WATER RESOURCES RESEARCH, VOL. 35, NO. 11, PAGES 3523-3530, NOVEMBER Effects of sediment supply on surface textures of gravel-bed rivers 1999 This file was created by scanning the printed publication. Errors identified by the software have been corrected; however, some errors may remain. John M. Buffington• and David R. Montgomery Departmentof GeologicalSciences,Universityof Washington,Seattle Abstract. Using previouslypublisheddata from flume studies,we test a new approach for quantifyingthe effectsof sedimentsupply(i.e., bed materialsupply)on surfacegrain sizeof equilibriumgravelchannels.Textural responseto sedimentsupplyis evaluated relativeto a theoreticalpredictionof competentmediangrainsize(D•0). We find that surfacemediangrain size(Ds0) variesinverselywith sedimentsupplyrate and systematically approachesthe competentvalue (D•0) at low equilibriumtransportrates. Furthermore,equilibriumtransportrate is a powerfunctionof the differencebetween appliedand critical shearstressesand is therefore a power functionof the difference betweencompetentand observedmediangrain sizes(D•0 andDs0). Consequently, we proposethat the differencebetweenpredictedand observedmediangrain sizescan be usedto determinesedimentsupplyrate in equilibriumchannels.Our analysisframework collapsesdata from different studiestoward a singlerelationshipbetweensedimentsupply rate and surfacegrain size.While the approachappearspromising,we cautionthat it has been testedonly on a limited set of laboratorydata and a narrow range of channel conditions. surfacetexturesof gravel-bedrivers.Althoughsurfacetextures can also respond to changesin the size distribution of the Laboratoryand field studiesdemonstratethat surfacetex- sedimentsupply[Jackson and Beschta,1984;Plattset al., 1989; tures and gravel-bedrivers respondto changesin the rate of Perkins,1989; Knighton,1991; Parker and Wilcock,1993; Posedimentsupply(i.e., bed material supply) [Leopoldet al., tyondyand Hardy, 1994],we concentratehere on textural re1964; Dietrich et al., 1989; Kinerson, 1990; Lisle and Hilton, sponseto alterred sedimentload of a fixedgrain-sizedistribu1992, 1999; Lisle et al., 1993; Rice, 1994; Nolan and Marron, tion. Furthermore,we focuson the specialcase of channels 1995].When deprivedof sediment,poorlysortedbed surfaces that are in equilibriumwith their sedimentsupply. typically coarsen through size-selectivewinnowing of fine 1. Introduction grains,as is commonbelowdams[Leopoldet al., 1964] and wood jams [Rice, 1994]. Winnowingresultsfrom transport ratesin excess of sedimentsupply.Conversely, wheninundated with sediment,bed load transportcapacitiesmaybecomeoverwhelmed,causingdepositionof bed load material (typically fine-grainedparticles[Leopold,1992]) and consequentreduction of surfacegrainsize[Rice,1994;Nolan andMarron, 1995]. On the basisof observedrelationshipsbetween sediment supplyand bed-surfacetexture, severalmethodshave been proposedfor qualitativelyassessing sedimentload (i.e., low, medium, and high) from inspectionof surface grain size [Dietrichet al., 1989;Lisle and Hilton, 1992].Lisle and Hilton [1999] recently demonstratedquantitativerelationshipsbetween sedimentyield and both the size and volume of fine sedimentin pools. However, the observedrelationshipsare sensitiveto basinlithologyand associated propensityfor productionof fine-grainedsediment.In this paper we use data .frompreviouslypublishedflume studiesto test a new model for quantitativelydeterminingrates of sedimentsupplyfrom observedbed-surfacegrainsize.We alsousethe modelframework to further examine the effects of sediment supply on 2. Theory Here we developa theoreticalrelationshipbetween sediment supplyrate and bed-surfacegrain size of equilibrium channels.For equilibriumconditionsthe bed load transport rate is equalto the sedimentsupplyrate (qb = qs). We define bed load transportrate as a power functionof the difference betweenthe applied and critical shearstresses qb = k(r' - rc)n-- qs, (•) where,' is the bed shearstress(that portionof the total shear stressapplied to the bed and responsiblefor sedimenttransport) [EinsteinandBarbarossa, 1952;Nelsonand Smith,1989], *c is the critical shearstressfor particle motion, and k and n are empirical values [du Boys, 1879; O'Brien and Rindlaub, 1934; Meyer-Peterand Mtiller, 1948; Chien, 1956; Ashmore, 1988;Wathenet al., 1995]. We define *c as the critical shearstressof the median bedsurfacegrain size(*cso). The mediangrainsize(D5o) is related to its criticalshearstressby the Shields[1936] equation, which is a force balancebetweenthe fluid stressesactingon a bed-surfaceparticleandthe resistinggrainweightper unit area •Nowat WaterResources Division, U.S.Geological Survey, Boul- * = •'c5o/(Ps-p)gD5o, der, Colorado. T c50 (2) Copyright1999by the American GeophysicalUnion. where p and Ps are the fluid and sedimentdensities,respec- Paper number 1999WR900232. 