Hydrodynamics

D’Entrecasteaux Channel Hydrodynamics

The D’Entrecasteaux Channel lies between the southern Tasmanian mainland andBruny Island in southern Australia (Figure 1). The Huon Estuary joins the D’entrecasteaux Channel near the southern limit of the channel and is a significant source of fresh water (over 1000 m³s^-1 during a flood). Increasing salmonid aquaculture activity in the region prompted the need for system characterization through modelling studies. The estuary/channel is highly stratified at times, and also characterized by complex geography, making modelling of the region challenging. Long period simulations were required (4 years) to assess the inter-annual impact of aquaculture on the aquatic environment, which required the parallel-processing capability of SHOC to be invoked. The model was forced with river flow from various sources (the largest being the head of the Huon Estuary) wind stress and surface elevations, temperature and salinity on the northern and southern limits of the channel.

Huon Estuary and D'Entrecasteaux Channel geography. — Figure 1: Huon Estuary and D’Entrecasteaux Channel geography.

A 3-tiered nesting strategy was used to propagate basin-scale circulation into the local domain. The largest scale regional model was nested within an eddy resolving, data assimilating global model (Oke et al., 2005). The nesting configuration is illustrated in Figure 2. The regional grid comprised a polar curvilinear grid, whereas the two higher resolution grids were constructed in generalized orthogonal curvilinear coordinates. The resolution of the local grid grid ranged from a minimum of 150 m in the Huon Estaury to a maximum of 700 m near the southern boundary, with 26 layers being used in the vertical. There exist 13000 total surface cells in this grid, only 1800 (13%) of which are wet; i.e. the majority of this grid is associated with dry land (as a consequence of the complex curvilinear grid employed). These grids are particularly well suited to the sparse coordinate system, which has the ability to entirely remove land from computations, and representation, in the model. Using a sparse representation has been shown to produce significant computational efficiencies on these grids with low ratios of wet to land cells (Herzfeld, 2006).

The global model provided low frequency sea level changes and temperature and salinity as boundary and initial conditions to the regional model. The amplitudes and phases of 14 tidal constituents were introduced at every open-boundary node in this model via the global tidal model of Eanes and Bettadpur (1995), using the methodology of Cartwright and Ray (1990). Surface momentum fluxes were prescribed from meteorological measurements obtained at various sites throughout the region and interpolated onto the grid in space and time; the wind has an annual average speed of speed of 7.0 ms^-1 from the south west. Heat-fluxes were computed from standard meteorological measurements, and applied as a surface boundary condition for vertical mixing of temperature (e.g. Figure 3). Short wave radiation had the option of being depth-distributed. Sea level, temperature and salinity from the regional model were saved on the boundary of the intermediate model, and subsequently used as boundary forcing for this grid. A similar procedure was used to obtain boundary conditions for the local model. This level of nesting is necessary, since downscaling boundary forcing becomes problematic above grid ratios of ~5:1, where boundary over-specification issues may produce stability and accuracy issues. Additionally, the effects of larger scale motions (boundary currents, coastally trapped waves) are allowed to propagate into the local domain. The full year of 2002 was simulated and calibrated to data collected within the Broad Scale Monitoring Program.

Several physical processes proved important in obtaining an acceptable calibration. Surface heat fluxes play a crucial role in regulating temperature in the region. The model proved sensitive to the type of bulk formulation used for surface sensible and latent heat fluxes, and to a lesser extent the depth to which short wave radiation is allowed to penetrate. Differential heating is apparent in the side bays, both in measured data and in the model, and this may contribute towards heating of the main channel.

Non-linear effects were important near the northern boundary of the domain. motivating the construction of the intermediate scale model which better resolved velocity in this region, hence was suitable for nesting the local model using boundary velocity forcing. The local model also proved sensitive to the background vertical diffusion coefficient, type of mixing scheme used and magnitude of the imposed Huon River flow.

Data collected within the Broad Scale Monitoring Program at sites illustrated in Figure 4 revealed that a temperature gradient (up to 1^oC) exists along the D’Entrecasteaux Channel during summer and autumn, with the northern end associated with higher temperature. The deeper waters at the southern end have the lowest temperature in the channel, presumably due to the sub-thermocline oceanic influence. Towards autumn the vertical temperature gradient at the southern end is less pronounced, as surface cooling decreases surface temperature heading into winter. In winter bottom waters become warmer than surface waters, but still several degrees cooler than the summer bottom temperature. This bottom temperature increase in winter is also observed at the northern end of the channel. Salinity is lower in the mid-channel region and attains the highest values in bottom waters at the ends of the channel throughout the year, thus density compensating the temperature distribution. Thin fresh water layers can be observed mid-channel during times of high HuonRiver flow.

