Suspended sediment transport in a semiarid watershed, Wadi Abd - Ird

16 Pages·2007·2.36 MB·English

JournalofHydrology(2007)343,187–202
available at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/jhydrol
Suspended sediment transport in a semiarid
watershed, Wadi Abd, Algeria (1973–1995)
Mohamed Achite a, Sylvain Ouillon b,*
a LaboratoireEau-Environnement, Universite´ Hassiba Ben Bouali,Chlef, Algeria
bInstitut de Recherchepour leDe´veloppement, IRDNoumea, BPA5, 98848Noumea, New Caledonia
Received 28 July2006; received in revisedform21 June2007; accepted 25June 2007
KEYWORDS Summary A quantification of the fine sediment budget of a wadi (dryland river) in NW
Watererosion; Algeriaispresentedforaperiodof22hydrologicalyears(1973–1995).TheclimateisMed-
Suspendedsediment iterraneanovertheWadiAbdbasin(2480km2),themeanannualprecipitationis250mm
concentration; and the mean annual discharge is 1.0m3s(cid:2)1 at the gauging station. Regression relation-
Sediment transport; shipsbetweenwaterdischargeQandsuspendedsedimentconcentrationCarecalculated
Rating curve; from1432pairedmeasurementsintheWadiAbd,leadingtopower-lawequations ofthe
Wadi; type C=aQb. Thevariability of coefficients a andb, calculated for 138 floodsandflood
Intermittent river; stages, is analyzed. The median value of b is 0.757, indicating that C is almost propor-
Algeria tionaltoQ3/4.Giventhatthe(a,b)pairsarecorrectlyaligned(r2=0.578),thecoefficients
a and b are not independent. Regression relationships between daily Q and daily sus-
pended sediment concentration and discharge Q are calculated from 702 input data.
s
The performances of these regression relationships are shown to be equivalent, leading
toover-estimationsof20–25%ofthesuspendedsedimentflux.Thenon-biasedC–Qsed-
iment rating curve is used to extrapolate a time series of C measurements, and thus to
analyze the long-term patterns in suspended sediment transport. Average sediment
wash-down (136tkm(cid:2)2yr(cid:2)1) is similar to the mean global value. The ratio of sediment
wash-downtotheriverwaterdischargeis10.7·106tkm3,20timesgreaterthantheaver-
age ratio in the Earth’s eastern hemisphere, and illustrates the highly erosive power of
wadis. Variability is shown to be significant at the seasonal scale (CV=89%) and higher
at the interannual scale (CV=139%). The fine sediment flux mainly occurs in autumn
(48.4%)andspring(32.7%).Althoughprecipitationdecreased,itwasmoreirregularfrom
oneyeartoanotherovertheperiod1985–1995thanduringtheperiod1973–1985,andthe
Wadi Abd, which was a perennial river, became intermittent in the late 1980s. This
increasingirregularityisaccompaniedby:anamplificationofthevariationsofdischarge,
anincreaseintheaveragedischargeofapproximately20%duringthesecondperiod,and
a higher and more irregular suspended particulate flux. The mean annual suspended
* Correspondingauthor.Tel.:[phone];fax:[phone].
E-mailaddresses:[email protected](M.Achite),[email protected](S.Ouillon).
0022-1694/$-seefrontmatter ª2007ElsevierB.V.Allrightsreserved.
doi:10.1016/j.jhydrol.2007.06.026
188 M. Achite,S.Ouillon
sedimentyieldisshowntobehighlycorrelatedwiththestandarddeviationofmeandaily
discharge calculated per year (r2=0.989). The highly significant interannual variability
pointstothedifficultyofdefiningasuitableperiodtocalculateareferencevalueforsed-
iment budgets. It also emphasizes the absolute necessity of continuing a series of mea-
surementsoverlongertimeperiodstostudyfluctuationsinthecontextofclimatechange.
ª2007 Elsevier B.V.All rights reserved.
Introduction and the hydraulic potential by 2–5% per year (Kassoul
et al., 1997). A correct estimation of the sediment volume
The volume and types of particles eroded and transported carried by a river is therefore necessary for an adapted
by the rivers exhibits great geographical and temporal var- water resources management strategy. A greater working
iability. The stakes associated with erosion and particulate knowledge of long-term variations of sediment loads in a
transport are numerous. These include impacts on ground variety of rivers is also necessary for an assessment of
fertility, transfer, storage and fate of nutrients and con- the current trends in global land-ocean sediment transfer
taminants, changes in water quality trends, aquatic habi- in the context of climate change (Walling and Fang,
tats, silting up of channels, reservoirs and harbours, and 2003). Sediment transport is well documented in perennial
reductions in hydroelectric equipment longevity (Williams, rivers in temperate or humid climates but is far less stud-
1989; Ouillon, 1998; Horowitz, 2003). In the Maghreb ied in dryland streams, also called wadis or oueds, despite
(North Africa), problems essentially relate to the silting their known high transport efficiency (Reid and Laronne,
up of the dams, which reduces the water storage capacity 1995).
Table1 Targetedpapersrelevanttosedimenttransportinwadis
Reference Riverbasinorregion Ephemeralor Durationof Suspensionand/ Maincontribution
intermittent monitoring orbedload
Schick(1977) Tunisia,Peru,Nile Both Review
SutherlandandBryan(1990) Katiorin,Kenya Ephemeral 1year Suspension Particle-sizeanalysis
ofsuspended
sediment
Laronneetal.(1994), Yatir,Negev Ephemeral Flashfloods Bedload Quantificationof
ReidandLaronne(1995), bedload,armouring
Powelletal.(2001) andbedloadgrainsize
distribution
ReidandFrostick(1997) Numerous Ephemeral Review
Clappetal.(2000) Yael,Israel Ephemeral 33years Both Long-termsediment
generationrates
versussedimentyield
Terfousetal.(2001) Mouilah,NWAlgeria Intermittent 16years Suspension Temporalvariability
ofsedimenttransport
Toua¨ıbiaetal.(2001) Haddad,Algeria Intermittent 22years Suspension Temporalvariability
ofsedimenttransport
Mina,Algeria
AchiteandMeddi(2004) Haddad,Algeria Intermittent 22years Suspension Temporalvariability
ofsedimenttransport
Megnounifetal.(2003) Sebdou,NWAlgeria 5years Temporalvariability
ofsedimenttransport
Alexandrovetal.(2003a) Eshtemoa,Negev Ephemeral 6years Suspension Temporalvariability
ofsedimenttransport
BenkhaledandRemini(2003) Wahrane,Algeria Intermittent 13floods Suspension C–Qratingcurves
Amosetal.(2004) Burdekin,Australia Intermittent 1flood Both Detailedsediment
(15days) transportduringone
flood
CohenandLaronne(2005) Rahaf-Qanna’im,Israel Ephemeral Severalfloods Both Sedimenttransport
duringhighdischarge
events
Alexandrovetal.(2007) Eshtemoa,Negev Ephemeral Suspension Relationshipbetween
C–Qcurveandrainfall
type
Suspendedsediment transport in asemiarid watershed, Wadi Abd,Algeria (1973–1995) 189
Few studies have been published on sediment transport quick events is often difficult (Schick, 1977; Reid and Lar-
in rivers in semiarid zones. Probst and Amiotte-Suchet onne, 1995; Serrat et al., 2001). For long observation peri-
(1992), in their review of suspended sediment transported ods (>15 years), average annual suspended sediment yield
by wadis in the Maghreb, underline the lack of available is equal to the arithmetic mean value of observed annual
dataforsuchrivertypes.Otherreviewshavealsobeenpub- suspended sediment yield.
