670Q 2003 Estuarine Research Federation

Estuaries Vol. 26, No. 3, p. 670–679 June 2003

A Modeling Study of the Satilla River Estuary, Georgia. II:

Suspended Sediment

LIANYUAN ZHENG1, CHANGSHENG CHEN2,*, MERRYL ALBER3, and HEDONG LIU2

1 College of Marine Science, University of South Florida, St. Petersburg, Florida 337012 School for Marine Sciences and Technology, University of Massachusetts-Dartmouth, New Bedford,

Massachusetts 027423 Department of Marine Sciences, University of Georgia, Athens, Georgia 30602

ABSTRACT: A three-dimensional (3-D) suspended sediment model was coupled with a 3-D hydrodynamic numericalmodel and used to examine the spatial and temporal distribution of suspended sediments in the Satilla River estuary ofGeorgia. The hydrodynamic model was a modified ECOM-si model with inclusion of the flooding-drying cycle overintertidal salt marshes. The suspended sediment model consisted of a simple passive tracer equation with inclusion ofsinking, resuspension, and sedimentation processes. The coupled model was driven by tidal forcing at the open boundaryover the inner shelf of the South Atlantic Bight and real-time river discharge at the upstream end of the estuary, with auniform initial distribution of total suspended sediment (TSS). The initial conditions for salinity were specified usingobservations taken along the estuary. The coupled model provided a reasonable simulation of both the spatial andtemporal distributions of observed TSS concentration. Model-predicted TSS concentrations varied over a tidal cycle;they were highest at maximum flood and ebb tidal phases and lowest at slack tides. Model-guided process studies suggestthat the spatial distribution of TSS concentration in the Satilla River estuary is controlled by a complex nonlinear physicalprocess associated with the convergence and divergence of residual flow, a non-uniform along-estuary distribution ofbottom stress, and the inertial effects of a curved shoreline.

Introduction

Georgia’s Satilla River has a well-mixed estuarywith a mean water depth of about 4 m and a max-imum tidal current of about 140 cm s21 (Blantonet al. 1999; Zheng et al. 2003). Tidally induced re-sidual currents are characterized by multiple eddy-like convergences and divergences. The maximumsubtidal velocity is over 15 cm s21 and occurs wherethe shoreline has a significant bend. This complexthree-dimensional (3-D) residual current feature ispredominately driven by tidal mixing, asymmetryof the tidal current, the along-river baroclinic pres-sure gradient, and centrifugal forcing associatedwith the curved shoreline (Zheng et al. 2003).Strong tidal currents in such a shallow estuarycause energetic vertical turbulent mixing and tidalmixing is generally stronger during the flood tidethan during the ebb tide (Dronkers 1986). Thefreshwater discharge of the river varies seasonally,with a 30-yr median value of 34 m3 s21 and an an-nual maximum flow of about 1,000 m3 s21 duringthe spring. Buoyancy-induced flow tends to accel-erate the offshore advection during ebb tide butslows down the landward movement during floodtide, causing an asymmetrical current pattern overthe course of a tidal cycle. The residual circulation

* Corresponding author; e-mail: c1chen@umassd.edu.

pattern is intensified significantly and shifteddownstream when freshwater discharged is includ-ed. In addition, the Satilla River estuary is bound-ed by shorelines that feature complex curvature.The cross-estuary residual current is generallystronger near the curved river shoreline, which isa result of the imbalance between the centripetaland cross-estuary pressure gradients (Fischer et al.1979; Geyer 1993; Ridd et al. 1998).

One of the most important features of the SatillaRiver estuary is the existence of extensive intertidalsalt marshes. These marshes are completelyflushed at slack high water and remain dry at slacklow water. They act like a water-absorber that di-rectly accelerates water movement inside the estu-ary over a tidal cycle. This can be seen in tidalsimulations using a fully 3-D primitive-equationsmodel (Zheng et al. 2003), which shows that tidalcurrents in the main channel of the estuary canbe increased 40–50% when the flooding-drying cy-cle over the intertidal salt marshes is included. Theflooding-drying process over intertidal salt marshesalso tends to enhance the asymmetry of tidal cur-rents over a tidal cycle, resulting in a relativelylarge residual flow along the estuary. The impor-tance of tidal asymmetry in transport and accu-mulation of sediment in an estuary has been welldocumented (Fitzgerald and Nummedal 1983),

Suspended Sediment Model in a Georgia Estuary 671

and we were interested in extending our hydro-dynamic model to examine sediment dynamics.

