Multiscale streamflow variations of the Pearl River basin and possible implications for the water resource management within the Pearl River Delta, China

lable at ScienceDirect

Quaternary International 226 (2010) 44–53

Contents lists avai

Quaternary International

journal homepage: www.elsevier .com/locate/quaint

Multiscale streamflow variations of the Pearl River basin and possibleimplications for the water resource management within the Pearl RiverDelta, China

Yongqin David Chen a,b, Qiang Zhang c,*, Chong-Yu Xu d, Xixi Lu e, Shurong Zhang e

a Department of Geography and Resource Management, The Chinese University of Hong Kong, Hong Kong, Chinab Centre of Strategic Environmental Assessment for China, The Chinese University of Hong Kong, Hong Kong, Chinac Department of Water Resources and Environment, Sun Yat-sen University, 135 Xingangxi Road, Guangzhou 510275, Chinad Department of Geosciences, University of Oslo, P O Box 1047 Blindern, N-0316 Oslo, Norwaye Department of Geography, National University of Singapore, Arts Link 1, 117570 Singapore

a r t i c l e i n f o

Article history:Available online 8 September 2009

* Corresponding author. Tel./fax: þ86 20 84113730E-mail address: [email protected] (Q. Zhang).

1040-6182/$ – see front matter � 2009 Elsevier Ltd adoi:10.1016/j.quaint.2009.08.014

a b s t r a c t

Long monthly streamflow series of three control hydrological stations of the Pearl River basin wereanalyzed by using the scanning t-test and the scanning F-test. Possible implications of the changingproperties of streamflow variations for the water resource management of the Pearl River Delta are alsodiscussed. The results indicated that: 1) more complicated changes were observed in terms of the secondcenter moment when compared to the first original moment. More significant abrupt changes of thesecond center moment imply more sensitive response of streamflow stability to climate changes andhuman activities; 2) abrupt behaviors of the first (second) center moment of the streamflow variationstend to be more sensitive to climate changes and/or human activities in the larger river basin whencompared to those in the smaller river basin. These phenomena are attributed to buffering functions ofmore storage space of longer river channel, and more complicated and longer runoff yield andconcentration processes in the river basin of larger drainage area; 3) annual minimum streamflow of thePearl River basin tends to be increasing. This will be helpful for better human mitigation of the salinityintrusion in dry seasons across the Pearl River Delta. Annual maximum streamflow, when compared toannual minimum streamflow, shows larger-magnitude variability reflected by larger standard deviation,implying unfavorable conditions for flood mitigation in the Pearl River Delta. The results of this paper areof scientific and practical merits for water resource management and sound human mitigation to waterhazards across the Pearl River Delta, and also are a good case study for similar researches in other riverdeltas in the world under the changing environment.

� 2009 Elsevier Ltd and INQUA. All rights reserved.

1. Introduction

The altered hydrological cycle and changed spatial and temporaldistribution of water resource as a result of the increasingtemperature attracted considerable concerns from hydrologists,meteorologists and also policy makers due to the tremendousimportance of water in both society and nature (e.g. Xu and Singh,2004), and which also poses a new challenge for the actual practiceof basin-scale water resource management under the changingclimate. Research results indicated that water resources are sensi-tive to climate changes, and this is particularly true for the ground

.

nd INQUA. All rights reserved.

surface water resources (WMO, 1987). Small perturbations inprecipitation frequency and/or quantity can result in significantimpacts on the mean annual discharge (Risbey and Entekhabi,1996). Any alterations in the hydrologic cycle will affect energyproduction and flood control measures (Xu and Singh, 2004) tosuch an extent that water management adaptation measures willvery likely be brought in (Minville et al., 2008). Therefore, climaticchanges and associate impacts on global/regional water resourcesare receiving increasing concerns from academic circles (Loukaset al., 2002; Camilloni and Barros, 2003; Zhang et al., 2008a).

Xu (2000) thoroughly investigated the influences of climatechanges on flow regimes of twenty-five catchments in centralSweden by using a conceptual monthly water balance model,suggesting that significant increase of winter flow and decrease ofspring and summer streamflow were resulted from most scenarios.

Y.D. Chen et al. / Quaternary International 226 (2010) 44–53 45

More and more hydrologists and meteorologists have focusedattention on variability and availability of regional water resourcesunder current climatic changes (e.g. Zhang et al., 2009a). Study ofimpacts of climate changes on hydrological processes heavily relieson trend detection of hydro-climatic variables such as precipitationand streamflow. Due to complex changing patterns of precipitation,temperature, and other meteorological variables, it is still unclearhow exactly these changes in meteorological variables may affectstreamflow variations (Birsan et al., 2005). Streamflow integratesthe various influences of atmospheric variables, human activitiessuch as land use changes, urbanization, over watershed, and allthese impacts are combined to reflect hydrological processes in theoutlet of the river basin. Thus, numerous studies attempted toaddress variations of streamflow of the river basins over the world(Pekarova et al., 2003; Kahya and Kalayci, 2004; Birsan et al., 2005;Zhang et al., 2006). Zhang et al. (2008) studied annual streamflowand sediment load series of the Pearl River basin. The current studyattempts to address statistical properties of streamflow variationsof the Pearl River basin and discuss associated implications forwater resource management within the Pearl River Delta based onlong monthly streamflow dataset.

