Integrating Coastal Models and Observations for Studies of Ocean Dynamics Observing Systems and Forecasting

John L. Wilkin, Weifeng G. Zhang, Bronwyn E. Cahill and Robert C. Chant

Abstract In coastal oceanography, simulation models are used to a variety of ends. Idealized studies may address particular dynamical processes or features of coastline and bathymetry; reproducing the circulation in a geographical region can compliment studies of ecosystems and geomorphology; and models may be employed to simulate observing systems and to forecast oceanic conditions for practical operational needs. Frequently, the interplay between multiple forcing mechanisms, geographic detail, stratification, and nonlinear dynamics, is significant, and this demands that ocean models for coastal applications are capable of representing a comprehensive suite of dynamical processes. Drawing on a series of recent modelbased studies of the inner to mid-shelf region of the Middle Atlantic Bight (MAB) we illustrate, by example, these methodologies and the breadth of dynamical processes that influence coastal ocean circulation. We demonstrate that the recent introduction of variational methods into coastal ocean simulation is a development that greatly enhances our ability to integrate models with data from the evolving coastal ocean observatories for the purposes of improved ocean prediction, adaptive sampling and observing system design.

19.1 Introduction

The discharge of rivers to continental shelf seas represents an important mechanism by which human activities in urban watersheds impact the neighbouring marine environment. Biogeochemical, sediment, and ecosystem processes that determine the ultimate fate of nutrients and pollutants delivered into the coastal ocean by river sources depends on the pathways and time scales of dispersal of these buoyant discharges. How coastal models, in conjunction with observations, can be used to study these circulation processes is illustrated here by example, by reviewing results

Institute of Marine and Coastal Sciences, Rutgers, The State University of New Jersey, New Brunswick, NJ, USA e-mail: [email protected]

A. Schiller, G. B. Brassington (eds.), Operational Oceanography in the 21st Century, 487

DOI 10.1007/978-94-007-0332-2_19, © Springer Science+Business Media B.V. 2011

from a recent series of model-based studies of the Hudson River outflow into New York Bight.

On many coasts, the flux of freshwater from rivers or groundwater first enters an estuary where it mixes with more salty waters of oceanic origin before reaching the adjacent shelf sea. The salinity of the estuary discharge can be sufficiently low that horizontal buoyancy gradients are a significant force influencing the plume circulation. A classical view of the ensuing dynamics is that the buoyancy force balances the Coriolis force, and the outflow turns to the right (in the northern hemisphere) and forms a narrow coastal current a few internal Rossby radii in width trapped against the coast. If the front that defines the outer extent of the low salinity water reaches the sea floor then the plume becomes bottom-attached and details of the coastal bathymetry strongly influence the plume trajectory. Alternatively, if the low salinity discharge is confined to a relatively thin surface layer the plume is described as surface-advected (Yankovsky and Chapman 1997) and may be more responsive to local wind forcing. Whether a plume falls into the surface-advected or bottom-trapped regime, or transitions from one regime to the other, depends on river discharge, bathymetry, and mixing within the surface and bottom boundary layers.

It is often the case that the freshwater transport of the coastal current is less than the freshwater flux out of the estuary, particularly during episodes of elevated river discharge, and this leads to the formation of a pronounced low salinity bulge near the estuary outflow. The across shelf scale of the bulge can be several times the width of the coastal current, especially for a surface-advected plume. The low buoyancy of the bulge evolves an anti-cyclonic circulation that significantly prolongs the duration that water discharged from the estuary is retained in the vicinity of the estuary mouth. Laboratory rotating tank experiments have shown that the coastal current can receive as little as one third of the estuary outflow (Avicola and Huq 2003), or in extreme circumstances the recirculation can pinch off from the coastal current and for a period of time direct all flow into the bulge (Horner-Devine et al. 2006). Numerical model studies show that the ratio of coastal current transport to estuary discharge decreases as the flow becomes increasingly non-linear as characterized by the Rossby number, i.e. the ratio of inertial to rotational forces (Fong and Geyer 2002; Nof and Pichevin 2001).

