The Influence of Wind and River Pulses on an Estuarine Turbidity Maximum: Numerical Studies and Field Observations in Chesapeake Bay

Transcription

1 Estuaries Vol. 7, No. 1, p February 004 The Influence of Wind and River Pulses on an Estuarine Turbidity Maximum: Numerical Studies and Field Observations in Chesapeake Bay E. W. NORTH*, S.-Y. CHAO, L.P.SANFORD, and R. R. HOOD University of Maryland Center for Environmental Science, Horn Point Laboratory, P. O. Box 775, Cambridge, Maryland 1613 ABSTRACT: The effect of pulsed events on estuarine turbidity maxima (ETM) was investigated with the Princeton Ocean Model, a three-dimensional hydrodynamic model. The theoretical model was adapted to a straight-channel estuary and enhanced with sediment transport, erosion, deposition, and burial components. Wind and river pulse scenarios from the numerical model were compared to field observations before and after river pulse and wind events in upper Chesapeake Bay. Numerical studies and field observations demonstrated that the salt front and ETM had rapid and nonlinear responses to short-term pulses in river flow and wind. Although increases and decreases in river flow caused downestuary and up-estuary (respectively) movements of the salt front, the effect of increased river flow was more pronounced than that of decreased river flow. Along-channel wind events also elicited non-linear responses. The salt front moved in the opposite direction of wind stress, shifting up-estuary in response to down-estuary winds and vice-versa. Modeled pulsed events affected suspended sediment distributions by modifying the location of the salt front, nearbottom shear stress, and the location of bottom sediment in relation to stratification within the salt front. Bottom sediment accumulated near the convergent zone at the tip of the salt front, but lagged behind the rapid response of the salt front during wind events. While increases in river flow and along-channel winds resulted in sediment transport downestuary, only reductions in river flow resulted in consistent up-estuary movement of bottom sediment. Model predictions suggest that wind and river pulse events significantly influence salt front structure and circulation patterns, and have an important role in the transport of sediment in upper estuaries. Introduction The estuarine turbidity maximum (ETM), or turbidity maximum zone (TMZ), is a region of elevated suspended sediment concentrations often found near the landward limit of salt intrusion in estuaries (Glangeaud 193; Postma and Kalle 1955; Schubel 196). In addition to developing near river fronts, ETMs may result from local availability of erodible sediments, local enhancement of bottom shear stress, convergence of bottom currents, and topographic and coastline irregularities. Although mechanisms and process affecting the formation of ETMs have been the subject of intensive field and modeling investigations, the effect of pulsed events on salt front structure, suspended sediment concentrations, and sediment transport associated with ETMs is less well understood. This research focuses on the influence of wind and river pulses on river-front ETMs. Both physical and chemical processes contribute to river-front ETM formation. Chemical processes such as salt flocculation (Krone 196) and organic coating (Eisma et al. 1991) enhance particle set- * Corresponding author; tele: 410/1-497; fax: 410/1-490; enorth@hpl.umces.edu tling near the front. Physical mechanisms promote trapping of these settling particles. A convergence zone occurs where freshwater from the river meets near-bottom intrusion of seawater at the landward margin of the salinity intrusion (referred to as the salt front here). Postma and Kalle (1955) first suggested that gravitational circulation, independent of tidal forcing, could trap particles in the convergence zone. This was verified using a numerical model with constant diffusivity by Festa and Hansen (197). Geyer (1993) demonstrated that reduced turbulent diffusion by stratification enhanced particle trapping near the landward limit of salinity intrusion in the same location as the convergence zone. Tidal forcing also contributes to particle trapping in the ETM. Near the salt front, bottom currents tend to be intensified during flood tide and surface currents tend to be intensified during ebb tide. Near-bottom intensified currents transport particles landward toward the tip of the salt front on flood tide, thereby enhancing particle trapping ( Jay and Musiak 1994). Burchard and Baumert (199) developed a hydrodynamic model with a sediment component to assess the relative importance of gravitational circulation, asymmetric tidal 004 Estuarine Research Federation 13

2 The Influence of Pulsed Events on an ETM 133 Fig. 1. Upper Chesapeake Bay, U.S. Circles represent the location of hydrographic survey stations in May, June, and July 001, and May 00. Symbols indicate the location of Conowingo Dam ( ), Tolchester Beach (#), and Thomas Point Light (3). advection, and differential turbulent diffusion in maintaining an ETM. They concluded that gravitational circulation and, most notably, asymmetric tidal advection were vital to ETM stability. Differential turbulent diffusion contributes to, but is not necessary for, ETM maintenance. The aforementioned studies focused on physical mechanisms that form ETMs. The theoretical investigations presented here examine the influence of wind and river pulses on an established ETM. Our focus is on longitudinal processes and the relationship between pulsed events, salt front structure, suspended sediment concentrations, and sediment transport to lowest order. The numerical model is loosely scaled to match upper Chesapeake Bay characteristics based on predicted tides, previous studies (North and Houde 001; Sanford et al. 001), and recent field observations reported here. The Chesapeake Bay (Fig. 1) is a partially mixed estuary that has a pronounced ETM found between N and 39 N (Schubel 196; Sanford et al. 001). The upper reaches of the Bay are characterized by a deep (1 m) navigation channel on the east side and broad shoals on the western side. Most freshwater and sediment delivery to the upper Chesapeake is from the Susquehanna River (average discharge rate 1,100 m 3 s 1 ). Winds are generally episodic with dominant periods of 7 d (Wang 1979) and tidal forcing is modest with a tidal range rarely exceeding 1 m. Historical data and numerical investigations indicate that the river-forced