The Pearl River (3�410N- 29�150N; 97�390E- 117�180E) (Fig. 1) isthe second largest river in terms of streamflow in China witha drainage area of 4.42�105km2 (PRWRC, 1991), which involvesthree major tributaries: West River, North River and East River. TheWest River is the largest tributary accounting for 77.8% of the totaldrainage area of the basin. The North River is the second largest witha drainage area of 4,6710 km2. The East River accounts for 6.6% of thetotal area of the Pearl River Basin. The annual mean temperatureranges between 14 and 22 �C and the precipitation mainly occursduring April–September (Zhang et al., 2008b), accounting for 72–88% of the annual precipitation (PRWRC, 1991). The streamflowvariations of the Pearl River basin have considerable influences onthe hydrological processes of the Pearl River Delta, one of the mostcomplicated deltaic drainage systems in the world (Chen and Chen,2002). Flat terrain at low-lying altitude and downstream location,together with rapid economic development and population growthover the past three decades have made the PRD region more andmore vulnerable to natural hazards such as flood, salinity intrusionand storm surge. In recent years, engineering facilities and othermodifications of the Pearl River network have been designed to

Fig. 1. Location of the study region a

strengthen flood protection and to cater for huge requirements ofbuilding materials. Since the mid-1980s, intensive channel dredgingand levee construction have significantly affected flood stages andcaused serious hydrologic alterations in the study region. Alteredstreamflow allocation between North River and West River was alsoseen as a major cause for the hydrological alterations of the PearlRiver Delta (Chen et al., 2008). Hydrological alterations led toabnormally-high water level in the flooding seasons and morefrequent salinity intrusion events in dry seasons, which posed newchallenges for effective water resource management across the PearlRiver Delta. However, better understanding of hydrologicalprocesses of the Pearl River basin will be of great scientific andpractical importance for effective water resource managementwithin the Pearl River Delta, and this is the major motivation for thisstudy. With these in mind, the objectives of this study are: 1) tounderstand abrupt behaviors of streamflow variations in terms offirst and second center moments; 2) to detect abrupt changes ofannual minimum and maximum streamflow series based onmonthly streamflow dataset; 3) to reveal possible implications forwater resource management in the Pearl River Delta.

This study analyzed long hydrological series, more than 50years. The results of this paper should be important for effectivewater resource management in the Pearl River Delta underchanging climate and also provide good case study for the waterresource management for other river deltas of the world.

2. Data and methodology

This study analyzed long hydrological series of three controlhydrological stations: Gaoyao, Shijiao and Boluo. More detailedinformation of these three hydrological stations is presented intoTable 1. Fig. 1 illustrates the location of these stations considered inthis study. The hydrological data are of good quality withoutmissing data. The hydrologic data before 1989 are extracted fromthe Hydrological Year Book (published by the Hydrological Bureauof the Ministry of Water Resources of China) and those after 1989are provided by the Water Bureau of Guangdong Province.

Abrupt changes of hydrological and meteorological series areoften of considerable importance (Lund et al., 2001; Lund andReeves, 2002). In this study, abrupt changes were analyzed usingthe scanning t-test technique on different time scales. Stable or

nd three hydrological stations.

Table 1Detailed information of hydrological stations along the mainstream and tributariesof the Pearl River basin.

Tributaries Station name Length of series Basin area(103 km2)

West River Gaoyao Jan. 1956–Dec. 2007 351.5North River Shijiao Jan. 1956–Dec. 2007 38.4East River Boluo Jan. 1954–Dec. 2007 25.3

Y.D. Chen et al. / Quaternary International 226 (2010) 44–5346

unstable status of hydrological variations is another importantstatistical property of the hydrological series since that unstablewater supply may seriously impact making procedure of waterresource management measures. Steady or unsteady variations ofthe hydrological series were evaluated by the change in the stan-dard deviation, which was analyzed by the F-test technique.

Jiang et al. (2002) extended the definition of student t-test and theF-test (Cramer, 1946) by identifying change points on different timescales. The scanning t-test attempts to detect significant changes inthe first moment (subseries mean or average) on different time scaleswithin a long time series; while the scanning F-test attempts toanalyze significant changes in subseries variance (the secondmoment) on various time scales. These two techniques were intro-duced in Jiang et al. (2007). For the sake of completeness of this study,a brief introduction of these two methods is presented here.

Statistic t(n, j) in the scanning t-test is defined as the differenceof the subsample averages between every two adjoining subseriesof equal subseries size (n):

tðn; jÞ ¼�xj2 � xj1

�� n1=2 �

�s2

j2 þ s2j1

��1=2(1)

where xj1 ¼ 1=nPj�1

i¼ j�n xðiÞ, xj2 ¼ 1=nPjþn�1

i¼ j xðiÞ,

1960 1965 1970 1975 1980

Time (ye

Tim

e sc

ales

(mon

ths)

Gaoyao station

8

16

32

64

128

256

1960 1965 1970 1975 1980

−2

0

2

4

6

Fig. 2. Contours of the normalized scanning t-test standardized by the ‘‘Table-Look-up’’ critand dashed lines denote abrupt changes significant at >95% confidence level. Solid linesstandardized streamflow, change points and episode average (solid line) from Fig. 2A.

s2j1 ¼ 1=n� 1

Pj�1i¼ j�nðxðiÞ � xj1Þ2,

s2j2 ¼ 1=n� 1

Pjþn�1i¼ j ðxðiÞ � xj2Þ2, in which subsample size n may

vary in this way as n¼ 2, 3,.,<N/2. The j¼ nþ 1, nþ 2,., N� nþ 1is the reference time.