A comprehensive interdisciplinary survey wasconducted in the Satilla River estuary duringspring tide in April 1995. Optical backscatter mea-surements at slack high water revealed a non-uni-form spatial distribution of total suspended sedi-ment (TSS) concentration along the estuary tran-sect (Blanton et al. 1999) with two maximum TSSconcentrations exceeding 150 mg l21 near the bot-tom: one 13 km upstream from the mouth of theestuary where the shoreline has a significant con-cave bend and the other near the estuary mouth(see Fig. 1a in Zheng et al. 2003). At an anchorsite, a distance of 16 km from the mouth of theestuary, the TSS concentrations varied consider-ably over a tidal cycle. They were highest at maxi-mum flood and ebb tides and lowest at slack highand low tides (see Fig. 1b in Zheng et al. 2003).Maximum TSS concentration occurred about 1.5h after maximum ebb current and 2 h after max-imum flood current (Blanton et al. 1999). Aroundmaximum ebb tide, TSS concentrations exceeding1,000 mg l21 were found throughout the water col-umn. Around maximum flood tide, TSS concen-trations exceeding 1,000 mg l21 were trapped nearthe bottom and decreased toward the surface. Atslack low water, TSS concentration near the surfacewas more than 100 mg l21. At slack high water, how-ever, it was trapped near the bottom and the near-surface concentration was only about 20 mg l21.

What are the physical processes that cause thenon-uniform spatial distribution and asymmetricaltemporal variation of TSS concentration in the Sa-tilla River estuary? Generally speaking, the distri-bution of TSS in an estuary is controlled by com-plex nonlinear processes associated with the inter-action between wind-induced waves, asymmetricaltidal horizontal advection and vertical mixing,short-term and long-term variations in estuarinecirculation, stratification, bottom stress (resuspen-sion), particle settling velocity (deposition), andflocculation-deflocculation processes (Postma1967; Dyer 1986; Sanford et al. 1991; Geyer 1993;Uncles and Stephens 1993; Jay and Musiak 1994;Lou and Ridd 1997). Which of these processes isdominant in the Satilla River estuary? To ourknowledge, this question has not been addressedin previous modeling studies in that area.

We used a coupled 3-D hydrodynamic and sus-pended sediment model to examine the physicalprocesses that control the non-uniform spatial dis-tribution of TSS concentration in the Satilla Riverestuary. This coupled model was developed basedon a 3-D hydrodynamic model of the Satilla Riverestuary with inclusion of numerical treatment forthe flooding-drying cycle over the intertidal salt

marshes (Zheng et al. 2003). The suspended sed-iment model consists of a 3-D passive tracer equa-tion with inclusion of sinking, resuspension, andsedimentation processes (Ariathurai and Krone1976). We used the model to perform a retrospec-tive simulation of the observations of April 1995described above. This simulation helped us verifythat the model captured the basic spatial and tem-poral distribution patterns of the observed TSS inthe Satilla River estuary. A series of process studieswas also carried out to explore the physical drivingmechanisms responsible for the observed distri-bution of TSS.

Suspended Sediment Model

The TSS model used in this study is a 3-D simplepassive tracer equation with inclusion of sinking,sedimentation, and resuspension processes. Ignor-ing the feedback effects of TSS on fluid motion aswell as on flocculation and deflocculation process-es, the TSS concentration in the water column canbe estimated by a concentration equation as fol-lows:

]C ]uC ]vC ](w 2 w )Cs1 1 1]t ]x ]y ]z

] ]C5 K 1 F (1)h c1 2]z ]z

where C denotes the suspended sediment concen-tration, u, v, and w are the x, y, and z componentsof the fluid velocity, Kh is the vertical eddy diffusiv-ity, ws is the settling velocity of suspended sedi-ment, and Fc represents horizontal diffusion pro-cesses. The vertical eddy diffusivity Kh is deter-mined by the turbulence intensity estimated fromMellor and Yamada’s (1982) level 2.5 turbulentclosure scheme.

The surface boundary condition for C is speci-fied as no sediment flux, i.e.,

]Cw C 1 K 5 0, at z 5 z(x, y, t) (2)s h ]z

where z is surface elevation. At the bottom, sedi-ment flux is specified as the difference betweenresuspension and sedimentation, so that

]C2w C 2 K 5 F 2 F , at z 5 2H(x, y) (3)s h e s]z

where

ce 1(zt z 2 t ) for B . 0 b cer0F 5 and (4)e 0 elsewhere

672 L. Zheng et al.

TABLE 1. Parameters of suspended sediments used in themodel in the Satilla River estuary. ws is settling velocity of thesediment and tce is critical stress for resuspension.

Sediment Type ws (cm s21) tce (kg s21 m22)

SandSettleableNonsettleable

2.818.8 3 1021 b

2.0 3 1022 c

0.196a

0.196b

0.25 3 1023 c

a Middleton (1976); b Blake et al. (2001); c Alexander (unpub-lished data).

w Cs b 1F 5 (t 2 zt z) (5)s cs btcs

Fe and Fs are the suspended sediment fluxes nearthe bottom caused by resuspension and sedimen-tation, respectively, ce is a proportionality factor(constant value of 0.5), Cb is near-bottom suspend-ed sediment concentration, tb is bottom shearstress, tce is critical shear stress for resuspension,and tcs is critical shear stress for sedimentation. Inthis study, tcs is given the same value as tce. Thesuperscript 1 is an indicator of Heaviside’s opera-tor. B is the bottom suspended sediment pool,which is refilled by sedimentation and emptied byresuspension as