Thus river flow rate, vertical turbulent mixing within the estuary and on the shelf, bathymetric detail, stratification, non-linear dynamics, and wind forcing are all factors that influence river plume dispersal characteristics. Shelf-wide along-shelf mean currents established by regional winds (Fong and Geyer 2002) or by upstream or offshore remote forcing further influence the circulation (Zhang et al. 2009a). Consequently, ocean models that seek to simulate interactions between river discharges and the adjacent inner shelf must be quite comprehensive in the suite of dynamical processes that they represent.

In this article we demonstrate the capabilities of one such model, the Regional Ocean Modelling System (ROMS; www.myroms.org), by summarizing results from a sequence of studies of the Hudson River's discharge into the coastal ocean based on efforts during the Lagrangian Transport and Transformation Experiment (LaTTE) (Chant et al. 2008). The Hudson River watershed is highly industrialized, and the LaTTE field program included observations—following the river plume associated with the spring freshest in the years 2004, 2005 and 2006—of phytoplank-ton and zooplankton assemblages, and natural and human-source nutrients, organic matter, and metal contaminants. An emphasis of the project was to investigate how the plume's physical structure influenced biogeochemical processes. Key processes in this regard include mixing that dilutes salinity and influences certain chemical reactions, light levels that affect photochemistry, and residence times and transport pathways that can impact rates of bioaccumulation and modify where regions of net export of particulate suspended matter might occur.

Dynamical and computational features of ROMS that are pertinent to the LaTTE simulations (and coastal processes in general) are described in Sect. 19.2, and revisited in Sect. 19.4 in a discussion of aspects of New York Bight (NYB) regional dynamics worthy of further analysis. Section 19.3 describes modelling approaches we have taken to address specific scientific objectives. Section 19.3.1 considers forward simulations initialized from climatology and forced with observed river flows and an atmospheric forecast model used for short-term forecasting for adaptive sampling during the LaTTE field experiments, and idealized studies of how the plume responds to the wind. Multi-year simulations to examine long-term transport and dispersal pathways, and the mean dynamics of the circulation, are presented in Sect. 19.3.2. Section 19.3.3 describes a reanalysis of the 2006 LaTTE season using Incremental Strong Constraint 4-Dimensional Variational Data Assimilation (IS4DVAR) to adjust initial conditions to each daily forecast cycle, and gives a brief overview of how variational methods might also be employed to assist observing system operation. In Sect. 19.5 it is summarized how the studies described here collectively illustrate how coastal models are being increasingly integrated with the growing network of regional coastal ocean observing systems to better understand coastal ocean processes, and improve ocean predictions.

19.2 Regional Ocean Modelling System 19.2.1 Dynamical and Numerical Core

ROMS solves the hydrostatic, Boussinesq, Reynolds-averaged Navier-Stokes equations in terrain-following vertical coordinates. It employs a split-explicit formulation whereby the 2-dimensional continuity and barotropic momentum equations are advanced using a much smaller time step than the 3-dimensional baroclinic momentum and tracer equations.

The ROMS computational kernel is described elsewhere (Shchepetkin and Mc-Williams 2005, 2009a, b) and will not be detailed here, but we do note several aspects of the kernel that are particularly attractive for coastal ocean simulation. These include a formulation of the barotropic mode equations that accounts for the non-uniform density field so as to reduce aliasing and coupling errors associated with the split-explicit method (Higdon and de Szoeke 1997) in terrain-following coordinates. Temporal-weighted averaging of the barotropic mode prevents alias ing of unresolved signals into the slow baroclinic mode while accurately representing barotropic motions resolved by the baroclinic time step (e.g. tides and coastal-trapped waves). Several features of the kernel substantially reduce pressure-gradient force truncation error that has been a long-standing problem in terrain following coordinate ocean models. A finite-volume, finite-time-step discretization for the tracer equations improves integral conservation and constancy preservation properties associated with the variable free surface, which is important in coastal applications where the free surface displacement represents a significant fraction of the water depth. A positive-definite MPDATA (multidimensional positive definite advection transport algorithm) advection scheme (Smolarkiewicz 1984) is available, which is attractive for biological tracers and sediment concentration. A monotonized, highorder vertical advection scheme for sinking of sediments and biological particulate matter integrates depositional flux over multiple grid cells so it is not constrained by the CFL criterion (Warner et al. 2008a).

