Journal of Hydrology 368 (2009) 96–104
Contents lists available at ScienceDirect
Journal of Hydrology
journal homepage: www.elsevier.com/locate/jhydrol
Multiscale variability of sediment load and streamflow of the lower Yangtze
River basin: Possible causes and implications
Qiang Zhang a,b,*, Chong-Yu Xu c, V.P. Singh d, Tao Yang e
a
State Key Laboratory of Lake Science and Environment, Nanjing Institute of Geography and Limnology, Chinese Academy of Sciences, Nanjing 210008, China
Institute of Space and Earth Information Science, The Chinese University of Hong Kong, Shatin, N.T., Hong Kong, China
c
Department of Geosciences, University of Oslo, P.O. Box 1047 Blindern, N-0316 Oslo, Norway
d
Department of Biological and Agricultural Engineering, Texas A&M University, College Station, TX 77843-2117, USA
e
State Key Laboratory of Hydrology-Water Resources and Hydraulics Engineering, Hohai University, Nanjing 210098, China
b
a r t i c l e
i n f o
Article history:
Received 9 September 2008
Received in revised form 20 January 2009
Accepted 22 January 2009
This manuscript was handled by G. Syme,
Editor-in-Chief
Keywords:
Sediment load
Streamflow variations
Scanning t-test
Scanning F-test
Yangtze River basin
s u m m a r y
Long monthly streamflow and sediment load series observed at the Datong station located in the lower
Yangtze River basin were analyzed using the scanning t-test, F-test and coherency analysis techniques.
The results indicated that: (1) different changing properties of the first and the second moments of the
hydrological series on different time scales were observed, reflecting different driving factors influencing
the hydrological processes of the lower Yangtze River basin; (2) a generally decreasing trend can be identified after the mid-1980s. Significant abrupt changes in sediment load were analyzed in the sediment
load series. However, more complicated changing patterns can be observed in the changes in streamflow.
Generally decreasing sediment load and increasing streamflow gave rise to anti-phase relations between
sediment load and the streamflow on longer time scales. In-phase relations between sediment load and
streamflow on shorter time scales may imply a considerable influence of the hydrological dynamics on
sediment transport; and (3) human activities, particularly the construction of water storage reservoirs,
exerted a massive influence on sediment load variations. Construction of a large amount of water reservoirs on the tributaries of the Yangtze River and the Gezhouba Dam on the mainstem of the Yangtze River
seem to be the main factors responsible for abrupt changes in the sediment load. Construction of the
Three Gorges Dam causes a sharp decrease and unstable variability in sediment load variations, which
may pose new challenges for the ecological environment conservation and the deltaic management of
the Yangtze Delta region.
Ó 2009 Elsevier B.V. All rights reserved.
Introduction
The reduction in the river-supplied sediment to coastal areas
and its consequent impact on the coastal environment has become
a global topic in recent years (e.g., Syvitski et al., 2005). This is because the population of the world is increasingly moving toward
coastal areas—about 60% (3.6 billion) of the world’s population
lives within 60 km (37 miles) of sea coasts (UNESCO, 1998). Human
activities and construction of dams and water reservoirs in particular should bear the major responsibility for the reduction in terrestrial sediment to coastal areas. Walling and Fang (2003) pointed out
that reservoir construction is probably the most important influence on land–ocean sediment fluxes. Therefore, a goal of the International Geosphere Biosphere Program (IGBP) and its core project,
* Corresponding author. Address: State Key Laboratory of Lake Science and
Environment, Nanjing Institute of Geography and Limnology, Chinese Academy of
Sciences, Nanjing 210008, China. Tel.: +852 26096639.
E-mail address: zhangqnj@gmail.com (Q. Zhang).
0022-1694/$ - see front matter Ó 2009 Elsevier B.V. All rights reserved.
doi:10.1016/j.jhydrol.2009.01.030
Land–Ocean Interaction in the Coastal Zone (LOICZ), has been to
survey the terrestrial sediment supply to the coast and to analyze
human perturbation of this flux (Syvitski, 2003).
A number of studies have investigated changes in sediment load
and streamflow, associated underlying causes and subsequent impacts on the coastal environment. Zhang et al. (2008b) analyzed
annual water discharge and sediment load series (1950s–2004)
at nine stations in the main channels and main tributaries of the
Zhujiang (Pearl River), demonstrating a significantly decreasing
sediment load at some stations in the main tributaries and more
stations have witnessed significantly decreasing sediment loads
since the 1990s. They also indicated that the decreasing sediment
load of the Pearl River basin was the result of hydrological regulations of water reservoirs. Similarly, tremendous changes in water
discharge and sediment load have also occurred in other rivers in
China, such as the Yellow River basin. The changes in streamflow
and sediment load of the Yellow River basin can also be beneficial.
The mean annual sediment load of the Yellow River in China has
declined markedly due to reduced precipitation, increased water
Q. Zhang et al. / Journal of Hydrology 368 (2009) 96–104
withdrawal and the implementation of comprehensive soil and
water conservation and sediment control programs within the
loess region of the Middle Yellow River (Mou, 1991; Milliman,
1997; Walling and Fang, 2003; Wang et al., 2003; Walling, 2006).