The Table-Look-Up Test (Von Storch and Zwiers, 1999) was usedto modify the significance criterion of statistic t(n, j) with lag-1autocorrelation coefficients of the pooled subsample and thesubsample size n in that hydrological series usually subject topersistence. Criterion t0.05 for the correction of the dependence wasaccepted as the significance level on time scales considered. Forshorter subsample sizes, the critical values are overly restrictive.The significance level varies with n and j, and so the test statisticwas normalized as:

trðn; jÞ ¼ tðn; jÞ=t0:05 (2)

When jtr(n,j)j> 1.0, the abrupt change is significant at the 0.05significance level. tr(n,j)<�1.0 denotes significant decrease andtr (n, j)> 1.0 significant increase.

The scanning F-test defines significant changes of subseriesvariances. Statistic Fr(n,j) is defined as:

Frðn; jÞ ¼

8>>><>>>:

��

S2j1=S2

j2

�.Fa; for Sj2 < Sj1;

0; for Sj2 ¼ Sj1 or Sj1 ¼ 0; Sj2 ¼ 0;�S2

j2=S2j1

�.Fa; for Sj2 > Sj1;

(3)

where the subsample standard deviations Sj1 and Sj2 are calcu-lated in the same way as in Eq. (1). Fa is a threshold valuebased on the effective degree of freedom after the correction of

1985 1990 1995 2000 2005

ar)

−1.1

−0.8

−0.5

−0.1

0.1

0.3

0.5

0.7

0.9

1.1

1985 1990 1995 2000 2005

B

A

ical value t0.05 for the standardized streamflow series at the Gaoyao station. Thick soliddenote positive values and dashed lines negative values (Fig. 2A). Fig. 2B indicates

1960 1965 1970 1975 1980 1985 1990 1995 2000 2005Time (year)

Tim

e sc

ales

(mon

ths)

−3.5−3−2.5−2−1.5−1−0.500.511.522.5

8

16

32

64

128

256 Gaoyao station

1960 1965 1970 1975 1980 1985 1990 1995 2000 2005−4

−2

0

2

4

6

B

A

Fig. 3. Contours of the normalized scanning F-test for the standardized streamflow of the Gaoyao station. Thick solid and dashed lines denote abrupt changes significant at >95%confidence level. Solid lines denote positive values and dashed lines negative values (Fig. 3A). Fig. 3B indicates standardized streamflow, change points and episode standarddeviation (solid line) from Fig. 3A.

1960 1965 1970 1975 1980 1985 1990 1995 2000 2005Time (year)

Tim

e sc

ales

(mon

ths)

0.10.30.50.70.91.1A

8

256

128

64

32

16

Shijiao station

1960 1965 1970 1975 1980 1985 1990 1995 2000 2005−2

−1.1−0.9−0.7−0.5−0.3−0.1

−1

0

1

2

3

4

5

B

Fig. 4. Contours of the normalized scanning t-test standardized by the ‘‘Table-Look-up’’ critical value t0.05 for the standardized streamflow series at the Shijiao station. Thick solidand dashed lines denote abrupt changes significant at >95% confidence level. Solid lines denote positive values and dashed lines negative values (Fig. 4A). Fig. 4B indicatesstandardized streamflow, change points and episode average (solid line) from Fig. 4A.

Y.D. Chen et al. / Quaternary International 226 (2010) 44–53 47

Y.D. Chen et al. / Quaternary International 226 (2010) 44–5348

dependence and in a normalized distribution for the time series.The effective degree of freedom was estimated as (Hammersley,1946):

Ef ðnÞ ¼ f ðnÞ �hXk

r2ðsÞi�1

; rðkÞ/0; (4)

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 significantchange towards a larger variance, i.e. the record becomes muchunsteadier (Jiang et al., 2007).

3. Results

3.1. Abrupt behavior of monthly streamflow variationsof West River

This study analyzed abrupt changes of first center moment andsecond center moment. For the sake of easy understanding, thecontours of scanning t-test and F-test results showing abruptbehaviors of monthly streamflow variations on different time scaleswere mapped. Standardized streamflow series and associatedepisode average, standard deviation and significant change pointswere mapped. To avoid redundancy, the contours of scanning t- andF-test with respect to annual minimum and maximum streamflowseries were not mapped. Fig. 2A shows that the monthly streamflowseries of the Gaoyao station is dominated by abrupt decrease inter-rupted by sporadic intervals characterized by abrupt increase ontime scales less than 32 months. A significant abrupt decrease ofaverage was identified at 2004. From the perspective of time scaleslonger than 32 months, abrupt increase of streamflow was detected