]B5 F 2 F (6)s e]t

The resuspension of bottom sediment only occurswhen both the bottom shear stress is larger thancritical shear stress for resuspension and the bottompool of sediment is available. With no flocculationand deflocculation processes, the interaction be-tween different sizes of sediments can be neglected.This assumption allows us to divide the TSS intothree individual groups: sand, settleable particles(silt and flocs), and nonsettleable (clay and smallparticles) particles (Table 1), and calculate theirconcentrations separately. The model-predicted TSSconcentration is equal to a sum of the concentra-tions of sand, settleable, and nonsettleable particles.The critical shear stresses for resuspension and sed-imentation of the three types of sediment shown inTable 1 are determined based on literature values(Middleton 1976; Blake et al. 2001) and unpub-lished data (Alexander personal communication).

To simplify our modeling experiments, two ad-ditional assumptions were made. No feedback ef-fects of suspended sediment on water density wereconsidered, which allowed us to easily couple thesediment model with the hydrodynamic modelwithout modifying estuarine circulation and waterdensity. The processes of sinking, resuspension,and sedimentation of particles over the intertidalsalt marsh area were also not taken into account.This assumption was made based on mass conser-

vation, by which the inflow and outflow of the sus-pended sediment over salt marshes should be bal-anced in a climatologically averaged sense. Directmodel-data comparisons under this simplificationhelped us identify and quantify the importance ofintertidal salt marshes on TSS concentrations inthe Satilla River estuary.

To examine how physical processes affect thespatial and temporal variation of TSS concentra-tion, we assumed that suspended sediment concen-tration was initially uniformly distributed through-out the computational domain. This simplificationmade it easy to test the hypothesis that the spatialand temporal distributions of TSS were caused byphysical processes associated with estuarine circu-lation, tidal mixing, and mixing between fresh andoceanic water. In the Satilla River estuary, obser-vations revealed that the TSS mainly consisted ofsand, settleable, and nonsettleable sediments (Al-exander personal communication). The averagesize of a grain of sand is about 177 mm and itssettling velocity is 2.81 cm s21 (Middleton 1976).Because sand sinks at a fast rate (it takes only 3–5min to sink from surface to bottom given a meandepth of 5 m), it is generally trapped near the bot-tom. Its spatial and temporal distribution shouldbe most affected by asymmetric tidal currents andstresses near the bottom. Settleable particles, suchas flocs, are about 200 mm in size and have a set-tling velocity of 0.88 cm s21 (Blake et al. 2001). Thesettling time scale for this kind of sediment in 5 mof water is only about 10–15 min, which is muchshorter than the tidal cycle. This type of sedimentwould also be trapped in the lower water columnwith a similar spatial distribution as the sand. Thenonsettleable particles are very small and their set-tling velocity was specified as 0.02 cm s21 in themodel (Alexander unpublished data). Since thesettling time scale of these sediments in 5 m ofwater is approximately 7 h, which is slightly longerthan half of a cycle of the semi-diurnal tide, theyshould be influenced significantly by asymmetrictidal advection and mixing and vertical stratifica-tion (Geyer 1993). Our numerical experimentstarted at neap tide slack low water on April 7,1995. Given the fast settling velocities and largecritical stresses for resuspension of sand and settle-able particles (Table 1), we assumed that these par-ticles would be found near the bottom at slack lowwater. The initial carbon (C) values were specifiedas zero for these two fractions. Several numericalexperiments were conducted to test the sensitivityof the model-predicted distribution of sedimentsto the initial conditions, and no significant differ-ences were found between sand and settleable par-ticles. Observations conducted in the Satilla Riverindicated that the average concentration of the

Suspended Sediment Model in a Georgia Estuary 673

nonsettleable particles was 15 mg l21 in the upper1 m below the surface and 45 mg l21 in the lower1 m above the bottom during neap tide (Alexan-der unpublished data). The initial C value of thenonsettleable particles was specified as 15 mg l21,the value observed near the surface.

One important parameter in the suspended sed-iment model is the bottom sediment pool, becauseits availability directly controls resuspension (i.e.,when the bottom sediment pool approaches zero,bottom sediment resuspension will not occur evenwhen bottom shear stress is larger than the criticalshear stress for resuspension). Observations in theSatilla River estuary in April 1995 showed that bot-tom sediments were spatially variable throughoutthe estuary. In the region between 6 and 10 kmfrom the estuary mouth, the bottom sediment wasdominated by sand. In the rest of the region, how-ever, it was dominated by settleable particles (Al-exander unpublished data). On average, 30% ofthe material was sand and the rest was settleableparticles (Alexander personal communication). Tosimplify, we specified initial bottom sediment poolsfor the three sediments’ fractions as constant val-ues over the entire computational domain. Thisspecification is consistent with our objective, whichwas to determine the extent to which the observedspatial variations in suspended sediment concen-tration in the Satilla River estuary are caused byphysical processes associated with tidal motion.The total concentration of sand and settleable par-ticles near the bottom after the maximum currentduring spring tide was observed to exceed 6,000mg l21 in the field measurement. This is 60 timeslarger than the maximum concentration observedin the water column of the along-estuary transect(Blanton et al. 1999). For this reason, an infinitebottom sediment pool was specified for both sandand settleable particles, which means that as longas bottom shear stress is larger than the criticalshear stress for the resuspension of sand and set-tleable particles, then sand and settleable particleswill be available in the bottom sediment pool forresuspension. No bottom sediment pool was spec-ified for the nonsettleable particles (Table 1) sincethey are considered to be in permanent suspen-sion in the water column.