The Yangtze River (Changjiang, Fig. 1), being the longest river in
China and the third longest river in the world, plays a vital role in
the economic development and ecological environmental conservation of China. Numerous water reservoirs have been built in the
Yangtze River basin. However, most of the reservoirs are located
in the tributaries. Based on Xu (2005), up to the end of 1980s, there
are 11,931 water reservoirs appeared in the upper Yangtze River
basin with a total storage capacity of 2.05 1010 m3. The construction of the Three Gorges Dam started in 1993 and will be ended in
2009 with the total storage capacity of 3.93 1010 m3. The Gezhouba Dam started its construction in 1970 and the operation started in
earlier 1980s with the total storage capacity of 1.58 109 m3. The
appearance of water reservoirs greatly altered the hydrological processes of the Yangtze River basin (Zhang et al., 2006b, 2008a). The
sediment load and water discharge changes directly affect the ecological environment of the river basin and also influence the propagation and recession as well as the wetland ecological
environment of the Yangtze Delta. A number of studies have investigated the changes in water discharge and sediment load. Zhang
et al. (2006a) analyzed annual maximum streamflow and water level, indicating increasing streamflow in the middle Yangtze River
and consistently increasing water level from the upper to the lower
reaches of the river. Based on annual streamflow and sediment load
data of hydrological stations along the mainstem and main tributaries of the Yangtze River, Zhang et al. (2006b) indicated that water
reservoirs exerted more influence on sediment transport than
water discharge and this influence was more significant in the tributaries than in the mainstem of the Yangtze River basin. Chen et al.
(2001) analyzed hydrological records (covering a 100-year period)
from the upper, middle, and lower portions of the Yangtze River.
97
They found increasing streamflow from the upper to the lower
reaches but identified decreasing sediment load. They also pointed
out quite stable sediment load changes at the Datong station. Yang
et al. (2006) analyzed the 1951–2004 time series of annual sediment supply and coastal bathymetric data, illustrating a significant
decreasing trend in riverine sediment supply since the late 1960s
and attributed this decreasing sediment to dam construction. They
also indicated that the subsequent result of the decreasing sediment load may be responsible for the recession of the deltaic coast
which poses a great challenge to coastal management. It should be
noted here that the hydrological data the previous study used were
annual sediment load and streamflow data. Therefore, higher resolution of the hydrological data should be analyzed to achieve deeper insight into variations of sediment load and streamflow of the
Yangtze River basin, and which is sure to be practically and scientifically significant in sound understanding of influences human activities exert on hydrological processes and in ecological
environmental conservation of the Yangtze River delta. Moreover,
more than one factor results in alterations of the hydrological processes and these driving factors have the potential to influence
changing properties of sediment load and streamflow on different
time scales. Thus, it is necessary to understand abrupt changes
and associated statistical properties of the hydrological series on diverse time scales with robust techniques. This is the major motivation of this study.
Better understanding of the dynamic processes of sediment
transport and water discharge needs more detailed datasets, such
as monthly data (Walling and Fang, 2003). Therefore, in the current
study, we analyzed long monthly sediment load and streamflow
data from the Datong station since this station is the last control
hydrological station reflecting the flux of sediment load and water
discharge from the Yangtze River basin to the East China Sea. Statistical techniques were used with the aim to obtain a better
understanding of the variations in streamflow and sediment load
Fig. 1. Location of the Yangtze River basin and the Datong hydrological station. TG Dam: Three Gorges Dam; GZB Dam: Gezhouba Dam. Shaded gray area denotes the Yangtze
River basin.
98
Q. Zhang et al. / Journal of Hydrology 368 (2009) 96–104
Fig. 2. Sediment load and runoff series of the Datong station, the Yangtze River
basin.
on different time scales. Thus, the objectives of this paper were: (1)
to analyze general trends within long sediment load and streamflow series at the Datong station; (2) to investigate streamflow
and sediment load variations on different time scales; and (3) to
discuss possible causes, particularly the influence of Three Gorges
Dam on the dynamic processes of sediment load and streamflow of
the lower Yangtze River, and implications for the wetland ecological environment of the deltaic coast.
Data
cesses on a longer time scale and human activities such as construction of water reservoirs can promptly cause decreasing
sediment transport and the abrupt changes occurred. Abrupt
changes occurred to the hydrological series are usually the results
of sharp alterations of the way the influencing factors work on the
hydrological processes or new driving factors available. Thus, it is
necessary to understand abrupt changes of the hydrological series
on diverse time scales. In this study, abrupt changes were analyzed
using the scanning t-test technique on different time scales. Stable
or unstable status of hydrological variations is another important
statistical property of the hydrological series since that unstable
sediment load and streamflow variations have the potential to seriously affect scouring and filling of river channels and also for the
growth and recession of the river delta. Stable or unstable variations of the hydrological series were evaluated by the change in
the standard deviation, which was analyzed by the F-test technique. Moreover, phase and anti-phase relations between streamflow and sediment load variations were determined by the
coherency analysis method. The scanning t-test, the scanning Ftest and the coherency analysis methods are now introduced.