1960 1965 1970 1975 1980 1Time (ye

Tim

e sc

ales

(mon

ths)

8

16

32

64

128

256 Shijiao station

1960 1965 1970 1975 1980 1−2

0

2

4

6B

Fig. 5. Contours of the normalized scanning F-test for the standardized streamflow of the Sconfidence level. Solid lines denote positive values and dashed lines negative values (Fig.deviation (solid line) from Fig. 5A.

during 1964–1975, but is not significant at >95% confidence level.Significant decrease of streamflow can be observed after the early1980s, and significant increase of streamflow was found in the early1990s. Fig. 2B demonstrates two periods characterized by decreasedstreamflow: 1983–1994 and 2004–2007 and two periods dominatedby increased streamflow: 1956–1982 and 1995–2003. Fig. 2A and Bindicate more frequent increase and decrease of streamflow varia-tions after 1990 when compared to those before 1990. Before the1980s, no significant abrupt streamflow changes were found. Abruptstreamflow variations were observed mainly after the early 1980s.

Stability of the streamflow variations of the Gaoyao stationshown by the scanning F-test was illustrated in Fig. 3. Morecomplicated patterns can be found in Fig. 3A when compared tothe scanning t-test (Fig. 2A). More significant change points ofstandard deviation can be detected on different time scales. Ontime scales shorter than 64 months, five time intervals wereidentified with different stability properties of streamflow varia-tions. The periods of 1956–1965, 1980–1990 and 2003–2007 werecharacterized by higher frequency of standard deviation changes;the remaining periods, i.e. 1965–1980 and 1990–2003 weredominated by lower variability of standard deviation. On timescales longer than 64 months, changing patterns of second centermoment were relatively simpler when compared to that on timescales shorter than 64 months. Higher stability was observedduring 1965–1975 and after 1983, and lower stability duringbefore 1965 and 1975–1983. The scanning F-test results deter-mined the centers with higher significance level, allowingmapping of the significant change points and associated episodeswith different stability properties (Fig. 3B). Subseries variancesshowed moderate changes. More frequent variations were foundafter the 1980s. After the 1980s, the streamflow changes of theGaoyao station tend to show variations of larger magnitude andfrequency with lower stability.

985 1990 1995 2000 2005ar)

00.511.522.533.5

A

B

985 1990 1995 2000 2005

−5−3.5−3−2.5−2−1.5−1−0.5

hijiao station. Thick solid and dashed lines denote abrupt changes significant at >95%5A). Fig. 5B indicates standardized streamflow, change points and episode standard

1955 1960 1965 1970 1975 1980 1985 1990 1995 2000 2005

Time (year)

Tim

e sc

ales

(mon

ths)

−0.8

−0.6

0

0.2

0.4

0.6

0.8

1

1.2

1.4

8

16

32

64

128

256 Boluo station

1955 1960 1965 1970 1975 1980 1985 1990 1995 2000 2005−2

−0.4

−0.2

0

2

4

6B

A

Fig. 6. Contours of the normalized scanning t-test standardized by the ‘‘Table-Look-up’’ critical value t0.05 for the standardized streamflow series at the Boluo station. Thick solid anddashed lines denote abrupt changes significant at >95% confidence level. Solid lines denote positive values and dashed lines negative values (Fig. 6A). Fig. 6B indicates standardizedstreamflow, change points and episode average (solid line) from Fig. 6A.

Y.D. Chen et al. / Quaternary International 226 (2010) 44–53 49

3.2. Abrupt behavior of monthly streamflow variationsof North River

The scanning t-test results of the streamflow series of the Shijiaostation are shown in Fig. 4. The streamflow series of the Shijiaostation displayed significant abrupt changes in the early 1970s,mid-1970s, mid-1980s and early 1990s. Frequent abrupt changesoccurred during the 1970s and 1980s, which differ from those of thehydrological processes of the Gaoyao station, the West River.Frequent abrupt changes of streamflow variations of the West Riveroccurred after the 1980s. Generally speaking, time intervals char-acterized by different hydrological episodes can be observed in thestreamflow series of the Shijiao station. Decreasing streamflow wasobserved during 1956–1966, 1976–1989 and after 1998; increasingstreamflow was found during 1967–1975 and 1990–1997. Morecomplicated changing patterns of decrease and/or increase ofstreamflow were detected on shorter time scales, i.e. shorter than16 months. Fig. 4A indicates that four abrupt changes after 1995 arenot significant statistically. The streamflow of the Shijiao stationafter 1995 is characterized by four episodes with different,changing properties. From a perspective of a longer time scale, e.g.longer than 32 months, the streamflow at the Shijiao station isdecreasing. Fig. 4B further indicates significant change points andepisodic average streamflow. Increased streamflow occurredmainly in the early 1970s and decreased after the late 1970s untilthe early 1990s. After the 1990s, a slight increase of streamflow canbe detected (Fig. 4B).