lishedbySchick(1977),andAlhamidandReid(2002) anda In this paper, regression relationships are built between
briefdescriptionoftheseandotherpapersisgiveninTable instantaneous and daily values of concentration and water
1. Nevertheless, data from these systems is still fragmen- discharge, and also between daily suspended sediment dis-
tary and further study of both the quantification of solid chargeandwaterdischargefromdatacollectedatthegaug-
transport in semiarid basins and itsvariability is clearly re- ingstation ofAinHamaraintheWadi Abdduringa22-year
quired. Therefore, the purpose of this paper is to quantify period(1973–1995).Theseregressionsarediscussedwitha
suspended sediment transport in a semiarid catchment primary focus on the coefficients involved in the sediment
area,theWadiAbdbasininAlgeria,andtoanalyzeitstem- ratingcurveandtheirtemporalvariability.Asecondarydis-
poralvariabilityovera22-yearperiod.WadiAbdisatribu- cussion focuses on the comparison of statistical perfor-
tary of the Wadi Mina and contributes to the silting of the mances of C–Q and Q–Q relationships. The regressions
s
Sidi M’hamed Benouada dam. arecomparedtothosefromotherrivers,andarethenused
For the gauged sites, suspended sediment yield is com- to (a) obtain a long-term estimate of suspended sediment
puted from rating curves established from long-term mea- discharge from daily water discharge values at the gauging
surement series (Walling, 1977; Ferguson, 1986; Jansson, station during the 1973–1995 period; and (b), analyze the
1996; Asselman, 2000; Achite, 2002; Horowitz, 2003). In variabilityofthemeansuspendedsedimentfluxatseasonal
ungauged rivers, suspended sediment yield may be com- andinterannual scalesovera22-year period.
puted from models where basic physiographic factors (cli-
mate, soils and top cover) are taken into account Study area: the Wadi Abd
(Poliakov,1935;Lopatin,1952;LisitisynaandAleksandrova,
1972;Merrittetal.,2003).IntheMediterraneanregion,the General information
quantification of sediment transport by a river is rendered
difficult by the temporal variability of water flow. In these
The Wadi Abd basin is located in the north western part of
semiarid environments, most of the river discharge occurs Algeria, draining an area of 2480km2 (Fig. 1). It is located
during flash floods, and obtaining samples during these
between 34(cid:2)400 N and 35(cid:2)250 N, and 0(cid:2)200 E and 1(cid:2)100 E.
Figure1 TheWadiAbdcatchmentarea.AinHamaragaugingstationisstation3.
190 M. Achite,S.Ouillon
45 2.0
Table2 MaincharacteristicsoftheWadiAbdbasin
Parameters Unit Value )mm 40 PWraetceipr idtaisticohnarge 1.6 1-s)
BMMMaeiansxaiiinnmmuaeurmlmeevaeaeltleeiovvanattiioonn mmmkm2 21728314831890 (notipatiecipry 3200 10..28 3m(egrhacdsi
MLeenagnthsloopfethgeramdaieinntwadi %km 01.1488 htlnoM 10 0.4 ertaW
Drainagedensity kmkm(cid:2)2 3.70
0 0.0
S O N D J F M A M J J A
Month
The main physiographic characteristics are reported in Ta- Figure2 Variabilityofmeanmonthlyprecipitationandwater
ble 2. The Wadi Abd is a tributary of the Wadi Mina; the discharge. (precipitation in S11, water discharge in 3, see
Wadi Mina is a main tributary of the Wadi Chelif, Algeria’s locationsinFig.1).
principal river. The basin can be divided into two major
zones:mountains(MountsofSaidaintheSouthandMounts
ofFrenda, Tiaret, inthe North), andthe plain ofMina. 350 2.5
TheWadiAbdhasbeenselectedforthisstudybecauseof
300
the availability of rainfall and hydrometric records. The m) 2.0 1-s)
availablehydrologicaldataconsistsofthemeandailywater m( 250 3m
cd(khisgacrmhgae(cid:2)r3g)(e.m(T3mhs3(cid:2)e1s)(cid:2)d1a)at,nadcohisnauscsipdbeeennedtnemdpersaoesvudiridmeemednetnbtycsotonhfcewenaNttareatritoidoninass-l notiatipciper 210500 11..05 (egrhcadsi
Agencyof Hydraulic Resources (ANRH,Alger). yrlea 15000 Precipitation 0.5 ertWa
Y
Geology, vegetation and topography Water discharge
0 0
73/74 75/76 77/78 79/80 81/82 83/84 85/86 87/88 89/90 91/92 93/94
TheWadiAbdcatchmentareaispredominantlymadeupof
UpperJurassic(marl-limestone,45.9%ofthesurface),large Figure3 Variabilityofannualprecipitationandmeanannual
zonesof MiddleJurassic (calcareous andkarsticdolomites, waterdischarge.(precipitationinS11,waterdischargein3,see
20.2%)andPliocene (7.4%) (Achite, 1999). locationsinFig.1)Horizontallinesareaveragevalues.
