Controls on suspended aggregate size in partially mixed estuaries

Transcription

1 Estuarine, Coastal and Shelf Science 58 (2003) Controls on suspended aggregate size in partially mixed estuaries David C. Fugate*, Carl T. Friedrichs College of William and Mary, School of Marine Science, Virginia Institute of Marine Science, Gloucester Point, VA 23062, USA Received 15 October 2002; accepted 26 March 2003 Abstract Knowledge of aggregate size in estuaries is important to determining the fate and transport of suspended sediment and particle adherent contaminants. We have used a suite of in situ instruments to determine the controls of aggregate size distributions in three muddy, partially mixed estuaries in the mid-atlantic USA. A novel method is presented to estimate turbulent kinetic energy (TKE) production and the resulting Kolmogorov microscale (k K ) using a profiling acoustic Doppler velocimeter that has been contaminated by boat motion. The physical processes that control particle size distribution differ in the three estuaries due to the different hydrodynamics and benthic characteristics. Controls within each estuary also vary with different depth regimes. Surface particle size dynamics in all the studied estuaries are affected by irregular advection events. In the hydrodynamically energetic York River, mid-depth regions are controlled tidally by the combined processes of small k K decreasing particle size at high TKE and differential settling increasing particle size during lower TKE, more stratified conditions. Mid-depth regions in the lower energy Elizabeth River are controlled by irregular resuspension and trapping at the pycnocline of large low density particles. Bottom regions in all estuaries are most strongly influenced by resuspension, tidally in the energetic estuaries and irregularly in the low energy estuary. Near-bed particle size distributions are controlled by both k K and the distribution of particles in the bed in the higher energy estuaries. Just above the bed, large porous particles survive resuspension in the lower energy Elizabeth River, particles become smaller with decreased k K in the more energetic York River, and biological aggregation causes large dense particles to resist turbulent breakup in the Chesapeake Bay, which has a more active benthic community. The net result just above the bed is that particle size and settling velocity are positively correlated to TKE production and sediment concentration in the estuary with higher currents and a biologically active bed, negatively correlated in the estuary with higher currents and a bed reworked by rapid erosion and deposition, and poorly correlated in the estuary with lower currents and a disturbed and contaminated bed. Ó 2003 Elsevier Ltd. All rights reserved. Keywords: particle aggregation; particle fall velocity; sediment transport; bioturbation; cohesive sediment 1. Introduction One of the major obstacles to predicting the fate and transport of sediment and particle adherent contaminants in estuaries is the determination of the size and settling velocity of particles. The settling velocity of the particles depends upon their size and density, qualities which are difficult to measure due to the fragile nature of aggregates in the water column and the different * Corresponding author. addresses: undave@vims.edu (D.C. Fugate) cfried@vims. edu (C.T. Friedrichs). densities of different types of particles. A further complication is that particle characteristics may vary over short and long time scales (Eisma & van Leussen, 1997; Fettweis, Sas, & Monbaliu, 1998; van Leussen, 1994). Important factors determining aggregation in estuaries include suspended sediment concentration, turbulent shear in the water, differential settling of the particles, and the amount of sticky organic compounds in the water (Eisma et al., 1991). The relative importance of each of these factors depends upon the physical and biological regimes of the specific estuaries. This paper will explore aggregate size dynamics in three estuaries located close to each other geographically but with different physical and biological regimes. The /03/$ - see front matter Ó 2003 Elsevier Ltd. All rights reserved. doi: /s (03)

2 390 D.C. Fugate, C.T. Friedrichs / Estuarine, Coastal and Shelf Science 58 (2003) first study area is the York River, VA, a physically energetic estuary (tidal currents on the order of 100 cm s ÿ1 ) where rapid erosion and deposition are the dominant mixing process in the bed. The second is the Cherrystone site in the lower Chesapeake Bay, which also is hydrodynamically energetic but where intense bioturbation is the dominant mixing process in the bed. Lastly, the Elizabeth River, VA, a low energy, dredged and polluted estuary will be considered Particle dynamics in estuaries Factors that affect the formation and breakup of particle aggregates in estuaries include the intensity and scale of turbulent shear, differential settling of particles, Brownian motion, suspended sediment concentrations, prevalence of sticky polymer compounds from biota, and salinity. Previous work has found that the physical processes in the water column are the most important of these in muddy estuaries, namely, turbulence, differential settling, and sediment concentration, although the organic constituent may also play an important role (ten Brinke, 1994; Chen, Eisma, & Kalf, 1994; van Leussen, 1999). Salinity has been postulated to affect the fragility of particle aggregates (Eisma et al., 1991). However, the effects of varying salinity on flocculation are only important at salinity of less than 1 3 (Gibbs, Tshudy, Konwar, & Martin, 1989), which is well below the levels at the estuarine sites under consideration in this paper. The relative importance of the physical factors listed above at different elevations in water column and in different environments will be used to identify separate depthdependent aggregation dynamic regimes and determine how they vary as a function of environmental setting. Suspended sediment concentration may be related to aggregate size through its effect on particle encounter frequency (Dyer, 1989; van Leussen, 1994). Variations in concentration might be expected to be important to particle size where collision frequency is low enough to become a constraining factor on the size which the particles may attain, for example near the surface of the water column. Alternatively, concentration and particle size may be correlated mainly due to their mutual dependence on turbulent shear. Examination of the disparate results of several investigations (Berhane, Sternberg, Kineke, Milligan, & Kranck, 1997; ten Brinke, 1994; Chen et al., 1994; Gibbs et al., 1989; Wolanski & Gibbs, 1995) shows that the effect of suspended sediment concentration on particle size is still unclear. The relationship between concentration and aggregate size will be examined further in this paper. Turbulent shear in the water column also works to increase aggregate size through encounter frequency but at the same time may limit the maximum attainable size. At low turbulence levels, particles may be in a growth stage where aggregate size increases with increasing turbulence. But as turbulent shear increases, it is postulated that the shear becomes limiting to the aggregate size when the length scale of the smallest turbulent eddies, characterized by the Kolmogorov microscale, reduces to the same magnitude as the aggregate diameter, at which point the aggregate is torn apart (Berhane et al., 1997; Hill, Nowell, & Jumars, 1992; van Leussen, 1994). The opposite effects of increased encounter frequency and increased shear on individual particles may work together to determine an equilibrium particle size for a given concentration and turbulence level (Chen et al., 1994). The mean particle size has been found to adjust rapidly to changes in turbulent shear (Chen et al., 1994, Gibbs et al., 1989). These paired processes of growth and destruction are likely to be important in the middle of the water column and near the bed where concentrations and turbulent shear are higher. Differential settling of particles may dominate the aggregation dynamics during times of reduced turbulent energy through an increase in encounter frequency. These falling particles may be concentrated in the pycnocline where their effective density approaches that of the higher density bottom water. Alternatively, if heavy enough, they may continue towards the bed where increased turbulent shear may disaggregate the particles, or they may be maintained at a constant size by the drag force produced by their sinking (Hill & McCave, 2001; Hill, Voulgaris, & Trowbridge, 2001). Consequently, the effects of differential settling are most likely to be seen in the middle of the water column, especially during stratified conditions. Resuspension of particles from the bed is an important process affecting size distribution in the lower water column. Loosely aggregated particles deposited on the bed may be disaggregated upon resuspension, so that higher bed shear stress and a correspondingly smaller Kolmogorov microscale result in smaller particles in suspension. Conversely, the bed may be dominated by consolidated large robust aggregates (e.g. fecal pellets), so that higher stresses will be associated with resuspension of larger denser aggregates in suspension (Komar & Taghon, 1985; Krasnow & Taghon, 1997). The above physical processes and factors are expected to have different degrees of importance at different depths depending upon the depth varying turbulent shear and proximity to different sources of particles. In addition, the importance and response of turbulent shear will be mediated by the overall hydrodynamic and biological conditions of the estuary. Particle size at the surface where suspended sediment concentration is low is expected to be dominated by the effect of concentration on encounter frequency. In the middle water column, at sufficiently high concentrations so that encounter frequency is not the limiting process, high turbulent shear is expected to limit aggregate growth. This was postulated

