Access to this full-text is provided by MDPI.
Content available from Remote Sensing
This content is subject to copyright.
remote sensing
Article
Spatial–Temporal Vegetation Dynamics and Their
Relationships with Climatic, Anthropogenic, and Hydrological
Factors in the Amur River Basin
Shilun Zhou 1, 2, †, Wanchang Zhang 1, *,† , Shuhang Wang 3, Bo Zhang 3and Qiang Xu 4
Citation: Zhou, S.; Zhang, W.; Wang,
S.; Zhang, B.; Xu, Q. Spatial–Temporal
Vegetation Dynamics and Their
Relationships with Climatic,
Anthropogenic, and Hydrological
Factors in the Amur River Basin.
Remote Sens. 2021,13, 684. https://
doi.org/10.3390/rs13040684
Academic Editor: Tal Svoray
Received: 6 January 2021
Accepted: 10 February 2021
Published: 13 February 2021
Publisher’s Note: MDPI stays neutral
with regard to jurisdictional claims in
published maps and institutional affil-
iations.
Copyright: © 2021 by the authors.
Licensee MDPI, Basel, Switzerland.
This article is an open access article
distributed under the terms and
conditions of the Creative Commons
Attribution (CC BY) license (https://
creativecommons.org/licenses/by/
4.0/).
1Aerospace Information Research Institute, Chinese Academy of Sciences, Beijing 100094, China;
zhousl01@radi.ac.cn
2University of Chinese Academy of Sciences, Beijing 100049, China
3National Engineering Laboratory for Lake Pollution Control and Ecological Restoration, Chinese Research
Academy of Environmental Sciences, Beijing 100012, China; wangsh@craes.org.cn (S.W.);
zhangbo@craes.org.cn (B.Z.)
4State Key Laboratory of Geohazard Prevention and Geoenvironment Protection, Chengdu University of
Technology, Chengdu 610059, China; xq@cdut.edu.cn
*Correspondence: zhangwc@radi.ac.cn; Tel.: +86-10-8217-8131
† The first two authors contributed equally to this work and should be considered as co-first authors.
Abstract:
Information about the growth, productivity, and distribution of vegetation, which are
highly relied on and sensitive to natural and anthropogenic factors, is essential for agricultural
production management and eco-environmental sustainability in the Amur River Basin (ARB). In
this paper, the spatial–temporal trends of vegetation dynamics were analyzed at the pixel scale in
the ARB for the period of 1982–2013 using remotely sensed data of long-term leaf area index (LAI),
fractional vegetation cover (FVC), and terrestrial gross primary productivity (GPP). The spatial
autocorrelation characteristics of the vegetation indexes were further explored with global and local
Moran’s I techniques. The spatial–temporal relationships between vegetation and climatic factors,
land use/cover types and hydrological variables in the ARB were determined using a geographical
and temporal weighted regression (GTWR) model based on the observed meteorological data,
remotely sensed vegetation information, while the simulated hydrological variables were determined
with the soil and water assessment tool (SWAT) model. The results suggest that the variation in
area-average annual FVC was significant with an increase rate of 0.0004/year, and LAI, FVC, and GPP
all exhibited strong spatial heterogeneity trends in the ARB. For LAI and FVC, the most significant
changes in local spatial autocorrelation were recognized over the Sanjiang Plain, and the low–low
agglomeration in the Sanjiang Plain decreased continuously. The GTWR model results indicate that
natural and anthropogenic factors jointly took effect and interacted with each other to affect the
vegetated regime of the region. The decrease in the impact of precipitation to vegetation growth over
the Songnen Plain was determined as having started around 1991, which was most likely attributed
to dramatic changes in water use styles induced by local land use changes, and corresponded to
the negative correlation between pasture areas and vegetation indexes during the same period. The
analysis results presented in this paper can provide vital information to decision-makers for use in
managing vegetation resources.
Keywords:
vegetation dynamics; climate changes; Amur River Basin; hydrological variables; land
use/cover changes
1. Introduction
As a major component of terrestrial ecosystems, vegetation plays an important role
in material cycling and energy flows, and provides irreplaceable service functions that
maintain the wellbeing of our planet and all the creatures that inhabit it. These function
Remote Sens. 2021,13, 684. https://doi.org/10.3390/rs13040684 https://www.mdpi.com/journal/remotesensing
Remote Sens. 2021,13, 684 2 of 25
services include food provision, climate regulation, carbon sequestration, timber produc-
tion, biodiversity preservation, and soil protection [
1
–
8
]. Vegetation growth affects the
ecological balance, the terrestrial carbon cycle, water circulation, and other biochemical
processes [
2
,
9
]. Thus, information about the spatial–temporal dynamics of regional and
global vegetation is a fundamental need in order to facilitate the better management of our
planet ecosystem, provide effective support for environmental sustainability and ensure
the safety of agricultural production; these have become emerging issues in the field of
environmental studies [2,5,10].
With the rapid advancement of remote sensing technology, global-scale land surface
parameters, derived from long time series of remote sensing observations, have provided
a good description of vegetation changes [
11
–
13
]. Leaf area index (LAI), defined as one-
half the total developed area of green leaves per unit horizontal surface area, can serve
as a proxy for detecting vegetation greenness and serves as a measure of the amount
of vegetation that is vertically distributed [
14
–
19
]. Fractional vegetation cover (FVC),
defined as the projected area of aboveground vegetation per unit ground area, can char-
acterize the horizontal density of land surface vegetation and reflect plant community
structure [
20
–
22
]. In addition, gross primary productivity (GPP) is the photosynthetic
carbon assimilation by land plants per unit space and time, and can serve as an indicator
of vegetation productivity [
23
–
25
]. Based on remote sensing products of these parameters
at regional and global scales, scholars around the world have used different methods to
study spatial–temporal changes in vegetation greenness (related to canopy structure) and
vegetation productivity (related to canopy function) [
26
–
28
], such as trend analysis and
spatial autocorrelation analysis.
Vegetation growth, production and distribution are highly reliant on—and sensi-
tive to—natural and anthropogenic factors, such as precipitation, temperature, water re-
sources, farming, and urban expansion. The condition of the global vegetation is constantly
changing at various spatial and temporal scales, driven by natural and anthropogenic
factors [29–31]
. The response of vegetation dynamics to differences in environmental fac-
tors varies significantly across regions, according to the regional climate conditions, water
availability, and land cover [32,33].
As the tenth largest river basin in the world, the natural conditions of the Amur River
Basin (ARB), a transboundary basin in eastern Eurasia, vary greatly from region to region.
In addition, the impact intensity of human activities on the ground vegetation has varied
considerably under the influence of different national policies, lifestyles, and economic
development [
34
–
38
]. In recent decades, significant climate changes have been observed
in the ARB, where an annual mean temperature warming rate of about
0.34 ◦C/10a
for
1975–2004 was reported, which is approximately 2.5 times higher than that for the 1891–
2004 period [
39
]. Regional climatic characteristics and dramatic climate change govern the
spatial–temporal changes of vegetation in the basin. In the context of changes in climatic
and hydrological factors, the continued intensification of agricultural activities, driven by
the large demand for food and economic development, affects the vegetation dynamics in
the ARB and threatens the ecosystem [40–42].
Most previous studies have used correlation analyses and regression models to reveal
the factors driving changes in vegetation dynamics in the whole basin or sub-basins of
the ARB. For example, the Pearson correlation analysis, applied to investigate the rela-
tionship between growing season and seasonal Normalized Difference Vegetation Index
(NDVI) and climatic variations in the ARB, was reported in [
37
] and it was concluded that
growing season NDVI was mainly regulated by precipitation. A linear trend analysis of
vegetation, temperature, and precipitation, and a comparison between the correlation of
NDVI–temperature and NDVI–precipitation in northeastern China was carried out by Mao,
et al. [
43
]. However, the analysis methods used in most of the previous studies focused
solely on a temporal scale or spatial scale, and therefore, they do not provide a compre-
hensive analysis to deal with variability between vegetation dynamics and environmental
factors at both spatial and temporal scales simultaneously [
44
–
46
]. Additionally, a few
Remote Sens. 2021,13, 684 3 of 25
efforts have been made to investigate the relationships between hydrological factors and
vegetation dynamics in the ARB, with close interactions between hydrological factors and
vegetation growth having been found in many studies [
47
–
49
]. Therefore, exploring the
spatial–temporal vegetation dynamics and their relationships with climatic, anthropogenic,
and hydrological factors in the ARB is essential for better understanding the mechanisms
driving vegetation changes, and can provide more detailed scientific support for envi-
ronmental sustainability and agricultural production management. The objectives of this
study are specified as follows:
1.
Analyze the spatial trends of three parameters (LAI, FVC and GPP) that can be
considered as representative of the growth condition of surface vegetation in the
study region using Mann–Kendall and Sen’s slope methods to understand the spatial
variability of vegetation growth conditions.
2.
Explore the spatial autocorrelation characteristics of vegetation indexes in the ARB,
based on the results derived from Moran’s I technique on remotely sensed veg-
etation information, for determining the rapid shift of eco-system changes over
previous decades.
3.
Utilize partial least squares regression (PLSR) and geographical and temporal weighted
regression (GTWR) models to further evaluate the relationships between land surface
parameters and climatic factors, land use/cover types, and hydrological variables
simulated with the soil and water assessment tool (SWAT) model in the ARB to
clarify spatial–temporal vegetation dynamics and their relationships with climatic,
anthropogenic, and hydrological factors in the Amur River Basin.
2. Study Area
As the 10th largest river basin in the world, the ARB (41
◦
–56
◦
north and 107
◦
–142
◦
east) is situated in northeastern Asia. Its tributary rivers flow through China, Russia,
Mongolia, and North Korea [
36
,
38
,
41
], covering a total land area of approximately 2 million
square kilometers with elevations ranging from 0 (the mouth of the Amur River into the
Tartary Strait) to 2565 m asl (the Khentii Mountains in northeastern Mongolia) [
50
,
51
] (See
Figure 1). As one of the most important transnational river basins, the ARB exhibits quite
different regional geographic and climatic characteristics: the topography is characterized
by high mountains over the west, with typical continental climate features and low hills in
east margin of the watershed, with an extensive plain in between, where the monsoonal
climate is predominant. Population density presents a north–south gradient over the basin,
with notable different anthropogenic effects [
34
,
35
,
38
]. Obtaining a better understanding
of the vegetation dynamics driven jointly by natural and anthropogenic factors in the ARB,
is therefore necessary for environmental sustainability studies. The hydrological variables
derived from SWAT model simulations were confined in the sub-basin of the ARB where
the Komsomolsk hydrological station gauged, covering 94% area of the whole ARB.
Remote Sens. 2021,13, 684 4 of 25
RemoteSens.2021,13,6844of27
Figure1.GeographicmapshowingthelocationoftheAmurRiverBasininEurasiawiththetopographicinformation
attachedandthelocationsof193meteorologicalstationsusedforthisstudy.
3.MaterialsandMethods
3.1.Datasets
Inthisstudy,Globallandsurfacesatellite(GLASS)LAI,FVCandGPPproductsfor
theperiodof1982–2013wereusedtoanalyzevegetationdynamics.TheseGLASSprod‐
uctswerereleasedusingtheHierarchicalDataFormat(HDF),accessibletothepublicfor
freeathttp://glass‐product.bnu.edu.cn.Thedetailedinformationoflandsurfaceparame‐
terproducts,representativeofthevegetationdynamicsusedinthisstudyarepresented
inTable1.ThemonthlyLAIandFVCvalueswerecomputedusingthemaximumvalue
composite(MVC)technique[52],whichiswidelyusedincompositingNDVIandother
vegetationindexesinordertominimizetheinterferenceofatmosphericeffects,thescan
angle,cloudcontamination,andsolarzenithangle[29,53–57].
