Transactions on Ecology and the Environment vol 2, © 1993 WIT Press, www.witpress.com, ISSN 1743-3541 The dispersion of treated effluent by tide and wind in a stratified bay A.M. Oakes, N.V.M. Odd, A.J. Copper, W. Roberts HR Wallingford, Howbery Park, Wallingford, Oxfordshire ABSTRACT The paper describes how 3D modelling techniques were used to simulate the interaction of tide and wind on circulations in St Andrews Bay in Scotland, which has a weak but well defined halocline at mid depth caused by the inflow from the River Tay. INTRODUCTION Fife Regional Council (FRC) intend to embark on a program of capital works to provide a new sewage treatment plant south of the town of St Andrews on the east coast of Scotland, Figure 1. In order to demonstrate that the proposed sewage treatment and disposal options meet Environmental Quality Objectives and Standards, FRC commissioned the preparation of marine hydrodynamic and bacterial dispersion models of St Andrews Bay. In vertically well mixed coastal waters, the standard method of predicting the physical dispersal of treated effluents by tide and wind from long sea outfalls and their impact on adjacent beaches, has been the use of two-dimensional depth averaged flow models and quasi3D random walk pollutant dispersion models (Mead'). In the case of semi-enclosed bays with weak or negligible tidal currents the interaction of wind stress and the Coriolis force generates complex 3D residual circulations governed by the geometry of the bay, with downwelling and upwelling near the shoreline, which can not be simulated by the above techniques. The situation is further complicated by the existence of a thermocline or halocline which can trap effluent below the surface and reduce the vertical exchange of momentum and solutes across the pycnocline. In these cases it is important to use 3D modelling techniques. St Andrews Bay outfall study is an example. Transactions on Ecology and the Environment vol 2, © 1993 WIT Press, www.witpress.com, ISSN 1743-3541 184 Water Pollution FLOW REGIME IN ST ANDREWS BAY The tidal regime is dominated by the effects of the main flood and ebb tides, which run from Arbroath to Fife Ness, the Firth of Tay, which discharges at the northern end of the Bay and wind stress. /I 2D far field mode 3D model area Figure 1 Location plan The Bay is relatively deep (-10m) and as a result the tidal currents are generally low, despite the 5m spring tidal range. In particular, the south western corner of the Bay has peak speeds rarely greater than 0.2 m/s. The wind, which is predominantly from a south westerly direction, interacts with the tide to produce complex 3D circulation. The currents are further complicated by a mid depth pycnocline, caused by the slightly brackish outflow of the Firth of Tay, with a salinity change of around 0.6 ppt over a 3m depth of water. FLOW MODELLING The use of a fine gridded, fully 3D flow model is limited to relatively small areas, due to the computing power required. It is often advantageous, therefore, to use a 2D far-field model to obtain accurate boundary conditions for a more local 3D model. 2D far-field model For the St Andrews study a uniform 200m gridded, 2D tidal flow model was set up covering the whole of the Bay, including the River Tay. Boundary conditions were extracted from a 2D model of the North Sea. Transactions on Ecology and the Environment vol 2, © 1993 WIT Press, www.witpress.com, ISSN 1743-3541 Water Pollution 185 Calibration The calibration of the 2D model was achieved by comparison with Admiralty diamond velocity data for mean spring tide conditions. Further data was available in the form of DRCM observations close to St Andrews, although the data was very variable both in terms of speed and direction and no comparable wind records were available. However, both model and observations showed an almost continuous anticlockwise circulation throughout the tidal cycle in the vicinity of the proposed outfall 3D model The 200m gridded, three-dimensional model consists of 10 discrete layers and uses mixing length theories (Odd and Rodger^) to define the vertical exchange of turbulent momentum and solutes. In St Andrews Bay, the layers were defined as equal divisions of the water column at each model cell, giving good vertical resolution throughout the model area and enabling the simple definition of a pycnocline at mid depth. The pycnocline was defined as a 0.6 ppt salinity change across 0.3 - 0.7 of the depth. This salinity gradient has the effect of severely damping the vertical turbulent exchange of momentum between the upper and lower layers of the water column. This "delinking" is particularly important when prevailing winds are applied to the surface, thereby generating completely different current patterns in the surface and bed layers. Boundary conditions for the 3D model with various applied wind stresses were obtained by first applying the wind conditions to the 2D model. By extracting the tidal levels from the 2D model throughout the tidal cycle a consistent set of boundary conditions for each wind condition was generated. Calibration The calibration of the 3D flow model was based on mean spring tidal level and velocity profile data at the proposed outfall site. Wind data was available for this period and was applied as a constant wind stress to the flow model. An important factor was the choice of the single vertical