The suspended sediment model was coupledwith the Satilla River estuary 3-D hydrodynamicmodel developed in Zheng et al. (2003), which in-cludes the flooding-drying cycle over intertidal saltmarshes. The hydrodynamic model is driven by M2,S2, and N2 tidal forcings at the open boundary overthe inner shelf of the South Atlantic Bight andreal-time freshwater discharge at the upstream endof the estuary. A detailed description of the designof the numerical experiments for tidal and salinity

simulation is given in Zheng et al. (2003). At first,the model was run with tidal forcing only. Whentidal elevations and currents reached a quasi-equi-librium state, the tidal currents, surface elevation,and turbulent mixing coefficients at the time ofneap tide slack low water were stored as an initialflow field for the coupled hydrodynamic and sed-iment model. Then the coupled model ran prog-nostically for an additional 9 d (until April 16,1995) under the initial physical condition specifiedusing the along-estuary distribution of salinity mea-sured on April 7, 1995. A spatially uniform sedi-ment concentration was specified as the initial con-dition for the sediment model. This is consistentwith our interest in determining the extent towhich the spatial structure of suspended sedimentis driven by physical processes.

Model ResultsThe coupled model reasonably reproduced the

along-estuary distribution of TSS observed in theSatilla River estuary in April 1995, as shown in Fig.1. The model predicted a non-uniform distributionof TSS in an along-estuary transect, with two max-ima near the bottom: one near the mouth of theestuary and the other 13 km upstream where theshoreline has a significant concave bend (Fig. 2).In these two maxima regions, TSS concentrationhas a large vertical gradient near the bottom andis uniform within 3 m of the surface. The TSS con-centration has a maximum value of more than 250mg l21 near the bottom and a minimum value ofabout 20 mg l21 near the surface. These locationsof high TSS concentrations are the same as thosefound in the observations (Fig. 1).

The model-predicted temporal variation of TSSconcentration was also in reasonable agreementwith the data taken at an anchor station over oneM2 tidal cycle on April 16, 1995 (see Fig. 1 inZheng et al. 2003 for the location and Figs. 1 and2 for the model-data comparison). The model-pre-dicted TSS concentration at that site varied peri-odically with the M2 tidal cycle; it showed two max-ima near the bottom 1 to 2 h after maximum ebband flood currents and two minima at slack highwater and slack low water, which is consistent withobservations. During maximum flood and ebbtides, the TSS concentration in the model reached1,000 mg l21 near the bottom and rapidly de-creased upward to 100 mg l21 near the surface.During slack waters, the TSS concentration in thewater column was less than 100 mg l21.

At maximum flood and ebb tidal currents, themodel-predicted TSS concentration near the bot-tom was about 30% sand and 70% settleable par-ticles. The model also showed that upward resus-pension was weaker during ebb tide than it was

674 L. Zheng et al.

Fig. 1. The distributions of TSS observed along the estuaryon April 15, 1995 (upper panel), and at an anchor site over atidal cycle (lower panel) on April 16, 1995, in the Satilla Riverestuary (ME: maximum ebb, MF: maximum flood, SLW: slacklow water, SHW: slack high water). Dots in the upper and lowerpanels represent measurement locations. During the measure-ment at the anchor station, the surface was selected as the ori-gin of the coordinate (z 5 0). The shaded area in the lowerpanel includes the temporal variation of sea elevation pluschange of bottom depth due to the boat’s shift.

Fig. 2. The model-predicted distributions of TSS along theestuary on April 15, 1995 (upper panel), and at an anchor siteover a tidal cycle (lower panel) on April 16, 1995, in the SatillaRiver estuary for the case with tidal forcing plus real-time fresh-water discharge. Abbreviations as described for Fig. 1. To com-pare with the observation data shown in Fig. 1, the surface wasselected as the origin of the coordinates in the lower panel, sothe shaded area indicates the temporal variation of surface el-evation.