circulation and the salt front develop preferably in the deep channel (Wang and Chao 1996), the location of the field observations presented here. The upper Chesapeake ETM is km in extent and is generally associated with the salt front but can be displaced from it by as much as 7 km (Boynton et al. 1997; Sanford et al. 001). This research is part of the BITMAX program (Bio-physical Interactions in the Turbidity Maximum unpublished material). BITMAX is a multiple-investigator, interdisciplinary program sponsored by the National Science Foundation that applies both field and modeling techniques to answer the question: How do estuarine turbidity maxima entrap particles, retain zooplankton, and promote fish recruitment? Research presented here provides the basis for understanding how episodic events affect the physical characteristics of the ETM and influence sediment dynamics as well as retention and survival of planktonic organisms. Field observations from BITMAX research cruises before and after river pulse and wind events in upper Chesapeake Bay are examined and compared to river pulse and wind scenarios from the theoretical numerical model. Model results of salt front behavior and sediment dynamics are used to gain insight on how pulsed events could affect ETM characteristics and sediment transport in upper estuaries. Methods MODEL FORMULATION The three-dimensional hydrodynamic model is based on the code of the Princeton Ocean Model (POM) (Blumberg and Mellor 197; Mellor 199) under hydrostatic and Boussinesq approximations. Eddy viscosity and diffusivity are determined by the level.5 turbulent closure scheme of Mellor and Yamada (1974). Our enhancements to POM include building suspended sediment and sediment transport components, adding constant loading of suspended sediment at the up-stream boundary, parameterizing bottom sediment burial with a Newtonian dampening term, and constructing a more reasonable formulation for background vertical diffusivity. The model domain (Fig. ) contains a 9-km long and 4.-km wide channel with a gradually widening seaward reservoir that is,37 km long and has a maximum width of 66 km. The entire basin is 1 m deep and contains 145 grid spacings in the x-direction (seaward) and 6 in the y-direction (across channel). Vertical resolution is provided by 1 sigma-coordinate layers. In the first 9 km of the channel, the longitudinal resolution ( x) is 1 km and the lateral resolution ( y) is 0. km. In the seaward reservoir, x and y gradually increase at

3 134 E. W. North et al. TABLE 1. Steady-state model parameters. Fig.. Schematic of model domain. Upper diagram represents the entire model domain. Lower diagram illustrates the grid line geometry of the channel, the region of interest. Bathymetry Upstream dissipation parameter (e) River inflow velocity (u 0 ) Max tidal current velocity (UAREF 1.5) Bottom roughness height Critical shear stress for erosion ( c ) M (constant in erosion equation) Sediment settling velocity (w s ) Sediment loading rate (L) Consolidation/burial rate ( ) Maximum background vertical diffusivity/viscosity ( max ) Flat, 1 m Deep 0.99 (modest) m s m 1 01 m 5 Pa kg m s m s 1 05 kg s 1 1/10/6,400 s m s 1 a rate of 1% and 5.5% per grid, respectively. The seaward reservoir, although of minor importance to the circulation and sediment transport within the channel, serves as a buffer zone to enhance the long-term computational stability of the model on the time scales of months. Mixing within the reservoir helps diffuse the river-induced density front and retard its seaward progression to the open boundary. Coefficients of horizontal viscosity and diffusivity are dependent on grid size (A A o x y ( x y ) 1 ) so that damping rates for grid scale noises are more or less uniform regardless of the mesh size. The constant A o is minimized to s 1, below which the computation becomes unstable. The temporal resolution of the model is split. Vertically averaged currents and sea level are resolved with a time step of 5 s, and salinity, sediment concentration, and vertically explicit current velocities are resolved with an internal time step of 40 s. A barotropic tidal generation force is included in the longitudinal momentum equation to produce semidiurnal tidal currents in the channel. This force is used instead of specifying sea-level oscillations on the seaward boundary because bottom dissipation and internal tide generation cause marked diminishment of tidal waves in the long reservoir when waves enter from the seaward boundary. The tidal generation force (F) oscillates in time (t) with a period (T) of 1 h, given by F (u t T 1 )sin( tt 1 ). The reference tidal current speed, u t, fluctuates with sea level but becomes constant in time when averaged over the water depth. Tidal current speed is parameterized so that modeled tidal current velocities are similar to those of predicted tides in upper Chesapeake Bay (Table 1). Because our objective is to examine longitudinal processes, a few measures have been taken to suppress lateral variability as much as possible. First, the free-slip boundary condition is used on the channel walls to eliminate sidewall friction and therefore the lateral shear of longitudinal currents. Second, the Coriolis force is removed to eliminate lateral variations associated with the earth s rotation. The formulation of the hydrodynamic portion of the model is the same as documented by Mellor (199) except for three modifications. Coefficients of horizontal viscosity and diffusivity are prescribed rather than varying with local flow divergence and vorticity. The water temperature is fixed at 0 C so that the water density varies only with salinity and suspended sediment concentrations. The background vertical diffusivity (for mass) and vertical viscosity (for momentum) are parameterized as a function of the Richardson number (Ri) using the Munk-Anderson equations. The Mellor-Yamada.5 (MY.5) turbulence closure (Mellor 199) in POM rapidly defaults to userspecified background viscosity and diffusivity levels when Richardson numbers exceed 0.. Because of this, changes in the background viscosity and diffusivity result in notable differences in stratification, salt front structure, and response to wind forcing when stratification is large (Garvine 1999). Garvine (001) set background vertical diffusivities (K b ) and vertical viscosities ( b )to m s 1 in numerical investigations of buoyant plumes. These values are too high to produce a realistic salt front with an appropriate response to wind forcing in our work. We parameterized K b and b by making them a function of the Richardson number (Ri) using the Munk-Anderson equations (Munk and Anderson 194): b max (1 10 Ri) 0.5 (1) K b max ( Ri) 1.5 () where max m s 1 and Ri (g/ o )( / z)( u/ z) (3) where g is gravity, is water density, u is alongchannel current velocity, and z is depth. Minimum values for K b and b are set at near-molecular levels ( and m s 1, respectively) to