Jiang et al. (2002) grafted the wavelet technique (Kumar and Foufoula-Georgiou, 1994) onto the Student t-test and the F-test (Cramer, 1946) with the aim to develop algorithms for scanning t-test
and F-test, respectively. The scanning t-test attempts to detect significant changes in the first moment (subseries mean or average)
on different time scales within a long time series; while the scanning
F-test attempts to analyze significant changes in subseries variance
(the second moment) on various time scales (Jiang et al., 2007).
In the scanning t-test, statistic t(n, j) is defined as the difference
of the subsample averages between every two adjoining subseries
with equal subseries size (n) expressed as:
tðn; jÞ ¼ ðxj2 xj1 Þ n1=2 ðs2j2 þ s2j1 Þ1=2
ð1Þ
3
In this study, we used the monthly streamflow data (m /s) and
monthly sediment load data (kg/s) observed at the Datong station
(Fig. 1). The time series of streamflow and sediment load data cover the period from January 1951 to December 2006 and January
1963 to December 2006, respectively (Fig. 2). All hydrological data
were provided by the Changjiang (Yangtze) Water Resource Commission (CWRC). The quality of the hydrological data was firmly
controlled before its release. The sediment load data used in this
paper refer to suspended sediment. Sediment concentration data
were collected on a monthly basis using equal width increment
sampling and sediment yields were determined using the relationship between runoff and sediment concentration (Zhang et al.,
2006b). There are no missing data in the streamflow series, and
missing data can be found within the sediment load series in
1965, 1966, 1968, 1976, and 1978. The existence of missing data
results in a decrease in the sample size available for the analysis.
To make full use of the data we have, and also to keep the statistical properties of the entire series, we replaced the missing data
with the multi-annual mean sediment load of the respective individual months. We believe that this processing procedure can largely keep the basic statistical properties of the series and has little
influences on the general trends and also on our analysis results.
Methodology
In this study, we used the difference of the hydrological series
from its mean value to show the relative changes of streamflow
or sediment load from its mean in different periods. A negative difference indicates a decrease and vice versa. We note that there is
more than one factor influencing the variations of streamflow
and sediment load series and the way the driving factors alter
the hydrological processes is different in terms of time scales.
Some factors like climate changes can alter the hydrological pro-
where
xj1 ¼
j1
1 X
xðiÞ;
n i¼jn
s2j2 ¼
jþn1
1 X
ðxðiÞ xj2 Þ2 ;
n 1 i¼j
xj2 ¼
jþn1
1 X
xðiÞ;
n i¼j
s2j1 ¼
j1
1 X
ðxðiÞ xj1 Þ2 ;
n 1 i¼jn
in which subsample size n may vary as n = 2, 3, . . . , <N/2, or may be
selected at suitable intervals. The j = n + 1, n + 2, . . . , Nn + 1 is the
reference time point.
It should be noted that hydrological series are usually auto-correlated. Thus, the Table-Look-Up Test (Von Storch and Zwiers,
1999) was adopted to modify the significance criterion of statistic
t(n, j) lag-1 autocorrelation coefficients of the pooled subsample
and the subsample size n. Criterion t0.05 for the correction of the
dependence was employed to determine significant changes on
time scales longer than 30 years. For shorter subsample sizes, the
critical values are overly restrictive. Since the significance level
varies with n and j, to make values comparable the test statistic
was normalized as:
tr ðn; jÞ ¼ tðn; jÞ=t 0:05
ð2Þ
When |tr(n, j)| > 1.0, the abrupt change is significant at the 95%
confidence level. tr(n, j) < 1.0 denotes a significant decrease and
tr(n, j) > 1.0 a significant increase.
The scanning F-test technique defines significant changes in
subseries variances. Statistic Fr(n, j) is defined as:
8
2
2
>
< ðSj1 =Sj2 Þ=F a ; for Sj2 < Sj1 ;
for Sj2 ¼ Sj1 or Sj1 ¼ 0;
F r ðn; jÞ ¼ 0;
>
: 2 2
ðSj2 =Sj1 Þ=F a ;
for Sj2 > Sj1 ;
Sj2 ¼ 0;
ð3Þ
99
Q. Zhang et al. / Journal of Hydrology 368 (2009) 96–104
where the subsample standard deviations Sj1 and Sj2 are calculated
in the same way as in Eq. (1). Fa is a threshold value based on the
effective degree of freedom after the correction of dependence
and in a normalized distribution for the time series. The effective
degree of freedom for the correction of the dependence was estimated as (Hammersley, 1946):
When statistic trc ðn; jÞ > 1:0 with both jtru ðn; jÞj; jtrv ðn; jÞj > 1:0,
the two series have abrupt changes in the same direction; while
if t rc ðn; jÞ < 1:0, the two series have abrupt changes in opposite
directions (Jiang et al., 2002). The coherency of abrupt changes between monthly streamflow series and monthly sediment load series can be taken as an indication of the interaction between these
two series on decadal and basin scales.
Two time intervals with obvious positive differences of sediment
load can be identified as: 1963–1970 and 1975–1985. After the late
1980s, sediment load turned to decrease. After about 2000, this decrease tendency became sharp, represented by larger changing
grades, i.e., more dense contours of sediment load difference. Comparatively, the sediment load of January, February, November, and
December experienced mild variations, and this was particularly
true for sediment changes in January and February. Fig. 3 shows
that the sediment load of January and February fluctuates around
the mean value. The sediment load of November and December
showed a positive difference before the mid-1970s and turned to
a negative difference after the mid-1970s. The increase and decrease can be observed before 1970 and after 2000, respectively.
