mm&#183;(10a)<sup>-1</sup>的速率减少, 径流以21.00 mm&#183;(10a)<sup>-1</sup>的速率减少.降水和径流变化在不同时期和不同生态-地理区间差异明显, 与区域气候和下垫面因素变化所致的流域能水通量过程的变异有关.

DOI

[ Li C B, Wang S B, Yang L S, et al.Spatial and temporal variation of main hydrologic meteorological elements in the Taohe River basin from 1961 to 2010[J]. Journal of Glaciology and Geocryology, 2013,35(5):1259-1266. ]

[26]
Qi W, Zhang Y L, Gao J G, et al.Climate change on the southern slope of Mt.Qomolangma (Everest) region in Nepal since 1971[J]. Journal of Geographical Sciences, 2013,23(4):595-611.Based on monthly mean, maximum, and minimum air temperature and monthly mean precipitation data from 10 meteorological stations on the southern slope of the Mt. Qomolangma region in Nepal between 1971 and 2009, the spatial and temporal characteristics of climatic change in this region were analyzed using climatic linear trend, Sen's Slope Estimates and Mann-Kendall Test analysis methods. This paper focuses only on the southern slope and attempts to compare the results with those from the northern slope to clarify the characteristics and trends of climatic change in the Mt. Qomolangma region. The results showed that: (1) between 1971 and 2009, the annual mean temperature in the study area was 20.0A degrees C, the rising rate of annual mean temperature was 0.25A degrees C/10a, and the temperature increases were highly influenced by the maximum temperature in this region. On the other hand, the temperature increases on the northern slope of Mt. Qomolangma region were highly influenced by the minimum temperature. In 1974 and 1992, the temperature rose noticeably in February and September in the southern region when the increment passed 0.9A degrees C. (2) Precipitation had an asymmetric distribution; between 1971 and 2009, the annual precipitation was 1729.01 mm. In this region, precipitation showed an increasing trend of 4.27 mm/a, but this was not statistically significant. In addition, the increase in rainfall was mainly concentrated in the period from April to October, including the entire monsoon period (from June to September) when precipitation accounts for about 78.9% of the annual total. (3) The influence of altitude on climate warming was not clear in the southern region, whereas the trend of climate warming was obvious on the northern slope of Mt. Qomolangma. The annual mean precipitation in the southern region was much higher than that of the northern slope of the Mt. Qomolangma region. This shows the barrier effect of the Himalayas as a whole and Mt. Qomolangma in particular.

DOI

[27]
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. ]

Options
Outlines

/

〈 〉