Table1.Landsurfaceparameterproductsusedinthisstudyasrepresentativeofvegetationdynamics.
ParameterProductSensorSpatial
Resolution
Temporal
ResolutionDataprocessingReference
LAIGLASSAVHRRabout5km8‐day
1)UseMVCtechniquetoproducemonthly
databasedon8‐dayproducts
2)Averagemonthlydatatoobtainannualdata
[19,58]
FVCGLASSAVHRRabout5km8‐day
1)UseMVCtechniquetoproducemonthly
databasedon8‐dayproducts
2)Averagemonthlydatatoobtainannualdata
[59]
GPPGLASSAVHRRabout5km8‐daySumming8‐dayproductstoobtainannual
data[60]
NOTE:LAI:Leafareaindex;FVC:Fractionalvegetationcover;GPP:Grossprimaryproductivity;GLASS:Globallandsurface
satellite;AVHRR:AdvancedVeryHighResolutionRadiometer;MVC:maximumvaluecomposite.
Figure 1.
Geographic map showing the location of the Amur River Basin in Eurasia with the topographic information
attached and the locations of 193 meteorological stations used for this study.
3. Materials and Methods
3.1. Datasets
In this study, Global land surface satellite (GLASS) LAI, FVC and GPP products for
the period of 1982–2013 were used to analyze vegetation dynamics. These GLASS products
were released using the Hierarchical Data Format (HDF), accessible to the public for free
at http://glass-product.bnu.edu.cn. The detailed information of land surface parameter
products, representative of the vegetation dynamics used in this study are presented in
Table 1. The monthly LAI and FVC values were computed using the maximum value
composite (MVC) technique [
52
], which is widely used in compositing NDVI and other
vegetation indexes in order to minimize the interference of atmospheric effects, the scan
angle, cloud contamination, and solar zenith angle [29,53–57].
Table 1. Land surface parameter products used in this study as representative of vegetation dynamics.
Parameter Product Sensor Spatial
Resolution
Temporal
Resolution Data processing Reference
LAI GLASS AVHRR about 5 km 8-day
(1) Use MVC technique to
produce monthly data
based on 8-day products
(2) Average monthly data
to obtain annual data
[19,58]
FVC GLASS AVHRR about 5 km 8-day
(1) Use MVC technique to
produce monthly data
based on 8-day products
(2) Average monthly data
to obtain annual data
[59]
GPP GLASS AVHRR about 5 km 8-day Summing 8-day products
to obtain annual data [60]
NOTE: LAI: Leaf area index; FVC: Fractional vegetation cover; GPP: Gross primary productivity; GLASS: Global land surface satellite;
AVHRR: Advanced Very High Resolution Radiometer; MVC: maximum value composite.
Remote Sens. 2021,13, 684 5 of 25
According to characteristics of climate changes during different periods in the ARB
that have described in previous studies [
61
], in addition to the land use/cover transfor-
mations reported in the literature [
36
,
40
,
62
], the entire study period was divided into
four periods to analyze the spatial–temporal relationships between environmental factors
and vegetation dynamics: 1982–1990 (Period 1), 1991–1999 (Period 2), 2000–2006 (Period
3), and 2007–2013 (Period 4). The detailed information of datasets used in this study as
representative of environmental factors are presented in Table 2. To investigate the driving
factors of climate on vegetation changes, temperature and precipitation data observed from
1982 to 2013 in 193 meteorological stations within the basin were used. The climate data
were interpolated at a spatial resolution of 8.8
×
8.8 km by using thin plate smoothing
splines and Anusplin software, and the LAI, GPP, and FVC datasets in the same duration
were resampled to the same spatial resolution. The combined datasets of Chinese land
use maps with global land use maps from the closest dates in the four historical periods
were used as inputs for the GTWR model to investigate the anthropogenic driving factors
responsible for regional vegetation changes (urbanization and wetland reclamation, etc.).
In addition, the hydrological variables simulated by the SWAT model, as presented in
our previous study [
61
], were adopted to analyze the relationship between hydrological
variables and vegetation dynamics.
Table 2. Datasets used in this study as representative of environmental factors.
Data Type Description Source Download Site
Climate
Daily temperature and precipitation
data from 1982 to 2013 in 193
meteorological stations
China Meteorological Data
Network (CMA) and National
Oceanic and Atmospheric
Administration (NOAA)’s
National Centers for
Environmental
Information (NCEI)
https://data.cma.cn/en
https:
//www.ncdc.noaa.gov/cdo-web/
Land-use
Chinese land-use map (1980, 1995,
2005, and 2010) and global land-use
map (1992–1993, 2000, 2005, and
2010) at spatial resolution of 1 km
Chinese land-use maps from
Resource and Environment Data
Cloud Platform;
Moderate Resolution Imaging
Spectroradiometer (MODIS) land
cover type product from U.S.
Geological Survey (USGS);
Global Land Cover
Characterization (GLCC)
from USGS;
and Global Land Cover 2000
database (GLC2000) from Joint
Research Centre,
European Commission
http://www.resdc.cn/Datalist1
.aspx?FieldTyepID=1,3
https://lpdaac.usgs.gov/
products/mcd12q1v006/
https://www.usgs.gov/centers/
eros/science/usgs-eros-archive-
land-cover-products-global-land-
cover-characterization-glcc
http://forobs.jrc.ec.europa.eu/
products/glc2000/data_access.php
Hydrology
Surface runoff (SURQ), lateral flow
(LATQ), snowmelt (SM), soil water
(SW), ground-water flow (GWQ),
and evapotranspiration (ET)
Simulated by the soil and water
assessment tool (SWAT) model authors’ previous study [61]
3.2. Methods
The paper presents the results of applying the Mann–Kendall, Sen’s Slope, and Global
and Local Moran’s I techniques to analyze vegetation changes and applying PLSR and
GTWR models to analyze the relationships between vegetation and climatic, anthropogenic,
and hydrological factors based on meteorological data, remotely sensed information and
hydrological variables simulated by the SWAT model. Figure 2shows an overview of the
general workflow followed in this study.
Remote Sens. 2021,13, 684 6 of 25
RemoteSens.2021,13,6846of27
informationandhydrologicalvariablessimulatedbytheSWATmodel.Figure2showsan
overviewofthegeneralworkflowfollowedinthisstudy.
Figure2.Flowchartofinvestigatingthespatial‐temporalvegetationdynamicsandtheir
relationshipswithclimatic,anthropogenic,andhydrologicalfactorsintheAmurRiverBasin
(ARB).PLSR:partialleastsquaresregression;GTWR:geographicalandtemporalweighted
regression;LISA:localindicatorsofspatialassociation.
3.2.1.TimeTrendAnalysis
ThemagnitudeofthetrendsinannualLAI,FVCandGPPwereestimatedusingSen’s
slopemethod[63].Themedianvalueoftheslopeseriescomputedfromtwoconsecutive
pointsoftheseriescanhelpreducetheinfluenceofoutliersormissingdata[63–65].The
formulaofSen’sslopemethodisasfollows:
Figure 2.
Flowchart of investigating the spatial-temporal vegetation dynamics and their relation-
ships with climatic, anthropogenic, and hydrological factors in the Amur River Basin (ARB). PLSR:
partial least squares regression; GTWR: geographical and temporal weighted regression; LISA: local
indicators of spatial association.
3.2.1. Time Trend Analysis
The magnitude of the trends in annual LAI, FVC and GPP were estimated using Sen’s
slope method [
63
]. The median value of the slope series computed from two consecutive
Remote Sens. 2021,13, 684 7 of 25
points of the series can help reduce the influence of outliers or missing data [
63
–
65
]. The
formula of Sen’s slope method is as follows:
Slope =Median(xj−xi)
(j−i),∀j>i(1)
where
xi
and
xj
are the values at times iand j(1
≤i<j≤n
), respectively, and a positive
(negative) value of Slope indicates an upward (downward) trend.
The Mann–Kendall [
66
,
67
] non-parametric statistical test, which is frequently used to ex-
amine the significance of trends in time series of data, remining insensitive to
outliers [23,68,69]
,
was used to analyze the significance of the annual LAI, FVC, and GPP trends in each grid
for the ARB during the 1982–2013 period. At a given significance level a= 0.05, the thresh-
old of the normal distribution is
Z1−a/2
(
Z1−a/2 =Z0.975 =
1.96). When
|Z|≤Z1−a/2
,
the null hypothesis can be accepted (the trend is nonsignificant), and when
|Z|>Z1−a/2
,
the trend is significant. The sequential version of the Mann–Kendall test proposed by
Sneyers [
68
] was used to test abrupt changes in data series. If the intersection point of
the forward sequence statistic UFk and the backward sequence UBk, which are calculated
using the same equation but in reverse data series order, is between the boundary lines of
the confidence zone (
±
1.96 at the confidence level of 95%), the critical point of change can
be detected [70–73].
3.2.2. Spatial Autocorrelation Analysis
Spatial autocorrelation is the term used to describe the dependency and covariance
of variables within a geographic area, and it reveals the phenomenon of geographic
proximity’s potential [
74
,
75
]. In this paper, both global and local measures of spatial
autocorrelation were calculated using the Python package GeoRasters to evaluate the
degree of spatial autocorrelation in vegetation growth patterns across the study region and
to indicate the variations in spatial autocorrelation throughout the study period [12,76].
Moran’s I [
77
] is one of the most commonly used indexes of global spatial autocor-
relation. Moran’s I values range from
−
1 to 1: the index value from 0 to 1 indicates a
positive autocorrelation, which implies a clustering state of spatial geographical phenom-
ena, and the index value from
−
1 to 0 represents a negative autocorrelation, which implies
a larger difference between neighbors, [
74
,
78
–
80
]. The formula for Moran’s I is given in
Equation (2):
I=
nn
∑
i=1
n
∑
j=1
wij (xi−x)xj−x
n
∑
i=1
n
∑
j=1
wij
n
∑
i=1
(xi−x)2(2)
where
xi
and
xj
are the values of variable x(vegetation indexes) at location iand location j,
respectively.
x
is the average value of x,nis the total number of locations, and
wij
represents
the spatial weight [4,44,81,82].
A local spatial autocorrelation analysis was performed using the local indicators of
spatial association (LISA) [
83
,
84
] to indicate the spatial heterogeneity of the geographic phe-
nomena at the local level across the study area, based on the following
equation [74,83,85]
:
Ii=(xi−x)
S2∑
j
wij (xj−x)(3)
where
S2
represents the variance of the observed values. In this study, the LISA map
shows the spatial distribution (clustered/random/dispersed) of each vegetation index
based on 9999 permutations at the significance level of p< 0.05. The spatial cluster/outlier
characteristics of regions can be divided into four categories: low–low clusters (LL), high–
high clusters (HH), low–high outliers (LH), and high–low outliers (HL). Positive and
Remote Sens. 2021,13, 684 8 of 25
negative LISA values represent the aggregation of similar observations—“spatial cluster”,
and different observations—“spatial outlier”, respectively [79,80,86–89].
3.2.3. GTWR
GTWR [
45
] was used in this study to reveal the spatial–temporal heterogeneity of the
relationships between vegetation growth (LAI, FVC and GPP) and environmental variables
(climatic, anthropogenic and hydrological factors) in the ARB at the sub-basin level [
90
].
Unlike the traditional geographically weighted regression model which considers only
the spatial dimension, the GTWR model can well reflect evolution in the spatial–temporal
relationships between vegetation dynamics and environmental variables and can simulta-
neously explore both spatial and temporal non-stationarity in vegetation changes [
91
,
92
].