The Jurassic zone of the Southern basin is eroded and
nearly 50% of its surface is covered by a varying density of
vegetation.Theforestscover5.8%ofthiszone,mainlywith (Fig.3).Thisdemonstratesanamplificationofthemeanan-
young plantations of Aleppo pines (Pinus halepensis). Two nualdischargecomparedtoannualprecipitation.Forexam-
formsofscrubcover32%ofthecatchmentarea:scrubwith ple, although annual rainfall was approximately 20% higher
PistaciaandOlea,andscrubwithTetraclinis.Inadditionto than average, mean yearly discharge varied between +61%
the natural vegetation, the herbaceous vegetation mainly (1986–87) and +112% (1994–95). In contrast, lower than
consistsofcerealcropswhichformtheprincipalpermanent average annual rainfall was accompanied by a highly dis-
cultures(Kouri, 1993;Mahieddine, 1997). persedmean annualdischarge(see Fig. 3).
Thebasinreliefexhibitsslopesbetween0.12%and0.25% Hydroclimatic fluctuations affected the basin producing
on the Jurassic mountains (Achite, 1999). Water erosion is a wetter period from 1973 to 1980 and a drier period from
mostly intensive in the northern and western mountainous 1981to1993(except1986and1990)(Fig.3).Itisofinterest
partsof theWadi Abdbasin. toplacethisvariabilityinthecontextofalongertimescale
and at a regional scale. An extensive comparison between
Climate, rainfall and runoff the study area in North Africa and Central and West Africa
during the second half of the 20th century shows that this
TheclimateisMediterraneanandischaracterizedbyawet hydrological variability is consistent. Several researchers
anddryseason.TherainyseasonisfromOctobertoMarch, (e.g. Laraque et al., 2001; L’Hote et al., 2002) show that
and the dry season lasts from April to August/September discontinuities occurred on a 10-year scale since the
(Fig.2).Annualprecipitationishighlyirregular,varyingbe- 1950s. There was a phase of surplus river discharge in the
tween 174mmyr(cid:2)1 and 303mmyr(cid:2)1 (Fig. 3). Mean annual 1960s in Central Africa, a discontinuity at the beginning of
precipitationis250mm,withacoefficientofvariation(de- the 1970s at the start of a period of dryness in the Sahel
notedhereafterCVanddefinedastheratioofthestandard and lower discharge in Central Africa. This was followed
deviation to the mean) of 17.6% between 1973–74 and by a second discontinuity, initiating a phase of further re-
1994–95.MeanannualdischargeattheAinHamaragauging duced discharge at the beginning of the 1980s in Central
station (see location in Fig. 1 and data in Fig. 3) was Africa,whilethedrynesscontinuedintheSahel.IntheWadi
1.0m3s(cid:2)1forthesameperiodofobservation.Theinteran- Abdbasin,thereductioninannualprecipitationfrom1981–
nualvariability,characterizedbyaCVof41%duringthe22- 82isclearcomparedtothepreviousyears:averageprecip-
year period, was higher than that of yearly precipitation itationwas287mmyr(cid:2)1overtheperiod1973–74/1980–81
Suspendedsediment transport in asemiarid watershed, Wadi Abd,Algeria (1973–1995) 191
and 225mmyr(cid:2)1 over the period 1981–82/1993–94. 1994 dry phase of varying duration. Rivers that flow less than
has been singled out as an anomalous year in the Sahel 20%oftheyearcanbeconsideredasephemeral,andthose
(L’Hote et al., 2002) and in the Wadi Abd basin increases that flow between 20% and 80% as intermittent (Matthews,
inprecipitationandflowweremeasuredduringtheautumn 1998).
of 1994 andspring1995 (Fig.3). Theflow regime of theWadi Abdhas been studied from
the daily river discharge at Ain Hamara station. From 1973
to 1985, water was always present in the Wadi except for
Hydrological and suspended sediment data
12 daysin 1979–1980 and1day in 1981–1982. There were
51 days without any water in the Wadi in 1985–86, 0 in
Suspended sediment concentrations (hereafter denoted C)
1986–87, 48 in 1987–88, 50 in 1988–89, 91 in 1989–90,
weremeasuredintheWadiAbdatAinHamarastationdur-
102 in 1990–91, 60 in 1991–92, 109 in 1992–93, 117 in
ing different hydrological conditions (low discharge to se-
1993–94 and96daysin 1994–95.
vere flood events). During each sampling, the
The Wadi Abd was thus a perennial river until the late
instantaneous water discharge (hereafter denoted Q, ex-
1980’s and thereafter became an intermittent river, with
pressed in m3s(cid:2)1)wasalso measured.
anaverageof72.4daysover10yearswithnoflowperyear
Atthegaugingstation,theriverflowsthroughinasingle
(i.e.20%oftime).Thenumberofdayswithnoflowwentup
46m wide channel. The flow is generally measured with a
and reached 96.8 days per year (26% of the year) over the
winchbygaugingasectionover5–8verticalswithbetween
last5consideredyears,i.e.from1990to95.Theheadwater
2 and 6 measurements per vertical. At night, during holi-
tributaries ofthe Wadi Abdshowthe same water regime.
days,orduringsomefloods,thedischargewasderivedfrom
Independentlyofthemainstudy,whichfocusesonsedi-
alimnimetricheightusingalocalabacus.Duringthe22con-
mentfluxes,theobservationofachangingflowregimedur-
sidered hydrological years, the Wadi Abd did not overflow
ing the 1980s is an important result that needs to be
the floodplain at thegauging station.
emphasizedinthecontextofclimatechange.Thepossibil-
Todetermine sedimentconcentrations, water wasman-
ity that these observations were not artificially introduced
uallysampledusingasimple0.5Lor1Ldipsample.Oneor
bychangesinsamplingormeasurementmethodswasexam-
twosamplesaretakenbymeasurementinthemiddleofthe
ined.Waterflowregimeresultsfromthreeinter-dependent
wadi and/or at its edge. Suspended matter concentrations
factors:intrinsicnatural variability,anthropogenic impacts
aredeterminedbyfiltrationofthesamplesonpre-weighted
on the catchment (such as changes in agriculture, vegeta-
fiberglass filters (0.45lm, Whatman GF/F), then dried and
tion cover, irrigation), and climate change. In the Wadi
weighed. The average concentration of the section is as-
Abd basin, the changes in water flow regime were due to
sumed to be the concentration. Other available long-term
a very long period of drought beginning in the early 1980s
data for the Algerian wadis were collected according to
whichresultedfromclimatechangeoralong-termintrinsic
thesameprotocol(Toua¨ıbiaetal.,2001;BenkhaledandRe- natural variability.