during flood tide. These temporal distribution pat-terns were in reasonable agreement with the ob-servations shown in Fig. 1. On the other hand,there were several differences between observa-tional and model data in terms of spatial variations.First, model-predicted TSS concentrations near thebottom on the along-estuarine transect were about50 mg l21 higher than the observations (Figs. 1 and2). This might be due to the vertical resolution ofthe measurements. Since observations were re-corded 1.5–3 m above the bottom, they might havefailed to resolve higher concentrations near thebottom. When we plotted model-predicted TSSconcentrations at the same depth as the field ob-servations, we obtained comparable TSS concen-trations. Another difference between observationaland model data occurred at the anchor site. Fieldobservations showed significant asymmetry in TSSdistribution over one M2 tidal cycle, with a narrowpeak in suspended sediment around maximum

flood tide and relatively high concentrations nearthe surface during both the ebb tidal phase and atslack low water. Neither of these differences wasresolved in our coupled model simulation. Thesediscrepancies were possibly the result of strong ver-tical mixing caused by the interaction of springtide and surface wind, as well as outflow of sus-pended sediment from the intertidal salt marshesduring ebb tide (Blanton et al. 1999). Since themodel did not include surface wind forcing andprocesses associated with resuspension and sedi-mentation in the intertidal salt-marsh area, we didnot expect good model-data comparison for thesetwo features.

We used information on the different sedimentfractions to explore the along-estuary distributionof model-predicted sand, settleable, and nonsettle-able particles. The TSS was dominated by the set-tleable particles near the bottom and by nonsettle-able particles in the upper water column. An ex-ample can be seen from the along-estuary distri-

Suspended Sediment Model in a Georgia Estuary 675

Fig. 3. Model-predicted, along-estuary distributions of (a)sand, (b) settleable particles, and (c) nonsettleable particlesconcentrations on April 15, 1995, for the case with tidal forcingplus real-time freshwater discharge.

bution of these three sediment fractions at slackhigh water (Fig. 3). The sand is trapped near thebottom and has a maximum concentration ofabout 20–30 mg l21. This amount of material onlyaccounts for 10–15% of the total near-bottom sed-iment concentration (Fig. 3a). These low sand con-centrations are likely due to its relatively fast set-tling velocity. With a settling velocity of 2.8 cm s21,no sand would be able to remain suspended in thewater column longer than 5 min. The settleableparticles are also trapped mainly near the bottomand exhibit two maxima of 200 mg l21 or higher;one occurs near the mouth of the estuary and theother is 13 km upstream (Fig. 3b). The concentra-tions of sand and settleable particles near the sur-face are very small and can be neglected. The mod-el-predicted nonsettleable particles dominated inthe upper 3 m of the water column (Fig. 3c). Theirconcentration is only about 25 mg l21 near the bot-tom and about 10 mg l21 near the surface.

The nonsettleable fraction exhibited spatial var-iation along-estuary, with the highest concentra-tions occurring at 6 and 23 km upstream. It shouldbe noted that these high concentrations are locat-ed in the regions of weaker vertical stratification

(Fig. 8 in Zheng et al. 2003), which is consistentwith a previous study by Geyer (1993). Geyer(1993) demonstrated that when the settling veloc-ity of sediment in stratified water is between 0.01and 0.1 cm s21 (intermediate-size such as nonset-tleable particles), it can accumulate in the verticalstratification region where a bottom convergencezone exists. Smaller particles are almost perma-nently suspended in the water column and coarserparticles (such as sand and settleable particles) cansink to the bottom quickly, so the effect of strati-fication on the spatial distributions of these smalleror coarser sediments becomes insignificant.

Mechanism StudiesBoth the observations and model results pre-

sented here show that TSS is distributed non-uni-formly along the Satilla River estuary, with highestconcentrations occurring near the mouth of theestuary as well as near the bend located 13 kmupstream. What are the physical driving mecha-nisms responsible for such a spatial distribution ofTSS? Are all the physical processes associated withhorizontal convergence and divergence, buoyancy-induced flow, shoreline curvature effects, asym-metrical tidal mixing, and the flooding-drying cy-cle over the intertidal salt marshes equally impor-tant or are some of these processes dominant? Toaddress these questions, a model-guided mecha-nistic study was carried out by examining the rolesof the individual physical forcings and vertical eddyviscosity in the spatial distribution of TSS. Threenumerical experiments were conducted underspecified conditions with no river discharge, con-stant vertical eddy viscosity, and no inclusion of thesalt-marsh intertidal zone, respectively. In the firstcase, salinity was specified at a constant value of 35psu throughout the model domain. In the secondcase, the vertical eddy viscosity was specified as 1.03 1022 m2 s21, which is equal to the tidally andspatially averaged value of the model-predicted ver-tical eddy viscosity coefficient from the Mellor andYamada’s (1982) level 2.5 turbulent closure model.

The model results show that removing freshwa-ter discharge does not alter the along-estuary dis-tribution pattern of TSS, i.e., two maxima near thebottom on the along-estuary transect. The concen-tration, however, becomes much smaller, with aTSS concentration of about 100 mg l21 near thebottom as compared to 250 mg l21 (Fig. 4 left pan-el). This suggests that freshwater discharge is notan essential physical factor that is responsible forthe occurrence of these two TSS maxima at thenear mouth of the estuary and a distance of 13 kmupstream. The decrease in magnitude of TSS isdue to the fact that in the absence of freshwaterdischarge-induced vertical stratification, the verti-

676 L. Zheng et al.

Fig. 4. Model-predicted, along-estuary distribution of total,sand, settleable particles, and nonsettleable particles concentra-tions on April 15, 1995, for the cases with no freshwater dis-charge and constant background salinity distribution (left pan-el) and constant vertical eddy viscosity (right panel).