3 D.C. Fugate, C.T. Friedrichs / Estuarine, Coastal and Shelf Science 58 (2003) by Chen et al. (1994) when they observed small particle dominance during periods of high velocity in the Elbe Estuary. At lower levels of turbulence, for example during slack tides, differential settling is expected to become the dominant process. Near-bed suspended size distributions are expected to more closely mirror the size distribution of particles lying on the bed. Schubel (1971) observed larger particles in suspension during periods of high velocity in the Chesapeake Bay and also attributed this to resuspension of progressively larger bottom material. It is postulated here that the near-bed particle size distribution in muddy partially mixed estuaries is generally dominated by resuspension, however, the response of this distribution to high bed stress is mediated by the type of benthic environment. High near-bed stress in muddy estuaries in which bottom sediment mixing is dominated by physical processes are likely to resuspend fine disaggregated particles, while high stress in muddy estuaries in which the bottom is biologically dominated is likely to resuspend more resilient consolidated aggregates. 2. Study sites and field methods 2.1. York River About 50 km long with a central channel 10 m deep, the York River is a partially mixed estuary that empties into the lower Chesapeake Bay just above the James River (Fig. 1). Salinity at the mouth is around 20 (Lin & Kuo, 2001); maximum salinity stratification during this study was 2.5 from top to bottom of the 10 m depth. Near surface tidal currents can reach over 100 cm s ÿ1 on spring tides, and the mean tidal range is 0.7 m at the mouth and 1 m at the head (Schaffner et al., 2001). The study site is at N and W inside the main channel. The width of the river in this region is about 3.5 km, and there is a wide (2.5 km) shoal to the south of the main channel. The bed sediments are almost exclusively mud and physical mixing from rapid deposition and erosion results in physical mixing lengths up to a meter (Dellapenna, Kuehl, & Schaffner, 1998). The dominant macrobenthos in the York River at the sampling site are the euryhaline opportunistic polychaetes Mediomastus ambiseta and Streblospio benedicti. These benthic invertebrates are of a smaller size than the dominant species found in the neighboring Lower Chesapeake Bay and bioturbate the sediments to a much lesser degree. Overall macrofaunal density in April 1999 was found to be around 1000 individuals m ÿ2 (Hinchey, 2003). Disturbance from rapid erosion and deposition results in a reduced species diversity which is nearly as low as the level of environmentally impacted sites (Hinchey, 2003) such as the Elizabeth River Chesapeake Bay The Chesapeake Bay is a large estuary with a length of 200 km and width of up to 30 km. Data were collected at the Cherrystone site in the lower Chesapeake Bay at N and W (Fig. 1). This site and the nearby Wolftrap site have been extensively studied with regard to benthic biology and bed characteristics (Lee, 1995; Schaffner, 1990; Thompson and Schaffner, 2001; Wright, Boon, Xu, & Kim, 1992; Wright, Schaffner, & Maa, 1997). The mean depth at the study site is 14 m and mean tidal range is 0.6 m. During this survey surface currents reached over 100 cm s ÿ1 and salinities ranged from about 17 to 23. Intense bioturbation in the sediment occurs down to 0.5 m and is the dominant bed mixing process (Dellapenna et al., 1998). The benthic community at this site is robust, consisting of communities of large burrowing, suspension, and deposit feeders. (Dauer, Stokes, Barker, Ewing, & Sourbeer, 1984; Dellapenna et al., 1998; Reinharz, Nilsen, Boesch, Bertelsen, & O Connel, 1982; Schaffner, 1990; Wright et al., 1997). Macrofauna density may reach 10,000 individuals m ÿ2 which consists of as much as 200 g wet weight of biomass. The dominant organism is the suspension feeder Chaetopterus variopedatus, but most bioturbation is caused by the large head down, deposit feeding polychaetes including Macroclymene zonalis and Clymenella torquata (Wright et al., 1997), and surface deposit feeders such as Loimia medusa (Schaffner, 1990). Bed sediments have been found to be primarily silts (40 50%) and fine sand sized particles (40 50%) (Schaffner et al., 2001), although these numbers may be misleading, since resilient fecal pellets survive sieving (Schaffner, personal communication). Wright et al. (1997) found that resuspension by currents is confined to a thin surface layer of only a few centimeters. Their analyses suggested that biological processes remove fine silts and clays from this thin surface layer leaving primarily large, low density, sand size pellets and aggregates. Divers frequently observed fecal coils and mounds being reworked by the current at the site Elizabeth River In contrast to the York River and Chesapeake, the Elizabeth River is geographically small with much lower current speeds and a dredged and contaminated bed. It is a partially mixed tributary of the James River located near the mouth of the James, which empties into the lower Chesapeake Bay (Fig. 1). The distance from the mouth of the main stem of the Elizabeth River to the head of its Southern Branch is 23 km, and the average depth of the main channel is about 12 m. Salinity during moderate freshwater flows in the Elizabeth River ranges from around 17 at the mouth to 13 near the head.

