The Mean Summertime Circulation along Australia s Southern Shelves: A Numerical Study

Transcription

1 2270 JOURNA OF PHYSICA OCEANOGRAPHY VOUME 33 The Mean Summertime Circulation along Australia s Southern Shelves: A Numerical Study JOHN F. MIDDETON School of Mathematics, University of New South Wales, Sydney, New South Wales, Australia GUENNADI PATOV Institute of Computational Mathematics and Mathematical Geophysics, Novosibirsk, Russia (Manuscript received 22 February 2002, in final form 29 April 2003) ABSTRACT For the first time, a high-resolution regional model is developed for the slope and shelf circulation within the Great Australian Bight and for the Gulfs region of South Australia. The results indicate the extent, nature, and dynamical interaction of a variety of circulation features that are most likely to be important for the region. In particular, the positive wind stress curl south of Australia leads to an equatorward Sverdrup transport in the deep ocean, westward Flinders Current along the slope, and upwelling of the (600 m) deep permanent thermocline. The wind stress curl also leads to a seaward topographic Sverdrup transport within the wide sloping shelf of the bight and results in an anticyclonic gyre that is intensified off the Eyre Peninsula and reduced in magnitude by the joint effect of baroclinicity and topographic relief. In the western half of the bight, the seaward surface Ekman and topographic transports are shown to converge with the onshore component of the Flinders Current leading to a ridge in sea level, eastward current over the shelf break, and downwelling to 100 m or so. The shelfbreak circulation is similar farther east but is driven by the anticyclonic gyre within the bight and a trough in sea level along the shelf slope: the latter results from the geostrophic adjustment to density anomalies that arise from wintertime downwelling and the Flinders Current. imited hydrographic, satellite, and current meter data support the existence of an eastward shelfbreak current. Off the Gulfs and Robe regions, the wind-forced coastal currents are to the northwest and both the model and observations indicate that upwelling occurs to depths of up to 150 m. 1. Introduction The aim of this work is to provide a first-order description of the mean summertime shelf and slope circulation and underlying dynamics of Australia s southern shelves between Esperance in the west and Robe in the east (Fig. 1). Few observational or modeling studies of this 2000-km region have been made. To this end, a high-resolution numerical model of the region will be nested inside the results of a global circulation model and forced with summertime-averaged winds. Results will be compared with both current and hydrographic observations for the region. This study is augmented by the companion wintertime study of Cirano and Middleton (2004, hereinafter CAM). Winds for the South Australian region are generally directed along the coast and can lead to significant upwelling events off Kangaroo Island, the west coast of the Eyre Peninsula, and Robe (Schahinger 1987; Corresponding author address: John F. Middleton, School of Mathematics, UNSW, Sydney 2052, NSW, Australia. john.middleton@unsw.edu.au Griffin et al. 1997). Indeed, the region here is one of the few along the Australian coast where a surface signature of upwelling is regularly found. The upwelling-favorable component of wind stress can be large with monthly extremes of Pa between January and March. However, as noted by Griffin et al. (1997), upwelling off the west coast of the Eyre Peninsula can occur even when the local winds are not upwelling favorable. The mean summertime wind stress is also upwelling favorable although the magnitudes are of order 0.05 Pa or about ¼ of the episodic upwelling-favorable stresses noted above. The upwelling-favorable winds also act to lower coastal sea level and, through geostrophy, should drive a westward nearshore current. Current-meter observations are relatively few, although those obtained off the Gulfs region and Robe (summarized below) do indicate a northwestward mean flow close to the coast but a southeastward flow over the shelf break. The mean currents are generally weak (10 cm s 1 ) and coexist with a strong coastal-trapped wave (CTW) signal (amplitude 25 cm s 1 ) over the shelf that is driven by the 10-day weather band forcing (e.g., Schahinger 1987; Evans and Middleton 1998) American Meteorological Society