mini, 2003; Megnounif et al., 2003). When the discharge
flow exhibited little variation, only one sampling was per-
Models and methods
formed per day at noon. The sampling rate arbitrarily in-
creased with the flow until reaching a sampling frequency
every 15mnor30mnatthe peakof aflood. Suspendedsedimentdischarge,whichisequaltoCtimesQ,
In our database (1973–1995, 1432 paired C and Q mea- is hereafter denoted Qs and is expressed in kgs(cid:2)1. Most
surements), water discharge varied from 0.09m3s(cid:2)1 to quantitative studies of suspended sediment load are based
352m3s(cid:2)1 and C varied from 0.14kgm(cid:2)3 to 118.5kgm(cid:2)3. onanempiricalrelationshipbetweenCandQafterlogarith-
Instantaneous flow was continuously recorded at Ain Ha- mic transformation of the data (Ferguson, 1987; Hasnain,
mara gauging station during the period 1973–1995. The 1996) or between Qs and Q (Restrepo and Kjerfve, 2000),
durationofrunoffeventsintheWadiAbdatAinHamarasta- leadingto equations of the type:
tion, which last at least 10 days, enabled us to study the C¼aQb ð1:aÞ
hydrology fromdaily discharges.
or
The size distribution of suspended particles demon-
strates a predominance of silts and clays. The mean class Q ¼a0Qb0 ð1:bÞ
s s
of grain diameter ranges from 4.35 to 5.82lm (very fine
Thea0-coefficientshouldbeequaltoaandthedifferencein
silt).Clayrepresentsmorethan20%ofthetotalpopulation
the b and b0 coefficients should be 1.0 if the mean of the
of particles. Sand ispresent in smallquantities (<5%).
productQ =C*Qisequaltotheproductofthemeans.How-
s
ever,thisisoftennotrigorouslyverifiedasCandQarenot
Flow regime of the Wadi Abd completely independent.
Here,twocomplementaryapproachesweresuccessively
Dryland rivers (i.e. the rivers of arid and semiarid regions, adopted. In a first part (Section ‘‘Instantaneous suspended
Davies et al., 1994) can be perennial, intermittent or concentrationversusriverdischarge’’),weanalyzethevar-
ephemeral (see terminology used in ecological studies in iabilityofcoefficientsaandbfloodbyfloodfromcoincident
Uys and O’Keeffe, 1997). Ephemeral rivers are defined as measurementsofCandQ.Theresultingvaluesmakeitpos-
those that run for short periods after rain has fallen high sibletoanalyzethevariabilityofthesedimentratingcurve
in their catchment area (Day, 1990). An intermittent river of Wadi Abd for 138 complete or partial floods. They also
is defined as having relatively regular, seasonally intermit- makeitpossibletocomparetheWadiAbd’sbehaviourwith
tent discharge, i.e. a river that experiences a recurrent other rivers.
192 M. Achite,S.Ouillon
In a second part (Section ‘‘Daily suspension flux versus 3
river discharge’’), as multiple pairs of (C, Q) values have
b = - 0.311 Ln a + 1.066
been measured for some days, the C, Q and Q values for r = 0.578
s 2
thesedaysareaveraged.Our1432pairedCandQmeasure- n = 138
amasgeen‘s‘t,mspewearenrdeadypa,eilrryefoscuromltneicnden7f0tor2ra7tvi0ao2lnu’ed’sa,yo‘sf‘.mCT,ehQaeniarndrdeasiQlpysercwetfaivetererraevddeitsro-- tneiciffe 1 AAWgveegarrakeegsrse it vhfleoa onfld oasov d(ensr=a (1gn0e=4 1f)l8o)ods (n=16)
oc 0
charge’’ and ‘‘mean daily solid discharge’’, respectively. - b
‘‘Mean’’ isdefined by:
-1
1 Xp
meanðxÞ¼x(cid:2)¼ x ð2Þ
p i -2
i¼1 0.1 1 10 100 1000 10000
Notethattheseaveragesarebasedoncoincidentavailable a - coefficient
CandQdata,andnotonatrueaveragevalueresultingfrom
Figure 4 Correlation between slope/intercept values of
continuous measurement throughout the day, as was the
suspended sediment rating curves fitted using power-law
case for the daily discharge. Regression relationships are
regressionrelationships.
built from mean daily values which are then compared
and discussed. The evaluation of the relationships is based
onthefollowing criteria: stdev isdefined by: Q) data are considered to form a single ‘‘stage’’ if they
are significantly correlated (r2>0.35, 36 flood stages) or
" #1=2
stdevðxÞ¼ 1 Xn ðx (cid:2)x(cid:2)Þ2 ð3Þ have been taken within a short duration (with n>6 and a
n(cid:2)1 i maximum of a few hours duration when flood duration is
i¼1
on average of ten days, 4 flood stages). We thus obtain
Fromtheseequations,themeannormalizedbias(MNB)and 138(a,b)pairscalculatedfrom,onaverage,8.2(C,Q)pairs
the normalized root mean square (rms) error, in percent, (minimum:3,median:7)withanaverager2of0.747,anda
arecalculated following: medianof 0.792.
(cid:2)y (cid:2)y (cid:3) The 138 values obtained are reported in Fig. 4. All epi-
MNB¼mean calc obs (cid:3) 100 ð4:aÞ sodes considered, the median value of (a, b) pairs is
y
obs (2.154, 0.757). The relative variation calculated with the
(cid:2)y (cid:2)y (cid:3)
rms¼stdev calc obs (cid:3) 100 ð4:bÞ whole of the instantaneous values (2.180, 0.626) is 1.2%
yobs for a and 20.9% for b. The value of b indicates that, from
where y is a variable calculated using a regression rela- a global point of view, C is almost proportional to Q3/4 in
calc
tionshipandy isthevaluemeasuredinsitu(forC)ordi- theWadi Abdduring the flood episodes.
obs
rectly derived from in situ measurements (for Q =C·Q). The(a,b)pairscalculatedfordifferentfloodsaresome-
MNBisanindicatorofsystematicerrorandrmsansindicator what correctly aligned (r2=0.578, n=138, see Fig. 4); the
ofrandom error. coefficients a and b are thus not independent. In the Wadi
Abd,aand bareconnected by theregression relationship:
Sediment rating curves b¼(cid:2)0:311lnaþ1:066 ð5Þ
Separating out the most aggressive floods from the least
Instantaneous suspended concentration versus aggressive floods, we can study their specific behaviour
river discharge and determine their seasonal variation and/or criteria of
occurrence. A flood which, for a given coefficient a (resp.