Fig. 5. Model-predicted, along-estuary distribution of total,settleable particles, and nonsettleable particles concentrationson April 15, 1995, for the case without inclusion of the flooding-drying process over intertidal salt marshes.

cal shear of the horizontal current decreases dra-matically. As a result, the bottom stress drops andless sediment can be resuspended from the bottominto the water column. When we look separately atthe along-estuary distributions of sand, settleable,and nonsettleable particle fractions (Fig. 4 left pan-el), we see that the contribution of sand is ex-tremely low and the TSS consists mostly of settle-able particles near the bottom and of nonsettleableparticles near the surface. The nonsettleable par-ticles are distributed uniformly along the estuarywith a value of about 10 mg l21. The fact that thetwo maxima of nonsettleable particles disappearunder these conditions suggests that the spatial dis-tribution of this type of sediment is effectively con-trolled by vertical stratification. This is, again, con-sistent with Geyer’s (1993) finding that the tem-poral and spatial distributions of fine sediments inestuaries are closely associated with vertical strati-fication. It should be noted that the concentrationof nonsettleable particles accounts for less than 5%of total TSS concentration near the bottom.

When the eddy viscosity coefficient was held con-stant during the model run, the model-predictedTSS again exhibited two maxima at the same lo-cations as in the observations (Fig. 4 right panel).

This implies that asymmetrical tidal mixing due totemporal and spatial variations of current shear isnot an essential mechanism responsible for the ob-served spatial distribution of TSS in the Satilla Riv-er estuary. Individually, the distributions of sandand settleable particles are similar to those ob-served in the full model run, where we used vari-able vertical viscosity values predicted from theMellor and Yamada’s (1982) level 2.5 model (Figs.3b,c and 4 right panel). However, the two maximaof nonsettleable particles observed near the bot-tom in the regions with relative strong verticalstratification disappear under this run, suggestingthat asymmetric tidal mixing is more essential fornonsettleable particles than for settleable particlesand sand.

Removing the flooding-drying cycle leads to a30% reduction of TSS concentration. It also causesthe location of the near-bottom maximum TSSconcentration to shift downstream, so that it is lo-cated about 9 km upstream from the estuary asopposed to 13 km upstream (Fig. 5). Such a sig-nificant reduction of TSS is due to a significantunderestimation of tidal currents and bottomshear stress in the case without inclusion of flood-

Suspended Sediment Model in a Georgia Estuary 677

Fig. 6. Synoptic distribution of near-surface residual currentvectors on April 15, 1995, for the case with tidal forcing plusreal-time freshwater discharge. The location of maximum ob-served TSS is marked as site A. Two surrounding sites B and Care marked for comparisons of model results.

ing-drying cycle (Zheng et al. 2003). This resultsin a large decrease in the concentration of settle-able particles (Figs. 3b and 5) and the disappear-ance of sand from the estuary. The nonsettleableparticles are trapped in the lower water columnunder these conditions. This results from the factthat vertical mixing is significantly reduced and thesettling process becomes relatively important whenthe flooding-drying cycle of the intertidal saltmarshes is removed (Fig. 5). The shift in the lo-cation of the TSS maximum is a result of the rel-ative importance of tidal and buoyancy flows.Freshwater discharge-induced buoyancy flow is oneorder of magnitude smaller than that of tidal cur-rents. When tidal currents become weak underfixed lateral boundary conditions, this offshorebuoyancy flow becomes relatively more importantand drives the TSS downstream.

DiscussionThe coupled 3-D hydrodynamic and suspended

sediment model has reasonably reproduced thespatial and temporal distribution of TSS observedin the Satilla River estuary. Several model experi-ments have been conducted to examine the phys-ical mechanism responsible for spatial and tem-poral distributions of TSS. These model experi-ments imply that neither asymmetric tidal mixing,freshwater discharge, nor the flooding-drying cycleover the intertidal salt marshes is the essentialphysical driving mechanism responsible for the twonear-bottom TSS maxima observed near themouth of the Satilla River estuary and 13 km up-stream where the shoreline has a significant bend.

Previous study has revealed that the distributionsof TSS in an estuary are closely related to estuarinecirculation patterns (Festa and Hansen 1978). Thehydrodynamic model (Zheng et al. 2003) predict-ed that residual currents in the estuary are char-acterized by multiple convergent and divergent cir-culation cells (Fig. 6). These complex residual flowpatterns have been recently demonstrated usingtowed-ADCP measurements collected from the Sa-tilla River estuary (Seim personal communication).The physical mechanisms responsible for the for-mation of these convergent and divergent flows arethe buoyancy-induced along-estuary pressure gra-dient, bottom topographic tidal rectification, in-ertial curvature shoreline effects, and asymmetryof tidal currents associated with the flooding-dry-ing cycle of intertidal salt marshes (Zheng et al.2003). It is interesting to find that the two maxi-mum TSS concentrations shown in Fig. 1 are lo-cated in a strong surface convergent flow zone(Fig. 6). In the surface-convergent flow, based onmass conservation there will be a downwellingzone that will enhance the sinking velocity of sus-

pended sediment. When suspended sediment iscarried from the surrounding area to the conver-gent zone, it will quickly sink and deposit on thebottom due to its increased settling velocity. Thehorizontal convergence of the surface residual cir-culation is probably one of the mechanisms re-sponsible for the formation of the two observedTSS maxima. This result is consistent with the find-ings by Jewell et al. (1993) on the Amazon conti-nental shelf, where they found that the sedimentaccumulation rate or near-bottom suspended sed-iment concentration was largest in a zone of con-vergence.