4 392 D.C. Fugate, C.T. Friedrichs / Estuarine, Coastal and Shelf Science 58 (2003) Fig. 1. Principle study sites in Virginia, USA. Vertical stratification in salinity under these conditions is about 5 between surface and bottom at the mouth and around 2 within the rest of the estuary (Neilson, 1975). The artificially deep channel reduces spring tide currents to around 50 cm s ÿ1 near the surface, and the mean tidal range is 0.8 m. The channel depth at our sampling site ( N and W) is maintained at 14 m, and the width of the river there is 300 m. The main stem and two of the three major branches of the Elizabeth River have undergone extensive industrial development since colonization of the area by Europeans in the 1600s (Nichols & Howard-Strobel, 1991). Dredging and infilling modified the geometry of the estuary from an irregular bottom profile with broad shoal and marsh areas to a regular, deeper, longitudinal stair step bottom profile with a narrower lateral extent. In addition, landing docks, extended piers, and narrow dredged tributaries create significant lateral roughness in some extents of the river. By 1982, maintenance dredging in the narrow Southern Branch was m 3 yr ÿ1 (Nichols & Howard-Strobel, 1991). Physical mixing from dredging overwhelms bioturbation in the muddy bed of this biologically degraded estuary. Sediments are generally clayey silt to silty clays (Francese & Dvorak-Grantz, 1998). The Southern Branch of the Elizabeth River receives pollution from sewage treatment plants and numerous industrial waste discharges (Hawthorne & Dauer, 1983). Dauer (2002) has determined that the condition of the benthic community in the Southern Branch of the Elizabeth River near the sampling site has been degraded by as much as 92% according to the benthic index of biotic integrity. Hawthorne and Dauer (1983) found macrofauna density of only 1300 individuals m ÿ2 near the site of our deployment. As in the York, the benthic macrofauna are dominated by the relatively small opportunistic polychaetes Mediomastus ambiseta and Streblospio benedicti Data collection and analysis Data from the three surveys discussed here were collected on the following dates: August 26 28, 1996 at the Cherrystone site, November 16 18, 1999 at the

5 D.C. Fugate, C.T. Friedrichs / Estuarine, Coastal and Shelf Science 58 (2003) Elizabeth River, and May 30 31, 2000 at the York River site. Hydrodynamic, physical, and geochemical data were collected similarly for each of the surveys. An instrument profiling lander, or ÔprofilerÕ, containing a suite of physical and hydrodynamic instruments took vertical profiles of the water column about eight times per tidal cycle. The sampling elevations in the Elizabeth River were bottom, 1 m above the bottom (mab), 3 mab, the middle of the water column (6 mab, depending on the tide), and 1 m below the surface. Sampling in the York was similar, but no data were collected at 3 mab. The Cherrystone site was sampled only at the bottom, and acoustic Doppler velocimeters (ADVs) were mounted on a nearby tripod sampling 11 min bursts every 15 min at 5 Hz. All the study sites were sampled for two or more tidal cycles. Each elevation was sampled for a few minutes after which the profiler was left to continually sample at 1 mab until time for the next vertical profile. The instruments on the profiler generally were a Sontek ADV sampling at 10 Hz, a Downing optical back scatter sensor (OBS) in conjunction with an Applied Microsystems conductivity temperature depth recorder (CTD) sampling at 1 2 Hz, a Sequoia Science laser in situ scattering and transmissometer 100 (LISST) sampling at 2 Hz, and a submersible pump intake at 30 cm above the bottom of the profiler. One liter bulk water samples were collected at each elevation and analyzed for total suspended solids (TSS). TSS was determined by passing a measured subsample through a preweighed 0.8 lm poresize glass fiber filter, followed by drying and weighing. Additional bulk water samples were taken for trace metal and nutrient analysis which are not considered in this paper. An upward looking Sontek acoustic Doppler profiler (ADP) was deployed nearby each profiling site sampling every 10 s with 0.25 m depth bins. The ADV backscatter was calibrated to mass concentrations using the 1 mab to surface TSS values from the York River 2000 experiment ðr 2 ¼ 0:93Þ LISST background Laser diffraction technology has been used and evaluated in various instruments to measure the size spectra of suspended particles (Agrawal & Pottsmith, 1994, 2000; Bale, 1996; Bale, Barrett, West, & Oduyemi, 1989; Beuselinck et al., 1999; Lynch, Irish, Sherwood, & Agrawal, 1994; McCave, Manighetti, & Robinson, 1995; Phillips & Walling, 1998; Traykovski, Latter, & Irish, 1999). Extensive descriptions of the LISST instrument and its operation principles can be found in Agrawal and Pottsmith (1994, 2000). The Sequoia Science LISST works well in resolving unimodal silicate particle distributions (Agrawal & Pottsmith, 2000; Battisto, 1999; McCave et al., 1995; Traykovski et al., 1999). However, the primary particle populations and loose aggregates found in biologically productive estuaries are more complex than sand particles with respect to their refractive indices and size distributions, which may be multimodal with wide and narrow peaks or may be uniform (Kranck, Smith, & Milligan, 1996). The presence of particles finer and coarser than the measured size range also affects the estimated size distributions from the LISST (McCave et al., 1995; Mikkelsen & Pejrup, 2000; Traykovski et al., 1999). An additional complication to estimating volume concentrations in estuaries with the LISST is the difficulty of determining the volume calibration constant for estuarine particles because of their size density dependence and fragility. Accurate measurements of aggregate sizes in estuarine environments may still best be made with high resolution video and still photographic techniques, while results from the LISST may be more successfully interpreted by examining relative changes in the size distributions and volume concentrations rather than relying on the exact values obtained from the instrument and the inversion results. Nevertheless, the LISST has been found to work reasonably well in examining aggregate size distributions (Fugate & Friedrichs, 2002; Mikkelsen & Pejrup, 2000). The LISST Model 100 used for this project measures particle size distribution in the range of lm. Because of potential problems with measuring estuarine aggregates with the LISST, the median grain sizes reported here should not be taken as precise measurements but rather as an index for comparing size distributions. 3. Results 3.1. Hydrodynamics and concentration Time series of along-channel velocities, salinities, and acoustic backscatter (a proxy for sediment concentration) for the three estuaries are shown in panels a c of Figs The York River and Chesapeake Bay velocities were much higher and more regular than those in the Elizabeth River. Along-channel ebb velocities in the York River reached 90 cm s ÿ1 ; in addition, there were significant cross-channel velocities up to 15 cm s ÿ1. Upper water column velocity gradients were negative at the beginning of flood tides, but in general the velocity profiles increased regularly upwards. Along-channel velocities at the Cherrystone site were even higher than those of the York, reaching 100 cm s ÿ1 during ebbs. In contrast to these two estuaries, velocities in the Elizabeth River were weak, peaking around 40 cm s ÿ1. The lower water column currents were quite irregular, reflecting the interaction of the mild currents with the narrow and unnaturally irregular geometry of the urbanized channel.