2 NOVEMBER 2003 MIDDETON AND PATOV 2271 Farther offshore, the currents again reverse direction flowing from east to west: the Flinders Current (Bye 1972, 1983). A study of this current (Middleton and Cirano 2002) has shown that it arises from the equatorward Sverdrup transport within the Southern Ocean. Upon reaching the continental slope of Australia s southern shelves, it is deflected to the west so that anticyclonic vorticity can be dissipated within a necessarily upwelled bottom boundary layer. In analogy with western boundary currents, it may be categorized as a northern boundary current and has a maximum core within the permanent upwelled thermocline (600-m depth) and adjacent to the continental slope. The deep upwelling associated with the Flinders Current is found in the data of evitus and Boyer (1994) and the CSIRO Atlas of Regional Seas (CARS; Ridgeway et al. 2002). The latter data will be used below to infer its existence assuming a level of no motion at a depth of 2000 dbar. Middleton and Cirano (2002) summarize further observational evidence. In the far west, the eeuwin Current is largely absent during summer (Church et al. 1989) and has little impact on the circulation or hydrography within the Great Australian Bight. (Herzfeld 1997). Herzfeld and Tomczak (1999) subsequently studied the shelf circulation within the bight using a numerical model with idealized topography, stratification, and wind field. Nonetheless, they were able to show that an anticyclonic gyre within the bight should exist during summer. The dynamics of this gyre result from the southward topographic Sverdrup transport that arises from the positive wind stress curl for the region and wide sloping shelf within the bight. This transport is then necessarily returned along the west coast of the Eyre Peninsula so as to dissipate vorticity acquired within the bight. These authors also suggested that upwelling off the Eyre Peninsula might be driven by the onshore Ekman transport within the bottom boundary layer, rather than by interior transports that arise from wind-forced upwelling. We will examine their conclusions below. The summertime circulation within Spencer Gulf is not well known although evaporation exceeds precipitation year-round, leading to the formation of anomalously salty (36.2 psu) water within the gulf. Nunes- Vaz et al. (1990) suggest that summertime heating leads to a density minimum at the gulf mouth and that little exchange occurs with the shelf. This is fortunate, because the modeling of such water mass formation is beyond the scope of this study. As noted above, the CTW variability of the region is large (25 cm s 1 ) and upwelling favorable wind stress events of 0.2 Pa are not uncommon during summer. The currents driven by the summertime mean wind stress of 0.05 Pa or less will turn out to be small (5 cms 1 ) when compared with the CTW activity. Indeed, an examination of altimeter data for the region shows that the rms variability is sufficiently large to obscure the long-term mean sea level signal so that such data will not be helpful in evaluating the model results below. However, the mean results obtained below are of interest since the current field obtained can advect density over several hundred kilometers over the seasonal (3 months) period of simulation. In addition, the seasonal upwelling will form a backdrop to any episodic upwelling and the results will also allow us to evaluate the interaction of the Flinders Current with the slope and shelf currents. In section 2, the details of the numerical model and thermohaline and forcing fields are adopted. In section 3 model results for the shelf slope circulation within the bight are discussed in the context of geostrophic adjustment to the initial density field and the convergence of the deep-ocean equatorward Sverdrup transport with currents over the shelf. The dynamics of an anticyclonic gyre within the bight are examined. In section 4, the circulation between the Eyre Peninsula and Robe is discussed in the context of upwelling. In section 5, a comparison of the model results is made with current-meter observations and with results from a global model. A summary and discussion of results are in section Numerical methodologies and forcing fields The numerical model, grid, topography, and surface forcing fields are first outlined. The global Ocean Circulation and Climate Advanced Modelling Project (OC- CAM) model is then briefly described along with the closure of the nested-model open boundary conditions and initial temperature and salinity fields. a. The nested model, grid, topography, and surface forcing The nested numerical model to be used is the Princeton Ocean Model (POM) with Smagorinsky horizontal diffusion and the Mellor Yamada turbulence closure scheme (Blumberg and Mellor 1987). Bottom stress is parameterized using a squared drag law with a drag coefficient of C D The model domain is illustrated in Fig. 1, and a total of 133 and 65 grid points were used in the along-shelf (y) and offshore (x) directions, respectively. The crossshelf grid spacing was determined so as to provide the highest resolution over the steep continental slope, (dx 3.5 km) and for the Gulfs region where upwelling is expected (dy 8 km). The resolution here is smaller than the 130-km internal deformation radius of the main thermocline. To help resolve both the surface mixed and bottom boundary layers, 39 sigma levels were adopted with finest resolution near the surface and bottom. The topography H(x, y) adopted was a combination of that obtained from the OCCAM model and that provided by the Australian Geological Survey Organization (AGSO). The minimum and maximum depths were set to 40 and 5000 m, and the resultant model topography is shown in Fig. 1. An external time step of 4 s is well