Interested readers are referred to Shchepetkin and McWilliams (2009b) for a thorough review of the choices of algorithmic elements that make ROMS particularly accurate and efficient for high-resolution simulations in which advection is strong, and currents, fronts and eddies are approximately geostrophic—characteristics of mesoscale processes in the coastal ocean and adjacent deep sea.

19.2.2 Vertical Turbulence Closure

ROMS provides users with several options for the calculation of the vertical eddy viscosity for momentum and eddy diffusivity for tracers. In the majority of recent ROMS coastal applications the choice of vertical turbulence-closure formulation has been either (1) a k-profile parameterization (KPP) for both surface and bottom-boundary layers (Large et al. 1994; Durski et al. 2004), (2) Mellor-Yamada level 2.5 (MY25) (Mellor and Yamada 1982), or (3) the generic length-scale (GLS) method (Umlauf and Burchard 2003) which encompasses a suite of closure and stability function options.

The KPP scheme specifies turbulent mixing coefficients in the boundary layers based on Monin-Obukhov similarity theory, and in the interior principally as a function of the local gradient Richardson number (Large et al. 1994; Wijesekera et al. 2003). The KPP method is diagnostic in the sense it does not solve a time evolving (prognostic) equation for any of the elements of the turbulent closure, whereas the MY25 and GLS schemes are of the general class of closures where two prognostic equations are solved—one for turbulent kinetic energy and the other related to turbulence length scale.

Warner et al. (2005) describe the implementation of the GLS formulation in ROMS, and contrast the performance of the various GLS sub-options (representing different treatments of the turbulent length scale) and the historically widely used MY25 scheme. They find that the differing schemes lead to differences in the vertical eddy mixing profiles, but the net impact on profiles of model state variables

(velocities and tracers) is relatively minor. Wijesekera et al. (2003) reach similar conclusions, but note that results for KPP tend to be less similar to GLS and MY25, which are quite alike. Warner et al. (2005) found that suspended sediment concentrations in their sediment transport model are much more sensitive to the choice of closure than is salinity in estuarine mixing simulations. In the LaTTE simulations we use the GLS k-kl closure option, which is essentially an implementation of MY25 within the GLS conceptual framework.

19.2.3 Forcing 19.2.3.1 Air-Sea Fluxes

Air-land-sea contrasts, orography, upwelling, fog, and tidal mixing over variable bathymetry in the coastal ocean can all contribute to creating wind and air temperature conditions at sea level that have much shorter time and length scales than typically occur further offshore or in the open ocean. Accordingly, coastal ocean simulations benefit from the availability of spatially and temporally well-resolved meteorological forcing and accurate parameterization of air-sea momentum and heat fluxes.

Surface atmospheric forcing in the LaTTE simulations made use of two sets of marine boundary layer products derived from atmospheric models. The short time scale simulations (Sect. 19.3.1) and IS4DVAR reanalysis (Sect. 19.3.3) used marine boundary conditions (downward long-wave radiation, net shortwave radiation, 10-m wind, 2-m air temperature, pressure and humidity) at 3-hourly intervals from the North American Mesoscale model (NAM; Janjic 2004)—a 12 km resolution 72-h forecast system operated by the National Centers for Environmental Prediction (NCEP). The multiyear simulations (Sect. 19.3.2) used marine boundary layer conditions taken from the North American Regional Reanalysis (Mesinger 2006)—a 25-km resolution 6-hourly interval data assimilative reanalysis product. Air-sea fluxes of momentum and heat were computed using standard bulk formulae (Fairall et al. 2003) from the atmospheric model based marine boundary layer conditions in conjunction with the sea surface temperature from ROMS.