The GTWR model can be expressed as follows:
Yi=β0(ui,vi,ti) + ∑
k
βk(ui,vi,ti)Xik +εi(4)
ˆ
β(ui,vi,ti) = hXTW(ui,vi,ti)Xi−1XTW(ui,vi,ti)Y(5)
where
(ui
,
vi
,
ti)
denotes the coordinates of location iin space and time;
Yi
and
Xik
are the de-
pendent variable and the kth independent variable, respectively;
β0(ui
,
vi
,
ti)
represents the
intercept value,
βk(ui
,
vi
,
ti)
is a set of parameter values at space–time location i;
ˆ
β(ui
,
vi
,
ti)
is the estimation of the parameters
βk(ui
,
vi
,
ti)
;
W(ui
,
vi
,
ti)
denotes the spatial–temporal
weight matrix, and εiis the residual error.
4. Results
4.1. Spatial–Temporal Variations in Vegetation in the ARB
To detect the spatial–temporal variations of vegetation in the ARB basin, the Mann–
Kendall test and Sen’s slope were employed to examine the trends and abrupt changes
in the LAI, FVC and GPP from 1982 to 2013 at the confidence level of 95%. Figure 3a–c
presents the Mann–Kendall Z value and Sen’s slope of the area-averaged annual vegetation
indexes across the study region. During the study period, the annual FVC in the study
region varied from a minimum of 0.327 in 1983 to a maximum of 0.352 in 2013 with a
significant increase rate of about 0.0004/year (Z = 3.57, p< 0.05). While, the changes in
area-averaged annual LAI in the same duration were not significant (Z =
−
1.67, p> 0.05),
these values varied from a minimum of 1.369 in 2003 to a maximum of 1.563 in 1999, with a
slight decreasing trend of
−
0.0013/year. Overall, an increasing trend at 0.19 gC m
−2
yr
−2
(Z = 0.21, p> 0.05) was found in annual GPP, however, large interannual fluctuations with a
maximum value of 900.37 gC m
−2
yr
−1
(1988) and a minimum value of 784.05 gC m
−2
yr
−1
(2003) were observed. The period averages of LAI, FVC, and GPP for the four periods in
Figure 3a–c indicated that the period averages of FVC gradually increased during the four
periods, while the period averages of LAI and GPP changed most dramatically between
Period 2 and Period 3 (before and after 2000).
The sequential Mann–Kendall tests for the annual LAI with a forward-trend UFk and
backward-trend UBk, as illustrated in Figure 3d–f, indicated an obvious abrupt change in
LAI in 2001. In addition, a significant abrupt change in GPP occurred in the study area
around 2001. Several abrupt changes in FVC occurred from 1999 to 2008, but most of
them were insignificant. These results suggested that all three types of indexes describing
different vegetation statuses had undergone a rapid change around the year 2001.
Figure 4a–i shows the spatial patterns of pixel-by-pixel Sen’s slope and the Mann–
Kendall Z value estimated over the ARB, as well as the significance of the trends computed
at a 95% significance level for the time series LAI, FVC and GPP images. Apparent
spatial heterogeneity for the changes of LAI from 1982 to 2013 over the ARB, as exhibited
in Figure 4a–c
, was carefully investigated. The results indicated that about 37.3% of the
basin area showed an increasing tendency in terms of LAI, in which approximately 7.7%
tended to increase significantly (p< 0.05), where the most obvious increasing trend occurred
Remote Sens. 2021,13, 684 9 of 25
over the Songnen Plain (an important food and oil production national base of China) in
the southern part of the ARB. On the contrary, a decreasing trend in LAI was significant
over about 18.9% of the ARB area, where the most significant reductions in LAI were
mainly distributed in Mongolia and the Inner Mongolia Autonomous Region of China in
the southwestern ARB area. Consistent with the area-averaged LAI trend, the LAI changes
at the regional scale were insignificant over a large portion of the study area.
RemoteSens.2021,13,6849of27
Figure3.Temporalchangesinarea‐averagedannualLAI(a),FVC(b)andGPP(c),andsequentialMann–Kendalltestsfor
theannualLAI(d),FVC(e)andGPP(f)withtheforward‐trendUFkandbackward‐trendUBkintheARBduring1982–
2013.p<0.05indicatessignificanceatthe95%confidentlevel.
ThesequentialMann–KendalltestsfortheannualLAIwithaforward‐trendUFkand
backward‐trendUBk,asillustratedinFigure3d–f,indicatedanobviousabruptchangein
LAIin2001.Inaddition,asignificantabruptchangeinGPPoccurredinthestudyarea
around2001.SeveralabruptchangesinFVCoccurredfrom1999to2008,butmostofthem
wereinsignificant.Theseresultssuggestedthatallthreetypesofindexesdescribing
differentvegetationstatuseshadundergonearapidchangearoundtheyear2001.
Figure4a–ishowsthespatialpatternsofpixel‐by‐pixelSen’sslopeandtheMann–
KendallZvalueestimatedovertheARB,aswellasthesignificanceofthetrends
computedata95%significancelevelforthetimeseriesLAI,FVCandGPPimages.
ApparentspatialheterogeneityforthechangesofLAIfrom1982to2013overtheARB,as
exhibitedinFigure4a–c,wascarefullyinvestigated.Theresultsindicatedthatabout37.3%
ofthebasinareashowedanincreasingtendencyintermsofLAI,inwhichapproximately
7.7%tendedtoincreasesignificantly(p<0.05),wherethemostobviousincreasingtrend
occurredovertheSongnenPlain(animportantfoodandoilproductionnationalbaseof
China)inthesouthernpartoftheARB.Onthecontrary,adecreasingtrendinLAIwas
significantoverabout18.9%oftheARBarea,wherethemostsignificantreductionsinLAI
weremainlydistributedinMongoliaandtheInnerMongoliaAutonomousRegionof
ChinainthesouthwesternARBarea.Consistentwiththearea‐averagedLAItrend,the
LAIchangesattheregionalscalewereinsignificantoveralargeportionofthestudyarea.
Figure 3.
Temporal changes in area-averaged annual LAI (
a
), FVC (
b
) and GPP (
c
), and sequential Mann–Kendall tests for
the annual LAI (
d
), FVC (
e
) and GPP (
f
) with the forward-trend UFk and backward-trend UBk in the ARB during 1982–2013.
p< 0.05 indicates significance at the 95% confident level.
Over past 32 years in the ARB, 84.3% of the total basin area has exhibited an increasing
trend in FVC, and pixels with a significant increasing trend accounted for about 62.5% of
this value (p< 0.05), mainly being concentrated in the Chinese and Russian parts of the
ARB. Only 2.0% of the total area showed a significantly decreasing trend, and pixels with
a decreasing trend were mainly distributed in the southwestern margin of the ARB. For
GPP, approximately 45.1% of the area in the ARB presented a decreasing trend, in which
about 10.5% of the area was decreasing significantly (p< 0.05). Furthermore, areas with
an increasing tendency accounted for 54.9% of the area, among which, 16.3% showed a
significant increasing trend over the ARB from 1982 to 2013 (slope > 0, p< 0.05).
4.2. Spatial Autocorrelation Analysis of Vegetation Dynamics in the ARB
The global Moran’s I was investigated for the four period-averaged LAI, FVC, and GPP
during 1982–2013 over the ARB. The global Moran’s I values of LAI for the four periods
were 0.898, 0.897, 0.913 and 0.917 with p-values all being less than 0.01, respectively, which
indicated a significant positive spatial autocorrelation and persistent high–high or low–low
aggregations of the LAI. Similarly, the global Moran’s I values of FVC and GPP for the four
stages were also all above 0.8, with p-values all less than 0.01, indicating positive spatial
autocorrelation and persistent high–high or low–low aggregations. However, the global
Moran’s I values do not indicate the possible spatial changes of spatial autocorrelation of
the studied indexes across the basin.
Remote Sens. 2021,13, 684 10 of 25
Figure 5presents the clustering maps of the local Moran’s I of LAI, FVC and GPP
in the ARB for the four studied periods (Figure 5a–d show Periods 1–4, respectively).
The local Moran’s I for LAI in the four periods ranged from
−
1.83 to
−
10.41,
−
1.82 to
9.47,
−
1.57 to 5.55 and
−
0.90 to 5.93, respectively. With respect to the local Moran’s
I of the LAI dataset for Periods 1–4, the areas of low–low agglomeration were always
concentrated in the Mongolian region and the Songnen Plain and the Inner Mongolian
Autonomous Region of China over four different phases. The clustering maps of the local
Moran’s I values further revealed that the local spatial autocorrelation of the Sanjiang
Plain demonstrated the most significant change and the low–low agglomeration in the
Sanjiang Plain decreased persistently. Among these changes, the most significant changes
were found between Period 2 and Period 3: i.e., the autocorrelation changed the most
around 2000 over the Sanjiang Plain, while the low–low agglomeration in the Sanjiang
Plain disappeared after 2000.
RemoteSens.2021,13,68410of27
Figure4.SpatialpatternsofSen’sslope,theMann–KendallZscoreandthesignificanceofchangesfortheannualLAI(a–
c),FVC(d–f)andGPP(g–i)from1982to2013intheARB.
Overpast32yearsintheARB,84.3%ofthetotalbasinareahasexhibitedan
increasingtrendinFVC,andpixelswithasignificantincreasingtrendaccountedforabout
62.5%ofthisvalue(p<0.05),mainlybeingconcentratedintheChineseandRussianparts
oftheARB.Only2.0%ofthetotalareashowedasignificantlydecreasingtrend,andpixels
withadecreasingtrendweremainlydistributedinthesouthwesternmarginoftheARB.
ForGPP,approximately45.1%oftheareaintheARBpresentedadecreasingtrend,in
whichabout10.5%oftheareawasdecreasingsignificantly(p<0.05).Furthermore,areas
withanincreasingtendencyaccountedfor54.9%ofthearea,amongwhich,16.3%showed
asignificantincreasingtrendovertheARBfrom1982to2013(slope>0,p<0.05).
4.2.SpatialAutocorrelationAnalysisofVegetationDynamicsintheARB
TheglobalMoran’sIwasinvestigatedforthefourperiod‐averagedLAI,FVC,and
GPPduring1982–2013overtheARB.TheglobalMoran’sIvaluesofLAIforthefour
periodswere0.898,0.897,0.913and0.917withp‐valuesallbeinglessthan0.01,
respectively,whichindicatedasignificantpositivespatialautocorrelationandpersistent
high–highorlow–lowaggregationsoftheLAI.Similarly,theglobalMoran’sIvaluesof
Figure 4.
Spatial patterns of Sen’s slope, the Mann–Kendall Z score and the significance of changes for the annual LAI (
a
–
c
),
FVC (d–f) and GPP (g–i) from 1982 to 2013 in the ARB.
Remote Sens. 2021,13, 684 11 of 25
RemoteSens.2021,13,68411of27
FVCandGPPforthefourstageswerealsoallabove0.8,withp‐valuesalllessthan0.01,
indicatingpositivespatialautocorrelationandpersistenthigh–highorlow–low
aggregations.However,theglobalMoran’sIvaluesdonotindicatethepossiblespatial
changesofspatialautocorrelationofthestudiedindexesacrossthebasin.
Figure5presentstheclusteringmapsofthelocalMoran’sIofLAI,FVCandGPPin
theARBforthefourstudiedperiods(Figure5a–dshowPeriods1–4,respectively).The
localMoran’sIforLAIinthefourperiodsrangedfrom−1.83to−10.41,−1.82to9.47,−1.57
to5.55and−0.90to5.93,respectively.WithrespecttothelocalMoran’sIoftheLAIdataset
forPeriods1–4,theareasoflow–lowagglomerationwerealwaysconcentratedinthe
MongolianregionandtheSongnenPlainandtheInnerMongolianAutonomousRegion
ofChinaoverfourdifferentphases.TheclusteringmapsofthelocalMoran’sIvalues
furtherrevealedthatthelocalspatialautocorrelationoftheSanjiangPlaindemonstrated
themostsignificantchangeandthelow–lowagglomerationintheSanjiangPlain
decreasedpersistently.Amongthesechanges,themostsignificantchangeswerefound
betweenPeriod2andPeriod3:i.e.,theautocorrelationchangedthemostaround2000
overtheSanjiangPlain,whilethelow–lowagglomerationintheSanjiangPlain
disappearedafter2000.