The 1432 instantaneous (C, Q) values lead to a regression b), has a coefficient b (resp. a) appreciably higher than an
type Eq. (1.a) where a=2.180, b=0.626, r2=0.445. The averageflood,carriesahighersedimentloadthananaver-
regression coefficient is low because the complete data age flood; this is referred to as an ‘‘aggressive’’ flood. On
set includes many (C, Q) pairs for very dispersed, very low thecontrary,afloodwhich,foragivencoefficienta(resp.
flowsandwhosesignificanceisnotimportantforthisstudy b) has a coefficient b (resp. a) appreciably lower than an
since the suspended sediment flux occurs primarily during average flood, carries a lower sediment load than an aver-
floods. age flood and is now referred to as a ‘‘weaker than aver-
Tofocusouranalysisonthemajortransportepisodes,we age’’ flood. A flood is considered as ‘‘aggressive’’ if its
havecalculatedthespecificregressionrelationshipforeach couple (a, b) fulfils the criterion: b+0.311 ln a(cid:2)1.066
flood. For floods where sufficient data are available, we >0.4,whilea‘‘weakerthanaverage’’floodfulfilsthecrite-
haveplottedallavailable(C,Q)datafloodbyfloodandhave rion:b+0.311lna(cid:2)1.066<(cid:2)0.4.Thecoefficient0.4isan
examined their evolution from beginning-to-end of each arbitrary choice; for example, floods #1 and #2 with
flood.For98floods,oneratingcurvewassufficienttorep- b +0.311 ln a =b +0.311 ln a +0.4 are such that b =
2 2 1 1 2
resentthevariationsin(C,Q),forfloodsthatdidnotpres- b +0.4ifa =a ,ora =3.6a ifb =b .Accordingtothese
1 1 2 2 1 1 2
ent any hysteresis (r2>0.35, median r2=0.777). For 13 arbitrary criteria, the 138 floods and flood stages are di-
floods, it was necessary to define distinct phases (40 on videdinto104‘‘average’’,18‘‘aggressive’’and16‘‘weaker
the whole) with separate (a, b) pairs: rising, temporary than average’’ floods(see Fig.4).
stagesifany(e.g.beginningoffallingbeforeanewrising), The rating parameters characteristic of each type of
falling. Todistinguish different flood stages, successive (C, floodareshowninTable3.Itisparticularlynoticeablethat
Suspendedsediment transport in asemiarid watershed, Wadi Abd,Algeria (1973–1995) 193
Table3 ParametersofthesuspendedloadversusdischargeregressionrelationshipforthefloodsorfloodstagesoftheWadiAbd
N Medianregression bversusa
coefficients
a b Regressionrelationship r2
Allofthefloods 138 2.154 0.757 b=(cid:2)0.311lna+1.066 0.578
‘‘Average’’floods 104 2.307 0.736 b=(cid:2)0.297lna+1.001 0.829
‘‘Aggressive’’floods 18 3.408 1.568 b=(cid:2)0.320lna+1.842 0.746
‘‘Weakerthanaverage’’floods 16 1.162 0.531 b=(cid:2)0.376lna+0.584 0.970
a3- )m. gk( C suo1186020000.... C =R 1 =1 1.02.887 Q2 17-19 Oct 78 3-b)m.gk( C su 64570000.... C =R 0 =.31 05.589 Q5 1-5 Mar 79
enatnatsnI 40. C = 302.96 Q C =R 2 =.9 401.326 Q0 oenatnatsnI 2300.. C =R 1 =6 230..016 Q6
20. R = 0.624 3 10. 3
C = 10.79 Q C = 0.838 Q
0. R = 0.877 0. R = 0.931
0. 5. 10. 15. 20. 25. 0. 10. 20. 30. 40. 50. 60.
Instantaneous Q (m3.s- 1 ) Instantaneous Q (m3.s-1)
Figure5 Suspendedsedimentratingcurvesfittedusingpower-lawregressionrelationshipsduringdifferentphasesofaflood.(a)
17th–19thOctober1978;(b)1st–5thMarch1979.Distinctstagesarechronologicallynumberedwithincircles.
the (a, b) pairs of aggressive floods are dispersed
(r2=0.746, n=18), while the (a, b) pairs of ‘‘weaker than 2
average’’ floods are remarkably linear (r2=0.970, n=16). 17-19 Oct 78
During aggressive floods, C is almost proportional to Q3/2 15-17 Feb 79
sinceb=1.568,whilefortheleastaggressivefloods,Cisal- 1-5 Mar 79
mostproportionaltoQ1/2(b=0.531).Averagefloodspracti- median (all years)
1 median (1978-79)
callybehavelikethemedianfloodcalculatedforthewhole
dataset,withCalmostproportionaltoQ3/4.Theremarkable nte
ci
linearityofthe(a,b)pairsforthe‘‘weakerthanaverage’’ ffie
o
floodsseemstoindicatethataminimalthresholdoferosion c
-
of thecatchment area exists. b
0
How do the phases of a flood follow one another in the
WadiAbdandisthereaseasonalvariationofthesephases?
To illustrate this question, Fig. 5 shows distinct stages of
twofloods.Amulti-peakfloodispresentedinFig.5a:stage
1,decreaseofflowon17thOct.1978from7.30to10h(the -1
riseduringthepreviousnightwasnotsampled);stage2,rise 0.1 1 10 100 1000
until 11.10h; stage 3, decrease until 14.15h. After an in- a - coefficient
creaseofflowwhichwasnotsampled,stage4corresponds
Figure6 Pairsof(a,b)valuesforthemainfloodsoftheWadi
tothefinalfallingwhichspreadoutbetween17thOct.17h
Abdin1978–79.
and 19th Oct. 15h. The flood presented in Fig. 5b is less
complex and shows a clockwise hysteresis: a rise of flow
(stage 1, from 1st March 1979 16h to 2nd March 12.30h), year and also to the median calculated over 22 years
a stage of high flow (stage 2, from 2nd March 13h to 21h, (1973–1995). The (a, b) pairs for the flood of October 78
isoneofthe4stagesofshortdurationandofpoorcorrela- (Fig. 5a) are as follows: stage 1 is ‘‘aggressive’’, stages 2
tionwithintheglobaldataset;theotherphasesinFig.5are to 4 are ‘‘average’’; for the flood of March 79 (Fig. 5b),
suchasr2>0.35), thenthe falling(stage3from2ndMarch stages1and2are‘‘average’’,stage3is‘‘weakerthanaver-
22h to 5thMarch 12h). age’’ (Fig. 6). The flood of March 1979 (Fig. 5b) shows a
InFig.6,the(a,b)pairsofthemainfloodsof1978–1979 clockwise hysteresis, with a higher particulate load during
are compared to the median pair of the same hydrologic the risingstage than during thefalling stage, at an equiva-
194 M. Achite,S.Ouillon
lent flow (see Williams, 1989 for a general typology of the ofrecentfloods,seasonalvariationsofvegetationcoverand
relationship between concentration and discharge during other parameters.