19.2.3.2 River Inflows and Open Boundary Conditions

In coastal Regions of Freshwater Influence (ROFI) (Hill 1998), lateral buoyancy input from rivers produces density gradients that are principally horizontal, which leads to relatively weak vertical stability compared to the vertical stratification generated from comparable surface air-sea buoyancy fluxes. Density stratification in ROFI subsequently arises from the baroclinic adjustment of these density gradients, and destratification and restratification can occur rapidly in response to changing rates of vertical mixing associated with wind forcing and tides (which may have significant spring-neap variability in intensity).

On some coasts, groundwater discharge directly to the coastal ocean or freshwater input from numerous small streams and rivers can be significant, but in the NYB terrestrial buoyancy input is overwhelmingly from large rivers, and predominantly from the Hudson. For river input to the LaTTE model we used daily average observations of river discharge from U.S. Geological Survey gauging stations on the Hudson and Delaware rivers, modified to include ungauged portions of the watershed following Chant et al. (2008).

At the open boundaries to the LaTTE model domain, simple Orlanski-type radiation conditions were applied to tracers (temperature and salt) and 3-D velocity. Our emphasis here on the buoyancy driven circulation associated with the Hudson River plume allows this simplification with its implicit neglect of the influence of remote sources of freshwater and heat. Open boundary sea level and depth-averaged velocity variability was set using the Chapman (1985) and Flather (1976) schemes to radiate surface gravity waves while also imposing tidal harmonic velocity variability derived from a regional tide model (Mukai et al. 2002). In the long multiyear simulations (Sect. 19.5), the boundary depth averaged velocity was augmented with the estimate of mean southwestward current on the shelf derived by Lentz (2008) based on long-term current-meter observations and momentum balance arguments.

19.2.4 Sub-Models for Interdisciplinary Studies

ROMS incorporates a set of sub-models for interdisciplinary applications that are integrated with the dynamical kernel. Among these are several ecosystem models formulated in terms of Eulerian functional groups wherein 3-D tracers representing nutrients, plankton, zooplankton, detritus, etc., expressed in terms of some common currency (usually equivalent nitrogen concentration), are advected and mixed according to the same transport equations as the dynamic tracers. Haidvogel et al. (2008) give an overview of examples of these models, which range in complexity from a four component nitrogen-based (NPZD) model (Powell et al. 2006; Moore et al. 2009) to a carbon based bio-optical model (EcoSim) (Bissett et al. 1999; Cahill et al. 2008) with a spectrally resolved light field and more than 60 state variables representing four phyto-plankton, five pigments, five elements, bacteria, dissolved organic matter, and detritus.

A Community Sediment Transport Model (CSTM; Warner et al. 2008a) and wave model (SWAN, Surface Waves in the Nearshore; Booij et al. 1999) are integrated with ROMS for studies of sediment dynamics and circulation in nearshore environments; wave radiation stresses are included in the momentum equations and wave-current interaction that enhances bottom stress is included in the bottom boundary layer dynamics. A user-defined set of non-cohesive sediment classes is tracked, with differential erosion and deposition of the various size classes contributing to the evolution of a multi-level sediment bed with varying layer thickness, porosity, and mass, which allows computation of bed morphology and stratigraphy. The application of the ROMS/ SWAN/CSTM to studies of sediment morphology, sorting and transport in an idealized tidal inlet and Massachusetts Bay are presented by Warner et al. (2008a).

19.3 ROMS Simulations of the New York Bight Region for LaTTE

19.3.1 Dispersal of the Plume During High River Discharge

The ROMS model domain for LaTTE (Fig. 19.1) extends from south of Delaware Bay to eastern Long Island, and from the New Jersey and New York coasts to roughly the 70-m isobath. The model has 30 vertical layers and horizontal grid resolution is 1 km.