Comparatively, streamflow changes are characterized by complex changing properties (Fig. 4). In terms of streamflow changes
in January, February, and March, negative differences can be detected before the mid-1980s. After 1985, streamflow turned to a
positive difference. Streamflow in April, May, and June had a negative difference before the late 1980s. Positive differences can be
detected during late the 1980s and 2002. After about 2002, streamflow changes turned to the negative difference. With respect to
streamflow changes in July–December, three time intervals with
positive streamflow differences can be identified: before 1970,
1980–1985, and 1990s–2002. Correspondingly, three time intervals with negative streamflow differences can be found: 1970–
1980, 1985–1990s, and 2002–2006. Generally, decreasing streamflow was observed in March–December after 2002.
Results
Scanning t- and F-test of streamflow variations
Difference analysis of sediment load and streamflow changes
We present the scanning t-test analysis for the long streamflow
series to define when the change points occurred in the lower Yangtze River basin (Fig. 5). The streamflow series from the Datong
station displayed an abrupt decrease in 1956 (on a time scale of
45 months), 1977 (on a time scale of 54 months), 1984 (on a time
scale of 45 months), and 2001 (on a time scale of 64 months).
Among these detected abrupt changes, only abrupt changes in
1956 and 1984 were significant at the 95% confidence level. Similarly, Fig. 5 also indicates an abrupt increase in 1963 (on a time
scale of 91 months), 1972 (on a time scale of 45 months), 1980
(on a time scale of 32 months), and 1988 (on a time scale of 38
months). The significance test indicated that only the abrupt
changes in 1980 and 1988 were significant at the 95% confidence
level. Generally, different abrupt changes can be observed in the
streamflow series from the Datong station which varies with different time scales. On a time scale longer than 180 months, only
Ef ðnÞ ¼ f ðnÞ "
k
X
#1
r 2 ðsÞ
;
rðkÞ ! 0;
ð4Þ
s¼0
where f(n) is the degree of freedom listed in the F table.
A local minimum in Fr(n, j) < 1.0 denotes a significant change
towards a smaller variance, i.e., the record becomes much steadier;
whereas a local maximum in Fr(n, j) > 1.0 indicates a significant
change towards a larger variance, i.e., the record becomes much
unsteadier (Jiang et al., 2007).
Finally, the coherency of abrupt changes between two series u
and v was defined as
t rc ðn; jÞ ¼ sign½tru ðn; jÞt rv ðn; jÞfjtru ðn; jÞt rv ðn; jÞjg1=2
ð5Þ
12
11
10
9
8
7
6
5
4
3
2
1
11
9
7
5
3
1
-1
-3
-5
-7
-9
-11
-13
-15
1965
1970
1975
1980
1985
1990
1995
2000
Sediment load (kg/s)
Months
To evaluate the changes in sediment load in comparison with
the mean for individual months, we computed the difference of
sediment load for each individual month. For the sake of easier
understanding of sediment load and streamflow changes of each
month and also for the concise presentation of our results, we
demonstrated the results in Fig. 3. This figure was made with Surfer software package. We compared all the interpolation methods
available in Surfer software package and found that local polynomial method is proper because it maximally illustrate the changes
of hydrological series as they are. Fig. 3 illustrates different changing properties of sediment load with respect to different months.
Similar changing properties can be identified for sediment load
changes in March–October: a positive difference was detected before the late 1980s and a negative difference after the late 1980s.
2005
Years
Fig. 3. Secular variations of monthly sediment load difference at the Datong station. Solid lines indicate positive difference and dashed lines denote minus difference of the
sediment load.
100
Q. Zhang et al. / Journal of Hydrology 368 (2009) 96–104
12
8
6
4
2
1965
1970
1975
1980
1985
Years
1990
1995
2000
2005
Streamflow (m3/s)
Months
10
5000
4000
3000
2000
1000
0
-1000
-2000
-3000
-4000
-5000
-6000
-7000
-8000
-9000
Fig. 4. Secular variations of the monthly streamflow difference at the Datong station. Solid lines indicate positive difference and dashed lines denote minus difference of the
streamflow series.
512
0.8
Time scale (months)
362
0.6
256
0.4
181
0.2
128
0
-0.2
91
-0.4
64
-0.6
45
-0.8
32
1955 1960 1965 1970 1975 1980 1985 1990 1995 2000 2005
Years
-1
Fig. 5. Contours of the normalized scanning t-test standardized by the ‘‘Table-Look-up” critical value t0.05 for the standardized streamflow series at the Datong station.
Contour interval is 0.2 and the zero-contour is hidden. Solid lines denote positive values and dashed lines negative values.
one significant increase was found in 1980. However, on a time
scale shorter than 180 months, four abrupt decreases and five
abrupt increases were identified without considering whether
the abrupt increase or decrease was significant at the 95% confidence level or not; wherein only abrupt changes in 1956, 1980,
1984, and 1988 were significant at the 95% confidence level.
Abrupt changes in streamflow in terms of the subseries mean
characterized the variation of mean streamflow. The standard deviation describes whether streamflow changes are stable or unstable.