Figure5.ClusteringmapsoflocalMoran’sIofLAI,FVCandGPPintheARBforfourstudiedperiods:(a–d)show
clusteringmapsofLAIforPeriods1–4,respectively.(e–h)and(i–l)arethesameas(a–d)offourstudiedperiods,butfor
clusteringmapsofFVCandGPP,respectively.
Figure5e–hpresentedtheclusteringmapsoflocalMoran’sIofFVCintheARBfor
fourstudiedperiods.ThelocalMoran’sIofFVCforthefourperiodsrangedfrom−0.67
to6.45,−0.69to6.67,0.57to6.21and−0.65to6.63,respectively.LiketheLAIspatial
autocorrelation,thelocalspatialautocorrelationsofFVCforthefourperiodswere
Figure 5.
Clustering maps of local Moran’s I of LAI, FVC and GPP in the ARB for four studied periods: (
a
–
d
) show
clustering maps of LAI for Periods 1–4, respectively. (
e
–
h
) and (
i
–
l
) are the same as (
a
–
d
) of four studied periods, but for
clustering maps of FVC and GPP, respectively.
Figure 5e–h presented the clustering maps of local Moran’s I of FVC in the ARB
for four studied periods. The local Moran’s I of FVC for the four periods ranged from
−
0.67 to 6.45,
−
0.69 to 6.67, 0.57 to 6.21 and
−
0.65 to 6.63, respectively. Like the LAI
spatial autocorrelation, the local spatial autocorrelations of FVC for the four periods were
relatively stable, especially for the high–high agglomeration in the northern part of the
basin. The clustering maps for the four periods all displayed strong high–high and low–low
agglomerations, whereas the spatial outliers (high–low clustering or low–high clustering)
were not conspicuous. A similar decreasing trend in low–low agglomeration areas in the
Sanjiang Plain was also found during the study period.
Figure 5i–l presents the clustering maps of local Moran’s I of GPP in the ARB for four
studied periods. The local Moran’s I of GPP for the four periods ranged from
−
1.77 to 8.24,
−1.94 to 8.59, −1.70 to 6.27 and −1.90 to 6.93, respectively. For the four periods, the areas
characterized with low–low agglomeration were always found in the Mongolia and the
Songnen Plain and the Inner Mongolian Autonomous Region of China. Unlike that of the
other two indexes, the spatial autocorrelation of GPP on the Sanjiang Plain changed very
slightly, with no obvious low–low clustering in this area. The most significant changes in
the clustering of GPP were found in Mohe County, the far most northern part of the Inner
Mongolia Autonomous Region of China and the nearby border region with Russia; these
areas exhibited a significant increase in high–high clustering.
4.3. The Relationships between Vegetation Dynamics and Changes in Climatic, Anthropogenic, and
Hydrological Factors
As previously mentioned in Section 3.1, in order to analyze the relationship between
vegetation status and climate changes, human activities and hydrological elements in
Remote Sens. 2021,13, 684 12 of 25
the ARB, area-average climatic factors, land use area ratios, and simulated hydrological
factors for 134 sub-basins—averaged annually over four phases—were used as explanatory
variables to represent the environmental status of the ARB at different geographic locations
and at different times. The statistical results of the explanatory variables in each sub-basin
at different time periods are shown in Table 3.
Table 3. Statistical results of period-average explanatory variables for four periods at the sub-basin scale.
Independent Variable Mean Standard
Deviation Minimum Maximum
Climate changes
Precipitation (Pcp, mm) 565.977 100.521 289.969 800.870
Mean temperature (Tavg, ◦C) 0.046 2.931 −6.253 6.051
Maximum temperature (Tmax, ◦C) 6.489 2.499 0.310 12.159
Minimum temperature (Tmin, ◦C) −6.263 3.394 −12.900 0.627
Anthropogenic
activities
Proportion of forest area (Forest) 0.460 0.324 0.000 0.997
Proportion of pasture area (Pasture) 0.181 0.234 0.000 0.988
Proportion of wetland area (Wetland) 0.049 0.081 0.000 0.542
Proportion of crop area (Crop) 0.253 0.231 0.000 0.887
Proportion of residential area (Residential) 0.014 0.035 0.000 0.500
Proportion of water area (Water) 0.032 0.101 0.000 1.000
Proportion of range area (Range) 0.011 0.065 0.000 1.000
Hydrological
processes
Surface runoff (SURQ, mm) 8.941 6.423 0.000 31.384
Groundwater flow (GWQ, mm) 2.532 3.332 0.000 32.949
Lateral flow (LATQ, mm) 0.087 0.154 0.000 1.288
Actual evapotranspiration (ET, mm) 33.769 7.152 14.279 76.573
Amount of water stored in the soil profile
(SW, mm) 59.386 27.181 0.000 136.027
Snowmelt (SM, mm) 5.052 3.034 0.000 20.291
However, analysis of the variance inflation factor (VIF) suggested a strong multi-
collinearity between these independent variables (i.e., VIF > 10). In order to obtain more
stable and accurate model results, a global analysis of relationships between vegetation in-
dexes and environmental factors in the ARB were conducted using the PLSR model [
93
,
94
];
an appropriate model suitable for cases where multicollinearity between the independent
or dependent variables exists. The independent variables were filtered using variable
importance in the projection (VIP) values. The PLSR model results are summarized in
Table 4, where strong (Adjusted R
2
>0.8) relationships between vegetation indexes and
environmental factors over the ARB can be observed.
Table 4. Summary of the partial least squares regression (PLSR) modeled results over the ARB.
Latent Factors (LF) Y Variance R2Adjusted R2
1 0.665 0.665 0.665
2 0.097 0.762 0.761
3 0.062 0.825 0.824
4 0.008 0.833 0.832
5 0.004 0.837 0.835
The VIP values of the independent variables are shown in Table 5. Independent
variables with VIP values greater than one are usually considered as relatively important in
a PLSR model. For further analysis, the independent variables with VIP values greater than
0.85 (Pcp, Tmax, Forest, Pasture, Surface runoff (SURQ), Amount of water stored in the soil
profile (SW), and Snowmelt (SM)) were used as inputs for the GTWR model. Regionally
averaged sub-basin LAI, FVC, and GPP values were used as dependent variable inputs.
Remote Sens. 2021,13, 684 13 of 25
Table 5.
Variable importance in the projection (VIP) values and beta coefficients of environmental factors in the PLSR
model analysis.
Independent
Variable
Beta Coefficient VIP
LAI FVC GPP Model LF 1 LF 2 LF 3 LF 4 LF 5
Pcp 0.0011 0.0002 0.4972 1.5346 1.6776 1.6435 1.5819 1.5751 1.5715
Tavg 0.0048 −0.0006 4.6085 0.7952 0.6384 0.7849 0.7568 0.7552 0.7535
Tmax 0.0078 −0.0009 6.4948 0.8564 0.7404 0.8523 0.8283 0.8274 0.8263
Tmin 0.0020 −0.0006 2.8969 0.7935 0.6432 0.7820 0.7528 0.7496 0.7481
Forest 0.6001 0.1132 237.3714 1.6577 1.7452 1.6676 1.7356 1.7357 1.7317
Pasture −0.6216 −0.1269 −
270.6097
1.4005 1.4170 1.5180 1.5098 1.5025 1.5008
Wetland −0.6399 −0.0686 −
272.2829
0.5943 0.5958 0.5673 0.5588 0.5665 0.5786
Crop −0.2707 −0.0376 −84.2908 0.6461 0.4925 0.5442 0.5552 0.5981 0.6009
Residential −0.9260 −0.2014 −
384.5357
0.4913 0.4026 0.3803 0.4589 0.4680 0.4679
Water −0.6076 −0.1382 −
221.1102
0.5024 0.3042 0.2963 0.5232 0.5220 0.5233
Range −0.6574 −0.1851 −
237.5839
0.4536 0.4024 0.3998 0.4715 0.4755 0.4807
SURQ −0.0052 −0.0001 −2.0949 1.2230 1.2454 1.1747 1.2252 1.2290 1.2262
GWQ 0.0020 0.0006 0.7943 0.8109 0.7918 0.7407 0.7975 0.7985 0.7967
LATQ 0.0825 0.0617 −32.9155 0.8351 0.8570 0.8210 0.8140 0.8138 0.8234
ET 0.0071 0.0008 4.4259 0.7231 0.4901 0.7509 0.7228 0.7193 0.7207
SW 0.0018 0.0003 1.1370 1.2293 1.3418 1.2865 1.2371 1.2401 1.2440
Snowmelt −0.0006 0.0018 −0.8766 1.2590 1.3398 1.2543 1.2591 1.2552 1.2529
The results of the GTWR models are summarized in Table 6. The results of the
ordinary least squares regression (OLS) model, temporally weighted regression (TWR)
model, and geographically weighted regression (GWR) model are also presented and
were comparatively analyzed to verify the effectiveness of the GTWR model. The results
indicated that the GTWR models of LAI, FVC, and GPP performed better than the other
three models.
Table 6.
The summarized results of the ordinary least squares regression (OLS) model, temporally weighted regression
(TWR) model, and geographically weighted regression (GWR) model for comparisons.
Variable LAI FVC GPP
OLS TWR GWR GWTR OLS TWR GWR GWTR OLS TWR GWR GWTR
AICc −
215.4
−
279.2
−
745.9
−
851.6
−
1955.4
−
2096.1
−
2328.2
−
2436.6
6442.5 6382.5 5665.5 5564.1
R20.838 0.866 0.955 0.970 0.871 0.908 0.952 0.970 0.781 0.818 0.962 0.975
Adjusted R20.836 0.864 0.955 0.970 0.870 0.907 0.952 0.970 0.778 0.816 0.961 0.974
NOTE: AICc: a version of Akaike information criterion (AIC) with a correction for small sample size.
4.3.1. Regional Impacts of Climate Changes on Vegetation
As shown in Table 5, the global results of the PLSR model of the drivers prone to vege-
tation changes revealed the important role of Pcp in vegetation growth. Pcp is one of the
factors among the independent variables that had a strong influence on LAI, FVC, and GPP
(VIP > 1.5); in addition, the effects of Pcp on LAI, FVC, and GPP were all generally positive.
Compared to Tmin and Tavg, Tmax and Pcp had a relatively strong effect on vegetation
growth status, and the global results suggested that Tmax had a positive effect on LAI and
GPP and a negative effect on FVC. Nevertheless, owning to significant spatial variability
of climate changes over the study area [
36
,
38
], the relationships between vegetation and
climate factors varied geographically, and thus, exhibited significant
spatial heterogeneity
.
The spatial distributions of the regression coefficients for Pcp with the LAI, FVC and
GPP GTWR models during the four studied periods are presented in the upper part of
Figure 6. Although the effects of Pcp on LAI, FVC, and GPP differed in time and space,
a general tendency for the coefficients is displayed on the figures: generally lower in the
northeast and relatively higher in the southwest. The areas with high coefficient values
were primarily concentrated at the borders of Mongolia, Russia, and China and near the
Remote Sens. 2021,13, 684 14 of 25
Argun River Basin and the Shilka River Basin. Areas with relatively low coefficient values
were found to be mostly distributed in the northeastern part of the basin, near the border
of the Sanjiang Plain in China and in Khabarovsk Krai and the Jewish Autonomous Oblast
in Russia. The area of negative Pcp regression coefficient values spread over time, or the
intensity of the negative effect of Pcp increased.