density (salinity) profile that was representative of conditions in the bay. In this case, the model was not used to simulate an evolving density field. It was also necessary to decide whether to base the damping function on the local or peak gradient Richardson Number. In the latter case, the heavy damping effect of the salinity gradient on vertical turbulent exchange in the pycnocline region also controls the level of turbulence in the whole of the upper and lower layers. This often happens in estuaries if the pycnocline is in the lower 0.7 of the depth. However, in St Andrews Bay, model calibration tests showed that the Transactions on Ecology and the Environment vol 2, © 1993 WIT Press, www.witpress.com, ISSN 1743-3541 186 Water Pollution pycnocline appears merely to form a zone of locally reduced turbulence separating the surface and bed layers. A detailed simulation of the observed velocity data was hindered by a great deal of variability in the observations, caused in part by the difficulty in obtaining accurate readings of very low current velocities. Local variability of the wind speed and direction also has an effect on the surface current velocities, which the 3D flow model is unable to reproduce, being only able to simulate the general flow pattern in the Bay from an averaged wind. Figure 2 shows comparisons between the observed and modelled velocities at surface, mid depth and bed. Mid depth -20 2 4 6 8 1 0 - 2 0 2 4 6 8 10 -2 Hours after High Water (Dundee) Hours after High Water (Dundee) 0 2 46 8 10 Hours after High Water (Dundee) Model Speed Model Direction Observed Speed Observed Direct,o Figure 2 Comparison of observed and modelled velocities Overall, reasonable agreement was achieved between model and observations, with the lower layers showing better agreement than the surface layers, which were affected by the local wind drift. Table 1 gives the calibration statistics for the observed and modelled velocities. Table 1 Depth Surface 0.3D 0.5D 0.7D Bed Spring tide velocity calibration statistics Correlation coefficient 0.5 0.6 0.7 0.8 0.7 Model current speed Clockwise deviation as a % of observed of model current current speed from observed 88 73 61 76 79 -3 -14 -13 -37 -4 RCM near bed velocity data was also available over a period of a month at two sites, together with daily wind information. A low pass filter was applied to the RCM data to investigate the magnitude and direction of the residual currents. Comparison with residual, bed layer Transactions on Ecology and the Environment vol 2, © 1993 WIT Press, www.witpress.com, ISSN 1743-3541 Water Pollution 187 currents from the 3D model with appropriate wind stress indicated a good degree of agreement both in speeds and directions. A good representation of the 3D residual circulation in the Bay is essential in order to be able to predict the dispersal of discharged effluent. The 3D model simulated a complex pattern of residual currents with upwelling and downwelling resulting from the interaction of the anticlockwise tidal residual circulation, the surface wind stress and coriolis forces, which has an over-riding influence on the pattern of dispersal of effluents in the Bay, as shown in Figure 3. This can not be simulated in a 2D model. Figure 3 Residual current pattern in bed and surface layers RANDOM WALK POLLUTANT DISPERSION MODEL Random walk dispersion models simulate the movement of pollutant plumes discharged from sea outfalls or storm water overflow, using a random walk representation of turbulent dispersion in both the horizontal and vertical planes. The 3D random walk model used for the St Andrews study was driven by the 3D tidal flow model, including 3D residual currents driven by winds crossing the whole Bay and used the same representation of the layers and the density gradient. The local Richardson number was used to evaluate the solute mixing length, assuming a critical flux Richardson Number of 0.15 (Odd and Rodger^). A finer grid resolution of 50m was used. Calibration In order to calibrate the random walk model a dye release in the surface layer was simulated and compared with field observations. Figure 4 shows a comparison between the modelled and observed surface dye plume, 10 hours after the start of the dye Transactions on Ecology and the Environment vol 2, © 1993 WIT Press, www.witpress.com, ISSN 1743-3541 188 Water Pollution release (HW+13.5hrs). The prevailing wind during the period of the dye release was simulated by applying a constant wind stress to the 3D flow model used to drive the random walk model. It was also decided to apply the actual time varying superposed wind drift to the surface layer of the random walk model, in order to simulate the local, short term surface drift fluctuations which the 3D flow model is unable to reproduce. These short term fluctuations tended to increase the local rate of dispersion of dye in the surface layer. The wind drift factor, which expresses the surface, wind-driven current as a percentage of wind speed, was considered a calibration parameter and a value of 1.5% was found to give the best results. This was additional to the wind driven currents simulated by the 3D flow model. Figure 4 Comparison of observed and predicted surface dye plume An objective statistical comparison between each observation and the model results within a 100m search radius of that observation is given in Table 2. It should