Fig. 5 illustrates scanning F-test results in time scale–time space.More significant change points were identified when compared tothose of the scanning t-test. Visual comparison between Figs. 4Aand 5A indicates that increase of streamflow largely corresponds to

lower stability of streamflow variations. Figs. 4 and 5 also indicatethat the stability of streamflow variations is more sensitive toclimate changes or human activities when compared to changes ofthe first center moment. This is also true for the streamflowvariations of the Gaoyao station of the West River. Significantchange points were determined by identifying the center of thesignificant regions circled by thick solid/dashed lines (Fig. 5B).Relatively lower stability of streamflow variations can be observedduring 1960–1965, 1973–1976, and after the early 1990s. The periodof 1970–1985 is characterized by higher frequency of abruptchanges in terms of the first center moment, and simultaneouslyfeatures higher frequency of abrupt changes with respect to thesecond center moment (Figs. 4B and 5B). When compared tostreamflow variations of the West River (Gaoyao station), higher-frequency appearance of abrupt changes of the first and the secondcenter moment was observed within the streamflow series of theNorth River (Shijiao River). Therefore, the streamflow variations ofthe river basin of smaller drainage area respond in a more sensitiveway to climate changes and/or human activities. Therefore, moreimportance should be attached to possible influences of climaticchanges and human activities on hydrological processes within theriver basin of smaller drainage area, e.g. the North River in thisstudy, and to the impacts of streamflow variations on hydrologicalprocesses of the Pearl River Delta.

3.3. Abrupt behavior of monthly streamflow variations of East River

The scanning t-test analysis for the long streamflow series wasused to define when the change points occurred at Boluo station,the control hydrological station of the East River (Figs. 1, 5). Thestreamflow series from Boluo displayed significant abrupt

1955 1960 1965 1970 1975 1980 1985 1990 1995 2000 2005

Time (year)

Tim

e sc

ales

(mon

ths)

00.511.522.5

A

B

256

128

64

32

16

8

Boluo station (A )

1955 1960 1965 1970 1975 1980 1985 1990 1995 2000 2005−2

−3.5−3

−2−2.5

−1.5−1−0.5

0

2

4

6

Fig. 7. Contours of the normalized scanning F-test for the standardized streamflow of the Boluo station. Thick solid and dashed lines denote abrupt changes significant at >95%confidence level. Solid lines denote positive values and dashed lines negative values (Fig. 7A). Fig. 7B indicates standardized streamflow, change points and episode standarddeviation (solid line) from Fig. 7A.

Y.D. Chen et al. / Quaternary International 226 (2010) 44–5350

decreases in 1975 (on a time scale of 128 months), 1985 (on a timescale of 128 months), 1993 (on a time scale of 64 months), and 2002(on a time scale of 32 months). Fig. 6A displays different behaviorsof abrupt changes of streamflow variations on different time scales.More complicated patterns can be observed on shorter time scales.On time scales longer than 64 months, the periods of 1954–1967,1984–1991, 1998–2005 are characterized by decreasing streamflow,and the time intervals of 1968–1983 and 1992–1997 are dominatedby increasing streamflow. More hydrological episodes character-ized by different magnitudes of increase and/or decrease can beobserved on time scales shorter than 64 months. Fig. 6B clearlyshows the time when change points occurred and associatedepisode average streamflow of the Boluo station. When compared

1960 1970 1980 1990 2000−3

−2

−1

0

1

2

3standardized streamflowmean streamflowstandard deviation

Gaoyao station (annual maximum streamflow)−1

−

−0

0

1

2

Fig. 8. t- and F-test of the annual maximum and minimum streamflow series of the Gaoyao sannual maximum streamflow are also not significant. Thus, we did not analyze the linearinspection.

to the scanning t-test results of the streamflow series of the Gaoyaostation and the Shijiao station, the frequency of change pointsappeared within the streamflow series seems to be higher than thatof the Gaoyao station, but lower than that of the Shijiao station. JustStreamflow variations here tend to be more sensitive to climatechanges or human activities. Moderate variations of the streamflowvariations of the East River can be attributed to hydrologicalregulations of the water reservoirs (Chen et al., 2009). As for thescanning F-test results of the streamflow series of the Gaoyao andShijiao station, more complex changing patterns of abrupt changesof the second center moment can be found in Fig. 7. The centers ofthe significant areas were checked and the significant changepoints were shown in Fig. 7B. Fig. 7B shows that more abrupt

1960 1970 1980 1990 2000.5

1

.5

0

.5

1

.5

2

.5

3standardized streamflowlinear fitstandard deviation

Gaoyao station (minimum annual streamflow

tation. The linear trend in the right panel is not significant. The linear trends of the resttrend of those streamflow series. The straight line in the right panel is only for visual

1960 1970 1980 1990 2000−2

−1.5

−1

−0.5

0

0.5

1

1.5

2

2.5standardized streamflowstandard deviation

Shijiao station (annual maximum streamflow)

1960 1970 1980 1990 2000−2

−1

0

1

2

3

4standardized streamflowstandard deviation

Shijiao station (annual minimum streamflow)

Fig. 9. t- and F-test of the annual maximum and minimum streamflow series of the Shijiao station.

Y.D. Chen et al. / Quaternary International 226 (2010) 44–53 51

changes of the second center moment can be found after the 1980s,when the hydrological processes show larger variability.