In spring 2005 and 2006 the model was used to forecast circulation in the NYB in support of LaTTE field observation programs (Foti 2007). Figure 19.2 shows vis

Fig. 19.1 The model domain (black line) and locations of observations used in the 4DVAR data assimilation (Sect. 19.3.4). Bathymetry of the New York Bight is in greyscale; black dash lines are model isobaths in metres; yellow star in the location of Ambrose Tower; green squares indicate the five HF radar stations

Longitude

Fig. 19.1 The model domain (black line) and locations of observations used in the 4DVAR data assimilation (Sect. 19.3.4). Bathymetry of the New York Bight is in greyscale; black dash lines are model isobaths in metres; yellow star in the location of Ambrose Tower; green squares indicate the five HF radar stations

Observations

Myroms Transect

Fig. 19.2 Left: Visible imagery from Ocean Colour Monitor (OCM) instrument aboard Indian IRS-P4 satellite, and MODIS instrument aboard NASA Terra satellite showing turbid waters associated with the Hudson River discharge, and vectors of surface current from HF radar (CODAR), on two days during the spring 2005 LaTTE experiment. Right: Modelled surface salinity and currents at the corresponding times

Fig. 19.2 Left: Visible imagery from Ocean Colour Monitor (OCM) instrument aboard Indian IRS-P4 satellite, and MODIS instrument aboard NASA Terra satellite showing turbid waters associated with the Hudson River discharge, and vectors of surface current from HF radar (CODAR), on two days during the spring 2005 LaTTE experiment. Right: Modelled surface salinity and currents at the corresponding times ible satellite imagery of the Hudson River plume as it enters the NYB on two days in 2005 overlaid with vectors showing surface current observed by HF-radar, and the modelled velocity and surface salinity and corresponding time—surface salinity being a proxy for the signature of the river source waters. A recirculating bulge of low salinity water is being over-run by a renewed ebb tide discharge of Hudson River estuary waters. Figure 19.3 compares satellite observed absorption at wavelength 488 nm from Oceansat-1 (a proxy for relative chlorophyll abundance and the presence of river source water) with the modelled equivalent freshwater thickness f = f-h (So - S(z))/S0dz, where S is salinity, h is the water depth, and z = Z is the sea surface. If it were possible to locally "unmix" the water column into two layers of salinities zero and So, the thickness of the fresh water layer would be &fw. This depicts the horizontal extent of freshwater dispersal more faithfully than sea surface salinity. Here we use a reference salinity So=32.

Figures 19.2 and 19.3, and further model-data comparisons in Zhang et al. (2009a), indicate that fundamental features of the river plume circulation such as the across and along-shelf length scales, the extent of the freshwater bulge, veloc-

Myroms Transect

Fig. 19.3 Top row: Modelled equivalent freshwater thickness in meters (left) and satellite observed absorption at wavelength 488 nm from Oceansat-1 (right) showing the patterns of influence of Hudson River source waters. Bottom: Observed and modelled salinity along the northernmost west-east transect indicated in the top right panel

Fig. 19.3 Top row: Modelled equivalent freshwater thickness in meters (left) and satellite observed absorption at wavelength 488 nm from Oceansat-1 (right) showing the patterns of influence of Hudson River source waters. Bottom: Observed and modelled salinity along the northernmost west-east transect indicated in the top right panel ity patterns, and the transport pathway from the harbor to the coastal current, are similar in model and observations.

Figure 19.4 shows the time evolution of simulated equivalent freshwater thickness during the spring freshet of 2005. From 1 to 7 April the river discharge exceeded 2,500 m3/s, or more than four times the annual mean, and peaked at 6,500 m3/s on 4 April. Initially, southward downwelling favourable winds drive the river plume rapidly southward along the New Jersey coast, but this flow is abruptly arrested on 4 April with the onset of northward upwelling favourable winds. This causes the river flow during peak discharge to form a large low-salinity recirculating bulge located predominantly on the northern side of the Hudson Shelf Valley. From 10 to 15 April a period of weak and variable winds associated with the sea breeze phenomenon enable the bulge to partially drain into a New Jersey coastal current. The return of upwelling winds on April 17 drives more low salinity water eastward and detaches the bulge

31-Mw 02-Apr 04-Apr 06-Aer M-Jkp. 1O-A0f li-Apr

31-Mw 02-Apr 04-Apr 06-Aer M-Jkp. 1O-A0f li-Apr

Fig. 19.4 Modelled equivalent freshwater thickness in meters during the spring freshet of 2005 and winds observed at Ambrose Tower in the New York Bight apex

from the estuary discharge that previously fed it. In the week that follows, sustained winds further disperse the plume as the river discharge drops and the freshet ends.