More evidence for the idea that convergencepatterns are important in determining TSS distri-bution can be seen in the cross-estuary distributionof residual vertical velocity at three selected sites(A, B, and C shown in Fig. 6). At site A, where thelargest shoreline bend is located, the horizontallyconvergent flow near the surface causes a down-ward motion that hastens sinking of suspendedsediment (Fig. 7a). At sites B and C, the horizon-tally divergent flow near the surface leads to a re-markable upward motion (upwelling zone) thatcancels the settling velocity of suspended sedimentand tends to carry material away from these tworegions (Fig. 7b,c).

The model-predicted along-estuary distributionof TSS concentration is also likely related to thespatially non-uniform distribution of bottom stress,which is largest at site A and smaller at sites B andC (Fig. 8a,b). One of the most important sourcesfor suspended sediment in the water column is theresuspension of bottom sediment. Equation 4 sug-

678 L. Zheng et al.

Fig. 7. Cross-estuary distribution of residual vertical velocity(3 1023 cm s21) at three selected sites A, B, and C on April 15,1995, for the case with tidal forcing plus real-time freshwaterdischarge. Filled triangles show the locations of field observa-tions. The contour interval is 1 3 1022 cm s21.

Fig. 8. Along-estuary distribution of depth-averaged TSSconcentration (upper) and bottom stress (lower) averaged overan M2 tidal cycle on April 15, 1995. Filled triangles show thelocations of sites A, B, and C shown in Fig. 6.

gests that the amount of suspended sediment re-suspension is proportional to the difference be-tween bottom stress and critical shear stress for re-suspension if bottom sediment is available. Ourmodel results suggest that TSS near the bottom isdominated by settleable particles. Since the initialpool for this sediment fraction was given as infi-nite, the location of maximum TSS should be con-sistent with the place where the largest bottomshear is observed. Thus the maximum TSS con-centration at 13 km upstream might be caused bysurface convergence as well as significant bottomstress that results in more resuspension.

The asymmetric distribution of TSS concentra-tion found in flood and ebb tides, i.e., a relativelyweaker upward resuspension of the sediment inthe water column (Fig. 2b) during ebb tidal phasecompared to that during flood tidal phase, is clear-ly caused by asymmetrical tidal mixing over a tidal

cycle. The model-predicted vertical eddy viscosityis stronger during flood tidal phase, which suggestsstronger vertical mixing (Fig. 9a). Weaker tidalmixing during ebb tidal phase is caused by the su-perimposition of gravitational and tidal flows, lead-ing to a relatively large vertical stratification and adecrease of vertical mixing. A relatively strongerbottom current speed is found during flood tidalphase that leads to a larger bottom stress and moreactive upward sediment resuspension during thistidal phase (Fig. 9b).

The 1–2-h time lag for the occurrence of maxi-mum TSS concentration relative to the maximumtidal current at ebb and flood tides is believed tobe due to the fact that the bottom shear stress with-in 1 to 2 h after maximum current is still largeenough to produce bottom sediment resuspensiondominant over sedimentation. This results in netsuspended sediment transport from bottom sedi-ment. When it is at the stage that the amount ofresuspension is the same as that of sedimentation,the TSS concentration reaches its maximum. Afterthis point, the sedimentation process becomesdominant over the resuspension process, resultingin net sediment deposition and a decrease of TSSconcentration. Due to a larger near-bottom veloc-ity during the flood tidal phase, a longer time lagoccurs during this period, which is consistent withboth observations (Fig. 1b) and model results (Fig.2b).

Suspended Sediment Model in a Georgia Estuary 679

Fig. 9. Time sequence of the distribution of vertical eddyviscosity Kh (3 1023 m2 s21) (upper panel) and bottom currentspeed (lower panel) over an M2 tidal cycle on April 15, 1996for the case with tidal forcing plus real-time freshwater dis-charge. The location of this plot was at the anchor site (see Fig.1 in Zheng et al. 2003). The surface was selected as the originof the vertical coordinate and the shaded area in the upperpanel represent the temporal variation of surface elevation overa tidal cycle.