3.4. Abrupt behavior of annual maximum/minimum streamflow ofthe Pearl River

Owing to hydrological alterations (Chen et al., 2008), thereoccurred more frequent abnormally-high water level in floodingseasons and abnormally-low water levels in dry seasons. Therefore,it is of great importance to understand the changing properties ofannual maximum and/or minimum streamflow based on monthlystreamflow dataset. Figs. 8, 9 and 10 display annual maximum (AM,left panel) and minimum (Am, right panel) streamflow variations ofthese three control hydrological stations respectively. Fig. 8 indi-cates that AM streamflow increased after the 1990s with lowerstability. No significant abrupt changes in terms of the first centermoment of Am streamflow can be detected. Scanning F-test indi-cates higher stability of streamflow variations after the 1980s.Linear fit line indicates increasing trend of Am streamflow of theGaoyao station. Generally, minimum streamflow occurs in winter(December, January, February) and maximum streamflow insummer (June, July, August). Changes of streamflow amount of theWest River mean much to the hydrological variations of the PearlRiver Delta in that the streamflow amount of the West Riveraccounts for more than 70% of the total streamflow amount of thePearl River basin. Increasing annual minimum streamflow of theWest River should be beneficial for human mitigation of salinityintrusion across the Pearl River Delta. Figs. 9 and 10 indicate nosignificant abrupt changes of the first center moment. ScanningF-test results indicate different abrupt changes of the second centermoment. Abrupt changes of standard deviation of the Amstreamflow seem to occur earlier when compared to AM: the

1960 1970 1980 1990 2000−2

−1

0

1

2

3

4Standardized streamflowStandardized deviation

Boluo station (annual maximum streamflow)

1

1

2

2

3

3

4

4

Fig. 10. t- and F-test of the annual maximum and m

former occurs in the 1980s and even earlier, and the later occursafter the 1990s. Figs 9 and 10 indicate that AM and Am streamflowseries of the Shijiao and Boluo station come to be lower stability.However, AM and Am of the Gaoyao station show different prop-erties. AM streamflow of the Gaoyao station tends to be lowerstability after 1990s and Am streamflow comes to be higherstability. Scanning F-test of AM and Am streamflow series of theShijiao and Boluo station indicates larger-magnitude of variationsin terms of AM streamflow changes when compared to those of Amstreamflow variations (Figs. 9 and 10). The Am streamflow varia-tions of the Shijiao and Boluo station have the increasing tendency.Therefore, the streamflow variations seem to be beneficial forbetter mitigation to salinity intrusion within the Pearl River Delta.Larger-magnitude of streamflow variability will not be helpful forflood mitigation in the Pearl River Delta.

4. Discussion

Hydrological processes are in close relation with climatechanges, particularly precipitation variations in space and time.Thorough analysis of precipitation variations (Zhang et al., 2009b)and dryness/wetness variations used the standardized precipita-tion index (SPI) and aridity index (AI) (Zhang et al., 2009c).However, the dry/wet tendency is not significant statistically at>95% confidence level. The wet tendency was observed mainly inthe west parts of the West River basin. The dry tendency was foundmainly in the east parts of the Pearl River basin, specifically theNorth and East Rivers. The non-significant wet and dry tendency inthe West River causes the streamflow series of the West River toshow moderate variation. Streamflow variations of the North andEast Rivers are subject to larger magnitude when compared tothose of West River basin, which can be attributed to the relative

1950 1960 1970 1980 1990 20000

50

00

50

00

50

00

50

00

50mean streamflowstandardized streamflowstandard deviation

inimum streamflow series of the Boluo station.

Y.D. Chen et al. / Quaternary International 226 (2010) 44–5352

drainage areas of the river basins. Study in the Yangtze River basin(Zhang et al., 2008c) indicated that the impacts of human activityand climatic changes on the sediment load and runoff changes aregreater in smaller river basins than in larger river basins. In thisstudy, higher sensitivity of streamflow changes to influences ofclimate changes and human activities was represented mainly bymore frequent abrupt changes of the first and second momentbased on the results of scanning t- and F-test technique. The resultsof the study on the trends and abrupt changes of precipitationmaximum over the Pearl River basin displayed the time when theabrupt changes of the precipitation maxima occurred in space(Zhang et al., 2009b). Generally, the abrupt changes of the precip-itation maxima mainly occurred in three time intervals, i.e. early1970s, early 1980s and early 1990s. Significant abrupt changes ofstreamflow series of the Pearl River basin mainly occurred in thesethree time intervals, which can be observed in the results by thescanning t-test technique. This result shows distinctly thetremendous influences of precipitation changes on the hydrologicalprocesses, though considerable impacts the human activities haveon the hydrological processes (Zhang et al., 2008c). Many studiestended to corroborate the overwhelmingly larger impacts ofprecipitation changes on the streamflow processes when comparedto human activities (e.g. Zhang et al., 2006, 2008c, 2009d). Zhanget al. (2008) also indicated that the decreasing sediment load of thePearl River basin was the result of reservoir construction andstreamflow variations are due to precipitation changes. Therefore,precipitation changes and human activities have different impactson sediment load and streamflow, and streamflow changes aremainly the result of precipitation variations. As for changes ofannual maximum streamflow, larger magnitude of changes can alsobe observed in the North and East Rivers, and moderate variationsin the West River basin, which shows the buffering effect of a largerriver basin on hydrological processes. In river channels close to thePearl River Delta, the streamflow variations could be heavilyinfluenced by topographical changes of river channel due to sandmining or dredging. However, these changes are mainly observed atthe Makou and Sanshui station. Based on previous studies in thePearl River basin, the streamflow changes of the Gaoyao, Shijiao andBoluo stations are mainly the results of precipitation variations.Besides, the results of changes of precipitation maxima (Zhanget al., 2009b) corroborated by the abrupt behaviors of precipitationmaxima are in line with those of streamflow series considered inthis study. For example, the abnormally-high streamflow occurredin about 1984 was in agreement with the occurrence of abnormallyintense precipitation event in 1984. The hydrological behaviors ofthese three hydrological stations, i.e. Gaoyao, Shijiao and Boluo, arethe integrated consequences of human activities and climatechanges. It is hard to say exactly to what degree the human activ-ities and climate changes influence the hydrological variations ofthese three control stations representing hydrological processes ofthree major tributaries of the Pearl River basin. Even so, precipi-tation changes should play the key role in the shifts of streamflowfrom one statistical condition to another. This result will benefit thefurther study on the streamflow changes and also helps to makesound water resource management policy.