00.43-1397/99/1999 WR900232509.00 tively,g is gravitational acceleration, and **csois the dimensionlesscritical shearstressfor incipientmotion of Dso. 3523 3524 BUFFINGTON AND MONTGOMERY: EFFECTS OF SEDIMENT SUPPLY ON GRAVEL-BED RIVERS Becausethe Shieldsequationis a force balance,it can be propriateconservative value for use with bed load transport invertedto determinethe competentmediangrain sizefor a data in general.There is alsohistoricalprecedentfor choosing given bed shear stressby definingthat stressas the critical this value.In their well-knownstudyof bed load transport, Tc50 Meyer-Peter andMtiller[1948,p. 54]propose * = 0ß03 asa value D•o = z'/Z*cso(psp)g, (3) where D}o is the competentmedian grain size for z'. Consequently,(1) canbe rewrittenin termsof a differenceof median grainsizes,ratherthan a differenceof shearstresses qb= k•(z' - Zc)n = k2(D•0- D50)n = qs, (4) whereDso is the observedmediangrain size(proportionalto *c) andTc50 * is assumed constant. Severaltestablehypothesesfor relating bed-surfacegrain sizeto sedimentsupplyrate can be developedfrom (4). First, we hypothesizethat if equilibriumtransportrate is a power functionof excessshearstress(,' - *c), it alsoshouldbe a functionof the differencebetweenD }o andDso (a surrogate for excessshearstress).If verifiable,ratesof sedimentsupply could be determined from the difference of observed and com- petent median grain sizes.Second,we hypothesizethat for givenvaluesof k, n, and ,', there shouldbe an inverserelationshipbetweensedimentsupplyrate and *c (and thusDso) for equilibriumtransport. 3. Data Sources and Methods Becauseit is difficultto demonstratechannelequilibriumin natural rivers,we testedthe abovehypotheses usingdata from flume studieswhere equilibriumtransportcould be verified. We used previouslypublisheddata from laboratory studies that met the following criteria: (1) Experimentswere con- lower limit for transportof mixedgrain sizes. Predictionof D}o from (3) alsorequiresestimationof the bed shearstress(,', alsoknown as the skin friction stressor fluid stressactingon the bed). The simplestand bestknown methodfor determining,' is that developedbyEinstein[1934, 1942, 1950;seealsoEinsteinand Barbarossa,1952]. However, the Einstein approachassumesa logarithmicvelocityprofile whichis inappropriatefor channelswith grainsthat are large relativeto the flow depth.Large grainscausesignificantparticle form drag [Andrews,1999]and developmentof a strongly nonlogarithmicvelocityprofile, producingbed stressesmuch smallerthan would be predictedfrom a logarithmicvelocity profile[Wibergand Smith,1991].Relativeroughness is a measure of potential particle form drag and is defined here as D84/h, where D84 is the surfacegrain sizefor which 84% of the particlesare smallerandh is flow depth.Valuesof D84/h for the compiledlaboratorystudiesrangefrom 0.07 to 0.55, with an averageof 0.28 (Table 1). Consequently, the Einstein methodis not justifiedfor the currentdata set.Instead,we use a more appropriate,albeit more complex,approachbasedon the work of Wibergand Smith [1991]. Wibergand Smith [1991] developeda model for simultaneoussolutionof the verticalvelocityprofileandthe form drag due to bed-surfacegrains;simultaneoussolutionis required because the two influence one another. Unlike the Einstein approach,the form of the velocityprofile is not specifieda priori, allowingmore accuratedeterminationof fluid forcesin conjunction with site-specific measuresof bed materialproperties. We modified the Wibergand Smith [1991] model to ducted in sediment-feed flumes with surface textures, bed includeother sourcesof roughness observedin the compiled slope,and water surfaceslopeallowedto equilibratewith an studies.We calculatedthe bed shearstress(,') in (3) as a imposeddischargeand sedimentfeed rate; (2) eachinvestiga- reach-averagedepth-slopeproduct correctedfor wall effects tor conductedexperimentsat severalfeed rates,usinga con- [Houjouet al., 1990], particleform drag [Wibergand Smith, stantsizedistribution;(3) experiments were allowedto run to 1991], and bed form drag where present[Nelsonand Smith, a stateof equilibriumtransport(definedasequalratesof input 1989] andoutput);and (4) grain-sizedistributions of the equilibrium ß = - [,w(Z) + + ,¾(z)], (5) bed surfacewere reported. We found only four studiesthat met these criteria [Parkeret al., 1982;Kuhnle and Southard, (equalto ,' at thebed),*r is the 1988, 1990;Dietrichet al., 1989;Lisleet al., 1993].In two of the where,f is the fluidstress total stress, *w is the stress associated withwallroughness, ,p investigations [Dietrichet al., 1989;Lisle et al., 1993],equilibis the stress caused by particle form drag, *of is the stress rium transportwas additionallydefinedby similarityof input associated with bed form drag,andz is heightabovethe bed. and outputgrain-sizedistributions. Furthermore,in thesetwo A finite differencemodel is usedto simultaneously determine studiesthe channelevolvedin responseto a progressivedeeach of thesereach-averageshearstresscomponentsin concreasein sedimentsupplyrate. In contrast,the channelwas junctionwith the reach-average verticalvelocityprofile [Houresetfor eachrun in the experimonts byParkeret al. [1982]and jou et al., 1990;Wibergand Smith,1991]. Kuhnleand Southard[1988,1990].The compiledstudiescover Followingthe approachof Wibergand Smith [1991], the a broad range of channelslopes(0.0035-0.031),width/depth vertical gradientof the downstreamvelocity(Ou/Oz) is exratios (2-23), mediangrain sizes(2.3-12.9 mm), equilibrium pressedin termsof the