The stability of streamflow changes may exert a considerable influence on the conservation of the wetland ecological environment.
Thus we analyzed the variation of standard deviation of the
streamflow series on different time scales using the scanning F-test
technique. The scanning F-test of the streamflow series (Fig. 6) was
calculated on the same time scales as in Fig. 5. It is seen from Fig. 6
that many frequent variations of standard deviation can be observed on short time scales. On time scales longer than 91 months,
however, two positive (increases in subseries variances) and three
negative (decreases in subseries variances) changes were detected,
with local maxima and minima in the contours (Fig. 6). Only the
positive change in 1991 was significant at the 95% confidence level.
On time scales shorter than 91 months, six negative and four positive significant changes, except the positive changes in 1963 which
was not significant at the 95% confidence level, were detected with
local minima and maxima in the contours.
In order to characterize the stable or unstable features of each
episode partitioned by abrupt changes of streamflow mean, we
combined the results of t- and F-test, as shown in Fig. 7. It can be
seen that episodes of high streamflow are usually characterized
by larger standard deviations, meaning unstable streamflow variations; and episodes of low streamflow are usually characterized by
smaller standard deviations, implying stable streamflow
variability.
Moreover, a larger standard deviation can also be identified in
the transition period from episodes of low to high mean streamflow (Fig. 7). After 2003, the standard deviation of streamflow becomes small (upper panel of Fig. 7) and the negative difference of
streamflow also occurred after 2003. The general trend of the standardized streamflow (seasonal variations have been removed by
SSFj,i = (SFj,i SFj,mean)/SFj,std, where SSF is the standardized streamflow series; SF is the raw streamflow series; j denotes the month,
i.e., January, February, and so on; i denotes the length of the series;
SFj,mean denotes the mean streamflow of month j; SFj,std means the
standard deviation of month j) indicated slightly increasing
streamflow, but the increase was not significant at the 95% confidence level. However the standard deviation obtained by the scanning F-test was decreasing. Therefore, the monthly streamflow at
the Datong station was slightly increasing and became stable over
time. Specifically, two time intervals were identified based on
Fig. 7 characterized by increasing monthly streamflow, i.e., from
mid-1950s to early 1970s and from mid-1980s to 2006.
Scanning t- and F-test of the sediment load variations
The results of scanning t-test of the sediment load series from
the Datong station are shown in Fig. 8. Visual comparison between
101
Q. Zhang et al. / Journal of Hydrology 368 (2009) 96–104
512
Time scale (months)
362
256
181
128
91
64
45
32
1955 1960 1965 1970 1975 1980 1985 1990 1995 2000 2005
Years
1.6
1.4
1.2
1
0.8
0.6
0.4
0.2
0
-0.2
-0.4
-0.6
-0.8
-1
-1.2
-1.4
Fig. 6. Contours of the scanning F-test on standardized streamflow series at the Datong station at 95% confidence level. Contour interval is 0.2 with the zero-contour lines
hidden. Solid lines denote positive values and dashed lines negative values.
Streamflow (m3/s)
Standard deviation
Figs. 5 and 8 reveals distinctly different patterns of scanning t-test
results of the sediment load series when compared with those of
the streamflow series. Based on the local maxima and minima of
the t-test values, on a time scale longer than 91 months two negative (decreasing sediment load) centers were identified (Fig. 8).
Significant abrupt changes were detected in 1976 and 1985. The
1.6
1.4
1.2
1
0.8
4
2
0
−2
1950 1955 1960 1965 1970 1975 1980 1985 1990 1995 2000 2005
Years
Fig. 7. Episode averages and standard deviations of the standardized streamflow
series at the Datong station. The thin gray solid line of the lower panel denotes
standardized streamflow series.
sediment load during 1976–1985 was 5.7% less than that before
1976; and the sediment load after 1985 was more than 32.6% less
than that during 1976–1985. On time scales shorter than 91
months, the sediment load series was partitioned into six negative
(decreasing sediment load) and two positive (increases in sediment load) centers based on the local maxima and minima of the
t-test statistic values (Fig. 8). These six negative abrupt changes
centered in 1969 (on a time scale of 64 months), 1976 (on a time
scale of 45 months), 1985 (on a time scale of 54 months), 1992
(on the time scale of 64 months), 1998 (on a time scale of 38
months), and 2003 (on a time scale of 45 months). However, the
abrupt changes in 1976, 1998, and 2003 were not significant at
the 95% confidence level. The sediment load after 2003 was
59.8% less than that before 2003. Two significant positive abrupt
changes centered in 1980 (on a time scale of 45 months) and
1988 (on a time scale of 38 months).
The scanning F-test results of standardized sediment load series
are shown in Fig. 9. For comparison, the scanning F-test of the sediment load was computed on the same time scale as that of the ttest (Figs. 8 and 9). Different properties of contours of the F-test
can be identified in Fig. 9 in comparison with those of Fig. 8. Many
frequent variations can be observed on shorter time scales. On time
scales longer than 91 months, one positive (increase in subseries
variance) and two negative (decreasing subseries variances) significant changes were detected by identifying the local maxima and
minima in the contours (Fig. 9). On time scales shorter than 91
months, four negative and two positive significant changes were
512
Time scale (months)
362
256
181
128
91
64
45
32
1965
1970
1975
1980
1985
1990
1995
2000
2005
1
0.8
0.6
0.4
0.2
0
-0.2
-0.4
-0.6
-0.8
-1
-1.2
-1.4
Years
Fig. 8. Contours of the normalized scanning t-test standardized by the ‘‘Table-Look-up” critical value t0.05 for the standardized sediment load series at the Datong station.