5. Conclusions

Thoroughly analysis was conducted of long streamflow series ofthree control hydrological stations of the Pearl River basin: Gaoyaostation of the West River, Shijiao station of the North River andBoluo station of the East River. Abrupt behaviors of the first and thesecond center moment were analyzed with robust statisticaltechniques, i.e. the scanning t-test and the scanning F-test. Possibleimplications of the changing properties of the hydrological

variations of the Pearl River basin with respect to water resourcemanagement of the Pearl River Delta were considered. The mostimportant conclusions are:

1) Specific time intervals characterized by increase or decrease ofaverage and standard deviations of the streamflow series wereidentified for the West River, North River and East Riverrespectively. Analysis results indicated that lower streamflowstability usually corresponds to higher streamflow average andvice versa. The first and the second center moment of thestreamflow variations of the river basin of smaller drainagearea are more sensitive to climate changes and/or humanactivities. This may be due to buffering functions of morestorage space of the longer river channels, and more compli-cated and longer runoff yield and concentration processes inthe river basin of larger drainage area.

2) In terms of monthly streamflow variations, after 2005, thestreamflow of the West River and East River decreased in termsof streamflow average. This is also true for the North River,however the abrupt changes are not significant at the >95%confidence level. Annual minimum streamflow of the PearlRiver basin tends to be increasing, and which is greatly helpfulfor better human mitigation of the salinity intrusion in dryseasons across the Pearl River Delta. Larger standard deviationof annual maximum streamflow when compared to that of theannual minimum streamflow implies an unfavorable conditionwith respect to flood mitigation in the Pearl River Delta.Extreme water events such as droughts or floods will havehigher probability of occurrence.

3) This study addressed abrupt behaviors of streamflow variationsof the Pearl River basin and discussed possible implications forthe water resource management of the Pearl River Delta. Theresults of this paper will be of considerable importance scien-tifically and practically for water resource management andsound human mitigation of water hazards across the PearlRiver Delta. Follow-up work will investigate possible correla-tions between precipitation, human activities such as waterreservoirs and other hydraulic facilities, and streamflow vari-ations of the Pearl River basin. This study also provides a goodcase study for the water resource management of other riverdeltas in the world under the influences of changing climateand intensifying human activities.

Acknowledgments

The work described in this paper was fully supported by a grantfrom the Research Grants Council of the Hong Kong SpecialAdministrative Region, China (Project No. CUHK405308), the ‘985Project’ (Grant No.: 37000-3171315), National Natural ScienceFoundation of China (Grant No.: 40701015; 50839005), and byProgram of Introducing Talents of Discipline to Universities – the111 Project of Hohai University. We are grateful to Prof. Jiang J.M. forhis constructive suggestions of the statistical techniques used inthis study. Last but not least, we would like to thank two anony-mous reviewers and also the editor, Norm Catto, for their profes-sional comments which greatly improved the quality of this paper.

References

Birsan, M.-V., Molnar, P., Burlando, P., Pfaundler, M., 2005. Streamflow trends inSwitzerland. Journal of Hydrology 314, 312–329.

Camilloni, I.A., Barros, V.R., 2003. Extreme discharge events in the Parana River andtheir climate forcing. Journal of Hydrology 278, 94–106.

Chen, X.H., Chen, Y.Q., 2002. Hydrologic change and its causes in the river networkof the Pearl River Delta. Acta Geographica Sinica 57 (4), 430–436 (in Chinese).

Y.D. Chen et al. / Quaternary International 226 (2010) 44–53 53

Chen, Y.D., Zhang, Q., Xu, C.-Y., Yang, T., 2008. Change-point alterations of extremewater levels and underlying causes in Pearl River Delta, China. River Researchand Application. doi:10.1002/rra.1212.

Chen, Y.D., Yang, T., Xu, C.-Y., Zhang, Q., Chen, X., Hao, Z.C., 2009. Hydrologicalalteration along the middle and upper East River (Dongjiang) basin, SouthChina: a visually enhanced mining on the results of RVA method.Stochastic Environmental Research Risk Assessment. doi:10.1007/s00477-008-0294-7.