abovestressesas transport rates(0.0028-1.07 kgm-• s-•), andbedform mor- phologies(planebed, dune,and alternatebar) (Table 1). The primarysourcesof roughness in thesestudiesare channelwalls Oz L •p , (6) and form drag due to bed-surfacegrainsand bed forms. To test the relationshipsindicatedby (4), valuesof the whereL is the lengthscaleof turbulentmixing(as definedby competentmediangrain size (D}o) were calculatedfrom (3) Wibergand Smith [1991]).Rearrangingtermsand integrating Ou_ 1(•)0.5 1[*r-(*w+ *p+ ,¾)]0.5 with Tc50 * equalto 0.03,a conservative valuefor visually based yields studiesof incipientmotion[Buffington andMontgomery, 1997]. Moreover,, cSovaluesnear 0.03 are commonlyreportedfrom studiesof bedloadtransport[e.g.,ParkerandKlingeman,1982; Andrewsand Nankervis,1995], suggesting that 0.03 is an ap- u(z) p o (7) BUFFINGTON AND MONTGOMERY: EFFECTS OF SEDIMENT SUPPLY ON GRAVEL-BED RIVERS 3525 3526 BUFFINGTON AND MONTGOMERY: EFFECTS OF SEDIMENT SUPPLY ON GRAVEL-BED RIVERS whereZo is the roughnesslengthscaleof the bed and u = 0 at proximatesthat of spheresresting on a bed of like grains z - Zo. Equation(7) is integratednumerically,with the stress [Coleman,1967]. Becausethe sizedistributionof the competentbed surfaceis componentsdefinedin terms of u and z, allowingiterative, simultaneoussolution of both the stressesand the velocity unknown,we approximatedCm using a lognormalgrain-size distribution profile [Wibergand Smith,1991]. The reach-averagetotal shear stressis written as a depthslopeproductthat variesas a linear functionof height above Cm= exp (12) the bed (8) where co is the maximumconcentrationof bed-surfaceparticles (set equal to 0.6), 4)rnand 4%0are the componentand median grain sizes,respectively,expressedin 4• units (4• = -log 2D, whereD is in millimeters[Kmmbein,1936]) and rr is where ro is the total boundaryshearstress(p#hS) and h and S are the reach-averageflow depth and water surfaceslope, the standard deviationof particlesizesin 4•.We chose'the respectively. smallestreportedvalue of rr in eachinvestigationas a conserForm drag causedby bed-surfacegrains,bed forms,or any vative estimatefor the competentbed surface,with the range other objectin the flow can be expressedas of grain sizeslimited by thoseusedin eachstudy.Here 4%0is determinedthroughsimultaneous solutionof (3) and (5) (disP cussedfurther below). Fd= 5 Cd(lg2)Ax' (9) Similarly, the stress associatedwith drag due to twodimensionalbedformsis definedbyNelsonand Smith[1989]as where F a is the drag force that resultsfrom pressuredifferencesacrossthe object, Ca is the drag coefficient,Ax is the flow-perpendicular areaof theobject,and(u2) isthesquare of PCd(U 2)•; gx Tbf(Z) =5 (13) the referencevelocityaveragedoverAx. The referencevelocity is definedas the velocitythat would existin the absenceof the objectof interest.Dividing the drag force by the flow-parallel area of the object(Ab) yieldsthe associated drag stress p 2 Ca•Ax p fzb• z¾ u2dz/( - Zo) , •Z0 whereZbSis the heightof thebedform.Here the dragcoeffi- Ax Td:5Cd(lg2) Ab' =- (IO) cientis 0.21 if flow separatesaroundthe bed form and is 0.84 The stressdue to grain form drag is defined accordingto Wibergand Smith[1991]asthe particledragforceper unit area summedover all grainsizes(smallestto largest,D i to D3•) at each level of z M if flow remainsunseparated[Smithand McLean, 1977]. Wall effects include both momentum diffusion causedby proximityof walls(i.e., width-to-deptheffects)[Leighly,1932; Parker,1978]and differencesbetweenbed andwall roughness [Einstein,1934; 1942; Houjou et al., 1990]. Using data from Houjouet al.'s [1990]studyof planarrectangularchannels,the ratio of wall to bed stress(rw/r') canbe expressed asa power functionof the width/depthratio (w/h) andthe ratio of bedto wallroughness ( zob/Zow) PZ Cd{ 112) C rn--t (lib) Tw ZOb ZOw/ r-7= 2.55 2 R = 0.99' (14) rn where all valuesare cross-sectional and reach averages.Rearranging(14) yieldsan expressionfor r w as a functionof r', P2 . Izm ] tg2 dz/(z m- Zo) , (11c) w/h, andZofZow o where ½mis the concentrationof a componentgrain size and Zm is the vertical height of that particle above the bed. We assumeellipticalgrainswith a Corey[1949]shapefactor of 0.6 (a reasonablevalueboth for naturalsedimentsand thoseused in the laboratorystudiesinvestigated here).The shapefactoris definedas D s/V'D•Dz•, wherethe subscripts S, I, and L ßw= 2.55' ZOw/ . Theeffective roughness of thebedin planarchannels isz%• 0.1D84 [•iting and DietHch, 1990; Wibergand Smith, 1991]. Thisdefinition of Zo•includes bothgrainskinfrictionand particleform drag [Wibergand Smith,1991].The effectivebed roughness in channelswith bed formsis foundby matchingthe inner and outer veloci• profilesat the height of the bed form indicate short, intermediate, and long axes,respectively.We assumethat the grainsare orientedwith their short axesver- (z•f) [Nelson andSmith,1989],yielding tical and that the intermediateand longdiametersare equalto one another (D s = 0.6D• - 0.6Dz•). Using the nominal graindiameter ((DsD•rDt)•/3), Cdm isapproximated bythe drag coefficientfor a sphere,which is a function of particle Reynolds number [e.g., Rouse, 1946]. While developedfor isolatedsphereswell awayfrom channelboundaries,this relationshipbetween drag coefficientand Reynoldsnumber ap- ( . (16) )-4(•'+•)/(•'+•+•) z¾ z0•= z¾ 0.1D84 We definethewall roughness fromNikuradse's[1950]equation for hydraulicallysmoothflow, appropriatefor the smoothwalled flumes studied here, BUFFINGTON AND MONTGOMERY: EFFECTS OF SEDIMENT SUPPLY ON GRAVEL-BED RIVERS 1 • Parker et al. [1982] [] Kuhnleand Southard[1988] 3527 [] ß Dietrich etal.