Contour interval is 0.2 and the zero-contour is hidden. Solid lines denote positive values and dashed lines negative values.
102
Q. Zhang et al. / Journal of Hydrology 368 (2009) 96–104
512
Time scale (months)
362
256
181
128
91
64
45
32
1965
1970
1975
1980
1985
1990
1995
2000
2005
1.8
1.4
1
0.6
0.2
-0.2
-0.6
-1
-1.4
-1.8
-2.2
-2.6
Years
Fig. 9. Contours of the scanning F-test on standardized sediment load series at the Datong station at the 95% confidence level. Contour interval is 0.2 with the zero-contour
lines.
1.68
1.09
Just as shown in Fig. 7, we combined the results of F-test with
those of t-test to characterize the changes in sediment load within
the episodes partitioned by the t-test as stable or unstable features
(Fig. 10). It can be seen that episodes with larger sediment load
mean values usually corresponded to episodes with a larger standard deviation (unstable) and vice versa. This result is similar to
that of streamflow variations. However, two episodes were exceptions: 1975–1980 and 2003–2006. These two episodes were characterized by low sediment load mean values but higher standard
deviations; in other words, these two episodes were characterized
by decreasing but unstable sediment load changes (Fig. 10).
Coherency relations between sediment load and streamflow series
0.49
Sediment load
(kg/s)
Standard deviation
observed, centering in 1971 (on a time scale of 45 months), 1985
(on a time scale of 54 months), 1991 (on a time scale of 76
months), 1993 (on a time scale of 32 months), 1997 (on a time
scale of 32 months), and 2002 (on a time scale of 38 months)
(Fig. 9). The change years in the second moment (F-test) were different from those in the first moment (t-test).
3
1
−1
1963 1968 1973 1978 1983 1988 1993 1998 2003
Years
Fig. 10. Episode averages and standard deviations of the standardized sediment
load series at the Datong station. The thin gray solid line of the lower panel denotes
standardized sediment load series.
What was discussed above characterized the abrupt changes
and stable or unstable properties of the sediment load and the
streamflow variations by using the scanning t- and F-test techniques. We also conducted coherency analysis of abrupt changes
of these two hydrological series (Fig. 11). It can be observed that
the coherency of significant changes in the streamflow series with
those in the sediment load series showed one negative (anti-phase)
center around 1988 on time scales of 128–181 months. This result
indicated that on a longer time scale, sediment load and streamflow variations were in anti-phase relations. On time scales shorter
than 91 months, six positive (in-phase) coherency centers were detected around the mid-1970s, the late 1980s, the early 1980s, the
mid-1980s, the late 1990s, and 2000–2005. Therefore, on shorter
512
1
Time scale (months)
362
0.8
256
0.6
181
0.4
128
0.2
0
91
-0.2
64
-0.4
45
-0.6
32
1965
1970
1975
1980
1985
Years
1990
1995
2000
2005
Fig. 11. Coherency between sediment load and streamflow series at the Datong station.
-0.8
Q. Zhang et al. / Journal of Hydrology 368 (2009) 96–104
time scales, sediment load variations mostly coincided with those
of streamflow, showing influences of hydrological dynamics on the
transport of suspended sediment in river channels. However, from
the viewpoint of a longer time scale, anti-phase relations between
sediment load and streamflow indicated other driving factors besides hydrological dynamics.
Discussions and conclusions
The Datong station is the last control station of the Yangtze River basin. The sediment load and streamflow variations measured at
the Datong station are taken as input of the terrestrial flux from the
Yangtze River basin into the East China Sea. In this study, we analyzed long monthly sediment load and streamflow series by using
scanning t-test and F-test techniques and coherency analysis with
the aim to understand variations in sediment load and streamflow
on different time scales. The following conclusions can be drawn
from this study.
The sediment load turns to decrease after the 1980s and this
decreasing tendency becomes sharp after 2000, which is mainly reflected by dense contour lines of the sediment load difference. Seasonal variations of sediment load show moderate variations in
January, February, November, and December. This is particularly
true for sediment changes in January and February. Larger variability can be identified in the changes of the sediment load in June,
July, August and September. Comparatively, streamflow series
exhibits complex changes. The negative difference of streamflow
was detected before the mid-1980s and positive difference after
1985 for January, February, and March. Streamflow in April, May,
and June shows a negative difference before the late 1980s, followed by a positive difference during the late 1980s and 2002.
After about 2002, streamflow changes turn to the negative difference. Generally, decreasing streamflow can be observed in
March–December after 2002. Streamflow shows an increasing
trend and sediment load a decreasing trend, although these trends
are not significant at the 95% confidence level. This may explain the
anti-phase relationship between the sediment load and
streamflow.
The results of scanning t-test indicate that significant abrupt
changes in sediment load occurred in 1976 and 1985. However,
the significant abrupt change in 1985 was detected at almost
all time scales, showing that this abrupt change in 1985 is the
principle abrupt change in sediment load at the Datong station.