3 2272 JOURNA OF PHYSICA OCEANOGRAPHY VOUME 33 FIG. 1. The model topography and geography of Australia s southern shelves. Isobath depths are in meters. The location of the Great Australian Bight is indicated as GAB. The coastline indicated is the actual coastline. within that inferred from the Courant Friedrich ewy criterion (8.7 s). The internal time step was 120 s. The wind stress field ( W ) to be used is the 2.5 climatological average for February (Fig. 2) provided by Trenberth et al. (1989) and is very similar to the Seifridt and Barnier (1993) average used to force the OCCAM model. The Trenberth average is adopted being obtained over a longer period. Surface fluxes of heat and freshwater were obtained from the climatology of Josey et al. (1996) and da Silva et al. (1994), respectively, and again for the month of February. Both fluxes are subject to error since few direct observations exist for the region (Hertzfeld 1997), although all of the flux estimates indicate that the surface layer of the ocean should become more stratified. It will turn out that the results for the shelf slope circulation change little with application of these fluxes. FIG. 2. The Feb average of wind stress obtained from Trenberth et al. (1989). The vector indicated has a magnitude of 0.1 Pa, and the contours of wind stress with magnitudes of and 0.05 Pa are indicated. b. The global model, open boundary conditions, and initial fields OCCAM is a ¼ global ocean model with 36 layers in the vertical (Webb et al. 1998). The model was initialized using evitus annual mean climatology. The summer mean transports, temperature (T), and salinity fields (S) to be used here were obtained by averaging over the OCCAM results from the last four summers (Jan Mar) of a 12-yr prognostic experiment. The last 4 years of output are used as the total model energy here oscillates in a smooth, periodic manner (Webb et al. 1998). OCCAM was not forced with meteorological fluxes of heat and freshwater. Rather, temperature and salinity in the top level were relaxed to the monthly evitus values over a scale of 25 days. A comparison of the OCCAM density fields was made with the ½ data from CARS. At depths below 300 m or so, both the CARS and OCCAM results are in good agreement. (The OC-

4 NOVEMBER 2003 MIDDETON AND PATOV 2273 FIG. 3. Results obtained for density (contour interval 0.1 kg m 3 ) at day 87.5 and at the 130E meridional transect. The results below 100 m are very similar to the OCCAM results used to initialize the regional model. FIG. 4. The volume-averaged kinetic energy for the entire domain (solid curves) and for depths less than 1500 m and near the coast (dashed curves). The curves labeled 1 and 2 denote results for the central case and case of enhanced friction, respectively. CAM density field, used to initialize the model, is well illustrated by that shown in Fig. 3.) However, the OC- CAM density field (Fig. 3) is about 0.3 kg m 3 denser than CARS at depths between 300 and 100 m, while nearer the surface (50 m), the CARS data indicate a stronger seasonal pycnocline than is found in the OC- CAM results. The discrepancies here are likely to result from the warm, summertime and cold, wintertime water formation in the shallow coastal regions that is not represented in the model. However, two important features of the CARS data are found in the OCCAM fields. The first is that water in the top 400 m is lighter nearer the shelf than farther offshore: a result of the deep winter mixing and downwelling. The second feature is the deep upwelling over the slope and at depths of m that is associated with the Flinders Current. As we will see, both features of the density field will effect the nearslope circulation. Attempts were made to use the CARS rather than OCCAM thermohaline fields, but the model was found to become numerically unstable: no explanation for this could be discovered. The OCCAM fields will therefore be used, and as an attempt to at least simulate the stronger pycnocline at 50 m, the meteorological fluxes noted above were applied to the model between days 60 and 72.5 during the model simulation. The results are shown below (Fig. 9b). While the surface waters are of similar density to CARS, vertical mixing by the mean wind field is too weak and the 50-m pycnocline is not much larger than found in OCCAM. However, the circulation obtained without the surface fluxes is found to be quantitatively similar to those obtained with the fluxes (see below). Results below will therefore be presented as those obtained using the OCCAM fields and with the meteorological fluxes applied between days 60 and 72.5 (the central case). To avoid spurious inertial oscillations, the pressure gradients of the OCCAM fields (and wind stress) were ramped in from zero over 3 days. In order to force the nested model along each of the open boundaries, the transports from the OCCAM model were first interpolated onto the POM grid and the values of the depth-averaged velocities obtained by division by the local depth of the POM model. The boundary condition adopted for the depth-averaged velocity follows Flather (1988). Further details of these conditions, pressure gradient errors, and two grid point instability smoothing are given in CAM and Platov and Middleton (2001). c. Model spinup Global ocean models typically take several years to spin up to a quasi-steady state from an initial state of rest. For the nested model here, the spinup of the deep ocean is much more rapid since the initial density field is adopted from the OCCAM model where spinup has largely been achieved. For the near coastal and shelf regions, the gross features of the sea level and circulation are also spun up over 20 days or so by the rapid passage of low-order coastal trapped waves and Kelvin waves. Results for the volume-averaged kinetic energy were obtained (Fig. 4) for the central case detailed below and for a case in which the effect of tidal and CTW variability on bottom friction was considered (enhanced friction). For both cases, the total kinetic energy plateaus after day 77.5 (Fig. 4). The energy of the shelf and slope circulation (depths 1500 m and near the shelf) was also determined and shows a slight rise and then fall in energy after the application of the surface fluxes of heat and freshwater (days ). While not completely steady, the circulation results presented below at days are qualitatively very similar. In general, results will be presented at day 87.5 and are most representative of the circulation and density fields that result from surface heating. Most important, the conclu-