hydrologicevents).BenkhaledandRemini(2003)haveana- The best-fit regression relationship between mean daily
lyzed the hysteresis episodes of C–Q relationships on an- sediment discharge and mean daily water discharge is a
other Algerian basin, the Wadi Wahrane basin, for 13 power-law relationship of the type of Eq. (1.b) given by
floods. They show a significant prevalence of hysteresis of (r2=0.826, n=702):
clockwisetype(8outof13).Asystematicanalysisoftheen-
Q ¼1:712 Q1:769 ð7Þ
tire data for the Wadi Abd was not conducted during this s
study.However,wehaveobservedthatthepair(a,b)dur- Theaanda0 coefficientsinEqs.(6)and(7)differby0.11%,
ingaflood(includingseveralfallingphasesasinOct.1978) and b0=b+1.019. To estimate Q from the entire Q data
s
isgenerallytranslatedinana–bdiagrambyashifteitherto available, one can apply Eq. (7) directly, or determine C
theleft(adecreases,asinOct.1978,seeFigs.5aand6),or using Eq. (6) then multiply the result by Q. In the relevant
towards the bottom (b decreases, as in March 1979, see literature, the second alternative is used more frequently
Figs. 5b and 6). This reduction in a or b probably indicates than the first because the coefficient of correlation of Q–
s
that thesuspended sedimentconcentration (resulting from QissuperiortothatofC–Q.Wehaveappliedthetwometh-
erosive power and sediment supply) of the Wadi Abd de- odsto our data set.
creases at the end of the flood. The successive positions Thefirstapplicationrelatestothedatasetbeingusedto
ofthe pair (a,b)onthe a–bdiagram also makeit possible establishtheregressionrelationships.TheQ–Qrelationship
s
to visually compare between floods. As an example, Fig. 6 provides Q values which are on average 48.5% higher than
s
showsaleft-shiftofthespringfloodascomparedtotheau- the measurement derived Q values (MNB=48.5%,
s
tumn flood, demonstrating that the autumn flood is more rms=133%). The C–Q relationship provides C values that
aggressive. These examples observed in 1978–1979 illus- are on average 48.1% higher than the C measured values
trate the tendency generally noted on Wadi Abd: the first (rms=132%) which, multiplied by Q, provide values of Q
s
stages of a flood are more ‘‘aggressive’’ than the last, and on average 45.7% higher than Q values derived from mea-
s
thefloodsofautumnaremore‘‘aggressive’’thatthefloods surements. Statistical comparison between 702 measured
ofspring. Q andC–QderivedvaluesofQ isbasedonlog-transformed
s s
We have tested the predictive capacity of the median data to encompass the several-orders-of-magnitude varia-
pairs of coefficients (a, b) obtained for the 3 flood types tion in Q. Statistics such as regression slope (1.011) and
s
(average, aggressive, weaker than average) to estimate intercept ((cid:2)0.0022), coefficient of determination (r2=
the instantaneous concentration from instantaneous dis- 0.827) and rms error (100%) provide a numerical index of
charge, if the type of flood is known. In the majority of modelperformancewhichcanbecomparedtothoseofother
‘‘aggressive’’ and ‘‘weaker than average’’ floods, the modelsandotherrivers(suchstatisticsarecommonlyusedto
choiceofanadaptedpairdecreasestheerrorinthecalcu- comparebio-opticalalgorithms(seetheexampleofO’Reilly
lation of C compared to the use of the total median pair. et al., 1998)). The difference between predicted and ob-
However,noneofthethreemedianpairsisadaptedtogive servedvaluesiscalculatedontheentiredatasetand,there-
acorrectestimationofCforfloodstagessuchasb<0.Neg- fore,doesnottakeintoaccounttheabsolutevaluesofQ.In
s
ative values of b correspond to a particular case where C fact,themostimportantaspectinthereconstitutionofthe
and Q evolve conversely, with C increasing while Q de- soliddischargeratesrelatestotheepisodesofhighestflow
creases and vice versa. None of the median pairs (all of whose contribution is essential both to estimation of total
which are b>0) can reproduce this generally transitory fluxofriver-drivenparticlesandfortheanalysisofitsvari-
behaviourwhichoccursduringanintermediatephaseinbe- abilityonseasonalandinterannualscales.Whenthecumu-
tween the rise and fall of the flood (see two examples in latedvaluesofsuspendedloadforthe702considereddays
Fig.5aandb).Ifonewantstodeterminepairsof(a,b)coef- are compared using the two methods of reconstitution (by
ficientsbyfloodtype,itisnecessarytorestrictthethreeal- summing the predicted mean daily solid discharges), an
ready defined types of floods by b>0 and to add a fourth over-estimationof25.7%isobtainedcomparedtothe‘‘mea-
type corresponding to the transitional stages where C and sured’’ daily suspended loads by the use of the Q–Q rela-
s
Qareinnegativecorrelation.Thecalculationofthemedian tionship,andof21.8%bytheuseoftheC–Qrelationship.
coupleforthe11 phasesofsuchfloodsoutof 138provides Several lessons canbe learnedfrom theseresults:
a=84.72andb=(cid:2)0.48.
1. Sincetheaveragerelativeerror((cid:4)48%)ishigherthanthe
Daily suspension flux versus river discharge absolute error of the cumulated solid discharge ((cid:4)25%),
therelativeerrorsareweakerduringhighflowsthandur-
The regression relationship between the mean daily C and ing low ones. This is easily explained by the increased
themeandaily Q is(r2=0.463,n=702): scattering of the concentrations during low discharge
relative to those during high discharge, and because
C¼1:710 Q0:750 ð6Þ
the episodes of high discharge, which are fewer, force
where C is expressed in gL(cid:2)1 and Q in m3s(cid:2)1. C is depen- the regression relationship more than do the points of
dent on two main factors: water discharge and supply. lowdischarge.