Studies (Zhang et al. 2008a) have shown that streamflow variations in the Yangtze River basin are mainly the result of climate
change, and precipitation changes in particular. The transport of
sediment load is heavily influenced by human activities, such as
construction of dams, land use, forestation/deforestation, and so
forth. Construction of the Gezhouba Dam started in 1970 and
the operation started in the early 1980s with a total storage of
1.58 109 m3, which exerts a tremendous influence on the sediment load transportation (Chen and Huang, 1991). Furthermore,
up to the end of the 1980s, there were 1880 water reservoirs constructed in the Jinshajiang River basin with a total storage of
2.813 109 m3 (Xu, 2005). The construction of these water reservoirs on the mainstem and tributaries of the Yangtze River
trapped large amounts of sediments and gave rise to significant
abrupt changes in sediment load in the mid-1980s. The streamflow at the Datong station is increasing, although this increasing
trend is not significant at the 95% confidence level. This result
is in agreement with that based on annual streamflow analysis
(Zhang et al., 2006b). Streamflow changes in the lower Yangtze
River are mainly attributed to the spatial and temporal distribution of precipitation (Zhang et al., 2005). The results of coherency
analysis, showing anti-phase relationships between sediment load
and streamflow on longer time scales, confirm the general trend
103
of sediment load and streamflow. However, in-phase relations between sediment load and streamflow seem to indicate the influence of hydrological dynamics on the transport of suspended
sediment in the river channel (Milliman and Syvitski, 1992). Dynamic transportation-sedimentation processes of sediment load
in the river channel also make the relations between abrupt
changes of sediment load and construction of water reservoirs
more complicated. Even so, our results still clearly show massive
influences of water reservoirs on sediment load variations in the
lower Yangtze River basin.
The scanning t-test results show a change point in 2003, but
this abrupt change is not significant at the 95% confidence level.
The scanning F-test shows that sediment changes become unstable
after 2003. These changing properties of the first and the second
moments may reflect the impact of the Three Gorges Dam on the
sediment transport in the lower Yangtze River. The period of
2003–2006 was decided as the post-TGD (Three Gorges Dam) by
Chen et al. (2008). Annual sediment load changes also indicate
decreasing sediment load after 2003 (Chen et al., 2008). The mean
sediment load during 2003–2006 is 59.8% less than that before
2003, showing a tremendous influence of the Three Gorges Dam
on sediment load changes in the lower Yangtze River. Sediment
load difference analysis indicates that the decrease of sediment
load occurred mainly during May–October. This may be due to
the operation scheme of the Three Gorges Dam (YRSRI, 1993; Chen
et al., 2008). The Three Gorges Dam usually stores water in dry seasons and releases extra water during wet seasons, particularly
when high flow events occur. Moreover, wet seasons are usually
the periods when production and transport of sediment load occur
in the Yangtze River basin. Therefore, human activities, particularly
the construction of water reservoirs, give rise to significant
decreasing trend of the sediment load transport in the lower Yangtze River. Moreover, the sediment load variations become unstable. These changing properties of the sediment load may imply
serious challenges for ecological environment conservation and
deltaic management.
This paper addressed an important scientific problem about
influences of hydrological regulations of the water reservoirs
and other influencing factors on hydrological processes such as
sediment load and streamflow. Moreover, the importance of this
study also lies in: (1) exploring sediment load and streamflow
variations on different time scales in that different influencing
factors usually alter the hydrological processes on diverse time
scales; (2) understanding stable/unstable properties of hydrological processes on different time scales, and which is practically
important in sound management practice of ecological environment conservation and the development of the Yangtze Delta;
(3) application of updated hydrological data and which is helpful
for deeper insight into influences of water reservoirs and other
factors on sediment load transportation and streamflow variations in a timely way; and (4) analogously, this study shows another way to study possible influences of water reservoirs on
hydrological processes besides wavelet technique as suggested
by White et al. (2005).
Acknowledgments
The research was financially supported by National Natural Science Foundation of China (Grant Nos. 40701015 and 40730635),
the innovative project from Nanjing Institute of Geography and
Limnology, CAS (Grant Nos. CXNIGLAS200814; 08SL141001; and
08YCZ11007), and by Program of Introducing Talents of Discipline
to Universities - the 111 Project of Hohai University. Thanks should
be extended to the Changjiang (Yangtze) Water Resources Commission (CWRC) for providing the hydrological data. We also
appreciate the comments and suggestions from two anonymous
104
Q. Zhang et al. / Journal of Hydrology 368 (2009) 96–104
reviewers and the editor, Prof. Geoff Syme. Their comments greatly
helped to improve the quality of this paper.
References
Chen, S.R., Huang, G.H., 1991. Water level changes of Yichang station after operation
of Gezhouba Dam. Yangtze River 1, 30–37 (in Chinese).
Chen, Z.Y., Li, J.F., Shen, H.T., Wang, Z.H., 2001. Yangtze River of China: historical
analysis of discharge variability and sediment flux. Geomorphology 41 (2-3),
77–91.