[1989] A LiMeetal. [1993] o. 1 //.4• • . 0.01 _ y=0.009 x1.3550.19, R2=0.82 • I .... I ' ' 0.001 0 5 10 15 D5O'- D5O(mm) Figure 1. Equilibriumtransportrate (q•, -- qs) asa function 1 of thedifference between competent mediangrainsize(D }o) andthe observed value(Dso). 10 Reach-averagebed stress(z', Pa) Zow = v/9u*w, (17) wherev is thekinematic velocityandu* is the shearvelocity w at thewall(rX/-•w/p). The competent mediangrainsize(D}o) is determined from simultaneous, iterativesolution of (3), (5), (8), (11), (13), and Figure2. Observed mediansurfacegrainsize(D so) versus bed shearstress(r') for experiments by Parkeret al. [1982] (diamonds), KuhnleandSouthard [1988](squares), Dietrichet al. [1989](circles), andLisleetal. [1993](triangles). The data are stratifiedby equilibriumtransportrates(q•, = qs) of <0.01,<0.1, and>1 kgm-• s-• shown asopen,shaded, and solid,respectively; specific equilibriumtransportratesfor each datapointarelistedin Table1. The dashedlinesarepredic- (15). BecauseD }o is a functionof bed shearstress,whichis,in tionsof Dso for equilibriumtransportratesof 0.01,0.1, and 1 onthecurve fitofFigure1, assuming Tc50 * --turn, a functionof the surfacegrain-sizedistributionand con- kgm- • s-• based 0.03 and Ps -2650 kg m -3. The bold solid line is the sequentparticleform drag,the competentmediangrainsize prediction of competent mediangrainsize(D•o, andthe bedshearstressmustbe solvedtogether.We empha- theoretical seeequation(3)). sizethat the competentmediangrainsizeis a hypothetical value,theoccurrence of whichdepends onthelocalsupplyand caliber of sediment. Channel characteristics and calculated averagemediansurfacegrainsize,(3) caliberof suppliedsedvaluesfor the compiledstudiesare presentedin Table 1. iment,and(4) sortingof suppliedsedimentin the four studies (Table1). Previous methods for assessing texturalresponse to sedimentsupplyresultedin a familyof curvesthat were de4. Use of Surface Grain Size to Infer Sediment Supply pendenton study-specific ratiosof bed shearstressto critical shearstressof thebedloadmaterial[Dietrichetal., 1989;Lisle Data fromthe compiledlaboratorystudiessupportthe form et al., 1993].The generalcollapse of the datatowarda single of (1) quitewell, demonstrating that equilibriumtransport functionsuggests that the analysis frameworkusedherepro(q•, = q•) is a powerfunctionof the differencebetweenthe videsa unifying relationship betweensediment supplyrateand appliedandcriticalshearstresses. The equilibriumtransport surfacegrain sizein equilibriumchannels. equation for thesedataisq•, = 0.025(r' - %)•.3•___o.•9 with R: = 0.82.Although theexponent of thisequation iswithin the rangeof typicallyreportedvalues(1.3-1.8 [O'Brienand 5. Effects of SedimentSupply on SurfaceGrain Rindlaub, 1934; Meyer-Peterand Maller, 1948; Chien, 1956; Size Ashmore,1988]), both the coefficientand exponentof the Figure2 comparesthe observedmediangrainsizeto the relationship are sensitiveto investigator choiceof (1) the predictedcompetentvalueasa functionof bed shearstressand methodfor determining%, (2) the characteristic grain size sedimentsupplyrate. The data are stratifiedby equilibrium represented by % (i.e., the particulargrain-size percentileand transport ratesof <0.01,<0.1, and>1 kg m-1 s-1, withspe- thepopulation it isdrawnfrom(surface, subsurface, or load)), cificvalueslistedin Table 1; contours of predictedgrainsize and (3) the methodof correctingr' for channelroughness. for thesetransportratesalsoare shownbasedon the curvefit Depending on the above choices,valuesof the coefficientk of Figure 1. The data of Lisleet al. [1993]andKuhnleand varybyanorderof magnitude (0.003-0.06),whilevaluesof the Southard[1988]showa wide rangeof mediansurfacegrain exponentn vary from 1.2 to 2.1 for this data set [Buffington, sizesfor a relativley smallrangeof bedshearstresses, withD so 19981. varyinginversely with sedimentsupplyas hypothesized from Becauseequilibriumtransportis a powerfunctionof excess shearstress,it followsthat it is alsoa powerfunctionof the competent andobserved mediangrainsizes(surrogates for the appliedand criticalshearstresses, (4)), as demonstrated by Figure 1. Plottedin this fashion,the data collapsetowarda singlefunctiondespitediffering(1) bedshearstress, (2) reach- (4). Furthermore,the compileddatashowthat mediansurface grain sizesystematically approaches the competentvalue at low equilibrium transportrates,as predictedfrom (4). This equationindicatesthat the observed surfacegrainsizeshould approach thetheoreticalcompetent valueasequilibriumtransport rates go to zero. 3528 6. BUFFINGTON Discussion AND MONTGOMERY: EFFECTS OF SEDIMENT and Conclusions Our analysisdemonstratesthat surfacetexturesrespondto changesin sedimentsupplyrate, as documentedby Dietrichet al. [1989] and Lisle et al. [1993]. Imbalancesbetweenrates