6 394 D.C. Fugate, C.T. Friedrichs / Estuarine, Coastal and Shelf Science 58 (2003) Fig. 2. Time series of York River physical data: (a) velocity (cm s ÿ1 ), flood positive, (b) salinity, (c) calibrated ADP backscatter (mg l ÿl ), and (d) log 10 (median particle size (lm)).

7 D.C. Fugate, C.T. Friedrichs / Estuarine, Coastal and Shelf Science 58 (2003) Fig. 3. Time series of the Cherrystone site physical data: (a) velocity (cm s ÿ1 ), flood positive, (b) salinity, (c) calibrated ADP backscatter (mg l ÿ1 ), and (d) log 10 (median particle size (lm)) at 10 cmab.

8 396 D.C. Fugate, C.T. Friedrichs / Estuarine, Coastal and Shelf Science 58 (2003) Fig. 4. Time series of Elizabeth River physical data: (a) velocity (cm s ÿ1 ), flood positive, (b) salinity, (c) calibrated ADP backscatter (mg l ÿ1 ), and (d) log 10 (median particle size (lm)).

9 D.C. Fugate, C.T. Friedrichs / Estuarine, Coastal and Shelf Science 58 (2003) Time series of salinities for the three estuaries are shown in Figs. 2b 4b. The York River showed typical stratification patterns of a partially mixed estuary, with ebb tides more highly stratified, about from the bottom to the top of the 10 m water column. The Cherrystone site in the Chesapeake Bay was also more stratified on ebb, around over a 12 m water column, but tending to become well mixed after flood. Stratification at the Elizabeth River site was similar to the lower Chesapeake Bay, around over a 12 m water depth. Maximum density gradients were centered around 4.5 mab during ebb and around 2 mab during flood. Time series of suspended sediment concentration determined from the calibrated ADP backscatter are shown in Figs. 2c 4c. Suspended sediment concentrations were highest in the York, with near-bottom concentrations ranging from 50 to over 300 mg l ÿ1. Surface concentrations peaked near 40 mg l ÿ1. Nearbottom concentrations in the lower Chesapeake ranged from 15 to over 70 mg l ÿ1. Concentrations in the Elizabeth River were much lower than the other estuaries in both surveys shown, usually below 10 mg l ÿ1 near the bottom, reaching near 20 mg l ÿ1 at peaks. The mid-depth water column showed occasional concentration peaks during slack phases and the surface showed higher concentrations during ebb. These surface peaks were corroborated by the OBS signal, and were not artifices of surface anomalies on the ADP. Bottom peaks in concentration were associated with boat wake rather than peak currents Estimating turbulent kinetic energy and Kolmogorov microscale with contaminated data Under ideal conditions, the high frequency sampling rate (10 Hz) of the ADV should allow measurement of the turbulent energy production in the water column through p ¼ s du dz (Nezu & Nakagawa, 1993) where P is the energy production, s the shear stress, and du=dz is the vertical velocity gradient. Shear stress can be determined from s ¼ qðÿu9w9þ where q is the density and ÿu9w9 is the Reynolds stress, or via the empirical relationship s ¼ C TKE (Gordon & Dohne, 1973; Kim, Friedrichs, Maa, & Wright, 2000; Soulsby, 1983; Stapleton & Huntley, 1995), where C is a constant and TKE is the turbulent kinetic energy measured by the ADV. With the profiler resting on the bottom, shear stress may be measured directly in the form of Reynolds stress as well as TKE. However, measurements further up in the water column are partially contaminated by boat motion and the dangling motion of the ADV on the deployment line. This motion can be seen as a low frequency (0.3 Hz) peak in the spectra of the velocity components (Fig. 5). In order to make estimates of the turbulent energy structure in the water column from the potentially contaminated ADV data, an alternate method is proposed and compared to estimates of energy dissipation made at the bottom using Reynolds stress. Because the bottom data were taken with the profiler resting on the bottom, they do not suffer from problems with boat or dangling motions. Turbulent kinetic energy (TKE) is usually calculated from the three fluctuating velocity components, u9, v9, and w9 as TKE ¼ 0:5ðu9 2 þ v9 2 þ w9 2 Þ. In our modified method, velocity components for each burst were first high pass filtered with a cutoff at 1.5 Hz and a three point taper. The cutoff frequency was determined empirically from observation of the velocity spectra, which, as a whole, contained no obvious peaks in spectral density due to profiler motion at frequencies higher than 1.5 Hz. Most peaks associated with profiler motion occurred at frequencies less that 1 Hz. The TKE contained in only the high frequency fluctuations of the three components was then calculated from these filtered series. An additional modification was made by using the median of fluctuating velocities instead of the mean. ADV data from the datasets in this study occasionally contain isolated short episodes of electrical spikes, whose bias can be minimized by using medians of the fluctuating velocity components. We term this quantity obtained from the sum of the median of the fluctuating high frequency velocity components TKE h. The velocity shear ðdu=dzþ is also Spectral Density, (cm 2 /s 2 )/Hz Field Signal Background Noise Signal Frequency (Hz) Fig. 5. Field signal spectral density from an ADV burst from Elizabeth River, 1 mab, bottom spectrum is from ADV noise experiment.