Chen, X.Q., Yan, Y.X., Fu, R.S., Dou, X.P., Zhang, E.F., 2008. Sediment transport from
the Yangtze River, China, into the sea over the post-Three Gorge Dam period: a
discussion. Quaternary International 186, 55–64.
Cramer, H., 1946. Mathematical Method of Statistics. Princeton University Press,
Princeton, NJ, USA.
Hammersley, J.M., 1946. Discussion of papers. Journal of the Royal Statistical
Society 8, 91.
Jiang, J.M., Mendelssohn, R., Schwing, F., Fraedrich, K., 2002. Coherency detection of
multiscale significant changes in historic Nile flood levels. Geophysical Research
Letters 29 (8), 112-1–112-4.
Jiang, J.M., Gu, X.Q., Ju, J.H., 2007. Significant changes in subseries means and
variances in an 8000-year precipitation reconstruction from tree rings in the
southwestern USA. Annales Geophysicae 25, 1–12.
Kumar, P., Foufoula-Georgiou, E., 1994. Wavelet analysis in geophysics: an
introduction. In: Foufoula-Georgiou, E., Kumar, P. (Eds.), Wavelets in
Geophysics. Academic Press, San Diego, pp. 1–43.
Milliman, J.D., 1997. Blessed dams or damned dams. Nature 386, 325–326.
Milliman, J.D., Syvitski, J.P.M., 1992. Geomorphic/tectonic control of sediment
discharge to the ocean: the importance of small mountainous rivers. Journal of
Geology 100, 525–544.
Mou, J., 1991. The impact of environmental change and conservation measures on
erosion and sediment loads in the Yellow River basin. In: Peters, N.E., Walling,
D.E. (Eds.), Sediment and Stream Water Quality in a Changing Environment.
Proc. Vienna Symposium, August, 1991. IAHS Press, Wallingford, UK, pp. 47–52
(IAHS Publication No. 203).
Syvitski, J.P.M., 2003. Supply and flux of sediment along hydrological pathways:
research for the 21st century. Global and Planetary Change 39 (1–2), 1–11.
Syvitski, J.P.M., Vörösmarty, C.J., Kettner, A.J., Green, P., 2005. Impact of humans on
the flux of terrestrial sediment to the global coastal ocean. Science 308, 376–
380.
UNESCO, 1998. Coasts and small islands home for two-thirds of world population.
1998 International Year of the Ocean Website. <http://www.education.
unesco.org/op/eng/98iyo/coastal.htm>.
Von Storch, H., Zwiers, F., 1999. Statistical Analysis in Climate Research. Cambridge
University Press, Cambridge. p. 116.
Walling, D.E., 2006. Human impact on land–ocean sediment transfer by the world’s
rivers. Geomorphology 79, 192–216.
Walling, D.E., Fang, D., 2003. Recent trends in the suspended sediment loads of the
world’s rivers. Global and Planetary Change 39, 111–126.
Wang, Z.Y., Hu, S.X., Wu, X.J., Shao, Y.S., 2003. Delta processes and management
strategies in China. International Journal of River Basin Management 1, 173–
184.
White, A.M., Schmidt, J.C., Topping, J.D., 2005. Application of wavelet analysis for
monitoring the hydrologic effects of dam operation: Glen Canyon Dam and the
Colorado River at Lees Ferry, Arizona. River Research and Application 21 (5),
551–565.
Xu, J.X., 2005. Variation in grain size of suspended load in upper Changjiang River
and its tributaries by human activities. Journal of Sediment Research 3, 8–16 (in
Chinese).
Yang, S.L., Li, M., Dai, S.B., Liu, Z., Zhang, J., Ding, P.X., 2006. Drastic decrease in
sediment supply from the Yangtze River and its challenge to coastal wetland
management. Geophysical Research Letters 33. doi:10.1029/2005GL025507.
YRSRI (Yangtze River Science Research Institute), 1993. Research on Key Techniques
of Sedimentation and Navigation Relevant to the Three-Gorge Project. Wuhan
Industrial University Press, Wuhan, China (in Chinese).
Zhang, Q., Jiang, T., Gemmer, M., Becker, S., 2005. Precipitation, temperature and
discharge analysis from 1951 to 2002 in the Yangtze catchment, China.
Hydrological Sciences Journal 50 (1), 65–80.
Zhang, Q., Liu, C.L., Xu, C.-Y., Xu, Y.P., Jiang, T., 2006a. Observed trends of annual
maximum water level and streamflow during past 130 years in the Yangtze
River basin, China. Journal of Hydrology 324, 255–265.
Zhang, Q., Xu, C.-Y., Becker, S., Jiang, T., 2006b. Sediment and runoff changes in
the Yangtze River basin during past 50 years. Journal of Hydrology 331, 511–
523.
Zhang, Q., Chen, G.Y., Su, B.D., Disse, M., Jiang, T., Xu, C.-Y., 2008a. Periodicity of
sediment load and runoff in the Yangtze Riverbasin and possible impacts of
climatic changes and human activities. Hydrological Sciences Journal 53 (2),
457–465.
Zhang, S.R., Lu, X.X., Higgitt, D.L., Chen, C.T., Han, J., Sun, H., 2008b. Recent changes
of water discharge and sediment load in the Zhujiang (Pearl River) basin, China.
Global and Planetary Change 60, 365–380.