RemoteSens.2021,13,68415of27
Figure6.ThespatialdistributionoftheregressioncoefficientsforPcpandTmaxwiththeLAI((a–d)presentthecoeffi‐
cientsforPeriods1–4,respectively),FVC((e–h)showthecoefficientsforPeriods1–4,respectively),andGPP((i–l)exhibit
thecoefficientsforPeriods1–4,respectively)GTWRmodelsforfourstudiedperiods.
Figure 6.
The spatial distribution of the regression coefficients for Pcp and Tmax with the LAI ((
a
–
d
) present the coefficients
for Periods 1–4, respectively), FVC ((
e
–
h
) show the coefficients for Periods 1–4, respectively), and GPP ((
i
–
l
) exhibit the
coefficients for Periods 1–4, respectively) GTWR models for four studied periods.
Remote Sens. 2021,13, 684 15 of 25
The spatial distributions of the regression coefficients for Tmax with the LAI, FVC, and
GPP GTWR models for the four studied periods are exhibited in the lower part of Figure 6.
The effects of Tmax on LAI and GPP showed similar spatial and temporal patterns and,
unlike the effects of Pcp, were roughly concentrated in high-value areas in the north and
low-value areas in the south of the basin. The impact of Tmax on FVC, in contrast, had a
distinct area of low values in the central part of the study area.
4.3.2. Regional Impacts of Land Use Changes on the Vegetation
In addition to climatic factors, anthropogenic activities that result in land use changes,
such as farming, urbanization, and reforestation, also obviously affect the vegetation
dynamics in the ARB. The PLSR model results (see Table 5) revealed that at the basin scale,
the proportions of forest and pasture are the most important drivers for vegetation changes.
Overall, the forest area posed a positive effect on all three indexes, while an increase in
pasture area most likely resulted in the decreases in all three indexes.
The upper part of Figure 7presents the spatial distribution of the regression coefficients
for forest with the LAI, FVC and GPP GTWR models for the studied four periods. From
the perspective of the temporal and spatial distribution of the regression coefficients, the
spatial distributions of the impacts of forest on LAI, FVC and GPP were very similar, and
the spatial pattern varied slightly in the four studied periods. Regression coefficients for
the forest over all sub-basins were greater than zero, and low values occurred in Huma
County and Tahe County in China and Amur Oblast in Russia.
According to the spatial distributions of the regression coefficients for pasture with
the LAI, FVC and GPP GTWR models, as shown in the lower part of Figure 7, the negative
values of the regression coefficients were mainly concentrated near the Songnen Plain in
China. In addition, areas of high pasture influence on the three vegetation indexes have
decreased in size and intensity over the four studied periods. Pasture had the strongest
impact on vegetation in the northeastern ARB in Period 1, i.e., 1982–1990.
4.3.3. Spatial–Temporal Heterogeneity of the Relationships between Vegetation Dynamics
and Hydrological Variables
Hydrological and ecological processes are interactive: on the one hand, the water
cycle strongly influences the component structure, spatial distribution, and dynamics
of plant communities; on the other hand, vegetation influences hydrological processes
directly through root water uptake and stomatal transpiration; while indirectly through
affecting water infiltration, slope runoff and evaporation processes via the vertical canopy
structure and horizontal community distribution [
95
]. As shown in Table 5, the PLSR
models predicted that SURQ, SW and SM were more closely related to vegetation growth
than the other hydrological processes considered, and had similar levels of importance.
Only SW was positively correlated with all three indexes, LAI, FVC, and GPP at the
global scale.
The spatial distribution of the regression coefficients for SURQ with the LAI, FVC and
GPP GTWR models for the four studied periods are presented in Figure 8. Comparing the
distributions of regression coefficients for SURQ with different vegetation indexes, it is
clear that SURQ has the smallest range of positive regression coefficients in the FVC model.
However, both the range and value of the positive coefficients in the FVC model showed
an overall increasing trend during the study period.
Remote Sens. 2021,13, 684 16 of 25
RemoteSens.2021,13,68417of27
Figure7.ThespatialdistributionoftheregressioncoefficientsforforestandpasturewithLAI((a–d)displaythecoeffi‐
cientsforPeriods1–4,respectively),FVC((e–h)showthecoefficientsforPeriods1–4,respectively),andGPP((i–l)present
thecoefficientsforPeriods1–4,respectively)GTWRmodelsforthefourstudiedperiods.
Figure 7.
The spatial distribution of the regression coefficients for forest and pasture
with LAI ((
a
–
d
) display the coefficients for Periods 1–4, respectively), FVC ((
e
–
h
) show
the coefficients for Periods 1–4, respectively), and GPP ((
i
–
l
) present the coefficients for
Periods 1–4, respectively) GTWR models for the four studied periods.
Remote Sens. 2021,13, 684 17 of 25
RemoteSens.2021,13,68419of27
Figure8.ThespatialdistributionoftheregressioncoefficientsforSURQwiththeLAI((a–d)showsthecoefficientsfor
Periods1–4,respectively),FVC((e–h)presentsthecoefficientsforPeriods1–4,respectively),andGPP((i–l)exhibitsthe
coefficientsforPeriods1–4,respectively)GTWRmodelsforthefourstudiedperiods.
ThespatialdistributionoftheregressioncoefficientsforSWwiththeLAI,FVCand
GPPGTWRmodelsforthefourstudiedperiodsarepresentedinFigure9.Intheregion
neartheArgunRiverBasininthewesternpartofthestudyarea,therelationshipbetween
SWandLAI,FVCandGPPgenerallyshowedaroughlypositivetonegativetrendforthe
fourstudiedperiods.DuringPeriod1(1982–1990),theregionwiththehighestpositive
valuesoftherelationshipbetweenSWandLAI/GPPappearedinthemiddleofthewater‐
shed,neartheChina–Russiaborder,anddifferedfromthedistributionofregressionco‐
efficientsintheFVCmodel.Thevaluesoftheregressioncoefficientsinthisregionde‐
creasedwithtime.
Figure 8.
The spatial distribution of the regression coefficients for SURQ with the LAI ((
a
–
d
) shows the coefficients for
Periods 1–4, respectively), FVC ((
e
–
h
) presents the coefficients for Periods 1–4, respectively), and GPP ((
i
–
l
) exhibits the
coefficients for Periods 1–4, respectively) GTWR models for the four studied periods.
The spatial distribution of the regression coefficients for SW with the LAI, FVC and
GPP GTWR models for the four studied periods are presented in Figure 9. In the region near
the Argun River Basin in the western part of the study area, the relationship between SW
and LAI, FVC and GPP generally showed a roughly positive to negative trend for the four
studied periods. During Period 1 (1982–1990), the region with the highest positive values of
the relationship between SW and LAI/GPP appeared in the middle of the watershed, near
the China–Russia border, and differed from the distribution of regression coefficients in the
FVC model. The values of the regression coefficients in this region decreased with time.
Remote Sens. 2021,13, 684 18 of 25
RemoteSens.2021,13,68420of27
Figure9.ThespatialdistributionoftheregressioncoefficientsforSWwiththeLAI((a–d)showsthecoefficientsforPeriods
1–4,respectively),FVC((e–h)illustratesthecoefficientsforPeriods1–4,respectively),andGPP((i–l)presentsthecoeffi‐
cientsforPeriods1–4,respectively)GTWRmodelsforthefourstudiedperiods.
Figure10showsthespatialdistributionoftheregressioncoefficientsforSMwiththe
LAI,FVCandGPPGTWRmodelsforthefourstudiedperiods.IncomparisonwithFig‐
ures8–10,theresultsindicatethattherelationshipbetweenthehydrologicalfactorsand
LAI,FVC,orGPPdifferedsignificantlyintermsofthespatialpattern.Thespatialdistri‐
butionofnegativevaluesoftheregressioncoefficientsforSURQwiththeLAIandGPP
modelsweremoresimilarthanthatwiththeFVCmodel,andtheregressioncoefficients
forSWhadthesimilarcharacteristics.Incontrast,theregressioncoefficientsforSMwith
theLAI,FVCandGPPmodelsweremoreconsistentintermsofspatialdistribution,with
asignificantincreaseintherangeofpositivevaluesinthestudyareabetweenPeriod1
andPeriod2(beforeandafter1991)andadecreaseintheareaofnegativevaluesinthe
middleofthestudyarea.
Figure 9.
The spatial distribution of the regression coefficients for SW with the LAI ((
a
–
d
) shows the coefficients for Periods
1–4, respectively), FVC ((
e
–
h
) illustrates the coefficients for Periods 1–4, respectively), and GPP ((
i
–
l
) presents the coefficients
for Periods 1–4, respectively) GTWR models for the four studied periods.
Figure 10 shows the spatial distribution of the regression coefficients for SM with
the LAI, FVC and GPP GTWR models for the four studied periods. In comparison
with Figures 8–10
, the results indicate that the relationship between the hydrological fac-
tors and LAI, FVC, or GPP differed significantly in terms of the spatial pattern. The spatial
distribution of negative values of the regression coefficients for SURQ with the LAI and
GPP models were more similar than that with the FVC model, and the regression coeffi-
cients for SW had the similar characteristics. In contrast, the regression coefficients for SM
with the LAI, FVC and GPP models were more consistent in terms of spatial distribution,
with a significant increase in the range of positive values in the study area between Period
1 and Period 2 (before and after 1991) and a decrease in the area of negative values in the
middle of the study area.
Remote Sens. 2021,13, 684 19 of 25
RemoteSens.2021,13,68421of27
Figure10.ThespatialdistributionoftheregressioncoefficientsforSMwiththeLAI((a–d)showsthecoefficientsfor
Periods1–4,respectively),FVC((e–h)exhibitsthecoefficientsforPeriods1–4,respectively),andGPP((i–l)presentsthe
coefficientsforPeriods1–4,respectively)GTWRmodelsforthefourstudiedperiods.
5.Discussion
Inthisstudy,Mann–KendalltestsandSen’sslopewereemployedtoexaminethe
trendsandabruptchangesintheLAI,FVCandGPPfrom1982to2013intheARB.The
resultssuggestthattheLAI,FVC,andGPPexhibitedstrongspatialheterogeneitytrends
withdifferentspatialpatterns.TheareademonstratingsignificantlyincreasedFVCvalues
(p<0.05),accountingforabout62.5%ofthebasinarea,waslargerthanthatpresenting
significantlyincreasedLAIvalues(7.7%)andGPPvalues(16.3%).TheareasofLAI,FVC
andGPPthatexhibitedsignificantdecreasingtends(p<0.05)accountedfor18.9,2.0and
10.5%ofthebasinarea,respectively.Thedifferencesinthevegetationdynamicsindicated
strongspatialvariationsintheimpactsofenvironmentalfactorsonvegetationgrowth
conditionsontheonehand,anddifferentdegreesofvegetationgreennessandproductiv‐
ity—influencedbytheenvironment—ontheotherhand.However,somesimilaritiesdid
existinthespatialpatternsofdynamictrendsofLAI,FVCandGPPfrom1982to2013in
theARB.Forexample,allthreeindexesshowedanoverallincreasingtrendintheSongnen
PlainandadecreasingtrendintheMongolianpartoftheARB.However,amuchmore
significantincreasingtrendwasfoundintheLAIandFVCovertheSongnenPlain,and
muchmoresignificantdecreasingtrendsinLAIandGPPwerefoundovertheMongolian
Figure 10.
The spatial distribution of the regression coefficients for SM with the LAI ((
a
–
d
) shows the coefficients for Periods
1–4, respectively), FVC ((
e
–
h
) exhibits the coefficients for Periods 1–4, respectively), and GPP ((
i
–
l
) presents the coefficients
for Periods 1–4, respectively) GTWR models for the four studied periods.