The r2 value means that the variations of C are explained 2. Withourdataset,thereconstitutedtimeseriesaremore
byupto46%bythoseofQ.Morethan50%ofthevariability likelytoover-estimatethelowsoliddischargesthanthe
inCisexplainedbythesupplyofsedimentstotheriversys- highones.Thiswill decreasethevariation ofsuspended
tem,withsupplydeterminedbytheoccurrenceorabsence loadofseasonal or interannualorigin.
Suspendedsediment transport in asemiarid watershed, Wadi Abd,Algeria (1973–1995) 195
3. An over-estimate of about 25% in the calculation of the loadoftheriver.Theirexplanationisthateachriverunder-
suspended load cumulated over 702 days can be easily goes higher-frequency variability that is controlled by
seen as high. Nonetheless, this performance, which one weather patterns and channel recovery from extreme pre-
may describe as ‘‘average’’, does not require any spe- cipitation events.
cific knowledge other than the liquid discharge which Ratingparametersaandbof ariverstronglydependon
caneasilybegauged,sinceasimpleregressionrelation- flowregimeanddonotvaryinthesameproportions(Table
shiphasnowbeenestablished.SincethevariabilityofC 4).Generally,bstaysinrange0.3–2.5,whileacanvaryby
intheWadiAbdisexplaineduptoonly46%bytheflow, severalordersofmagnitude (e.g.ReidandFrostick, 1997).
we estimate that the performance of the reconstitution Frostick et al. (1983) have shown that the coefficient b is
is acceptable. To achieve a finer precision, the integra- smallerinaridzoneflashfloods(usuallyb<1)ascompared
tionofotherexplanatoryfactorsthanthedischargeonly to typical perennial systems draining temperate regions
is needed which would require the building of a more where b>1. Intermittent rivers in semiarid environments
complexmodel. seemtohaveb-coefficientslowerthan1,similartoephem-
4. AlthoughtheQ–Qcorrelationishigherthanthatofthe eral rivers in arid zones (Table 4). The rating parameters
s
C–Q relationship, the performance of the two methods calculated for the Katiorin basin in Kenya (Sutherland and
of prediction of Q is equivalent. This is due to the fact Bryan, 1990) and for the Eshtemoa in the northern Negev
s
that the former is biased, andthe latter isnot. The use Desert (Alexandrov et al., 2003a) are such that b+0.311
of a non-biased C–Q relationship introduces an uncer- lna(cid:2)1.066gives0.40ontheKatiorinand0.24ontheEste-
tainty in C, but when Q is calculated from C, no addi- moa,respectively.Thismeansthatthesefloodsareslightly
s
tional error is introduced because the value of Q more ‘‘aggressive’’ than the average floods of the Wadi
multipliedbyCwasmeasured.TheuseoftheC–Qrela- Abd.Anotherinterestingcomparisonisthatinthethreeba-
tionshipandthemultiplicationofCbyQthusappearsto sins, the variations in discharge explain only about 50% of
introduce less errors than using the law Q–Q, even the variance in suspended sediment discharge (49% for the
s
though the Q–Q relationship presents a higher Eshtemoa,52%fortheKatiorinand46.3%fortheWadiAbd).
s
correlation. Eqs. (6) and (7) provide little understanding of the
physical processes that underlie the sediment transport.
Moreover, we may also note that the pair (a, b) estab- However, physical explanations on shapesof C–Q relation-
lishedusingmeandailyconcentrationandwaterdischarge, ships, either linear or non-linear (concave or convex), can
(1.710, 0.750), differs slightly from that calculated for the be found in the literature (e.g. Williams, 1989; Sichinga-
flood phases, (2.154, 0.757). While the b-coefficient re- bula,1998;Moreheadetal.,2003).Inparticular,theavail-
mains unchanged, the a-coefficient is reduced by 20.6% ablestudies underlinetheinfluenceofpreceding discharge
when considering the mean daily variables. The different or antecedent moisture, which tends to generate quick or
methods of calculation can partly explain the discrepancy delayed runoff. This then causes rapid or slow increases in
between both pairs, since the pair (a, b) relative to the sediment concentration in concert with discharge changes
meandailyvariablesresultsfromabest-fitregressionrela- as well as affecting the variability in the amount of easily
tionship, while the other is the median of the (a, b) pairs erodible sediment stored in thechannel.
calculatedfromacollectionoffloodepisodes.Thefactthat Thehigh-frequencyvariability(atthefloodevent)ofthe
atlowdischarge,concentrations,whicharegenerallylower rating parameters is also examined in the present study. It
than those estimated by the regression relationship, are appears that their alignment demonstrates that a and b
consideredwithindailyvaluesbutnotasfloodphasevalues are not independent and that one cannot interpret one
(whichincludeonlyhighsedimentloadepisodes)isprobably without the other, as is sometimes found in the literature.
themainreasonfordiscrepancy betweenboth(a,b)pairs. The alignment of the (a, b) pairs for the sediment rating
curveseitheratastation,oratseveralstationsofthesame
Comparison to other rivers catchment area, has already been described by Asselman
(2000) for the perennial Rhine River and its tributaries.
Comparisonsofaandbamongsitesaredelicatebecausethe Asselman obtains the following relationships: b=(cid:2)0.296 ln
dimensionofa[hereexpressedin(kgm(cid:2)3)(sm(cid:2)3)bforevery a+0.467 (Andernach station, Rhine), b=(cid:2)0.393 ln a+
given example] depends on the value of b (dimensionless). 0.577 (Cochem station, Mosel) and b=(cid:2)0.462 ln a+0.736
Nevertheless, these comparisons make it possible to have (Rockenau station, Neckar). The slope for the relationship
a greater understanding of the respective physical signifi- intheWadiAbd(Eq.(5),seeFig.4)isconsistentwiththese
cancesofaandbwhichdependonmultiplefactors(Frostick values, although the interceptis higher.