The influence of wind direction and strength on Hudson River plume dispersal has been considered in some detail (Choi and Wilkin 2007) using the same model but for idealized winds and freshet river discharge. Figure 19.5 contrasts the plume behaviour commencing from the same initial conditions (Fig. 19.5a) in response to winds from differing directions (Fig. 19.5d-g) sustained for 3 days. The sensitivity described for the April 2005 simulations is confirmed. Southward winds, and to a lesser extent eastward winds, favour New Jersey coastal current formation. Northward winds eliminate the buoyancy-driven coastal current, disperse the bulge eastward and drive flow along the Long Island coast. Westward winds hamper the discharge from the Hudson River estuary, leading to a build up of low salinity water in New York Harbor. In the absence of wind forcing, the low salinity bulge continues to grow in volume in agreement with the modelling and tank experiments noted in Sect. 19.1. In the LaTTE region then, winds play a crucial role in determining the fate of material transported by the Hudson River to the inner shelf.

Choi and Wilkin (2007) also considered the influence of river discharge magnitude on the relative contribution of buoyancy and wind forcing to the momentum balance of the river plume. They found that relatively modest wind speeds of order 5 m/s are sufficient to overwhelm buoyancy forcing during typical non-freshet conditions.

It follows then that relatively short timescale variability in river discharge and weather conditions could lead to different dispersal patterns for the freshet in any

Jeux Des Familles Imprimer

Plume moves north of Plume water forms broad Freshwater moves to northern Freshwater accumulates in

Hudson Shelf valley and southward flow and drains rip el Hudson Shelf valley and P.vic.il Bay and is arrested at toward Long Island. to the south. accumulates- south of Raritan Bay mouth.

Fig. 19.5 Surface salinity of the Hudson River plume showing sensitivity of plume trajectory to wind during a high discharge event (3,000 m3/s)

Plume moves north of Plume water forms broad Freshwater moves to northern Freshwater accumulates in

Hudson Shelf valley and southward flow and drains rip el Hudson Shelf valley and P.vic.il Bay and is arrested at toward Long Island. to the south. accumulates- south of Raritan Bay mouth.

Fig. 19.5 Surface salinity of the Hudson River plume showing sensitivity of plume trajectory to wind during a high discharge event (3,000 m3/s)

given year, and this was indeed found to be the case in the three LaTTE field seasons (Chant et al. 2008). In 2004, river waters were first transported southward in a modest coastal current, and then dispersed eastward in the surface Ekman layer associated with strong upwelling winds; 2005 was characterized by strong bulge formation and sea breeze activity as described above; while in 2006 unusually large river discharge fed a coastal current that flooded the New Jersey inner shelf with low salinity water, but this flow subsequently detached from the coast leading to significant across-shelf transport in the region south of the Hudson Shelf valley.

19.3.2 Shelf-Wide Transport and Dispersal Pathways

The preceding studies revealed that while some processes act to trap river plume water near the apex of the NYB (i.e. the recirculating bulge, and coastal current flow reversals) others disperse it widely (i.e. fast coastal currents and offshore wind-driven Ekman transport). Therefore the duration that river source waters dwell in the vicinity of the coastline can be quite variable, and questions arise as to where these waters eventually go.

To examine the ultimate fate of Hudson River source waters on time scales much longer than the spring freshet, we conducted multi-year simulations using the same model configuration but with modified open boundary inflow/outflow transport conditions and meteorology forcing from NARR.

The open boundary conditions were adapted to acknowledge that on inter-annual timescales the mid and outer New Jersey shelf is flushed by a southwestward along-shelf mean flow. An analysis of long term current meter observations and the mean momentum balance (Lentz 2008) indicates the depth-averaged along-shelf current is roughly proportional to water depth; this provides a convenient relationship upon which to base the time mean boundary transports to which we add the tidally varying currents.