Cramer, H., 1946. Mathematical Method of Statistics. Princeton University Press,Princeton, N.J., USA.

Hammersley, J.M., 1946. Discussion of papers. Journal, Royal Statistical Society 8, 91.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 thesouthwestern USA. Annales Geophysicae 25, 1–12.

Jiang, J.M., Mendelssohn, R., Schwing, F., Fraedrich, K., 2002. Coherency detection ofmultiscale significant changes in historic Nile flood levels. Geophysical ResearchLetters 29 (8), 112-1–112-4.

Kahya, E., Kalayci, S., 2004. Trend analysis of streamflow in Turkey. Journal ofHydrology 289, 128–144.

Loukas, A., Vasiliades, L., Dalezios, N.R., 2002. Potential climate change impacts onflood producing mechanisms in southern British Columbia, Canada using theCGCMA1 simulation results. Journal of Hydrology 259, 163–188.

Lund, R., Reeves, J., 2002. Detection of undocumented changepoints: a revision ofthe two-phase regression model. Journal of Climate 15, 2547–2554.

Lund, R., Seymour, L.B., Kafadar, K., 2001. Temperature trends in the United States.Environmetrics 12, 1–18.

Minville, M., Brissette, F., Leconte, R., 2008. Uncertainty of the impact of climatechange on the hydrology of a Nordic watershed. Journal of Hydrology 358,70–83.

Pearl River Water Resources Committee (PRWRC), 1991. Pearl River WaterResources Committee (PRWRC). In: The Zhujiang Archive, vol. 1. GuangdongScience and Technology Press, Guangzhou (in Chinese).

Pekarova, P., Miklanek, P., Pekar, J., 2003. Spatial and temporal runoff oscillationanalysis of the main rivers of the world during the 19th–20th centuries. Journalof Hydrology 274, 62–79.

Risbey, J.S., Entekhabi, D., 1996. Observed Sacramento Basin streamflow response toprecipitation and temperature changes and its relevance to climate impactstudies. Journal of Hydrology 184, 209–223.

Von Storch, H., Zwiers, F., 1999. Statistical Analysis in Climate Research. CambridgeUniversity Press, Cambridge, p. 116.

World Meteorological Organization (WMO), 1987. Water Resources and ClimateChange. Sensitivity of Water Resources Systems to Climate Change and Vari-ability. WMO/TO.

Xu, C.-Y., 2000. Modeling the effects of climate change on water resources in centralSweden. Water Resources Management 14, 177–189.

Xu, C.-Y., Singh,V.P.,2004. Review onregionalwaterresources assessment models understationary and changing climate. Water Resources Management 18, 591–612.

Zhang, Q., Xu, C.-Y., Becker, S., Jiang, T., 2006. Sediment and runoff changes in theYangtze River basin during past 50 years. Journal of Hydrology 331, 511–523.

Zhang, S.R., Lu, X.X., Higgitt, D.L., Chen, C.T., Han, J.T., Sun, H.G., 2008. Recentchanges of water discharge and sediment load in the Zhujiang (Pearl River)Basin, China. Global and Planetary Change 60, 365–380.

Zhang, Q., Xu, C.-Y., Zhang, Z.X., Chen, Y.D., Liu, C.-L., 2008a. Spatial and temporalvariability of precipitation maxima during 1960–2005 in the Yangtze Riverbasin and possible association with large-scale circulation. Journal of Hydrology353, 215–227.

Zhang, Q., Xu, C.-Y., Gemmer, M., Chen, Y.D., Liu, C.-L., 2008b. Changing properties ofprecipitation concentration in the Pearl River basin, China. Stochastic Environ-mental Research Risk Assessment. doi:10.1007/s00477-008-0225-7.

Zhang, Q., Chen, G.Y., Su, B.D., Disse, M., Jiang, T., Xu, C.-Y., 2008c. Periodicity ofsediment load and runoff in the Yangtze River basin and possible impacts ofclimatic changes and human activities. Hydrological Sciences-Journal-desSciences Hydrologiques 53 (2), 457–465.

Zhang, Q., Xu, C.-Y., Yang, Y., 2009a. Variability of water resource of the Yellow Riverbasin of past 50 years, China. Water Resources Management. doi:10.1007/s11269-008-9320-2.

Zhang, Q., Xu, C.-Y., Becker, S., Zhang, Z.X., Chen, Y.D., Coulibaly, M., 2009b. Trendsand abrupt changes of precipitation maxima in the Pearl River basin, China.Atmospheric Science Letters. doi:10.1002/asl.221.

Zhang, Q., Xu, C.-Y., Zhang, Z.X., 2009c. Observed changes of drought/wetnessepisodes in the Pearl River basin, China, using the standardized precipitationindex and aridity index. Theoretical and Applied Climatology. doi:10.1007/s00704-008-0095-4.

Zhang, Q., Xu, C.-Y., Sing, V.P., Yang, T., 2009d. Multiscale variability of sediment loadand streamflow of the lower Yangtze River basin: possible causes and impli-cations. Journal of Hydrology 368, 96–104.