be noted that the observations were nonsimultaneous and were batched into hourly groups for ease of comparison with the model results. A search radius of 100m was not considered unreasonable since the 3D flow model, which drives the random walk model, had a grid size of 200m. Table 2 Spring tide plume calibration statistics Hours after high water 9.5 11.5 12.5 13.5 14.5 mean % of observations within a factor of 2 of model results 73 48 95 92 83 71 Transactions on Ecology and the Environment vol 2, © 1993 WIT Press, www.witpress.com, ISSN 1743-3541 Water Pollution 189 Similarly accurate results for the dispersal of dye in the surface layer were obtained by driving a quasi-3D random walk model with the 2D depth averaged tidal flow model. In this case the flow model has no applied wind and the observed winds are directly applied to the random walk model assuming a logarithmic decrease in wind influence with depth and a return current near the bed. A higher wind drift factor of about 3% was found to give the best results, to compensate for the lack of a wind driven circulation in the flow model. THE DISPERSAL OF A TRAPPED PLUME The proposed new outfall in St Andrews Bay is to release treated effluent from a diffuser on the sea bed. The predicted impact of the effluent on the beaches of St Andrews would depend on whether the plume was trapped below the mid depth pycnocline, since the surface and bed residual circulations differed so greatly. In order to estimate the initial rise height of the effluent plume, the EPA plume model ULINE (Schuldt et aP) was used. The model utilises the diffuser design, pollutant output flow and density, together with the ambient cross-flow velocity and density structure to determine the trapping level. This trapping level can then be used as the effluent input level in the random walk model. In the case of St Andrews, the effluent will be trapped at the base of the pycnocline. The random walk model predicted that in this case, the effluent would mix downwards throughout the lower layers, but very little upwards dispersion would take place because of the damping of vertical turbulent exchange by the pycnocline. Without the pycnocline, the random walk model would mix the effluent throughout the water column, resulting in pollutant in the surface layers travelling in the opposite direction to the pollutant in the lower layers. The trapping of effluent below the pycnocline causes worst case contamination of beaches to occur with offshore winds, when the near bed current carries the pollutant onshore. In this case the use of a quasi-3D random walk model, driven by 2D depth averaged flow results, would give totally misleading predictions. This is mainly due to the wind drift profile in the quasi-3D random walk model, where an unrealistic return current in the opposite direction to the wind is assumed. In fact, the wind driven current in the bed layer may be in any direction, depending on the geometry of the bay. In St Andrews Bay, the tide averaged residual current speed at the proposed outfall site from the depth averaged, 2D model is 0.2 m/s in a southerly direction, whereas that for the bed layer of the 3D model is 0.02 m/s in a north westerly direction. The 2D flow model results would be seriously misleading in predicting the environmental effects of an effluent plume trapped in the lower layers. Transactions on Ecology and the Environment vol 2, © 1993 WIT Press, www.witpress.com, ISSN 1743-3541 190 Water Pollution CONCLUSIONS 1. In St Andrews Bay, the accuracy of the 3D flow model was improved by obtaining consistent boundary conditions from a calibrated far-field 2D model with appropriate wind stress. 2. Representation of the damping of vertical turbulent exchange of momentum across a pycnocline is essential in reproducing the 3D pattern of residual circulation in the Bay, 3. Wind effects play a significant part in the dispersal of effluent plumes due to their impact on residual circulation in the Bay. 4. Short term variations in the local wind cause a secondary drift and lateral dispersal of a surface plume over that caused by 3D far field wind driven current circulations. 5. Effluent plumes discharged near the sea bed may be trapped by a pycnocline resulting in very different patterns of movement to that of a surface plume. An accurate representation of the effect of the pycnocline in the random walk model is essential to give valid predictions of beach contamination. 6. 2D flow models and quasi-3D random walk models are inadequate in modelling effluent plumes trapped below the surface by a pycnocline. ACKNOWLEDGEMENTS The authors would like to thank Fife Regional Council and Environmental Management Ltd for giving permission to show results from a study of St Andrews Bay. REFERENCES 1. Mead, C.T. Random walk simulations of the dispersal of sewage effluent and dredged spoil, Proceedings of the 10th Australasian Conference on Coastal and Ocean Engineering, pp477-480, 1991. 2. Odd, N.V.M and Rodger, J.G. Vertical Mixing in Stratified Tidal Flows, Journal of the Hydraulics Division, ASCE, Vol 104, No HY3, Paper 13599, 1978. 3. Schuldt, M.D., Davis, L.R., Frick, W.E., Muellenhoff, W.P., Soldate, A.M. and Baumgartner, D.J. Initial Mixing Characteristics of Municipal Ocean Discharge - Volume 1: Procedures and applications, Environmental Protection Agency, EPA/SW/MT-86/012a, 1985.
© Copyright 2026 Paperzz