4 The Influence of Pulsed Events on an ETM 135 ensure model stability. This approach also provides a more realistic parameterization of the relationship between background viscosity and background diffusivity, which POM lacks. The absolute magnitude of the neutral viscosity ( max ) is used to tune the model, introducing some uncertainty into model results (Garvine 1999). Our model is tuned to resemble field observations of responses to wind and discharge pulses, but the uncertainty introduced by max restricts our results to qualitative, rather than quantitative, interpretations. VERTICAL AND HORIZONTAL BOUNDARY CONDITIONS At the water surface of the ETM model, fluxes of water and salt are set to zero. If the seaward or landward wind forcing is included, shear stress is applied at the water surface. At the bottom, salt and water fluxes are set to zero. The frictional boundary layer is logarithmic over a bottom roughness height of z cm. Freshwater inflow from the upstream boundary is characterized by a depth-averaged reference speed of u 0 that fluctuates about u 0 due to propagation of tidal waves and wind-induced currents through the upstream boundary. We developed a mixed boundary condition that accommodates river inflow as well as upstream transmission of waves and currents. The mixed boundary condition requires that u b t 1 eu 0 (1 e) u t (4) where u b is the longitudinal flow at the upstream boundary, u is the longitudinal flow one grid spacing into the interior domain, denotes vertical averaging, and e is the upstream dissipation parameter, a positive number less than 1. Model sensitivity to e lies mostly between 01 and 0.1 for our time step. For e 01, barotropic disturbances transmit through the upstream boundary freely and result in a delay in river inflow (u 0 ) that is proportional to the time step chosen. For e 0.1, the upstream boundary behaves like a vertical barrier to time-varying barotropic disturbances and depth-averaged river inflow fluctuates little about u 0. A modest upstream dissipation parameter (e 1) is used so that modeled sea level heights are similar to predicted tides in upper Chesapeake Bay. Lateral walls bounding the channel are impenetrable, impermeable, and free-slip. At the upstream boundary, the normal gradient for longitudinal currents (less the depth-averaged inflow) is zero. At the seaward boundary, the normal gradient for longitudinal currents is set to zero. On both landward and seaward boundaries, normal gradients of sea level are set to zero, the longitudinal gradients for salinity and suspended sediment concentrations are set to zero, and the weak tangential flow is advected by the longitudinal flow. Currents across landward and seaward boundaries are assumed to be strictly horizontal with zero vertical velocity. In the case of inflow, the incoming current does not contain a tangential component. SEDIMENT MODEL The sediment transport model is embedded in the hydrodynamic model and has the same temporal and spatial resolutions. The parameterization of sediment concentration and transport is kept as simple as possible to explore the relationship between physical forcing and sediment transport to lowest order. Elaborate treatments of flocculation for cohesive sediments such as in Cancino and Neves (1999a,b) are excluded for simplicity. As noted by Burchard and Baumert (199), processes like aggregation or destruction of flocs and consequent changes of settling velocity will modify the result quantitatively but not qualitatively. The density of sediment-water mixture is the sum of water density and the density of suspended sediment concentration per unit volume (C). The density correction caused by suspended sediment concentration is cosmetic in the application because C is too small to affect the water density significantly. The suspended sediment concentration is subject to advection and diffusion in three dimensions, while sinking with a characteristic settling velocity (w s ). The governing equation for C is C/ t (uc) H (A H C) / z(w s C K v C/ z) (5) where u is the three-dimensional velocity vector, is the three-dimensional gradient operator, H is the horizontal gradient operator and K v is the vertical eddy diffusivity. When integrated in time, the last term on the right-hand side of Eq. 5 is treated implicitly to cope with small vertical grid spacing ( z) in shallow waters. In the three (past, present, and future) time steps of interest, the implicit treatment evaluates C at the future time. Other terms in Eq. 5 are evaluated in the conventional manner. The implicit treatment stabilizes the computation considerably for large values of w s and K v. Without the implicit treatment, time steps must be reduced substantially to stabilize the computation at the expense of computational costs. No sediment flux is allowed at the water surface, so that w s C K v C/ z 0 (6) Sediment flux at the bottom is the difference between the deposition rate (D) and erosion rate (E), so that

5 136 E. W. North et al. w s C K v C/ z D E (7) The amount of erodible sediment per unit bottom area (B) is increased by deposition but decreased by erosion and is buried over time, so that db/dt D E B A s B () where D w s C, is a first order consolidation/ burial rate, and A s is the horizontal diffusivity of bottom sediment. A s, approximately 1% of the water column diffusivity, smoothes large differences in accumulation between adjacent cells. Bottom shear stress ( b ) is calculated following the commonly used quadratic law. The dimensionless bottom drag coefficient is defined as C D max{ [ln(z/z o )], 05}, where 0.4 is the von Karman constant and z is the vertical distance between the bottom and the first grid point for horizontal velocities nearest the bottom. The bottom erosion rate is E M[ b c ]H[ b c ]H[B db/dt t] (9) where b is the applied shear stress, c is the critical stress for erosion, M is the erosion rate proportionality constant, and H[x] 1ifx 0and H[x] 0ifx 0. The first Heaviside step function term (H) is the