10 398 D.C. Fugate, C.T. Friedrichs / Estuarine, Coastal and Shelf Science 58 (2003) needed to obtain a proxy estimate of the TKE production via TKE h du=dz. It was assumed that the near-bed ADV sensor (0.3 mab) was in the constant stress layer so that du=dz hu9w9i 1=2 ðjzþ ÿ1, where j is the von Karman constant. Fig. 6 shows that there is a good relationship between turbulent energy production estimated with a Reynolds stress at the bottom and the quantity TKE h du=dz (r 2 ¼ 0:60 for the Cherrystone dataset, and >0.82 for the other three datasets). All the relationships are statistically significant from zero at p < 0:001. TKE production estimates were made for the three estuaries using this alternate method by calculating TKE h du=dz and calibrating each dataset separately against the measures of TKE production made using the Reynolds stress, as shown in Fig. 6. For estimates of TKE production above the near-bed region (z > 0:3 m), du=dz was estimated by finite differencing the ADP observations. In order to get a gross estimate of the background noise generated by the ADV in still water, an experiment was set up according to the procedure suggested by Nikora and Goring (1998). Spectral plots of the noise generated were flat and very low, about two orders of magnitude lower than those measured in field conditions (see Fig. 5). Consequently, background ADV noise turns out to be insignificant to the measures of TKE production. This result is consistent with Nikora and Goring s findings, which found an average ratio of corrected to uncorrected mean velocity fluctuations of Finally, the Kolmogorov microscale, k K, is calculated by assuming that TKE production is equal to dissipation and substituting into k K ¼ðm 3 =eþ 0:25, where m is kinematic viscosity and e is the TKE dissipation rate (e.g. Hill et al., 1992). Consequently, increases in TKE production are associated with decreasing Kolmogorov microscales as the TKE drives velocity shear into smaller length scales. This assumed balance between TKE production and dissipation is valid as long as there is no significant contribution by buoyancy to TKE either as a sink associated with mixing away of stable stratification or as a source associated with straininduced convective overturning (Simpson, Burchard, Fisher, & Rippeth, 2002). Recent observations from the Hudson River estuary indicate that production and dissipation are well balanced up to at least 3 m above the bed (Trowbridge, Geyer, Bowen, & Williams, 1999). For tidal currents in Liverpool Bay, Simpson et al. (2002) found the balance to be satisfied through most of the tidal cycles more than 10 m above the bed Particle size distribution Time series of median particle diameters (D50) vs. depth as measured by the LISST are shown in Fig. 2d for the York River. Relative particle sizes are compared with the Kolmogorov microscale and sediment concentration as measured by the ADV backscatter in Fig. 7a, b. The line superimposed on Fig. 7a is the least squares best-fit in log space to D50 ¼ constant k K. Even though there is a reasonable correspondence between observed D50 and calculated k K, it is worth reemphasizing that the median particle diameter as reported by the LISST should be thought of as a relative index of the 10 1 Elizabeth R. York R. Cherrystone 10 2 TKE Production from Reynolds Stress (m 2 /s 3 ) TKE Production from TKE (m 2 /s 3 ) h Fig. 6. TKE production as measured by Reynolds stress at the bottom from the Elizabeth River (*), York River (hexagon), and Cherrystone site (.) compared with TKE production as measured by TKE h du=dz. These relations are used to calibrate the TKE h du=dz values separately for each estuary.