5. Discussion
In this study, Mann–Kendall tests and Sen’s slope were employed to examine the
trends and abrupt changes in the LAI, FVC and GPP from 1982 to 2013 in the ARB. The
results suggest that the LAI, FVC, and GPP exhibited strong spatial heterogeneity trends
with different spatial patterns. The area demonstrating significantly increased FVC values
(p< 0.05), accounting for about 62.5% of the basin area, was larger than that presenting
significantly increased LAI values (7.7%) and GPP values (16.3%). The areas of LAI, FVC
and GPP that exhibited significant decreasing tends (p< 0.05) accounted for 18.9, 2.0
and 10.5% of the basin area, respectively. The differences in the vegetation dynamics
indicated strong spatial variations in the impacts of environmental factors on vegetation
growth conditions on the one hand, and different degrees of vegetation greenness and
productivity—influenced by the environment—on the other hand. However, some similar-
ities did exist in the spatial patterns of dynamic trends of LAI, FVC and GPP from 1982
to 2013 in the ARB. For example, all three indexes showed an overall increasing trend in
the Songnen Plain and a decreasing trend in the Mongolian part of the ARB. However, a
much more significant increasing trend was found in the LAI and FVC over the Songnen
Plain, and much more significant decreasing trends in LAI and GPP were found over the
Remote Sens. 2021,13, 684 20 of 25
Mongolian part of the ARB. In general, changes in the environmental factors of the Songnen
Plain led to an increasing trend in the greenness and productivity of the vegetation in
the region.
Despite the overall increasing trend of vegetation indexes in the Songnen Plain, local
Moran’s I results showed that low–low aggregation of three indexes occurred in most
areas of the region during the four studied phases. For LAI and FVC, the most significant
changes in local spatial autocorrelation were recognized over the Sanjiang Plain, and
the low–low agglomeration in the Sanjiang Plain decreased continuously. The different
characteristics and stage changes in the spatial autocorrelation in the Sanjiang Plain and
Songnen Plain, both of which are the food production bases of China, may be due to
different planting structures.
The PLSR results suggested that Pcp, Tmax, Forest, Pasture, SURQ, SW, and SM
were the major factors affecting LAI, FVC and GPP in the entire study region. The spatial
distributions of regression coefficients for the GTWR models showed that the relationships
between vegetation indexes and climate factors varied with geography, and thus, exhibited
significant spatial heterogeneity. In general, areas with low Pcp coefficients were found to
be mostly distributed in the northeastern part of the basin, near the border of the Sanjiang
Plain in China and in Khabarovsk Krai and the Jewish Autonomous Oblast in Russia. The
areas with relatively high coefficient values were primarily concentrated on the borders of
Mongolia, Russia, and China and near the Argun River Basin and the Shilka River Basin.
A temperate humid monsoon climate predominates the eastern part of the ARB, while
the Mongolian region and the Argun River Basin—as well as the Shilka River Basin in the
headwaters region of the western ARB—are less influenced by the Pacific coastal monsoon
and characterize a continental arid–semi-arid climate [
34
]. Therefore, the area near the
Argun River Basin and the Shilka River Basin showed a high degree of positive influence of
Pcp during all four phases, which was most likely attributable to the relatively dry climate
of this region.
In addition, comparisons among the spatial distributions of the regression coefficients
for Pcp with the LAI, FVC and GPP GTWR models indicated that the area and intensity of
the region of high Pcp coefficient values in northeastern China decreased from Period 1 to
2 (before and after 1991). Previous studies [
96
,
97
] have reported that large areas of pasture
and wetland were converted into agricultural land—especially paddy fields—in this region
during this period. With significant increases in rice cultivation in this region, the extensive
river runoff interception and groundwater extraction were conducted to meet the demands
of irrigation [
96
], which probably resulted in the weakened relationship between Pcp and
regional vegetation growth. In addition, the continuous conversion of pasture to paddy
field in the Songnen Plain may be one of the reasons for the negative impact of pastures on
LAI, FVC and GPP in the region. These results suggested that natural and anthropogenic
factors jointly took effect and interacted with each other, affecting the vegetation regime of
the ARB.
The variations in the spatial patterns of the regression coefficients for the hydrological
variables were generally more pronounced in time than those of the climatic factors, which
is most likely attributable to more dramatic spatial–temporal changes in the hydrological
processes. For example, the relationship between SW and GPP shifted from positive to
negative in a large area of the southwestern part of the basin during Period 2 and Period
3 (before and after 2000). In addition, a significant increase in the range of positive SM
coefficient values in the study area was found between Period 1 and Period 2 (before
and after 1991). To some extent, the changes represent the increasingly apparent positive
effect of SM runoff on vegetation growth in the study area during the period of 1982–
1999. Comparing the distributions of regression coefficients for SURQ with different
vegetation indexes, the smallest range of positive regression coefficients in the FVC model
indicated that the relationship between surface runoff and vegetation productivity or
vegetation vertical canopy distributions may be more positive than that with vegetation
horizontal cover in the ARB. The impact of a single hydrological variable on the vegetation
Remote Sens. 2021,13, 684 21 of 25
canopy structure and canopy function may vary in both direction (positive or negative)
and intensity.
6. Conclusions
This study analyzed the spatial–temporal variations and spatial autocorrelation of
vegetation growth conditions in the ARB at the pixel scale and investigated the spatial–
temporal heterogeneity of the relationships between vegetation indexes and climatic,
anthropogenic and hydrological factors at the sub-basin scale, using GTWR models from
1982 to 2013.
The trends of LAI, FVC and GPP in the ARB from 1982 to 2013 were quite different,
but only the variation of area-average annual FVC proved to be significant with an increase
rate of 0.0004/year (p< 0.05). All the annual vegetation indexes underwent a rapid change
around the year 2001. Additionally, all three indexes showed an overall increasing trend in
the Songnen Plain and a decreasing trend in the Mongolian part of the ARB during 1982–
2013. The local Moran’s I results of LAI, FVC, and GPP suggested that similar decreasing
trends in low–low agglomeration area of LAI and FVC on the Sanjiang Plain occurred
during the study period.
According to the results of the GTWR models, the relationships between vegetation
indexes and driving factors exhibited significant spatial–temporal heterogeneity. For
example, a general tendency of the coefficients for Pcp existed: generally lower in the
northeast and relatively higher in the southwest. The area near the Argun River Basin
and the Shilka River Basin showed a high degree of positive influence on Pcp during
all four phases, which was most likely attributed to the relatively dry climate in this
region. Additionally, the negative values of regression coefficients for pasture were mainly
concentrated near the Songnen Plain. The continuous conversion of pasture to paddy field
may be one of the reasons for the negative impact of pastures on the vegetation dynamics
in this region. SURQ had the smallest range of positive regression coefficients in the model
for FVC among the three vegetation indexes, indicating that the relationship between
surface runoff and vegetation productivity or vegetation vertical canopy distributions may
be more positive than that with vegetation horizontal cover. Additionally, a significant
increase in the range of positive regression coefficients for SM between Period 1 and Period
2 may suggest the increasingly apparent positive effect of SM runoff on vegetation growth
in the study area during the period of 1982–1999.
Natural and anthropogenic factors jointly took effect and interacted with each other
to affect the vegetated regime of the region. The decrease in the impact of precipitation on
vegetation growth in the Songnen Plain was determined as having commenced around
1991, which was most likely attributable to dramatic changes in water use styles induced by
local land use changes, and this corresponded to the negative correlation between pasture
areas and vegetation indexes during the same period.
Author Contributions:
Conceptualization, W.Z.; data curation, S.Z., S.W., B.Z. and Q.X.; formal
analysis, S.Z.; methodology, W.Z.; software, S.Z.; Writing—original draft, S.Z.; writing—review and
editing, W.Z. All authors have read and agreed to the published version of the manuscript.
Funding:
This research was funded by the National Key R&D Program of China, grant numbers
(2016YFA0602302) and (2018YFB0605603-04).
Institutional Review Board Statement: Not applicable.
Informed Consent Statement: Not applicable.
Data Availability Statement:
The data presented in this study are openly available in FigShare at
https://doi.org/10.6084/m9.figshare.13498542.v2, reference number [13498542].
Acknowledgments:
Open discussions in weekly seminars with the graduate students in Wanchang
Zhang’s group are acknowledged.
Conflicts of Interest: The authors declare no conflict of interest.
Remote Sens. 2021,13, 684 22 of 25
References
1.
Peng, W.F.; Kuang, T.T.; Tao, S. Quantifying influences of natural factors on vegetation NDVI changes based on geographical
detector in Sichuan, western China. J. Clean Prod. 2019,233, 353–367. [CrossRef]
2.
Liu, X.X.; Tian, Z.X.; Zhang, A.B.; Zhao, A.Z.; Liu, H.X. Impacts of Climate on Spatiotemporal Variations in Vegetation NDVI
from 1982–2015 in Inner Mongolia, China. Sustainability 2019,11, 768. [CrossRef]
3.
Zhang, Y.H.; Ye, A.Z. Spatial and temporal variations in vegetation coverage observed using AVHRR GIMMS and Terra MODIS
data in the mainland of China. Int. J. Remote Sens. 2020,41, 4238–4268. [CrossRef]
4.
Karasiak, N.; Dejoux, J.F.; Fauvel, M.; Willm, J.; Monteil, C.; Sheeren, D. Statistical Stability and Spatial Instability in Mapping
Forest Tree Species by Comparing 9 Years of Satellite Image Time Series. Remote Sens. 2019,11, 2512. [CrossRef]
5.
Akujärvi, A.; Lehtonen, A.; Liski, J. Ecosystem services of boreal forests—Carbon budget mapping at high resolution. J. Environ.
Manag. 2016,181, 498–514. [CrossRef] [PubMed]
6.
Li, B.; Chen, D.; Wu, S.; Zhou, S.; Wang, T.; Chen, H. Spatio-temporal assessment of urbanization impacts on ecosystem services:
Case study of Nanjing City, China. Ecol. Indic. 2016,71, 416–427. [CrossRef]
7.
Zhao, A.Z.; Zhang, A.B.; Lu, C.Y.; Wang, D.L.; Wang, H.F.; Liu, H.X. Spatiotemporal variation of vegetation coverage before and
after implementation of Grain for Green Program in Loess Plateau, China. Ecol. Eng. 2017,104, 13–22. [CrossRef]
8.
Du, Z.Q.; Zhao, J.; Pan, H.H.; Wu, Z.T.; Zhang, H. Responses of vegetation activity to the daytime and nighttime warming in
Northwest China. Environ. Monit. Assess. 2019,191, 10. [CrossRef] [PubMed]
9.
Gribovszki, Z.; Kalicz, P.; Balog, K.; Szabó, A.; Tóth, T.; Csáfordi, P.; Metwaly, M.; Szalai, S. Groundwater uptake of different
surface cover and its consequences in great Hungarian plain. Ecol. Process. 2017,6, 39. [CrossRef]
10. Yang, Y.J.; Erskine, P.D.; Lechner, A.M.; Mulligan, D.; Zhang, S.L.; Wang, Z.Y. Detecting the dynamics of vegetation disturbance
and recovery in surface mining area via Landsat imagery and LandTrendr algorithm. J. Clean Prod.
2018
,178, 353–362. [CrossRef]
11.
Wu, T.; Feng, F.; Lin, Q.; Bai, H.M. Advanced Method to Capture the Time-Lag Effects between Annual NDVI and Precipitation
Variation Using RNN in the Arid and Semi-Arid Grasslands. Water 2019,11, 1789. [CrossRef]
12.
Zhang, M.; Wang, J.M.; Li, S.J. Tempo-spatial changes and main anthropogenic influence factors of vegetation fractional coverage
in a large-scale opencast coal mine area from 1992 to 2015. J. Clean. Prod. 2019,232, 940–952. [CrossRef]
13.