traditional critical stress threshold, while the second prevents completely emptying or overemptying the active sediment box so that erosion is switched off until the sediment layer is replenished again. For critical sheer stress for erosion, a modest value of 5 Pa is used. This choice depends on the locale and sediment type and is within the range ( Pa) used by Sanford et al. (1991) for the northern Chesapeake Bay. Sediment settling velocity is set at 0.3 mm s 1 to simulate settling velocities in the upper Chesapeake ETM region during spring (Sanford et al. 001). The sediment consolidation-burial rate ( ) is parameterized as the inverse of e-folding time scale for sediment loss (Table 1). Suspended sediment concentrations are input to the model at the upstream boundary with a concentration profile (similar to a Rouse profile) that allows constant loading. The concentration profile is specified as C(z) c Z (10) where c is depth-averaged suspended sediment concentration, and w s w s/au* 1 zw s/au*h Z (1 e ) e (11) au* TABLE. Model Run Pulse Pulse 1 Pulse 3 Pulse 5 Pulse Pulse 9 Pulse 0.11 Pulse 0.13 Pulse 0.15 Pulse 0.17 Model parameters for river pulse scenarios. Wind Stress (dyne cm ) River Flow (m s 1 ) River Flow Duration (days) where u * is bed shear velocity, a is a constant that controls the shape of the concentration profile and h is water column height. We specified a 5 to approximate a Rouse profile. Sediment concentrations predicted by Eq. 10 are dependent on sea level height (h), near bottom current velocity (u * ), and settling velocity (w s ). The profile of suspended sediment concentration enters the basin through advection by the boundary-imposed incoming current which varies with time. To maintain constant loading, c is adjusted at each time step so that the total loading entering the basin through the upstream boundary (L in kg s 1 ) is constant in time. Sediment loading rate is set so that the ratio of modeled sediment loading to river inflow is similar to the ratio of average loading to flow of the Susquehanna River in upper Chesapeake Bay (Sanford et al. 001; Table 1). MODEL SIMULATIONS All model scenarios presented in this paper started from a steady-state solution forced by constant river inflow, constant sediment loading, and semidiurnal tidal currents. The channel was initially filled with motionless clear water of 1 practical salinity units (psu) and sediment was absent from the channel. The model was run with river inflow m s 1 and constant sediment loading until the model reached a quasi steady-state at day 350. In this quasi steady-state, sediment input equaled sediment burial and the location of the salt front was stable: it oscillated with the tidal current in a repetitive cycle but did not progress up or down estuary. After the model reached steady-state conditions, perturbation experiments were conducted to examine the influence of river pulses and wind events on salt front dynamics and sediment distributions. In river pulse scenarios, river inflow was modified from day to day 360 by reducing inflow (0, 0.1, 3, and 5 m s 1 ), maintaining steady-state conditions ( m s 1 ), or increasing inflow (9, 0.11, 0.13, 0.15, and 0.17 m s 1 ) for the -d period (Table ). In wind event scenarios, different levels of along-channel wind stress (0.4, 0.7, 1.0, 1.3, and 1.6 dyne cm ) were imposed in either the up- or down-estuary direction (Table 3).

6 The Influence of Pulsed Events on an ETM 137 TABLE 3. Model parameters for wind event scenarios. Model Run Wind Stress (dyne cm ) Wind Duration (days) Wind Start (day) River Flow (m s 1 ) Down-estuary wind 1.6 Down-estuary wind 1.3 Down-estuary wind 1.0 Down-estuary wind 0.7 Down-estuary wind 0.4 Up-estuary wind 1.6 Up-estuary wind 1.3 Up-estuary wind 1.0 Up-estuary wind 0.7 Up-estuary wind Wind events began on day, had a -d duration, and were applied while maintaining the constant inflow rate ( m s 1 ). In addition, scenarios with simultaneous river and wind pulses were conducted to examine interactive effects. In these scenarios, down-estuary wind stress (1.6 dyne cm ) was applied at different river inflow rates (, 3,, 0.11, and 0.15 m s 1 ; Table 4). Changes in river flow occurred from day to 360 and the -d down-estuary wind event occurred from day to 356. In all model scenarios, salinity, sea level height, current velocities, suspended sediment concentrations, and bottom sediment concentrations were tracked over time. The salt front location (the intersection of the 1 psu isohaline with bottom) and bottom sediment distributions after modified river flow and/or added wind stress were compared to those under steady state conditions. FIELD OBSERVATIONS Axial surveys of the upper Chesapeake Bay (Fig. 1) were conducted on board the R/V Cape Henlopen (May, 13, and 14, 001; July 6, 001; and May 6 7, 00) and a small fleet research vessel ( June 1, 001) as part of the BITMAX field program. On board the R/V Cape Henlopen, a Seabird 911 CTD equipped with a calibrated 5-cm pathlength Seatech transmissometer was used to determine hydrographic conditions in the upper Bay. A Seabird SBE5 Sealogger CTD equipped with the same transmissometer was used on June 1. During all cruises, water samples were collected with a pump attached to the CTD frame to calibrate turbidity measurements to total suspended solids (TSS) following the methods described in Sanford et al. (001). All axial surveys were conducted from south to north in the direction of tidal propagation and were completed within half of a tidal cycle. Contour plots of salinity and TSS were created with Surfer 7 (Golden Software Inc.). Long-term records of wind, freshwater discharge, and water level data were examined to identify environmental conditions that influenced the ETM and salt front location in the upper Bay during the spring-summer of 001 and spring of 00. Wind speed and direction at Thomas Point Light (-min averages), daily freshwater discharge for the Susquehanna River at Conowingo, and hourly water level height at Tolchester Beach (Fig. 1) were obtained from the National Oceanic and Atmospheric Administration (NOAA) National Buoy Data Center, the United States Geological Survey (USGS) National Water Information System, and the NOAA National Ocean Service, respectively. Results STEADY-STATE MODEL After 350 d, the steady-state model produced a stratified salt front that had a tidal excursion of.9 TABLE 4. Model parameters for simultaneous wind and river pulse scenarios. Model Run River Inflow (m s 1 ) River Flow Start (day) Wind Stress (dyne cm ) Wind Start (day) Wind Duration (days) 0 Flow, 1.0 Wind 0 Flow, 1.6 Wind 3 Flow, 1.0 Wind 3 Flow, 1.6 Wind Flow, 1.0 Wind Flow, 1.6 Wind 0.11 Flow, 1.0 Wind 0.11 Flow, 1.6 Wind 0.15 Flow, 1.0 Wind 0.15 Flow, 1.6 Wind