11 D.C. Fugate, C.T. Friedrichs / Estuarine, Coastal and Shelf Science 58 (2003) Fig. 7. York River particle size distributions: (a) median particle size by k K and (b) median particle size by concentration. () 0.3 mab, (+) 1 mab, and () 6 mab. median particle size rather than as a precise measure. There are also significant assumptions associated with our estimates of k K. First, it is assumed that TKE h represents stress throughout the water column even though the method was calibrated only at the bed. Second, the contributions of buoyancy production to TKE are neglected. Peak surface and mid-depth concentrations in the York River lag peak velocities. We might expect these conditions to have led to a pattern of small particles at the onset of high currents, followed by larger particles as concentration increases and turbulent shear decreases with decreasing velocity shear. However, this relation, if it existed, was overwhelmed by the unvarying presence of large particles starting with the second flood cycle (Fig. 2d). These particles may have been advected into the surface region either from the head of the estuary or from the shoals with cross-channel currents. Excluding the surface, particle size distributions in the rest of the water column were consistent with the combined processes of (i) intense turbulent shear (represented by small k K ) reducing maximum aggregate size and (ii) collisions during differential settling increasing aggregate size (p < 0:001, r 2 ¼ 0:52, n ¼ 28 for bottom and 1 mab, p ¼ 0:005, r 2 ¼ 0:50, n ¼ 14 for middle water column; Fig. 7a). The former is likely to be dominant during high TKE production and the latter during low TKE production. Near the bed there was a bias towards larger particles during the second tidal cycle, but the near-bottom relationship between size and k K persisted. It is likely that the large particles that were introduced at the surface had sunk to the lower elevations where the higher local population of suspended sediment diluted their effect on particle size distribution. The inverse relation between size and concentration near the bed (p < 0:001, r 2 ¼ 0:46, n ¼ 28; Fig. 7b) is also indicative of aggregate breakup at high shear. Loosely bound aggregates at the bottom were resuspended and immediately disaggregated by high bed stress. Median particle sizes near the bed at the Cherrystone site are shown in Fig. 8a. In contrast to the relationship in the York River, smaller k K tended to be associated with larger particles (p < 0:001, r 2 ¼ 0:35, n ¼ 25). Like the York, this is a relatively energetic hydrodynamic regime, but the dominant benthic processes are different. Resuspension in the York inputs fragile aggregates into the water column that are immediately broken up by high stress, whereas strong resuspension at Cherrystone inputs larger biologically aggregated and pelletized particles into the water column. Fig. 8b displays the positive relation seen at the Cherrystone site between particle size and concentration (p < 0:001, r 2 ¼ 0:49, n ¼ 22), which is also consistent with higher stress suspending increasingly larger aggregates. Fig. 4d shows the time series of median particle size by depth from the Elizabeth River. In general, the largest particles tended to be found in the middle of the water column, the smaller particles tend to be near the surface, and the near-bed region exhibits intermediate size particles. Local areas of large particles were seen to migrate vertically about 2 m in concert with the tidally moving pycnocline. Relationships among k K, particle size, and concentration in the Elizabeth River were weak (Fig. 9). This may be partly due to the difficulty of estimating TKE production via our empirical method in the weak and irregular currents. Furthermore, the assumption that TKE dissipation equals production may be violated here. At the very low bed shear conditions in the Elizabeth River, k K was relatively large overall (Fig. 9a). Consequently, resuspended loosely bound aggregates were not completely torn apart at the maximum stresses, and particle size was not tightly coupled to k K. In addition, the size of aggregates which sank from the surface may have been constrained by the higher stresses associated with sinking rather than those from TKE in the water column (Hill et al., 2001). A similar survey in the Elizabeth River in May of 1999, after spring summer season recruitment of benthic macrofauna (Hawthorne & Dauer, 1983) produced similar relationships among k K, particle size, and

12 400 D.C. Fugate, C.T. Friedrichs / Estuarine, Coastal and Shelf Science 58 (2003) Fig. 9. Elizabeth River particle size distributions: (a) median particle size by k K and (b) median particle size by concentration. () 0.3 mab, (+) 1 mab, and () 3 mab. Fig. 8. Cherrystone site particle size distributions: (a) median particle size by k K and (b) median particle size by concentration. concentration, suggesting that there is no seasonal effect from changes in biological productivity Near-bed settling velocity Knowledge of the settling velocity of near-bed suspended particles is useful to understanding particle type and aggregation dynamics. Together with an idea of the relative particle size distributions from the LISST, we can gain insight into the relative densities of near-bed suspended material. The settling velocity may be indirectly estimated by making the common assumption of a lowest order sediment concentration balance between gravitational settling and upward turbulent diffusion (Fugate & Friedrichs, 2002; Glenn & Grant, 1987; Sherwood et al., 1994; Sleath, 1984): ÿw sn C n ¼ K dc n =dz where w sn is the settling velocity of particle type n, C n the concentration of particle type n, and K is the eddy diffusivity. Using an ADV, turbulent diffusion can be measured from the Reynolds diffusive flux: K dc n =dz ¼ÿhw9C n 9i where w9 is the vertical fluctuating component of velocity and C9 is the deviation from mean concentration as estimated by the ADV backscatter. One can then solve for settling velocity simply by dividing both sides by C (Fugate & Friedrichs, 2002). Fig. 10 displays plots of hw9c9i vs. C for the York River and Cherrystone site with w s determined from the slope of the best-fit regression. In each case, the x-axis intercept of the best-fit regression at hw9c9i ¼0 is interpreted as the background non-settling component of concentration.

13 D.C. Fugate, C.T. Friedrichs / Estuarine, Coastal and Shelf Science 58 (2003) consistent with decreasing fall velocity at higher concentrations and higher shear and smaller particle size due to aggregate breakup. A best-fit of a quadratic curve to the Cherrystone data (Fig. 10b), in contrast, would be concave downward, consistent with increasing fall velocity and larger particle size at higher concentrations and higher shear due to resuspension of resilient aggregates. It should be noted, however, that our interpretation of the curvature in the York River data set is highly speculative, given the small number of data points. The apparently low settling velocity in combination with the lack of steady state conditions imposed by sporadic resuspension from boat wakes did not permit a statistically significant indirect measurement of settling velocity at the Elizabeth River. 4. Conceptual model of controls on size distributions in partially mixed estuaries Fig. 10. Reynolds flux ðw9c9þ by concentration (mg l ÿ1 ) for bottom elevations of (a) York River and (b) Cherrystone site. The estimated settling velocity of near-bed particles in the York River was mm s ÿ1 and the estimated background concentration was mg l ÿ1 (r 2 ¼ 0:89, n ¼ 14). The estimated settling velocity at Cherrystone was mm s ÿ1, and the estimated background concentration was 16 2mgl ÿ1 (r 2 ¼ 0:69, n ¼ 250). Both these best-fit fall velocities are consistent with muddy aggregates. If the particles of the order of 100 lm were fine mineral sand, the fall velocity would be larger. The higher settling velocity of the particles at Cherrystone kept the background concentration lower than that in the York, even though both are physically energetic. It is worth noting the opposite curvature in the trend of the w9c9 vs. concentration scatter plots for the York and Cherrystone sites. Best-fit of a quadratic curve to the York data in Fig. 10a would be concave upward, The different hydrodynamic regimes and particle sources at different levels in the water column suggest examination of the particle aggregation processes as a function of height above the bed. Conditions near the surface are characterized by a range of TKE production and dissipation rates, which may be low even though velocities are high, since velocity shear may be reduced. Suspended sediment concentration is low at the surface. The higher velocities at the surface and remoteness from the bed make the immediate source of particles most likely to be from horizontal advection. Near-bed conditions are characterized by lower velocities, higher TKE production from velocity shear, and an associated smaller k K. Suspended sediment concentrations are high, and resuspension from the bed provides the major source of sediments. Mid-depth regions experience a range of TKE and shear conditions, and particle sources may be either from resuspension from the bed, horizontal advection, or sinking from the surface. The relevant processes that determine aggregate size in each of these depth categories will be discussed with regard to their respective hydrodynamic regime and sediment concentrations and sources. These processes are summarized in Table Surface regime Low values of turbulent shear, and large k K at the surface are not accompanied by large particle formation because the surface is generally bereft of particles. Low TKE production and dissipation values occur not only at slack, but also during peak currents when velocity shear is low and relative concentration is high. The primary processes affecting size distribution and concentration at the surface in both the York River and Elizabeth River occur on a subtidal time scale. In the