The modelled mean circulation for 2005-2006 (Zhang et al. 2009a) is shown in Fig. 19.6. Buoyancy input from the Hudson River dominates flow in the apex of the NYB by driving the anticyclonic recirculation (a local maximum in sea surface height, SSH) associated with the low salinity bulge. This feature is sustained in the annual mean because it is the consequence not only of the spring freshet but also

Fig. 19.6 Mean SSH (sea surface height) contours (a, top), and velocity at sea surface (b, centre) and 20-m depth (c, bottom) over the 2-year period 2005-2006

Fig. 19.6 Mean SSH (sea surface height) contours (a, top), and velocity at sea surface (b, centre) and 20-m depth (c, bottom) over the 2-year period 2005-2006

of high discharge events that can occur throughout the year. In the 3 years of the LaTTE program, the peak discharge actually occurred in July 2006 following heavy rains across all of New York State.

Transport is eastward along the Long Island coast, but this current ultimately detaches from the coast and reverses in the face of the mean flow that enters from the eastern open boundary.

On the mid to outer shelf the flow is to the southwest, largely parallel to isobaths, and deflected by the Hudson Shelf Valley as evidenced by the currents at 20 m (Fig. 19.6c). The influence of the valley extends throughout the water column and affects SSH. In the very apex of NYB the flow at 20 m is toward New York Harbor, indicating that the HSV serves as a conduit for shoreward flow that is vertically mixed and entrained into the estuary outflow and bulge recirculation. Away from the coast the surface currents (Fig. 19.6b) are dominated by southward wind-driven Ekman flow.

A New Jersey coastal current is not readily apparent in the annual mean. Zhang et al. (2009a) show it is prominent in spring and fall, moderate in winter, but overwhelmed by upwelling winds in the summer.

To avoid the ambiguity of reference salinity in lengthy simulations and to distinguish the Hudson River from other freshwater sources, Zhang et al. (2009a) introduce a passive tracer with unit concentration in the modelled Hudson River source and follow it to obtain an unambiguous measure of the dispersal pathways. Figure 19.7 shows the flux of Hudson River source water identified by its tracer signature across a set of arcs centred on the Harbor entrance. The qualitative features noted above are again evident. The New Jersey coastal current is clearly very tightly trapped against the coast, which partly explains why it is not conspicuous in Fig. 19.6a, b. Figure 19.7 quantifies the volume transports across sectors of the

Fig. 19.7 Left: Two-year averaged, vertically integrated freshwater flux (thick black lines) across arcs of radius 20, 40, 60, 80, 100, and 120 km (numbered 1-6) centred at the entrance to New York Harbor (star). Right: Freshwater transport (m3/s) across the segments of the arcs on either side of the Hudson Shelf Valley (gray dashed-dotted line), and across the valley itself

Fig. 19.7 Left: Two-year averaged, vertically integrated freshwater flux (thick black lines) across arcs of radius 20, 40, 60, 80, 100, and 120 km (numbered 1-6) centred at the entrance to New York Harbor (star). Right: Freshwater transport (m3/s) across the segments of the arcs on either side of the Hudson Shelf Valley (gray dashed-dotted line), and across the valley itself arcs split at the HSV. In this 2-year mean, we see that river discharge is entirely to the shelf north of the HSV but that the majority of this flow subsequently crosses the valley within the general region of the recirculating bulge. Once south of the valley, the outflow is partitioned between the coastal current and a weaker but much broader across-shelf pathway guided by the south flank of the HSV. The latter current feature has been noted from HF radar surface current observations (Castelao et al. 2008). Despite initially entering the coastal ocean along the New York coast, the Hudson River discharge is thus ultimately dispersed to the mid and outer shelf on the south side of the Hudson Shelf Valley.

Biogeochemical observations during LaTTE (Moline et al. 2008) support the notion that the coastal current is typically supplied with biogeochemically processed water that has circulated around the bulge's perimeter rather than newly discharged water from the estuary.