7 13 E. W. North et al. Fig. 3. Model output. Along-channel contour plots of salinity (psu, line contours) and suspended sediment concentrations (kg m 3, shaded contours with scale on right) from the steadystate model during a) ebb and b) flood tide. Suspended sediment concentrations were averaged across the channel. Arrows are current velocity vectors (m s 1 ). km at the location where the 1 psu isohaline intersected the bottom (Fig. 3). Although the front oscillated with the tides, it remained centered at.0 river kilometer (rkm) and did not progress up- or down-estuary. High suspended sediment concentrations occurred up-estuary of, and within, the tip of the salt front, forming an estuarine turbidity maximum. Suspended sediment concentrations peaked up-estuary of the salt front during ebb tide (Fig. 3a) and within the tip of the salt front during flood tide (Fig. 3b). Although the model reproduced general characteristics of a salt front, near-bottom mixing in the model was more intense, and near-surface mixing was less intense, in comparison to field observations. The total amount of sediment in the steady-state model became constant after 00 d (Fig. 4a), indicating that sediment input to the model at the up-stream boundary equaled the amount of sediment lost due to burial. This sediment balance was possible because the sub-tidal location of the salt front was fixed, creating a stable spatial distribution of sediment erosion, deposition, and burial rates. At day 350, bottom sediment concentrations had two pronounced accumulation zones (Fig. 4b). One zone was near the up-stream boundary and resulted from sediment sinking to the bottom after it entered the model. The second accumulation zone occurred just up-estuary of the salt front, the result of convergent circulation in this area. Little sediment was found seaward of 30 rkm because near-bottom currents within the salt front were stronger on flood tide than ebb tide, continually moving sediment up-estuary toward the tip of the salt front, and because the only source of sediment in our model was riverine input. Fig. 4. Model output. Sediment distribution in steady-state model. a) Total sediment (kg) in model domain over time. b) Bottom sediment concentration (kg m ) at day 350 in the first 60 km of the model domain. RIVER PULSES In the field, a pronounced pulse in Susquehanna River discharge occurred from June, 001 (Fig. 5), 10 d after a hydrographic survey on June 1 and d before a hydrographic survey on July 6 (Fig. 6). Prior to the river pulse event, the intersection of the 1 psu isohaline with the bottom occurred at 16 rkm. After the river pulse, the intersection of the 1 psu isohaline with the bottom Fig. 5. Field observations. Wind velocity at Thomas Point Light (upper panel) and freshwater discharge from the Susquehanna River (lower panel) in June and July 001. Gray bars indicate the June 1 and July 6 hydrographic surveys, results of which are presented in Fig. 6. Wind velocity vectors pointing up (positive) indicate winds blowing to the north.

8 The Influence of Pulsed Events on an ETM 139 Fig. 6. a b) Field observations. Along-channel contour plots of salinity (psu, line contours) and total suspended solids concentrations (mg l 1, shaded contours with scale on right) from hydrographic surveys of upper Chesapeake Bay on a) June 1, 001, and b) July 6, 001. c d) Model output. Along-channel contour plots of salinity (psu, line contours) and suspended sediment concentrations (kg m 3, shaded contours with scale on right) after d of c) decreased river flow (1 m s 1 ), and d) increased river flow (0.13 m s 1 ) compared to the steady-state flow rate ( m s 1 ) (Fig. 3). Suspended sediment concentrations were averaged across the channel. Arrows are current velocity vectors (m s 1 ). Fig. 7. Model output. a) Location of the intersection of the 1 psu isohaline with the bottom versus time at different river inflow velocities. The location of the 1 psu isohaline at the steady-state river inflow rate ( m s 1 ) oscillates with the tides but does not progress up or down estuary. b c) Bottom sediment concentration (kg m ) at day 360 in the first 50 km of the model domain after d of b) decreased river flow, and c) increased river flow. was displaced down-estuary by 10 km and stratification within the salt front intensified. Down-estuary movement of the salt front in response to short-term river pulses has been observed in the upper Chesapeake Bay (Elliott et al. 197; Schubel and Pritchard 196; North and Houde 001), the Tamar estuary (Uncles and Stephens 1993), and the Weser estuary (Grabemann and Krause 001). Seasonal movement of the salt front or ETM downestuary or up-estuary in response to increased (or decreased) freshwater discharge also has been documented in studies of many other estuaries, including Chesapeake Bay (Schubel and Pritchard 196), San Francisco Bay-Delta ( Jassby et al. 1995), Mobile Bay (Noble et al. 1996), Elbe (Kappenberg and Grabemann 001), and Hudson River (Geyer et al. 001) estuaries. Modeled salt front behavior was similar to field observations in the river pulse scenarios. Decreased river flow resulted in up-estuary movement of the salt front and ETM (Fig. 6) and increased river flow resulted in down-estuary movement. A plot of the location of the intersection of the 1 psu isohaline with the bottom over time indicated that the salt front responded to changes in river inflow within 6 h (Fig. 7). Salt front displacement was approximately proportional to the change in river flow, except that the down-estuary response to increases in flow was stronger than the up-estuary response to decreases in flow. Changes in sea surface height near the river boundary likely accounted for this discrepancy. Mean sea surface height at 3 river km over the -d period was higher in increased flow scenarios than in decreased flow scenarios (by as much as 0.4 m). This appears to have created an additional pressure gradient force in increased flow scenarios that enhanced the down-estuary response of the salt front. The highest modeled river inflow rate (0.17 m s 1 ) resulted in 0.34 m s 1 down-estuary sub-tidal surface current velocity at river km 50, a value less than net flows (0.60 m s 1 ) observed during a record pulse in discharge associated with Tropical Storm Eloise in upper Chesapeake Bay (Elliott et al. 197). In the model, peak suspended sediment concentra-