14 402 D.C. Fugate, C.T. Friedrichs / Estuarine, Coastal and Shelf Science 58 (2003) Table 1 Dominant factors and times scales controlling particle size distribution in partially mixed estuaries York River, large particles are irregularly advected into the surface region, possibly from transverse currents during slack. Elizabeth River concentrations are erratic and dominated by boat wakes and horizontal advection Mid-depth regime High energy high concentration Mid-depth particle aggregation dynamics in the York River are consistent with conventional models. Differential settling allows large particles to aggregate during reduced levels of TKE production and correspondingly large k K. High levels of TKE dissipation and associated small k K are accompanied by smaller disaggregated particles. In contrast, aggregation dynamics in the Elizabeth River cannot be attributed to these processes. The low energy levels and low concentration levels permit stochastic processes such as boat wake, advection, and possibly industrial discharge to dominate the particle size regime. Stresses associated with particle sinking may provide an upper limit on aggregate size in this low energy environment (Hill et al., 2001). Nevertheless, differential settling does appear to allow large particles to aggregate at the pycnocline, although their appearance is not strictly associated with tidal currents. When these larger particles are present, they migrate vertically with the pycnocline Near-bottom regimes Low energy low concentration Surface Advection Advection Sub-tidal frequency Sub-tidal frequency Mid-depth Turbulent shear Differential settling Differential settling Sub-tidal frequency Tidal frequency Near-bed Turbulent shear Resuspension Resuspension Benthic environment Benthic environment Sub-tidal frequency Tidal frequency Near-bottom regimes in each of the three estuaries are different, reflecting their different particle types and hydrodynamic regimes. In the mild currents of the Elizabeth River (median k K ¼ 480 lm), loosely bound porous particles are resuspended and disaggregated irregularly by episodic boat wake. Near the bed of the York River (median k K ¼ 239 lm) high currents and associated shear disaggregate particles as they are suspended, as evidenced by the decreasing particle size with decreasing k K and increasing concentration. In contrast to both these estuaries are the particles at the Cherrystone site in the Chesapeake Bay. Like the York River, this is also a high energy region (median k K ¼ 167 lm) but particle size increases with increased stress here. These particles have the highest settling velocity and are likely dense aggregates formed by biological aggregation and pelletization. Consequently, their relationships to TKE production and k K is opposite to those in the York; higher turbulent energy resuspends larger particles and leads to higher fall velocities. 5. Conclusions An alternate method for determining TKE production with ADV data potentially contaminated by spurious instrument motion proved to be a good estimator when compared with measurements made with Reynolds stress. By assuming a balance between production and dissipation of TKE, the Kolmogorov microscale for the smallest turbulent eddies was then estimated. This approach may also be useful for estimating TKE and the Kolmogorov microscale with an ADV in the presence of surface or internal gravity waves. Along with other more conventional methods, this method was used to reach the following conclusions regarding particle aggregation dynamics in partially mixed estuaries. The processes that determine particle aggregation dynamics in estuaries are a function of the elevation in the water column, as well as the overall hydrodynamic and biological environment. For low energy, low concentration estuaries such as the Elizabeth River, stochastic events such as advection, boat wake, or other forcings dominate the processes associated with tidally varying TKE production and dissipation. Despite the control of discrete irregular events on particle dynamics, vertical profiles tended to be consistent, with the largest particles found in the middle of the water column at the pycnocline, the smallest at the surface, and intermediate sizes near the bed. A result from the Elizabeth River relevant to characterizing particle dynamics in partially mixed estuaries is that important advective events occur at time scales longer than the tidal period. Even in the more energetic York River, particle sizes in the low concentration surface region that were expected to respond to changes in concentration were instead controlled by stochastic events. Aggregation dynamics in the mid-depth region of the York River, a relatively high energy and high concentration estuary, were determined by the combined processes of turbulent shear and differential settling. Turbulent shear over lengths governed by k K tore aggregates apart, limiting the maximum aggregate size. As shear decreased and k K increased, collisions produced by differential settling presumably caused aggregates to grow. When samples were pooled from the mid- and lower-water column of the York River, median