etal.,1983;Syvitskietal.,2000;Asselman,2000).Syvitski Concerning the wadis in Algeria, Toua¨ıbia et al. (2001)
etal.(2000)havecomparedratingparametersat57gauging present rating curves based on a long-term series of flow
stationsinNorthAmerica,1inEuropeand1inChina.They and concentration measurements at two stations close to
examine the rating equations in comparison to the long- Ain Hamara, one of which is located downstream from Ain
termcharacterofthesuspendedsedimentloadintherivers Hamara,theotherislocatedintheneighbouringsub-catch-
and show that the rating coefficient a is inversely propor- ment of the Wadi Haddad. All the mean monthly relation-
tional to the long-term mean discharge and is secondarily ships that they obtain from instantaneous values have a
related to the average temperature and the basin’s topo- higher factor a and a smaller factor b than in the present
graphic relief. They also show that the rating exponent b study. The high values of b (together with low values of a)
correlates most strongly with the average air temperature have been obtained for the months where there is an in-
andbasinreliefandisweaklycorrelatedwiththelong-term crease in precipitation after 3 months of low water level,
196 M. Achite,S.Ouillon
Table4 Examplesofcoefficients(a,b)intheratingcurveC=aQb,whereC(concentrationofsuspendedsediment)isexpressed
ingL(cid:2)1andQ(waterdischarge)inm3s(cid:2)1
Reference River Semiarid Rainfall a b Complementary
basin (mmyr(cid:2)1) information
Chikita(1996) Ikushunbetsu,Japan No 1418 4.28·10(cid:2)3 2.11 Sedimentdischarge
dominatedbythe
riverbankerosionin
snowmeltrunoff
Moreheadetal.(2003) NorthSaskatchewan, No Notgiven 1.3·10(cid:2)5 1.58
Alberta
Asselman(2000) Rhine,Germany No 600–2500 7.7·10(cid:2)7 1.22–1.44 From4stationsalong
(cid:2)1.7·10(cid:2)6 theRhineRiver
AchiteandOuillon,thispaper Abd,Algeria Yes 250 1.71 0.75 Fromdailyvalues
Toua¨ıbiaetal.(2001) Haddad,Algeria Yes 249 5.79–25.13 0.36–0.74 Monthlyvalues
3.62 0.44 Medianofmonthly
values
Alexandrovetal.(2003a) Eshtemoa,Israel Yes 220–350 16.98 0.43
SutherlandandBryan(1990) Katiorin,Kenya Yes 640 43 0.30
(a,b)providedbyChikita(1996);Moreheadetal.(2003)andAsselman(2000)resultedfrominstantaneousCandQmeasurements.
inJulyintheWadiHaddadandinFebruaryintheWadiMina (Eisma, 1998; Syvitski et al., 2003; see also ref. quoted in
(El-Abtal station). Conversely, the low values of b, which Coynel et al., 2004). However, bedload transport seems to
arenotassociatedwiththehighestvaluesofa,areobtained be extremely variable in dryland rivers. Reid and Laronne
for the monthsthat followa period of at least two months (1995) have reported very high bedload rates in a gravel
of high flow (in December and March in the Wadi Haddad, bed ephemeral river, that are orders of magnitude higher
in December in the Wadi Mina). It should be noted that than in its more humid counterparts. On the other hand,
thesevariationsofbonamonthlyscaleareconsistentwith Serrat(1999)notedthatbedloadsedimentvolumewasless
theshort-termvariationsofbthatwehaveobservedonthe than 1% of the suspended load in an intermittent river of
C–Qcurvesduringaflood(Fig.5).Thisisalsoinagreement Mediterranean type. In the Wadi Abd, the small proportion
withthevariationsbetweenrisingandfallingdischargeepi- ofsandinsuspension(<5%ofdryweightofsuspendedparti-
sodesobservedin theRhineRiverbyAsselman(2000).This cles) does not allow us to conclude on the relative impor-
reinforces the assumption of a strong relation between tance of transport in suspension and bedload transport.
coefficients aandb. The value of 136tkm(cid:2)2yr(cid:2)1 may only be considered as an
estimate of the fine sediment flux for the Wadi Abd Basin.
Annual sediment wash-down and its variability The value of 136tkm(cid:2)2yr(cid:2)1 is similar to the global mean
sediment wash-down, of the same order as the sediment
Annual suspended sediment budget wash-downinSouthAmerica,andfarhigherthanthemean
sedimentwash-downinAfrica(seeTable5).Theseregional
The average value of the suspended sediment flux for the values are average values andit must be remembered that
WadiAbdbasincalculatedfromthereconstructedtimeser- the specific sediment yield is known to decrease with the
ies of suspended sediment transport is 339·103tyr(cid:2)1 for drainage basin area (see Fig. 1 in Jiongxin and Yunxia,
the period 1973–1995. This corresponds to a specific sus- 2005).ThesuspendedsedimentyieldfortheWadiAbdbasin
pendedsedimentyield,whichistheratioofsuspendedload hasbeenalsocomparedtovaluesprovidedforothercatch-
to basin area, of 136tkm(cid:2)2yr(cid:2)1. The specific suspended ments in the Mediterranean basin (Table 5); it is similar to
sediment yield is also called ‘‘sediment wash-down’’ or thevalues reported for other basins in Algeria andIsrael.
‘‘suspended solidsyield’’ in the literature. The specific sediment yield Y, which is generally ex-
Suspendedsedimentfluxshouldnotbeconfusedwiththe pressed in tkm(cid:2)2yr(cid:2)1, can also be expressed in kg
rateofsoilerosionassedimenterodedfromuplandsoilsis km(cid:2)2day(cid:2)1to permitcross-sitecomparisonusingquantiles
progressivelystoredintherivervalley(MillimanandMeade, ofitsdistributionthatcannotbeexpressedonayearlybasis
1983).Moreover,itseemsthat,globally,thetwoquantities (Meybeck et al., 2003). In the Wadi Abd, the average daily
are inversely affected by anthropogenic impacts, with soil suspendedsolidsyieldis374kgkm(cid:2)2day(cid:2)1andthemedian
erosionacceleratingwhilesedimentfluxtothecoastalzone daily sediment fluxes over the period 1973–1995 (Y ) is
50
isglobally decreasing (Syvitski, 2003). 28kgkm(cid:2)2day(cid:2)1. The resulting flux variability as defined
Norshouldsuspendedsedimentfluxbeconfusedwiththe bytheirratiois13.4,whichistypicalofhighsuspendedsol-
totalsedimentdeliverywhichresultsfrombedloadcharge, idsfluxvariabilityonthedaily-scalefollowingtheclassifica-
suspendedsediment,anddissolvedsediment(Serrat,1999). tion proposedby Meybeck et al.(2003).
Bedloadisgenerallyassumedto beaminorpartofpartic- Lastly, it is of interest to jointly analyze river flow and
ulatematterflux,accountingforonlyapproximately5–10% sediment wash-down. The average water discharge calcu-