9 140 E. W. North et al. Fig.. Field observations. Wind velocity at Thomas Point Light (upper panel) and water level at Tolchester Beach (lower panel) in May 001. Gray bars indicate the May, 13, and 14 hydrographic surveys, results of which are presented in Fig. 9. Wind velocity vectors pointing up (positive) indicate winds blowing to the north. tions were associated with the tip of the salt front, moving up- or down-estuary with the salt front in response to decreased or increased river flow. Model predictions indicated that bottom sediment distributions were affected when river inflow rates changed (Fig. 7). Compared to steady-state conditions, bottom sediment moved up-estuary when river inflow decreased and down-estuary when river inflow increased. Because sediment loading to the model was constant, these changes in bottom sediment distribution over an -d period represented the response of sediment to modifications in river inflow alone. Increased river flow resulted in more scour near the up-stream boundary as well as transport to, and deposition within, the accumulation zone near the salt front as it moved seaward. Decreased river flow resulted in less transport away from the up-stream boundary and an up-estuary shift in sediment distributions as the accumulation zone at the tip of the salt front progressed up-estuary. The response to increased river flow was more pronounced than the response to decreased river flow, likely because of the increased scour and because the convergence zone associated with the salt front was displaced a greater distance in increased river flow scenarios, resulting in a more discernible movement of sediment. DOWN-ESTUARY WIND EVENTS In the field, a pronounced down-estuary wind event occurred on May 13 15, 001, that resulted in decreased water levels in the upper Chesapeake Bay (Fig. ). Two hydrographic surveys framed the down-estuary wind event (Fig. 9) and an abbreviated survey occurred during the event. During the wind event, the salt front became more stratified Fig. 9. a c) Field observations. Along-channel contour plots of salinity (psu, line contours) and total suspended solids concentrations (mg l 1, shaded contours with scale on right) from hydrographic surveys of upper Chesapeake Bay on a) May, 001, b) May 13, 001, and c) May 14, 001. d e) Model output. Along-channel contour plots of salinity (psu, line contours) and suspended sediment concentrations (kg m 3, shaded contours with scale on right) from d) day.3 and e) of a downestuary wind-event scenario. Down-estuary wind stress (1.6 dyne cm ) was applied from day to day. Suspended sediment concentrations were averaged across the channel. Arrows are current velocity vectors (m s 1 ). and was located at 3 rkm. By the end of the down-estuary wind event, in less than one day, the tip of the salt front moved 10 km up-estuary. An up-estuary intrusion of the salt front in response to a down-estuary wind event also was documented in upper Chesapeake Bay in May 199 (North and Houde 001). Studies of wind-forced subtidal circulation provide supporting evidence. Indications of a baroclinic response to down-estuary wind stress (enhanced up-estuary current velocities near bottom and down-estuary currents near surface) were observed in the Providence River (Weisberg 1976), Potomac River (Elliott 197), upper Chesapeake Bay (Wang 1979), Mobile Bay (Noble et al. 1996), and Childs River (Geyer 1997) estuaries.

10 The Influence of Pulsed Events on an ETM 141 Modeled wind event scenarios also demonstrated an up-estuary intrusion of the salt front in response to down-estuary wind stress (Fig. 9). At the beginning of a down-estuary wind event, a large sediment resuspension event occurred up-estuary of the salt-front, the salt front was initially pushed down estuary, and internal waves were observed in the pycnocline. The large resuspension event occurred during ebb tide due to increased energy added by the wind. In addition, the initial downestuary movement of the salt front may have facilitated resuspension of bottom sediment accumulated near the front by removing stratification-induced suppression of turbulence (Geyer 1993). By the end of the wind event, surface waters were well mixed, the salt front had moved up-estuary, isohalines at the tip of the salt front were compressed, and sediment resuspension within the tip of the salt front was pronounced. This resuspension event was confined within the salt front by stratification and was enhanced by the intensified near-bottom currents during the up-estuary intrusion of the salt front on flood tide. The response of the salt front was rapid: the initial down-estuary movement of the salt front began 7 h, and up-estuary movement began 1 4 h, after wind stress was applied (Fig. 10). The degree of salt front movement in response to down-estuary winds depended upon the magnitude of wind stress. The stronger the wind-stress, the further the front was initially pushed down-estuary, and the larger the subsequent up-estuary response. At the end of the wind event, the strongest down-estuary wind (1.6 dyne cm ) actually resulted in slightly less up-estuary intrusion than the second strongest wind (1.3 dyne cm ). But, the upestuary response of the salt front continued for 15 h after the end of the wind event. Although the salt front location in the highest wind stress scenario was down-estuary of the salt front in the 1.3 dyne cm wind stress scenario, the total upestuary movement of the salt front (from the maximum down-estuary location during the wind event to 15 h after the wind stress was removed) was greatest in the highest wind stress scenario. The up-estuary movement of the salt front likely resulted from down-estuary transport of surface waters that lowered sea surface height in the upper estuary and created a pressure gradient that drove the up-estuary response in the lower layer. Model predictions indicated that -d down-estuary wind events transported bottom sediment down-estuary from the accumulation zone near the up-stream boundary (Fig. 10). Although some sediment was transported up-estuary from the accumulation zone near the tip of the salt front during the weak wind event (0.4 dyne cm ), strong wind Fig. 10. Model output. a b) Location of the intersection of the 1 psu isohaline with the bottom versus time in the