Hua, L.; Wang, H.D.; Sui, H.G.; Wardlow, B.; Hayes, M.J.; Wang, J.X. Mapping the Spatial-Temporal Dynamics of Vegetation
Response Lag to Drought in a Semi-Arid Region. Remote Sens. 2019,11, 1873. [CrossRef]
14.
Chen, J.M.; Black, T.A. Measuring leaf-area index of plant canopies with branch architecture. Agric. For. Meteorol.
1991
,57, 1–12.
[CrossRef]
15. Chen, J.M.; Black, T.A. Defining leaf area index for non-flat leaves. Plant Cell Environ. 1992,15, 421–429. [CrossRef]
16.
Li, X.J.; Qu, Y. Evaluation of Vegetation Responses to Climatic Factors and Global Vegetation Trends using GLASS LAI from 1982
to 2010. Can. J. Remote Sens. 2018,44, 357–372. [CrossRef]
17.
Reygadas, Y.; Jensen, J.L.R.; Moisen, G.G.; Currit, N.; Chow, E.T. Assessing the relationship between vegetation greenness and
surface temperature through Granger causality and Impulse-Response coefficients: A case study in Mexico. Int. J. Remote Sens.
2020,41, 3761–3783. [CrossRef]
18.
Verger, A.; Camacho, F.; García-Haro, F.J.; Meliá, J. Prototyping of Land-SAF leaf area index algorithm with VEGETATION and
MODIS data over Europe. Remote Sens. Environ. 2009,113, 2285–2297. [CrossRef]
19.
Xiao, Z.; Liang, S.; Wang, J.; Xiang, Y.; Zhao, X.; Song, J. Long-Time-Series Global Land Surface Satellite Leaf Area Index Product
Derived From MODIS and AVHRR Surface Reflectance. IEEE Trans. Geosci. Remote Sens. 2016,54, 5301–5318. [CrossRef]
20.
Liu, D.Y.; Jia, K.; Wei, X.Q.; Xia, M.; Zhang, X.W.; Yao, Y.J.; Zhang, X.T.; Wang, B. Spatiotemporal Comparison and Validation of
Three Global-Scale Fractional Vegetation Cover Products. Remote Sens. 2019,11, 2524. [CrossRef]
21.
Mu, S.J.; Yang, H.F.; Li, J.L.; Chen, Y.Z.; Gang, C.C.; Zhou, W.; Ju, W.M. Spatio-temporal dynamics of vegetation coverage and its
relationship with climate factors in Inner Mongolia, China. J. Geogr. Sci. 2013,23, 231–246. [CrossRef]
22.
Yunhao, C.; Peijun, S.; Xiaobing, L.; Jin, C.; Jing, L. A combined approach for estimating vegetation cover in urban/suburban
environments from remotely sensed data. Comput. Geosci. 2006,32, 1299–1309. [CrossRef]
23.
Xu, H.J.; Zhao, C.Y.; Wang, X.P. Spatiotemporal differentiation of the terrestrial gross primary production response to climate
constraints in a dryland mountain ecosystem of northwestern China. Agric. For. Meteorol. 2019,276, 15. [CrossRef]
24.
Yao, Y.T.; Wang, X.H.; Li, Y.; Wang, T.; Shen, M.G.; Du, M.Y.; He, H.L.; Li, Y.N.; Luo, W.J.; Ma, M.G.; et al. Spatiotemporal pattern
of gross primary productivity and its covariation with climate in China over the last thirty years. Glob. Chang. Biol.
2018
,24,
184–196. [CrossRef]
25.
Schaefer, K.; Schwalm, C.R.; Williams, C.; Arain, M.A.; Barr, A.; Chen, J.M.; Davis, K.J.; Dimitrov, D.; Hilton, T.W.; Hollinger, D.Y.;
et al. A model-data comparison of gross primary productivity: Results from the North American Carbon Program site synthesis.
J. Geophys. Res. Biogeosci. 2012,117, 15. [CrossRef]
26.
Zhong, Q.Y.; Ma, J.; Zhao, B.; Wang, X.X.; Zong, J.M.; Xiao, X.M. Assessing spatial-temporal dynamics of urban expansion,
vegetation greenness and photosynthesis in megacity Shanghai, China during 2000–2016. Remote Sens. Environ.
2019
,233, 12.
[CrossRef]
27.
Zhang, X.; Liao, C.; Li, J.; Sun, Q. Fractional vegetation cover estimation in arid and semi-arid environments using HJ-1 satellite
hyperspectral data. Int. J. Appl. Earth. Obs. Geoinf. 2013,21, 506–512. [CrossRef]
Remote Sens. 2021,13, 684 23 of 25
28.
Tong, S.Q.; Zhang, J.Q.; Ha, S.; Lai, Q.; Ma, Q.Y. Dynamics of Fractional Vegetation Coverage and Its Relationship with Climate
and Human Activities in Inner Mongolia, China. Remote Sens. 2016,8, 776. [CrossRef]
29.
Zhao, A.Z.; Zhang, A.B.; Liu, J.H.; Feng, L.L.; Zhao, Y.L. Assessing the effects of drought and “Grain for Green” Program on
vegetation dynamics in China’s Loess Plateau from 2000 to 2014. Catena 2019,175, 446–455. [CrossRef]
30.
Zhou, Y.; Pei, F.S.; Xia, Y.; Wu, C.J.; Zhong, R.; Wang, K.; Wang, H.L.; Cao, Y. Assessing the Impacts of Extreme Climate Events on
Vegetation Activity in the North South Transect of Eastern China (NSTEC). Water 2019,11, 2291. [CrossRef]
31.
Meng, Y.Y.; Liu, X.N.; Wu, L.; Liu, M.L.; Zhang, B.Y.; Zhao, S. Spatio-temporal variation indicators for landscape structure
dynamics monitoring using dense normalized difference vegetation index time series. Ecol. Indic. 2019,107, 11. [CrossRef]
32.
Vicente-Serrano, S.M.; Azorin-Molina, C.; Pena-Gallardo, M.; Tomas-Burguera, M.; Dominguez-Castro, F.; Martin-Hernandez, N.;
Begueria, S.; El Kenawy, A.; Noguera, I.; Garcia, M. A high-resolution spatial assessment of the impacts of drought variability on
vegetation activity in Spain from 1981 to 2015. Nat. Hazards Earth Syst. 2019,19, 1189–1213. [CrossRef]
33.
Sebastian, D.E.; Ganguly, S.; Krishnaswamy, J.; Duffy, K.; Nemani, R.; Ghosh, S. Multi-Scale Association between Vegetation
Growth and Climate in India: A Wavelet Analysis Approach. Remote Sens. 2019,11, 2703. [CrossRef]
34. Simonov, E.A.; Dahmer, T.D. Amur-Heilong River Basin Reader; Ecosystems Hongkong: Hong Kong, China, 2008.
35.
Haruyama, S.; Shiraiwa, T. Environmental Change and the Social Response in the Amur River Basin; Springer: Berlin/Heidelberg,
Germany, 2014; Volume 5.
36.
Dai, C.; Wang, S.; Li, Z.; Zhang, Y.; Gao, Y.; Li, C. Review on hydrological geography in Heilongjiang River Basin. Acta Geogr. Sin.
2015,70, 1823–1834. [CrossRef]
37.
Chu, H.; Venevsky, S.; Wu, C.; Wang, M. NDVI-based vegetation dynamics and its response to climate changes at Amur-
Heilongjiang River Basin from 1982 to 2015. Sci. Total Environ. 2019,650, 2051–2062. [CrossRef]
38.
Yan, B.; Xia, Z.; Huang, F.; Guo, L.; Zhang, X. Climate Change Detection and Annual Extreme Temperature Analysis of the Amur
River Basin. Adv. Meteorol. 2016,2016, 1–14. [CrossRef]
39.
Novorotskii, P. Climate changes in the Amur River basin in the last 115 years. Russ. Meteorol. Hydrol.
2007
,32, 102–109. [CrossRef]
40.
Zou, Y.C.; Wang, L.Y.; Xue, Z.S.; E, M.J.; Jiang, M.; Lu, X.G.; Yang, S.D.; Shen, X.J.; Liu, Z.M.; Sun, G.Z.; et al. Impacts of
Agricultural and Reclamation Practices on Wetlands in the Amur River Basin, Northeastern China. Wetlands
2018
,38, 383–389.
[CrossRef]
41.
Cui, M. Status quo of Wetlands of Heilongjiang River Valley and the Protection (in Chinese). For. Inventory Plan.
2006
,1, 75–78.
[CrossRef]
42.
Mao, D.H.; He, X.Y.; Wang, Z.M.; Tian, Y.L.; Xiang, H.X.; Yu, H.; Man, W.D.; Jia, M.M.; Ren, C.Y.; Zheng, H.F. Diverse policies
leading to contrasting impacts on land cover and ecosystem services in Northeast China. J. Clean Prod.
2019
,240, 11. [CrossRef]
43.
Mao, D.H.; Wang, Z.M.; Luo, L.; Ren, C.Y. Integrating AVHRR and MODIS data to monitor NDVI changes and their relationships
with climatic parameters in Northeast China. Int. J. Appl. Earth Obs. Geoinf. 2012,18, 528–536. [CrossRef]
44.
Tu, M.Z.; Liu, Z.F.; He, C.Y.; Fang, Z.H.; Lu, W.L. The relationships between urban landscape patterns and fine particulate
pollution in China: A multiscale investigation using a geographically weighted regression model. J. Clean Prod.
2019
,237, 16.
[CrossRef]
45.
Huang, B.; Wu, B.; Barry, M. Geographically and temporally weighted regression for modeling spatio-temporal variation in
house prices. Int. J. Geogr. Inf. Sci. 2010,24, 383–401. [CrossRef]
46. Li, J.; Heap, A.D. A Review of Spatial Interpolation Methods for Environmental Scientists; Geoscience Australia: Canberra, Australia, 2008.
47.
Le Maitre, D.C.; Scott, D.F.; Colvin, C. A review of information on interactions between vegetation and groundwater. Water SA
1999,25, 137–152.
48.
Ludwig, J.A.; Wilcox, B.P.; Breshears, D.D.; Tongway, D.J.; Imeson, A.C. Vegetation patches and runoff-erosion as interacting
ecohydrological processes in semiarid landscapes. Ecology 2005,86, 288–297. [CrossRef]
49.
Yang, Z.F.; Li, C.H.; Huang, G.H.; Cai, Y.P. Analysis of relationships between NDVI and climatic/hydrological parameters in the
Yellow River basin. Int. J. Environ. Pollut. 2010,42, 166–183. [CrossRef]
50. Zha, F.; Liu, C.; Shi, R. Elevation Cluster Dataset Covering the Amur River Basin. Global Chang. Data Repos. 2016,1. [CrossRef]
51.
Yuefei, H.; Tiejian, L.; Enze, L.; Jiaye, L.; Rui, B.; Chuang, L.; Ruixiang, S. Boundary Data of the Amur River Basin. Glob. Chang.
Data Repos. 2016,1. [CrossRef]
52.
Holben, B.N. Characteristics of maximum-value composite images from temporal AVHRR data. Int. J. Remote Sens.
1986
,7,
1417–1434. [CrossRef]
53.
Knyazikhin, Y.; Glassy, J.; Privette, J.L.; Tian, Y.; Lotsch, A.; Zhang, Y.; Wang, Y.; Morisette, J.T.; Votava, P.; Myneni, R.B.; et al.
MODIS Leaf Area Index (LAI) and Fraction of Photosynthetically Active Radiation Absorbed by Vegetation (FPAR) Product (MOD
15) Algorithm Theoretical Basis Document-1999. Available online: https://modis.gsfc.nasa.gov/data/atbd/atbd_mod15.pdf
(accessed on 18 February 2020).
54.