15 D.C. Fugate, C.T. Friedrichs / Estuarine, Coastal and Shelf Science 58 (2003) particle size over a broad range of energies tracked close to the estimated k K. Particle size relationships near the bed were different in each of the three study sites, reflecting their different hydrodynamic and benthic environments. Near-bed aggregate sizes in the energetic, physically dominated York River were positively correlated with k K due to aggregate break up at high shear. In contrast, particle sizes at the energetic, biologically influenced Cherrystone site were negatively correlated with k K due to resuspension of large resilient aggregates at high shear. Thus, though situated in a generally muddy environment, the majority of regularly resuspended material near the bed water interface responds to shear stress in the same manner as sand particles. Near-bed particles in the low energy Elizabeth River appeared to be large porous aggregates, whose size was unconstrained by the generally large k K. Concentration did not determine particle size near the bed in any of the three estuaries. Rather, concentration and particle size were simultaneously influenced by hydrodynamic forcing. Acknowledgements The authors gratefully acknowledge the assistance of Grace Battisto in collection of these data. Funding for collection and analysis of data from (i) the Chesapeake Bay, (ii) the Elizabeth River, and (iii) the York River was provided by: (i) the Office of Naval Research Harbor Processes Program Grant N , (ii) the EPA through the Chesapeake Bay Environmental Effects Committee, and (iii) the National Science Foundation Division of Ocean Sciences Grant OCE Support for analysis was also provided by National Science Foundation Division of Ocean Sciences Grant OCE This paper has benefited greatly from the helpful suggestions of John Trowbridge and two anonymous reviewers. This is contribution number 2534 of the Virginia Institute of Marine Science, College of William and Mary. References Agrawal, Y. C., & Pottsmith, H. C. (1994). Laser diffraction particle sizing in STRESS. Continental Shelf Research 14(10/11), Agrawal, Y. C., & Pottsmith, H. C. (2000). Instruments for particle size and settling velocity observations in sediment transport. Marine Geology 168, Bale, A. J. (1996). In situ laser optical particle sizing. Journal of Sea Research 36(1 2), Bale, A. J., Barrett, C. D., West, J. R., & Oduyemi, K. O. K. (1989). Use of in-situ laser diffraction particle sizing for particle transport studies in estuaries. In J. McManus, & M. Elliott (Eds.), Developments in estuarine and coastal study techniques (pp ). Fredensborg, Denmark: Olsen & Olsen. Battisto, G. M. (1999). Field measurements of mixed grain size suspension in the nearshore under waves. Master s thesis, Virginia Institute of Marine Science, Gloucester Point. Berhane, I., Sternberg, R. W., Kineke, G. C., Milligan, T. C., & Kranck, K. (1997). The variability of suspended aggregates on the Amazon Continenetal Shelf. Continental Shelf Research 17(3), Beuselinck, L., Govers, G., Steegen, A., Hairsine, P. B., & Poesen, J. (1999). Evaluation of the simple settling theory for predicting sediment deposition by overland flow. Earth Surface Processes and Landforms 24(11), ten Brinke, W. B. M. (1994). In situ aggregate size and settling velocity in the Oosterschelde Tidal Basin (The Netherlands). Netherlands Journal of Sea Research 32(1), Chen, S., Eisma, D., & Kalf, J. (1994). In situ size distribution of suspended matter during the tidal cycle in the Elbe Estuary. Netherlands Journal of Sea Research 32(1), Dauer, D. M. (2002). Benthic biological monitoring program of the Elizabeth River Watershed (2001) with a study of Paradise Creek. Virginia Department of Environmental Quality. Dauer, D. M., Stokes, T. L., Barker, H. R., Ewing, R. M., & Sourbeer, J. W. (1984). Macrobenthic communities of the lower Chesapeake Bay. 4. Bay-wide transects and the inner continental shelf. Internationale Revue der Gesamten Hydrobiologie 69(1), Dellapenna, T. M., Kuehl, S. A., & Schaffner, L. C. (1998). Sea-bed mixing and particle residence times in biologically and physically dominated estuarine systems: a comparison of lower Chesapeake Bay and the York River subestuary. Estuarine and Coastal Marine Science 46, Dyer, K. R. (1989). Sediment processes in estuaries: future research requirements. Journal of Geophysical Research 94(10), Eisma, D., Bernard, P., Cadee, G. C., Ittekkot, V., Kalf, J., Laane, R., Martin, J. M., Mook, W. G., Put, A., van Schuhmacher, T. (1991). Suspended-matter particle size in some west-european estuaries; Part 2: a review on floc formation and break-up. Netherlands Journal of Sea Research 28(3), Eisma, D., Dyer, K. R., & van Leussen, W. (1997). The in situ determination of the settling velocities of suspended fine-grained sediment a review. In N. Burt, R. Parker, & J. Watts (Eds.), Cohesive sediments: fourth nearshore and estuarine cohesive sediment transport conference, INTERCOH 94, July 1994 Wallingford, England, UK (pp ). Chichester: Wiley. Fettweis, M., Sas, M., & Monbaliu, J. (1998). Seasonal, neap spring and tidal variation of cohesive sediment concentration in the Scheldt Estuary, Belgium. Estuarine and Coastal Shelf Science 47, Francese, R. S., & Dvorak-Grantz, A. L. (1998). Evaluation of sediment: test results for the Southern Branch. Norfolk Harbor Federal Navigation Project, Norfolk District, Corps of Engineers. Fugate, D. C., & Friedrichs, C. T. (2002). Determining concentration and fall velocity of estuarine particle populations using ADV, OBS and LISST. Continental Shelf Research 22, Gibbs, R. J., Tshudy, D. M., Konwar, L., & Martin, J. M. (1989). Coagulation and transport of sediments in the Gironde Estuary. Sedimentology 36, Glenn, S. M., & Grant, W. D. (1987). A suspended sediment stratification correction for combined wave and current flows. Journal of Geophysical Research 92(8), Gordon, C. M., & Dohne, C. F. (1973). Some observations of turbulent flow in a tidal estuary. Journal of Geophysical Research 78(12), Hawthorne, S. D., & Dauer, D. M. (1983). Macrobenthic communities of the lower Chesapeake Bay. 3. Southern Branch of the Elizabeth River. Internationale Revue der Gesamten Hydrobiologie 68(2), Hill, P. S., & McCave, I. N. (2001). Suspended particle transport in benthic boundary layers. In B. P. Boudreau & B. B. Jorgensen (Eds.), The benthic boundary layer: Transport processes and biogeochemistry (pp ). Oxford: Oxford University Press.