Meaurement sites for the Broad Scale Monitoring Program. — Figure 4: Measurement sites for the Broad Scale Monitoring Program.

The model results confirm these trends and validate that the Huon Estuary behaves as a salt wedge estuary with marine flow in bottom waters directed upstream in the estuary and a fresh water surface flow heading downstream. The head of the salt wedge is located near Huonville under low flow and is pushed downstream under high flow conditions. The downstream surface flow generally favours the northern bank of the river, heading northwards up-channel upon entering the D’Entrecasteaux. Under high flow conditions fresher water may be found as far north as North West Bay, and may be advected north as much as 24km in just over 2 days.

On diurnal timescales the tidal flow dominates the region, with flow directed up-river and up-channel during the flood tide, and vice versa during the ebb. Strongest currents exist in the narrowest point in the channel near Gordon, where they approach 0.5 ms^-1. The tide undergoes a neap-spring cycle of the order of 14 days, with maximum tidal ranges approaching 1m. The tide is predominantly of diurnal (daily) mixed character with a form factor F ~ 1.5. Maximum tidal excursions are of the order of 4km mid-channel. In the southern channel the excursion decreases and in the northern channel and Huon Estuary the excursions are less than 1 km.

The mean seasonal flow for the D’Entrecasteaux–Huon Estuary system consists of bottom water entering the region at the southern end of the channel and moving up into the Huon Estuary in the salt wedge, favouring the southern bank. Entrainment occurs from the salt wedge into the surface downstream freshwater flow, the majority of which then turns north upon entering the channel and exits into Storm Bay at the northern end of the channel. A smaller proportion of Huon flow exits the channel through the southern boundary. A schematic of the residual flow is provided in Figure 5.

The calculation of flushing times can be subjective depending on the method used to compute the flushing. Using an e-folding rate based on depletion of total mass in a region the flushing times varied from around 3 days for the lower Huon Estuary under high flow conditions to ~20 days for the whole domain in winter. A flushing estimate for the whole domain based on the average time for neutrally buoyant particles to exit the domain was computed as ~26 days.

Distributions of passive tracers resulting from release in the top 14 m of the water column at locations corresponding to selected farm sites showed significant variability with release location. Generally those sites in the northern channel result in distributions confined to the northern D’Entrecasteaux (e.g. Figure 6). Release sites in the channel below Gordon and in the lower Huon Estuary resulted in relatively uniform concentrations throughout the domain outside a well defined mixing zone of high concentration (e.g. Figure 7). For release sites further up the Huon the largest concentrations are confined to the upper Huon and uniform concentrations of lower magnitude are found throughout the rest of the domain. These general distributions were also observed in results obtained via particle tracking of neutrally buoyant particles released from the respective farm sites. The southern channel and Huon Estuary can be characterized as well connected to the whole domain, whereas the northern channel has relatively poor connectivity with the southern channel.

Figure 6: Median distribution of tracer for the annual simulation with the release site located in North West Bay.

Figure 7: Median distribution of tracer for the annual simulation with the release site located in Port Esperence.

Particle tracking results also confirmed the diurnal dominance of tidal forcing, with particles exhibiting up-channel and up-river movement on the flood tide, and down-channel / river on the ebb. During flood events the favoured trajectory out of the Huon was up-channel (Figure 8). The freshwater plume also favoured the northern bank of the Huon due to the influence of Coriolis forces (Figure 9). The location of the freshwater plume was, however, sensitive to wind direction, with north-easterly winds pushing the freshwater plume southwards.

Figure 8: Particle trajectories under the influemce of a flood 15 August 2002. Flow ~1000 m3s-1, tidal range 0.88 m with south-westerly winds ~5-10 ms-1.

Figure 9: Surface salinity after the flood event 15 August 2002. Lower salinity water can be found as far north as North West Bay.