In an example of the type of controlled dynamical analysis one can conduct with a model, Zhang et al. (2009a) separately withdrew individual forcing processes to examine the effect of each on the circulation. Their results are shown in Fig. 19.8, which should be compared to Fig. 19.6a, b for the full physics solution.

Fig. 19.8 Mean SSH (sea surface height) contours (left) and surface currents and magnitude (right) over the 2-year period 2005-2006 for three simulations with changes to the full physics configuration shown in Fig. 19.6. Top row: Outer shelf boundary forcing removed. Middle row: Wind stress removed. Bottom row: Bathymetry of Hudson Shelf Valley filled in

Fig. 19.8 Mean SSH (sea surface height) contours (left) and surface currents and magnitude (right) over the 2-year period 2005-2006 for three simulations with changes to the full physics configuration shown in Fig. 19.6. Top row: Outer shelf boundary forcing removed. Middle row: Wind stress removed. Bottom row: Bathymetry of Hudson Shelf Valley filled in

Without the remotely forced along-shelf mean flow the bulge recirculation remains, but the across-shelf surface flow is more eastward being the result solely of Ekman transport and not combined with geostrophic southward flow.

In the absence of wind forcing the bulge is more intense, in accordance with the results of Fong and Geyer (2002) who found that along-shore transport driven by wind arrests continuous growth of bulge recirculation. As in the full physics case, part of this recirculation feeds flow on the south side of the HSV, but without winds the downstream flow is largely at mid-shelf parallel to the coast and does not disperse to the outer shelf.

Zhang et al. (2009a) explored whether the Hudson Shelf Valley impacts circulation by simply removing the valley from the model bathymetry. Figure 19.8 shows that in the No Valley case the SSH signature of the bulge is substantially weakened, and surface velocity shows far more of the estuary outflow enters the NJ coastal current.

In an extension of their passive tracer approach for following Hudson River waters, Zhang et al. (2010a) employ the concept of 'mean age' (Deleersnijder et al. 2001) to determine the transit time from river source to shelf ocean. If we denote the equation governing the transport of a passive tracer with concentration C by

dt then an 'age concentration' tracer a can be introduced satisfying

— + V ■ (ua) = V ■ (K ■ Va) + C dt where the last term on the right causes a to increase in proportion to the concentration of river source water present. The concentration of the tracers in the river source are C = 1 and a = 0. The 'mean age' (Deleersnijder et al. 2001) is given by a(x, t) = a(x, t)/C(x, t) and describes the duration it has been on average since the waters at a given position and time (x, t) entered the domain at the river source. Figure 19.9 illustrates how mean age evolves in a simulation where the river tracer release commenced on 13 March. It takes some 4-5 days for river water to reach the bulge circulation, and water on the southwest side of the bulge is clearly older than water to the north. On March 18 an increase in river discharge a few days previously introduces a surge of younger water that forms a sharp gradient in mean age across the western edge of bulge. In 7 days none of the river water has escaped the bulge. In regions the passive tracer has not reached the mean age is undefined.

Zhang et al. (2010a) show that mean age patterns in the 2005 LaTTE period mimic an age proxy determined from a ratio of satellite observed water leaving radiance that expresses the relative concentration of CDOM (Coloured Dissolved Organic Matter) to phytoplankton. CDOM is the dominant optical constituent in river source waters and has high absorption at 490 nm but it subsequently photo-degrades whereas phytoplankton concentration (with chlorophyll-a spectral peak at 670 nm) increases as the plume ages, so the CDOM decrease and phytoplankton iar ojxyr'. v^-vvl

1 7 Mar J^^^Vl

J55H

J55H

18-Mar Kkar-^p^ ^

Was this article helpful?

0 0
Guide to Alternative Fuels

Guide to Alternative Fuels

Your Alternative Fuel Solution for Saving Money, Reducing Oil Dependency, and Helping the Planet. Ethanol is an alternative to gasoline. The use of ethanol has been demonstrated to reduce greenhouse emissions slightly as compared to gasoline. Through this ebook, you are going to learn what you will need to know why choosing an alternative fuel may benefit you and your future.

Get My Free Ebook


Post a comment