of sedimentsupplyand bed load transportcausesize-selective depositionor erosionthat altersbed roughness which,in turn, likely alters both the critical shear stressfor grainstraveling over the bed [Kirchneret al., 1990; Buffin•on et al., 1992; Johnstonet al., 1998] and the effectivebed shearstress[Naot, 1984; Wibergand Smith, 1991]. These responsesfeed back on one another,partiallycounteractinglocal imbalancesbetween sedimentsupplyand transportcapacity[Dietrichet al., 1989; Lisle et al., 1993]. For example, transport capacitiesoverwhelmedby highsedimentsuppliesshouldcausefine-sediment deposition,creating smootherbed surfaceswith lower intergranularfriction anglesand higherbed stresses, both of which will promote increased transport rates acrossthat surface (equation(1)), reducingthe differencebetweensupplyand transport rates. Parkerand Klingeman[1982,p. 1422]hypothesizedthat surfacetextureand degreeof armoringin gravel-bedchannelsare controlledby the ratio of bed stressto criticalstress(an alternativeexpression for excessshearstress).They suggested that "when stressesare well above critical stress,mobility differenceson a grain-by-grainbasis are reduced;the differences disappearasymptotically. Thus a coarsepavementshouldnot form. The low-stressrange for which pavementis required is typical of gravel bed streams;the high-stressrange where it shouldnot occuris typicalof sandbed streams."Subsequent laboratory and field studies corroborate their hypothesis [Parkeret al., 1982;Wilcockand Sourhard,1989;Pitlick,1992]. We further suggestthat bed-surfacetexturerepresentsa feedback between rates of both sediment supply and bed load transport(the latter being a functionof surfaceroughness, excessshearstress,and availabilityof transportable sediment). Althoughbed-surfacetexturemay respondto rates of both sedimentsupplyand bed load transport(i.e., excessshear stress),our use of equilibriumconditionsrendersthesetwo factorsequal and inseparable,allowinginterpretationof equilibrium rates of sediment supply from inspectionof bedsurfacegrain size.The analysispresentedhere extendsprevious studiesof the relationshipbetweensedimentsupplyrate and surfacegrain size [Dietrichet al., 1989;Lisle et al., 1993] and showsthat there may be a singlerelationshipwhen sediment supplyis relatedto the differencebetweenD •o andD5o. We find that this difference is an indirect measure of excess shearstress(z' - z½) and is well correlatedwith rates of equilibriumtransportacrossdifferentstudies(Figure 1). Althoughthere are investigation-specific trendswithin the data, our analysisframeworkproducesa reasonablecollapseof the compileddata. SUPPLY ON GRAVEL-BED RIVERS responseto eliminationof sedimentsupply.Temporalchanges in bed-surfacegrain size and sortingwill also affect friction angledistributions andconsequent valuesof z*•[Kirchner etal., 1990;Buffingtonet al., 1992;Johnstonet al., 1998].Our constant valueof Tc50 * = 0ß03 is generallyappropriate for nonimbricatedplanebeds,but highervaluesof z*eare advisedfor channelswith more copmlexbed materialstructures(e.g.,imbricatedgrains,particle clusters,or stone cells) that reduce particlemobility[Churchet al., 1998]. Despitetheuseof a constant valueof z *½50we find a strong relationshipbetweensurfacegrain size and equilibriumtransport rate for the data examined here. On the basis of our findingswe suggest that ratesof sedimentsupplyin equilibrium channels can be determined from the difference between ob- servedmedian grain size and the predictedcompetentvalue (Figure1). Regardless of whetherthe caliberof the observed surfacegrain size reflects textural responseto excessshear stress,sedimentloading(Figure 2), or the particularsizedistribution of the supply,the differencebetweenpredictedand observedgrain sizeis a measureof excessshearstressand is therefore a measureof both q•, and qs for equilibriumchannels. Assessingrates of sediment supply in natural gravel-bed rivers using our approach requires conditions of quasiequilibrium. For equilibrium channelswith well-developed floodplainsandself-formedbeds(i.e., all sizesmobileat typical high flows),D•o shouldbe predictedfrom the bank-fullbed stress.An expedientapproachfor applyingFigure 1 is to confine its use to hydraulicallysimple,plane-bedreacheswhere correctionfor bank effectsand particleform drag may be the only roughness correction needed. Hydraulically simple reachesare recommendedbecausethe grain-sizeresponseto sedimentsupplyobservedin our analysisis fairly small(-17 mm over3 ordersof magnitudeof sedimentloading)andmay be overwhelmedby parameter uncertainty in formulas for stresscorrectionof hydraulicallycomplexreaches.Lisle and Hilton [1999] also report fairly small magnitudesof textural responseto sedimentloadingin gravel-bedriversof northern California and southernOregon.They found that the D5o of fine pool materialdecreases approximately9 mm over3 orders of magnitudeof sedimentyield. The sensitivityof equilibrium transportratesto smallchangesin grainsizealsohighlightsthe importance of accurategrain-sizesampling,insuring a high degreeof confidencefor the grain-sizestatistics(further discussionof sedimentsamplingproceduresand accuracyis given by Churchet al. [1987],Riceand