A 4-year simulation using the larger scale model was used to assess the transport of nitrate into D’Entrecasteaux Channel. A seasonal cycle of nitrate flux into southern D’Entrecasteaux was observed, with maximum flux of ~100 T/month in spring, and minimum fluxes in late summer and early autumn. The increased fluxes during spring are due to the instruction of high nitrate water from depth onto the slope and shelf, which is then available for transport into the coastal zone via local processes. The up-slope intrusion is hypothesized to be due to the interaction of large scale currents with the slope, resulting in upwelling in the bottom boundary layer (Figure 10, 11). Local uplift of high nitrate water on the slope and shelf is possible during March and April, confined to the southern Tasmanian coast and southern tip of Bruny Island. This also increases the flux of nitrate into southern D’Entrecasteaux. This uplift is hypothesized to be the result of onshore flow in the bottom boundary layer over the shelf and slope due to the passage of the EAC over the slope south of Tasmania. The EAC exhibits considerable variability in its passage down the west coast of Tasmania, and the upwelling events of this nature probably only coincide with strong EAC intrusions into southern Tasmanian waters during February to April. It is only during these times that there is the possibility of the EAC interacting with the shelf south ofTasmania; other times of the year the shelf is dominated by eastward transport of water in the Zeehan Current.

Figure 10: Nitrate and currents at 100 m depth for 10 Oct 2003 showing increased initrate concentration over the slope south of the Tasman Peninsula.

Figure 11: Cross section of nitrate along the transect identified in Fig. 10 showing upwelling of nitrate over the slope.

High resolution models were developed to investigate the mixing zone characteristics around a fish farm cage. These models were capable of resolving the farm cage at scales of metres, including the floating pontoon surrounding the cage. These analyses revealed that a continuous release of tracer from the cage results in a plume emanating from the cage in the form of a long narrow ribbon. The position and concentration of the plume is dependent on the in situ flow conditions, and possesses large variability in space and time. It is possible for a plume having concentration of 10% (i.e. 10:1 dilution) of the source concentration to exist up to 0.5 km from the cage on occasion. Instantaneous releases of tracer from the cage site result in the tracer’s transport with the flow away from the cage in the form of a coherent pool, rather than simple diffusion around the cage site. This pool slowly diffuses horizontally as it is advected. The rate of advection of this pool is dependent on the flow conditions at the time, hence subject to considerable variability. Sub-surface currents often determine the trajectory of this pool, with the surface signature of tracer being the result of vertical mixing of tracer from depth. In these cases the surface signature of the pool may move in the opposite direction to the surface currents.

Particle tracking was used to investigate the ‘footprint’, or spatial distribution of particles in the sediment, for various settling velocities of particles. For settling rates typical of feed pellets or faecal material, e.g. of the order of 10 cms^-1, the footprint is restricted to the cage site. However, settling rate an order of magnitude less than this result in footprints in the form of an elongated plume with typical length ~0.5 km and width ~200 m emanating from the cage site. The direction of this plume is controlled by sub-surface currents, and may be in the opposite direction to surface currents.

This study shows that the one-way nested modelling approach can be successfully applied to advance the understanding of the physical dynamics of a coastal region. Central to the success of this methodology is the availability of good field measurements to calibrate and validate the model, and use as forcing data in the absence of suitable global products. The curvilinear grid capability is also desirable so that the narrow regions of the domain (e.g. river sections) may be represented in reduced dimension so as not to compromise stability constrains and model run-times. This is even more important when coupling to sediment transport or biogeochemical modules, which can further decrease run times by several orders of magnitude. Using complex curvilinear grids often results in a small percentage of cells in the domain being wet (10-20%), and models that effectively cope with this situation can also lead to run-time efficiency (Herzfeld, 2006). Finally, calibration of the model more often involves application of differing numerical schemes or process representation than changing the values of parameters, hence a comprehensive suite of these methods and processes within the model is an advantage.

The results of this study have clear implications for the salmonid aquaculture industry, where the release of nutrients from farms located in the southern channel are likely to have system-wide effects, whereas the effects of those located in the northern channel are likely to be local. However, this study only considers the combined Huon Estuary/D’Entrecasteaux Channel system, and potential system-wide impacts may need to be considered in the far-field context including the Derwent Estuary and Storm Bay.

References.

Cartwright, D.E., Ray, R.D. (1990) Oceanic tides from Geosat altimetry. J. Geophys. Res., 95 C3, 3069-3090.

Eanes, R., Bettadpur, S. (1995) The CSR 3.0 global ocean tide model. Center for Space Research, Technical Memorandum, CST-TM-95-06.

Herzfeld, M. (2006) An alternative coordinate system for solving finite difference ocean models. Ocean Modelling, 14, 174 – 196.

Oke, P.R., Schiller, A., Griffin, D.A., Brassington, G.B. (2005) Ensemble data assimilation for an eddy-resolving ocean model of the Australian region. Q.J.R. Meteorol. Soc., 131, 3301-3311.