ACKNOWLEDGMENTS

This research was supported by the Georgia Sea Grant Col-lege Program under grant numbers NA26RG0373 andNA66RG0282 for Dr. Changsheng Chen, and by the NationalScience Foundation Land-Margin Ecosystems Research ProgramGrant No. DEB-9412089 for Dr. Merryl Alber. Dr. LianyuanZheng was supported by Dr. Chen’s Sea Grant funds. We thankDr. Jack Blanton of Skidaway Institute of Oceanography (SKIO)for providing the observational data used for validating themodel. These observations were collected through support toSKIO provided by the Georgia Coastal Zone Management Pro-gram (Grant No. RR100-279/9262764), the National ScienceFoundation (LMER Grant No. DEB-9412089), (LTER Grant No.OCE-9982133), and a grant from the Georgia General Assembly.We thank Dr. Mac Rawson for his encouragement and help inproject organization. We would like to thank Dr. Clark Alexan-der for permission to use his observed suspended sedimentdata. The 2-m elevation line was obtained from the GIS databasecreated by Dr. Alice Chalmers; this work would be impossiblewithout her kind help. We also want to thank Mr. George Da-vidson and anonymous reviewers for their editorial help.

LITERATURE CITED

ARIATHURAI, R. AND R. B. KRONE. 1976. Mathematical modelingof sediment transport in estuaries, p. 98–106. In M. Wiley(ed.), Estuarine Processes. Academic Press, New York.

BLAKE, A. C., G. C. KINEKE, T. G. MILLIGAN, AND C. R. ALEXAN-DER. 2001. Sediment trapping and transport in the ACE basin,South Carolina. Estuaries 24:721–733.

BLANTON, J., C. R. ALEXANDER, M. ALBER, AND G. KINEKE. 1999.The mobilization and deposition of mud deposits in a coastalplain estuary. Limnologica 29:293–300.

DRONKERS, J. 1986. Tidal asymmetry and estuarine morphology.Netherlands Journal of Sea Research 20:117–131.

DYER, K. R. 1986. Coastal and Estuarine Sediment Dynamics.John Wiley and Sons, New York.

FESTA, J. F. AND D. V. HANSEN. 1978. Turbidity maxima in par-tially mixed estuaries: A two-dimensional numerical model.Estuarine, Coastal and Shelf Science 7:347–359.

FISCHER, H. B., B. E. LIST, R. C. Y. KUO, J. IMBERGER, AND N. H.BROOKS. 1979. Mixing in Inland and Coastal Waters. Academ-ic Press, New York.

FITZGERALD, D. M. AND D. NUMMEDAL. 1983. Response charac-teristics of an ebb-dominated tidal inlet channel. Journal ofSedimentary Petrology 53:833–845.

GEYER, W. 1993. The importance of suppression of turbulenceby stratification on the estuarine turbidity maximum. Estuaries16:113–125.

JAY, A. J. AND J. D. MUSIAK. 1994. Particle trapping in estuarinetidal flows. Journal of Geophysical Research 99:445–461.

JEWELL, P. W., R. F. STALLARD, AND G. L. MELLOR. 1993. Numer-ical studies of bottom shear stress and sediment distributionon the Amazon continental shelf. Journal of Sedimentary Petrol-ogy 63:734–745.

LOU, J. AND P. V. RIDD. 1997. Modelling of suspended sedimenttransport in coastal areas under waves and currents. Estuarine,Coastal and Shelf Science 45:1–16.

MELLOR, G. L. AND T. YAMADA. 1982. Development of a turbu-lence closure model for geophysical fluid problems. Review ofGeophysics 20:851–875.

MIDDLETON, G. V. 1976. Hydraulic interpretation of sand sizedistributions. Journal of Geology 84:405–426.

POSTMA, H. 1967. Sediment transport and sedimentation in theestuarine environment, p. 158–179. In G. H. Lauff (ed.), Es-tuaries, American Association Advanced Scientific Publication83. Washington, D.C.

RIDD, P. V., T. STIEGLITZ, AND P. LARCOMBE. 1998. Density-drivensecondary circulation in a tropical mangrove estuary. Estua-rine, Coastal and Shelf Science 47:621–632.

SANFORD, L. P., W. PANAGEOTOU, AND J. P. HALKA. 1991. Tidalresuspension of sediments in northern Chesapeake Bay. Ma-rine Geology 97:87–103.

UNCLES, R. J. AND J. A. STEPHENS. 1993. The freshwater-saltwaterinterface and its relationship to the turbidity maximum in theTamar estuary, United Kingdom. Estuaries 16:126–141.

ZHENG, L., C. CHEN, AND H. LIU. 2003. A modeling study of theSatilla River estuary, Georgia. Part I: Flooding-drying processand water exchange over the salt marsh-estuary-shelf com-plex. Estuaries 26:651–669.

SOURCES OF UNPUBLISHED MATERIALS

ALEXANDER, C. Unpublished Data. Skidaway Institute of Ocean-ography, Savannah, Georgia 31411.

SEIM, H. Personal Communication. Department of Marine Sci-ences, University of North Carolina, Chapel Hill, North Car-olina 27599.

Received for consideration, December 1, 2000Revised, June 5, 2002

Accepted for publication, July 9, 2002