Church[1996],andBuffin•on and Montgomery [1999]). While Figure 1 may proveusefulfor assessing ratesof local bed load sedimentsupply,it is cautionedthat our resultsare based on a limited set of flume data with small grain sizes (D5o < 13 mm) andchannels not entirelyself-formed; except for the studyof Lisle et al. [1993] the laboratorychannelsused However,ouranalysis assumes a constant valueof To50 * and in our compilationwere constrainedto specificwidths. Furso doesnot fully describetexturalresponseto alteredratesof thermore,the caliberof sedimentsupplywasheld constantin equilibrium transport. Dimensionless criticalshearstress ( eachexperiment,and althoughthe data collapsefairlywell into will vary with site-specificdifferencesin bed material proper- a singlefunction despite differing size distributionsof sedities (e.g., grain shape,rounding,protrusion,imbrication,and ment supply,the limited amount of availabledata does not distribution of intergranular frictionangles). Moreover, •*•val- allow a full explorationof the effectsof theseparameters.For ues will covary with changesin bed-surfacegrain size and example,everythingelsebeing equal,bed surfaceresponseto sorting that occur in responseto altered rates of bed load a supplyof fine-grainedsedimentmaybe very differentthan to transportor sedimentsupply.For example,Churchet al. [1998] a supplyof coarser-grained material.Use of a constantvalueof observed temporalchanges in effective valuesof ß*½5oas bed Tc50 ß alsomaycauseerrorsin calculated valuesof D 5o • and surfacescoarsenedand as bed material structurechangedin corresponding estimatesof equilibriumtransportrate (as dis- BUFFINGTON AND MONTGOMERY: EFFECTS OF SEDIMENT cussedabove).Becauseof theseconcerns, furtherinvestigation of our findingsis warrantedbefore they are appliedas a land managementtool. Moreover, the condition of near-equilibrium transportmay not be a realisticassumptionfor actively disturbedlandscapes,such as watershedsthat have been intenselyclear-cutand are experiencinghigh sedimentloadsdue acceleratedlandslidefrequency,road runoff, or destabilized river banks.Where it is knownthat channelsare aggrading,our proposedmethodprovidesa minimumestimateof local sediment supplyrate; the difference between predicted and observedD5o can providean estimateof the bank-fullqb, which must be lessthan qs for an aggradingsystem. The requirementof equilibrium transport for use of our approach motivates an interesting question about natural channels:What is the time for bed-surfaceadjustmentto perturbationsof sedimentsupplyor transportcapacity?That is, how long doesit take for surfacegrain sizeto equilibratewith imposedchannel hydraulicsand sedimentload? Gravel-bed channelsoccurin a wide rangeof environmentsthat are characterizedby different magnitudesand frequenciesof discharge eventsand sedimentinputs,offering a range of timescalesfor channeladjustmentto theseperturbations[Pitlick, 1994]. In arid regions,channelsexperienceinfrequent,intensefloodsof short duration (on the scaleof severalhours) that may not offer sufficienttime for morphologicadjustmentto the imposeddischarges and sedimentloads[LaronneandReid, 1993]. In mountaindrainagebasinsof the westcoastof North America, hydraulic dischargeis typically perennial, with rainfalldriven flood eventsoccurringone or more times a year and having durationsof the order of one or more days,allowing significantly more time for morphologicadjustments.In intermontaine regionsof the United States,flood flows typically result from spring snowmeltand are characterizedby long, graduallyvaryingdischargesthat occurof the order of tens of days to months and allow even more time for morphologic adjustmentsand attainmentof quasi-equilibrium.To put these environmentaldifferencesin context,the 100 year flood may be less than 2 times the mean annual flow in snowmelt basins, while it can be more than 10 times the mean annual flow in watershedswith flashy, storm-drivenhydrographs[Pitlick, 1994]. The likelihoodof channelmorphologyequilibratingwith the imposedhydrologyand sedimentload dependson the shape and durationof the localhydrograph,aswell asthe timescale of adjustmentfor the morphologicfeature of interest.Moreover, alluvialchannelsare nonlineardynamicsystemsthat can exhibit a variety of responsesto perturbationsof hydraulic dischargeor sediment supply. For example, many alluvial channelsare free to adjusttheir width, depth,slope,bed form morphology,channelpattern, and bed-surfacetexture in responseto changesin dischargeand sedimentsupply(see review by Montgomery and Buffington[1998]). Severalmorphologic adjustments may occur simultaneously for a given perturbation,with many responsesfeeding back on one another. The specificmorphologicresponseto a particularchannel perturbationdependson localchannelcharacteristics, geomorphic setting,and historicaleventswithin the watershed. Nevertheless,changesin bed-surfacetexturesmay be an active and potentiallyrapid, first-orderresponseto alteredsediment supply or hydraulic discharge. Consequently, observing changesin bed-surfacetexture may be a more sensitiveand effective means of monitoring channel responseto altered sedimentsupplyand hydraulicdischargeover shorttimescales. SUPPLY