Shabanov, N.V.; Huang, D.; Yang, W.Z.; Tan, B.; Knyazikhin, Y.; Myneni, R.B.; Ahl, D.E.; Gower, S.T.; Huete, A.R.; Aragao, L.; et al.
Analysis and optimization of the MODIS leaf area index algorithm retrievals over broadleaf forests. IEEE Trans. Geosci. Remote
Sens. 2005,43, 1855–1865. [CrossRef]
55.
Jiang, L.; Kogan, F.N.; Guo, W.; Tarpley, J.D.; Mitchell, K.E.; Ek, M.B.; Tian, Y.; Zheng, W.; Zou, C.-Z.; Ramsay, B.H. Real-time
weekly global green vegetation fraction derived from advanced very high resolution radiometer-based NOAA operational global
vegetation index (GVI) system. J. Geophys. Res. Atmos. 2010,115, D11114. [CrossRef]
Remote Sens. 2021,13, 684 24 of 25
56.
Hashimoto, H.; Wang, W.L.; Milesi, C.; White, M.A.; Ganguly, S.; Gamo, M.; Hirata, R.; Myneni, R.B.; Nemani, R.R. Exploring
Simple Algorithms for Estimating Gross Primary Production in Forested Areas from Satellite Data. Remote Sens.
2012
,4, 303–326.
[CrossRef]
57.
Zhang, S.Q.; Chen, H.; Fu, Y.; Niu, H.H.; Yang, Y.; Zhang, B.X. Fractional Vegetation Cover Estimation of Different Vegetation
Types in the Qaidam Basin. Sustainability 2019,11, 864. [CrossRef]
58.
Xiao, Z.; Liang, S.; Wang, J.; Chen, P.; Yin, X.; Zhang, L.; Song, J. Use of General Regression Neural Networks for Generating the
GLASS Leaf Area Index Product From Time-Series MODIS Surface Reflectance. IEEE Trans. Geosci. Remote Sens.
2014
,52, 209–223.
[CrossRef]
59.
Jia, K.; Liang, S.; Liu, S.; Li, Y.; Xiao, Z.; Yao, Y.; Jiang, B.; Zhao, X.; Wang, X.; Xu, S. Global land surface fractional vegetation cover
estimation using general regression neural networks from MODIS surface reflectance. IEEE Trans. Geosci. Remote Sens.
2015
,53,
4787–4796. [CrossRef]
60.
Yuan, W.P.; Liu, S.G.; Yu, G.R.; Bonnefond, J.M.; Chen, J.Q.; Davis, K.; Desai, A.R.; Goldstein, A.H.; Gianelle, D.; Rossi, F.; et al.
Global estimates of evapotranspiration and gross primary production based on MODIS and global meteorology data. Remote
Sens. Environ. 2010,114, 1416–1431. [CrossRef]
61.
Zhou, S.L.; Zhang, W.C.; Guo, Y.D. Impacts of Climate and Land-Use Changes on the Hydrological Processes in the Amur River
Basin. Water 2020,12, 76. [CrossRef]
62.
Pan, X.F.; Yan, B.X.; Muneoki, Y. Effects of land use and changes in cover on the transformation and transportation of iron: A case
study of the Sanjiang Plain, Northeast China. Sci. China Earth Sci. 2011,54, 686–693. [CrossRef]
63. Sen, P.K. Estimates of the Regression Coefficient Based on Kendall’s Tau. J. Am. Stat. Assoc. 1968,63, 1379–1389. [CrossRef]
64.
Tong, S.Q.; Zhang, J.Q.; Bao, Y.H.; Lai, Q.; Lian, X.; Li, N.; Bao, Y.B. Analyzing vegetation dynamic trend on the Mongolian
Plateau based on the Hurst exponent and influencing factors from 1982–2013. J. Geogr. Sci. 2018,28, 595–610. [CrossRef]
65.
Mekonnen, D.F.; Duan, Z.; Rientjes, T.; Disse, M. Analysis of combined and isolated effects of land-use and land-cover changes
and climate change on the upper Blue Nile River basin’s streamflow. Hydrol. Earth Syst. Sci. 2018,22, 6187–6207. [CrossRef]
66. Mann, H.B. Nonparametric Tests Against Trend. Econometrica 1945,13, 245–259. [CrossRef]
67. Kendall, M.G. Rank Correlation Methods; Charles Griffin & Co. Ltd.: London, UK, 1975; p. 160.
68.
Sneyers, R. Sur l’Analyse Statistique des Séries d’Observations; Secrétariat de l’Organisation Météorologique Mondiale: Geneva,
Switzerland, 1975.
69.
Bei, X.Y.; Yao, Y.J.; Zhang, L.L.; Xu, T.R.; Jia, K.; Zhang, X.T.; Shang, K.; Xu, J.; Chen, X.W. Long-Term Spatiotemporal Dynamics
of Terrestrial Biophysical Variables in the Three-River Headwaters Region of China from Satellite and Meteorological Datasets.
Remote Sens. 2019,11, 25. [CrossRef]
70.
Yang, Y.; Tian, F. Abrupt change of runoff and its major driving factors in Haihe River Catchment, China. J. Hydrol.
2009
,374,
373–383. [CrossRef]
71.
Moraes, J.M.; Pellegrino, G.Q.; Ballester, M.V.; Martinelli, L.A.; Victoria, R.L.; Krusche, A.V. Trends in Hydrological Parameters of
a Southern Brazilian Watershed and its Relation to Human Induced Changes. Water Resour. Manag.
1998
,12, 295–311. [CrossRef]
72.
Gerstengarbe, F.W.; Werner, P.C. Estimation of the beginning and end of recurrent events within a climate regime. Clim. Res.
1999
,
11, 97–107. [CrossRef]
73.
Miao, C.Y.; Ni, J.R.; Borthwick, A.G.L. Recent changes of water discharge and sediment load in the Yellow River basin, China.
Prog. Phys. Geogr. 2010,34, 541–561. [CrossRef]
74.
Li, J.; He, J.W.; Liu, Y.; Wang, D.J.; Rafay, L.; Chen, C.; Hong, T.; Fan, H.L.; Lin, Y.M. Spatial Autocorrelation Analysis of Multi-Scale
Damaged Vegetation in the Wenchuan Earthquake-Affected Area, Southwest China. Forests 2019,10, 21. [CrossRef]
75.
Walker, I.J.; Eamer, J.B.R.; Darke, I.B. Assessing significant geomorphic changes and effectiveness of dynamic restoration in a
coastal dune ecosystem. Geomorphology 2013,199, 192–204. [CrossRef]
76. Nelson, T.A.; Boots, B. Detecting spatial hot spots in landscape ecology. Ecography 2008,31, 556–566. [CrossRef]
77. Moran, P.A. Notes on continuous stochastic phenomena. Biometrika 1950,37, 17–23. [CrossRef] [PubMed]
78.
Fu, Z.; Li, R. The contributions of socioeconomic indicators to global PM2.5 based on the hybrid method of spatial econometric
model and geographical and temporal weighted regression. Sci. Total Environ. 2020,703, 135481. [CrossRef] [PubMed]
79.
Gong, Z.W.; Gu, L.; Yao, S.B.; Deng, Y.J. Effects of bio-physical, economic and ecological policy on forest transition for sustainability
of resource and socioeconomics development. J. Clean Prod. 2020,243, 10. [CrossRef]
80.
Zhang, D.J.; Ge, W.Y.; Zhang, Y. Evaluating the vegetation restoration sustainability of ecological projects: A case study of Wuqi
County in China. J. Clean Prod. 2020,264, 17. [CrossRef]
81.
Zhang, F.; Sun, X.X.; Zhou, Y.; Zhao, C.J.; Du, Z.H.; Liu, R.Y. Ecosystem health assessment in coastal waters by considering
spatio-temporal variations with intense anthropogenic disturbance. Environ. Model. Softw. 2017,96, 128–139. [CrossRef]
82.
Moons, E.; Brijs, T.; Wets, G. Hot Spot Analysis: Improving a Local Indicator of Spatial Association for Application in Traffic Safety;
Springer: Berlin/Heidelberg, Germany; pp. 221–231.
83. Anselin, L. Local Indicators of Spatial Association—LISA. Geogr. Anal. 1995,27, 93–115. [CrossRef]
84. Anselin, L. The Moran scatterplot as an ESDA tool to assess local instability in spatial. Spat. Anal. 1996,4, 111.
85.
Bartesaghi-Koc, C.; Osmond, P.; Peters, A. Spatio-temporal patterns in green infrastructure as driver of land surface temperature
variability: The case of Sydney. Int. J. Appl. Earth. Obs. Geoinf. 2019,83, 17. [CrossRef]
Remote Sens. 2021,13, 684 25 of 25
86.
Gao, J.; Wang, L.C. Embedding spatiotemporal changes in carbon storage into urban agglomeration ecosystem management—A
case study of the Yangtze River Delta, China. J. Clean Prod. 2019,237, 12. [CrossRef]
87.
Zhang, C.S.; Luo, L.; Xu, W.L.; Ledwith, V. Use of local Moran’s I and GIS to identify pollution hotspots of Pb in urban soils of
Galway, Ireland. Sci. Total Environ. 2008,398, 212–221. [CrossRef]
88.
Yuan, Y.M.; Cave, M.; Zhang, C.S. Using Local Moran’s I to identify contamination hotspots of rare earth elements in urban soils
of London. Appl. Geochem. 2018,88, 167–178. [CrossRef]
89.
Chaikaew, N.; Tripathi, N.K.; Souris, M. Exploring spatial patterns and hotspots of diarrhea in Chiang Mai, Thailand. Int. J. Health
Geogr. 2009,8, 36. [CrossRef]
90.
Zhang, Y.; Wang, W.; Liang, L.W.; Wang, D.P.; Cui, X.H.; Wei, W.D. Spatial-temporal pattern evolution and driving factors of
China’s energy efficiency under low-carbon economy. Sci. Total Environ. 2020,739, 12. [CrossRef]
91.
Li, E.K.; Lu, Y.Q.; Chen, Y. Geographic pattern evolution of China’s merchandise export and its influencing factors: Based on the
analysis of merchandise export distance and the GTWR model. Geogr. Res. 2019,38, 2624–2638.
92. Shi, T.; Yang, S.Y.; Zhang, W.; Zhou, Q. Coupling coordination degree measurement and spatiotemporal heterogeneity between
economic development and ecological environment—Empirical evidence from tropical and subtropical regions of China. J. Clean
Prod. 2020,244, 11. [CrossRef]
93. Cramer, R.D. Partial Least Squares (PLS): Its strengths and limitations. Perspect. Drug Discov. Des. 1993,1, 269–278. [CrossRef]
94.
Wold, S.; Martens, H.; Wold, H. The multivariate calibration problem in chemistry solved by the PLS method. In Matrix Pencils;
Springer: Berlin/Heidelberg, Germany, 1983; pp. 286–293.
95.
Lu, X. Responses of Eco-Hydrological Process to Changing Environment in Yanwachuan Basin Based on Green Water Theory.
Ph.D. Thesis, Xi’an University of Technology, Shaanxi, China, 2019.
96.
Luan, Z.-Q.; Deng, W. Study on Hydrological Effects of Human Activities in Sanjiang Plain (in Chinese). Bull. Soil Water Conserv.
2003,23, 11–14. [CrossRef]
97.
Chen, H.; Zhang, W.C.; Gao, H.R.; Nie, N. Climate Change and Anthropogenic Impacts on Wetland and Agriculture in the
Songnen and Sanjiang Plain, Northeast China. Remote Sens. 2018,10, 356. [CrossRef]
Available via license: CC BY 4.0
Content may be subject to copyright.
Content uploaded by Wanchang Zhang
Author content
All content in this area was uploaded by Wanchang Zhang on Feb 14, 2021
Content may be subject to copyright.