downestuary wind event scenarios. The responses of the 1 psu isohaline to wind stresses of a) 0.4 to 1.0 dyne cm, and b) 1.0 to 1.6 dyne cm are depicted. The -d duration of the down-estuary wind event is indicated by the shaded box in each panel. c d) Bottom sediment concentration (kg m ) in the first 40 km of the model domain c) at day after dofdown-estuary wind stress (dyne cm ) and d) at day 360, 6 d after wind stress was removed. events resulted in down-estuary transport of sediment. The stronger the wind stress, the more sediment was scoured from the accumulation zone near the up-stream boundary, and the greater the movement of sediment down-estuary after d. Although sediment was transported down-estuary after d of wind stress, bottom sediment distributions continued to evolve for at least 6 d after the wind event, moving up-estuary toward the tip of the salt front (Fig. 10). UP-ESTUARY WIND EVENTS In the field, an up-estuary wind event occurred on May 6, 00, that resulted in increased water levels in the upper Chesapeake Bay (Fig. 11). Two hydrographic surveys occurred during the up-estuary wind event (Fig. 1). In the 3 h between the two surveys, the salt front moved 7 km down-estuary. Although Grabemann and Krause (001) did not find evidence of down-estuary movement of the salt front and ETM in response to storm events, studies of wind-forced subtidal circulation indicate that residual up-estuary currents near bottom can be reduced or reversed in the presence

11 14 E. W. North et al. Fig. 11. Field observations. Wind velocity at Thomas Point light (upper panel) and water level at Tolchester Beach (lower panel) in May 00. Gray bars indicate the May 6 7 hydrographic surveys, results of which are presented in Fig. 1. Wind velocity vectors pointing up (positive) indicate winds blowing to the north. of up-estuary wind stress (Weisberg 1976; Elliott 197; Wang 1979; Geyer 1997), suggesting that down-estuary movement of the salt front is possible. Model scenarios showed that up-estuary wind stress resulted in down-estuary movement of the Fig. 13. Model output. a) Location of the intersection of the 1 psu isohaline with the bottom versus time in the up-estuary wind event scenarios. The responses of the 1 psu isohaline to different levels of wind stresses (dyne cm ) are depicted. The -d up-estuary wind event occurred from day to. b c) Bottom sediment concentration (kg m ) in the first 50 km of the model domain b) at day after d of down-estuary wind stress (dyne cm ) and c) at day 360, 6 d after wind stress was removed. Fig. 1. a b) Field observations. Along-channel contour plots of salinity (psu, line contours) and total suspended solids concentrations (mg l 1, shaded contours with scale on right) from hydrographic surveys of upper Chesapeake Bay on a) May 6, 00, and b) May 7, 00. c) Model output. Along-channel contour plot of salinity (psu, line contours) and suspended sediment concentrations (kg m 3, shaded contours with scale on right) from day of an up-estuary wind-event scenario. Upestuary wind stress (1.6 dyne cm ) was applied from day to day. Suspended sediment concentrations were averaged across the channel. Arrows are current velocity vectors (m s 1 ). salt front (Fig. 1) that increased in magnitude with increasing wind stress (Fig. 13). The up-estuary transport of surface water resulted in increased sea surface height near the up-stream boundary of the model, setting up an adverse pressure gradient that forced down-estuary flow in the lower layer, pushing the salt front down-estuary. During the wind event, sediment was suspended up-estuary of the salt front on ebb tide (Fig. 1) but little was resuspended within the salt front on flood tide,

12 The Influence of Pulsed Events on an ETM 143 most notably after the salt front had moved downestuary of peak bottom sediment accumulations. Model predictions indicated that bottom sediment was redistributed down-estuary after d of up-estuary wind stress (Fig. 13), and that the magnitude of this redistribution increased with increasing wind stress. Bottom sediment was broadly distributed down-estuary at the end of the wind event, but became focused near the tip of the new salt front location after 6 d of tidal resuspension, constant river flow, and zero wind stress. SIMULTANEOUS WIND AND RIVER PULSES Simultaneous wind and river pulse scenarios were conducted to examine potential interactive effects of pulsed events. In these scenarios, downestuary wind stress enhanced the up-estuary movement of the salt front when river flow was reduced (compare lines for and 3 m s 1 in Fig. 7 and Fig. 14). Down-estuary wind stress opposed the down-estuary movement of the salt front when river flow was increased (compare lines for 0.11 and 0.15 m s 1 in Fig. 7 and Fig. 14). Increased flow dampened the response of the salt front to downestuary wind stress (compare lines for 0.11 and 0.15 m s 1 with the steady-state flow rate of m s 1 in Fig. 14). Down-estuary wind stress enhanced bottom sediment transport in the presence of both reduced and increased river flow compared to changes in river flow alone (Fig. 14b). When river flow was reduced, down-estuary wind stress enhanced sediment transport up-estuary because the strong upestuary intrusion of the salt front intensified upestuary near-bottom currents within the salt front, enhancing resuspension and up-estuary transport. Strong down-estuary wind stress coupled with high river inflow enhanced sediment transport down-estuary (Fig. 14) because near-bottom current velocities up-estuary of the salt front were intensified, increasing scour and rapidly transporting sediment down estuary. Discussion Numerical studies and field observations demonstrate that the salt front and ETM have dynamic and non-linear responses to short-term pulses in river flow and wind. Model results also suggest that pulsed events have a significant effect on the transport and redistribution of sediment in upper estuaries. In numerical studies and field observations, increased river flow or up-estuary wind events resulted in down-estuary movement of the salt front. In contrast, down-estuary wind events or reduced river flow resulted in up-estuary movement. Although wind and river pulses caused similar movements of Fig. 14. Model output. a) Location of the intersection of the 1 psu isohaline with the bottom versus time in simultaneous river and wind pulse scenarios. A constant down-estuary wind stress was applied during different river inflow rates (m s 1 ). Changes in river flow occurred from day to 360. The -d down-estuary wind event (1.6 dyne cm ) occurred from day to 356. b) Bottom sediment concentration (kg m ) at three river inflow rates: reduced flow ( m s 1 ), steady-state flow (0.7 ms 1 ), and increased flow (0.15 m s 1 ). The first set of lines represent bottom sediment at the start of the wind event (day ), the second set represents bottom sediment after d of constant river inflow, and the third set represents bottom sediment after d of constant river inflow and down-estuary wind stress. the salt front, the flow-induced response was depth-independent while the wind-induced response was depth-dependent. The rapid response of the salt front to both hydrological and meteorological forcing is notable, suggesting that shortterm changes in wind and river flow can significantly impact circulation patterns and sediment transport in upper estuaries. The quick response of the salt front to wind and discharge events and the lag in bottom sediment accumulation near the salt front after wind events may help explain why

13 144 E. W. North et al. the location of salt front and ETM do not always coincide in field observations (Sanford et al. 001). In model simulations, the location of peak concentrations of suspended sediment in relation to the tip of the salt front depended upon tidal stage as well as circulation patterns driven by river flow and wind events. In steady-state conditions, peak suspended sediment concentrations during flood tide occurred within the salt front, while those on ebb tide occurred just up-estuary of the salt front. The pulsed river and wind events affected suspended sediment distributions by modifying the location of the salt front, near-bottom shear stresses, and the location of bottom sediment in relation to stratification within the salt front. The large resuspension event up-estuary of the salt front on ebb tide at the beginning of the down-estuary wind event was especially notable. Although field observations of this event would provide validation of model predictions, it is important to recognize that a simplified model such as this cannot reproduce all the intricacies of sediment processes in the field. Despite this drawback, the modeling results of Brenon and Le Hir (1999) suggest that our model likely captures first-order sediment dynamics. Brenon and Le Hir found that qualitative sediment patterns resulting from interactions with a salinity front were independent of sediment characteristics such as settling velocity and critical shear stress. Model results demonstrated the dynamic nature of sediment transport during pulsed events. The model was parameterized with constant sediment loading to isolate physical processes responsible for redistribution of sediment in the upper estuary. In reality, peak discharge events deliver up to 90% of annual total suspended sediment input to an estuary (Vale et al. 1993). Model results suggest that, in addition to adding more sediment to the basin, peaks in discharge will scour sediment from within the basin and transport sediment down-estuary as the salt front is pushed down-estuary. Wind events influenced sediment distributions by increasing scour and shifting the position of the salt front. Throughout model simulations, the convergent zone at the tip of the salt front was the primary feature controlling the accumulations of bottom sediment, most notably after the relaxation of wind stress (Figs. 10 and 13). At the end of the wind event, both down- and up-estuary winds had caused down-estuary transport of bottom sediment, sediment that was eventually transported toward the tip of the salt front after wind stress was removed. Only reductions in river flow resulted in consistent up-estuary movement of bottom sediment. Results of simultaneous river and wind pulse scenarios showed that the response of the salt front to wind depended on the strength of river inflow. During increased flow, the effect of the down-estuary wind on the salt front was dampened, while during reductions in flow, the effect of wind was enhanced. This suggests that the effect of simultaneous wind and river pulses depends upon a balance between the strength of depth-independent (river-induced) and depth-dependent (wind-induced) responses. Although the response of the salt front differed, sediment transport was enhanced by down-estuary wind stress in both the up- (reduced river flow) and down- (increased river flow) estuary directions due to added scour from wind-enhanced near-bottom current velocities. In the theoretical model presented here, the combination of suspended sediment input from the up-stream boundary, symmetrical tides, and gravitational circulation associated with saline waters near the tip of the salt front created and maintained a steady-state ETM. This model includes asymmetrical tidal resuspension (Sanford et al. 001) and satisfies the theoretical mechanisms necessary for ETM formation (Burchard and Baumert 199): gravitational circulation (Festa and Hansen 197) and asymmetric tidal advection ( Jay and Musiak 1994). Asymmetric tidal advection and resuspension were induced by gravitational circulation coupled with symmetrical tides. In addition, model results highlight the significant influence of stratification within the salt front on suspended sediment concentrations (Geyer 1993) during wind and river pulse events. A single source of sediment was sufficient to produce an ETM in the theoretical model, but in reality there are multiple sources of sediment from both up- and down-estuary as well as from shoals and shoreline erosion. In the model, sediment was input from the up-stream boundary and erodible bottom sediment only occurred where resuspension, transport, and deposition patterns caused sediment accumulation at time scales shorter than the consolidation-burial rate. This model formulation is consistent with evidence that ETMs are often associated with a spatially limited pool of erodible particles (e.g., Uncles and Stephens 1993; Jay and Musiak 1994; Geyer et al. 199; Brenon and Le Hir 1999) and that ETMs trap river-borne particles, most of which are accumulated and buried in the bottom sediments (Biggs 1970; Colman et al. 199; Sanford et al. 001). The current model configuration confines the scope of inference of model results to the effect of pulsed events on river-borne sediment. Adding multiple sediment sources would expand model inference, as would enhancements such as building a more sophisti-