ON GRAVEL-BED RIVERS 3529 Over longer timescales,other morphologicfeatures may be better indicatorsof channelresponse,particularlywhere textural changesare short-livedor counteractedby subsequent, larger-scalemorphologicadjustments. Acknowledgments. Financial supportwas provided by the WashingtonStateTimber, Fishand Wildlife agreement(TFW-SH10-FY93004, FY95-156), the PacificNorthwestResearchStationof the USDA Forest Service(cooperativeagreementPNW 94-0617),and the National ResearchCouncil(USGS ResearchAssociateship). We thank Mike Church,Rob Ferguson,JohnPitlick, Jim Pizzuto,Peter Wilcock, and an anonymousreviewerfor insightfulcriticismsof earlier draftsof this work. References Andrews, E. D., Bed-material transportin the Virgin River, Utah, WaterResour.Res., in press,1999. Andrews, E. D., and J. M. Nankervis, Effective dischargeand the designof channelmaintenanceflowsfor gravel-bedrivers,in Natural and AnthropogenicInfluencesin Fluvial Geomorphology: The Wolman Volume,edited by J. E. Costaet al., Geophys.Monogr.Ser.,vol. 89, pp. 151-164, AGU, Washington,D.C., 1995. Ashmore, P. E., Bed load transportin braided gravel-bedstream models,Earth Su• Processes Landforms,13, 677-695, 1988. Buffington,J. M., The use of streambedtextureto interpretphysical and biologicalconditionsat watershed,reach,and subreachscales, Ph.D. dissertation,147 pp., Univ. of Wash., Seattle, 1998. Buffington,J. M., and D. R. Montgomery,A systematicanalysisof eight decadesof incipientmotion studies,with specialreferenceto gravel-beddedrivers, WaterResour.Res.,33, 1993-2029, 1997. Buffington,J. M., and D. R. Montgomery,A procedurefor classifying textural facies in gravel-bedrivers, Water Resour.Res., 35, 19031914, 1999. Buffington,J. M., W. E. Dietrich, and J. W. Kirchner,Friction angle measurements on a naturallyformedgravelstreambed:Implications for criticalboundaryshearstress,WaterResour.Res.,28, 411-425, 1992. Chien,N., The presentstatusof researchon sedimenttransport,Trans. Am. Soc. Civ. Eng., 121, 833-884, 1956. Church,M. A., D. G. McLean, and J. F. Wolcott, River bed gravels: Samplingand analysis,in SedimentTransportin Gravel-BedRivers, editedby C. R. Thorne, J. C. Bathurst,and R. D. Hey, pp. 43-88, John Wiley, New York, 1987. Church,M., M. A. Hassan,andJ. F. Wolcott,Stabilizingself-organized structuresin gravel-bedstreamchannels:Field and experimental observations, Water Resour. Res., 34, 3169-3179, 1998. Coleman,N. L., A theoreticaland experimentalstudyof drag and lift forcesactingon a sphererestingon a hypotheticalstreambed,Proc. Int. Assoc.Hydraul. Res.,3, 185-192, 1967. Corey,A. T., Influenceof shapeon the fall velocityof sandgrains,M.S. thesis,102 pp., Colo. StateUniv., Fort Collins,1949. Dietrich, W. E., J. W. Kirchner, H. Ikeda, and F. Iseya, Sediment supplyand the developmentof the coarsesurfacelayer in gravelbedded rivers, Nature, 340, 215-217, 1989. du Boys, P., Le Rh6ne et les rivi•rs a lit affouillable,Ann. Ponts Chauss•es,Mem. Doc. 5, 18, 141-195, 1879. Einstein,A., Der hydraulische oderproill-radius,Schweiz. Bauztg.,103, 89 -91, 1934. Einstein, H. A., Formulasfor the transportationof bed load, Trans. Am. Soc. Civ. Eng., 107, 561-597, 1942. Einstein,H. A., The bed-loadfunctionfor sedimenttransportationin open channelflows,Tech.Bull. 1026, 73 pp., Soil Conserv.Serv., U.S. Dep. of Agric., Washington,D.C., 1950. Einstein,H. A., and N. L. Barbarossa,River channelroughness, Trans. Am. Soc. Civ. Eng., 117, 1121-1146, 1952. Folk, R. L., Petrology of Sedimentary Rocks,182pp., Hemphill,Austin, Tex., 1974. Houjou, K., Y. Shimizu,and C. Ishii, Calculationof boundaryshear stressin open channelflow, J. Hydrosci.Hydraul. Eng., 8, 21-37, 1990. Iseya,F., and H. Ikeda, Pulsationsin bedloadtransportrates induced by a longitudinalsedimentsorting:A flume studyusing sand and gravelmixtures,Geogr.Ann., 69A, 15-27, 1987. 3530 BUFFINGTON AND MONTGOMERY: EFFECTS OF SEDIMENT Jackson,W. L., and R. L. Beschta, Influences of increasedsand deliv- ery on the morphologyof sand and gravelchannels,WaterResour. Bull., 20, 527-533, 1984. Johnston, C. E., E. D. Andrews, and J. Pitlick, In situ determination of particle friction angles of fluvial gravels, Water Resour.Res., 34, 2017-2030, 1998. Kinerson, D., Bed surfaceresponseto sedimentsupply,M.S. thesis, 420 pp., Univ. of Calif., Berkeley, 1990. Kirchner,J. W., W. E. Dietrich, F. Iseya,and H. Ikeda, The variability of criticalshearstress,friction angle,and grain protrusionin water worked sediments,Sedimentology, 37, 647-672, 1990. Knighton,A.D., Channelbed adjustmentalongmine-affectedriversof northeastTasmania, Geomorphology, 4 205-219, 1991. Krumbein, W. C., Application of logarithmicmomentsto size frequency distributionsof sediments,J. Sediment.Petrol., 6, 35-47, 1936. Kuhnle, R. A., and J. B. Southard,Bed load transportfluctuationsin a gravel bed laboratory channel, Water Resour.Res., 24, 247-260, 1988. Kuhnle,R. A., andJ. B. Southard,Flume experimentson the transport of heavy minerals in gravel-bedstreams,J. Sediment.Petrol., 60, 687-696, 1990. Laronne, J. B., and I. Reid, Very high rates of bedload sediment transportby ephemeraldesertrivers,Nature,366, 148-150, 1993. Leighly, J. B., Toward a theory of the morphologicsignificanceof turbulencein the flow of water in streams,Univ.Calif Publ. Geogr., 6, 1-22, 1932. Leopold,L. B., Sedimentsizethat determineschannelmorphology,in Dynamicsof Gravel-BedRivers,editedby P. Bil!i et al., pp. 289-311, John Wiley, New York, 1992. Leopold, L. B., M. G. Wolman, and J.P. Miller, Fluvial Processes in Geomorph9logy, 522 pp., W. H. Freeman,New York, 1964. Lisle, T. E., and S. Hilton, The volumeof fine sedimentin pools:An indexof sedimentsupplyin gravel-bedstreams,WaterResour.Bull., SUPPLY ON GRAVEL-BED RIVERS Parker, G., Self-formed straight rivers with equilibrium banks and mobile bed, 2, The gravelriver, J. Fluid Mech., 89, 127-146, 1978. Parker, G., and P. C. Klingeman,On why gravel bed streamsare paved, WaterResour.Res.,18, 1409-1423, 1982. Parker,G., andP. R. Wilcock,Sedimentfeed andrecirculatingflumes: Fundamentaldifference,J. Hydraul.Eng., 119, 1192-1204, 1993. Parker, G., S. Dhamotharan, and H. Stefan, Model experimentson mobile,pavedgravelbed streams,WaterResour.Res.,18, 1395-1408, 1982. Perkins,S. J., Interactionsof landslide-supplied sedimentwith channel morphologyin forestedwatersheds,M.S. thesis,104 pp., Univ. of Wash., Seattle, 1989. Pitlick,J., Flow resistance underconditionsof intensegraveltransport, Water Resour.Res., 28, 891-903, 1992. Pitlick,J., Relationbetweenpeak flows,precipitation,and physiographyfor fivemountainousregionsin the westernUSA, J. Hydrol.,158, 219 -240, 1994. Platts, W. S., R. J. Torquedada,and M. L. McHenry, Changesin salmonspawningand rearinghabitatfrom increaseddeliveryof fine sediment to the South Fork Salmon River, Idaho, Trans.Am. Fish. Soc., 118, 274-283, 1989. Potyondy,J.P., and T. Hardy, Use of pebblecountsto evaluatefine sediment increasein stream channels,WaterResour.Bull., 30, 509520, 1994. Rice, S., Towards a model of changesin bed material texture at the drainagebasinscale,in Process Modelsand Theoretical Geomorphology,editedby M. J. Kirkby,pp. 160-172,JohnWiley,New York, 1994. Rice, S., and M. Church,Samplingsurficialfluvialgravels:The precision of size distributionpercentileestimates,J. Sediment.Res.,66, 654-665, 1996. Rouse,H., ElementaryMechanicsof Fluids,376 pp., Dover, Mineola, N.Y., 1946. Shields,A., Anwendungder Aehnlichkeitsmechanik und der Turbu28, 371-383, 1992. lenzforschungauf die Geschiebebewegung, Mitt. Preuss.Versuchsanst.Wasserbau Schiffbau,26, 26 pp., 1936. Lisle,T. E., andS. Hilton, Finebedmaterialin poolsof naturalgravel bedchannels, Water Resour. Res.,35,1291-1304, 1999. Smith,J. D., and S. R. McLean, Spatiallyaveragedflow over a wavy Lisle, T. E., H. Ikeda, and F. Iseya,Formation of stationaryalternate surface,J. Geophys.Res., 82, 1735-1746, 1977. bars in a steepchannelwith mixed-sizedsediment:A flume exper- Wathen, S. J., R. I. Ferguson,T. B. Hoey, and A. Werritty, Unequal iment,Earth Surf.Processes Landforms,16, 463-469, 1991. mobilityof gravelandsandin weaklybimodalriver sediments,Water Resour. Res., 31, 2087-2096, 1995. Lisle, T. E., F. Iseya, and H. Ikeda, Responseof a channelwith alternate bars to a decreasein supplyof mixed-sizebed load: A Whiting, P. J., and W. E. Dietrich, Boundaryshearstressand roughflume experiment,WaterResour.Res.,29, 3623-3629, 1993. ness over mobile alluvial beds,J. Hydraul. Eng., 116, 1495-1511, 1990. Meyer-Peter, E., and R. MUller, Formulasfor bed-loadtransport,in Proceedings of the 2nd Meetingof the InternationalAssociation for Whiting, P. J., W. E. Dietrich, L. B. Leopold,T. G. Drake, and R. L. HydraulicStructures Research,pp. 39-64, Int. Assoc.for Hydraul. Shreve,Bedload sheetsin heterogeneoussediment,Geology,16, Res., Delft, Netherlands, 1948. 105-108, 1988. Montgomery,D. R., and J. M. Buffington,Channelprocesses, classi- Wiberg,P. L., andJ. D. Smith,Velocitydistributionandbedroughness fication,and response,in RiverEcologyand Management,editedby in high-gradientstreams,WaterResour.Res.,27, 825-838, 1991. R. Naiman and R. Bilby, pp. 13-42, Springer-Verlag,New York, Wilcock,P. R., and J. B. Southard,Bed-loadtransportof mixed-size 1998. sediment:Fractionaltransportrates, bed forms, and the developNaot, D., Responseof channel flow to roughnessheterogeneity,J. ment of a coarsebed-surfacelayer, WaterResour.Res., 25, 16291641, 1989. Hydraul.Eng., 110, 1568-1587, 1984. Nelson,J. M., and J. D. Smith, Flow in meanderingchannelswith naturaltopography,in RiverMeandering,WaterResour.Monogr.,vol. J. M. Buffington,Water ResourcesDivision,U.S. GeologicalSur12, editedby S. Ikeda and G. Parker, pp. 69-102, AGU, Washing- vey, 3215 Marine Street, Building RL6, Boulder, CO 80303. ton, D.C., 1989. (jbuffing@usgs.gov) Nikuradse,J., Laws of flow in roughpipe, Tech.Memo. 1292, 62 pp., D. R. Montgomery,Department of GeologicalSciences,University Natl. Advis. Comm. for Aeronaut.,Washington,D.C., 1950. of Washington, Seattle, WA 98195. (dave@bigdirt.geology. Nolan, K. M., and D.C. Marron, History, causes,and significance of washington.edu) changesin the channelgeometryof Redwood Creek, northwestern California,1936to 1982,U.S.Geol.Surv.Prof.Pap.,1454-N,22 pp., 1995. O'Brien,M.P., andB. D. Rindlaub,The transportationof bed-loadby streams,Eos Trans.AGU, 15, 593-603, 1934. (ReceivedApril 12, 1999;revisedJuly 19, 1999; acceptedJuly22, 1999.)