5 2274 JOURNA OF PHYSICA OCEANOGRAPHY VOUME 33 FIG. 5. Results obtained at day 87.5 from the model for (a) sea level (cm). The bold contour indicates the 200-m isobath. The location of the Albany high (AH) and Adelaide trough are indicated. (b) The depth-averaged velocity U (a vector of length 2 cm s 1 is indicated). The curved tails of the velocity vectors are obtained by local interpolation of the instantaneous velocity. (c) Results obtained for the depthaveraged velocity U at day 47.5 (a vector of length 4 cm s 1 is indicated). The curved tails of the velocity vectors are obtained by local interpolation. The boldface contour indicates the 200-m isobath. (d) As in (c) but for day sions drawn below apply to all results obtained between days 47.5 (before heating) and days (after heating). 3. The large-scale circulation and circulation within the bight In the following, an overview is given of the largescale circulation features that are identified to include the Flinders Current, a cyclonic trough located over the shelf slope and an anticyclonic gyre within the bight. The nature and cause of these features are then detailed along with the unexpected eastward current and associated downwelling over the shelf break. a. An overview Solutions for sea level and the depth-averaged circulation at day 87.5 are presented in Figs. 5a and 5b. The boundary transports and anticyclonic wind stress lead to the high in sea level that extends from the western boundary to around 130E. For convenience we will refer to this feature as the Albany high. Associated with this high is a generally equatorward transport throughout the domain and the westward Flinders Current near the shelf slope (Fig. 5b). As noted, Middleton and Cirano (2002) have shown that the equatorward transport is principally determined by Sverdrup dynamics and that the bottom boundary layer of the current over the slope must be upwelling favorable. The latter leads to the upward tilting of isopycnals that is found both in the model results at 130E (Fig. 3) and in the CARS data. The magnitude of the Flinders Current increases from about 5 cm s 1 at 140E to16cms 1 at 120E. Near the coast, the generally offshore Ekman transport acts to lower sea level, leading to a westward coastal current of up to 10 cm s 1. Typical values of coastal sea level are around 26 cm while values of 50 cm are found for this region in the winter study by CAM. The seasonal change of 25 cm is broadly consistent with the 20-cm seasonal variation estimated from coastal sea level observations for the region (Pariwono et al. 1986). Between the coast and the Flinders Current, a notable trough in sea level is apparent over the shelf slope and extends from Robe to the midbight (Fig. 5a). This feature will be referred to as the Adelaide trough and results in a cyclonic circulation off the Eyre Peninsula and an east to southeast flow along the shelf break (the 200-m

6 NOVEMBER 2003 MIDDETON AND PATOV 2275 FIG. 6. A summertime average (Jan Mar) of dynamic height (cm) relative to the 2000-m isobath (source CARS). Regions of high (H) and low () sea level are indicated and the arrows indicate the direction of the surface geostrophic flow. isobath). Details of the flow are presented at day 47.5 (Fig. 5c) and day 97.5 (Fig. 5d). As shown, the circulation changes little between these times, although the southeast and northwest currents south of the western tip of Kangaroo Island are more intense at day Results (not presented) also show that the vertical structure of the current field generally changes little after day An anticyclonic gyre is also found within the shallow waters of the bight and the east to southeast arm of this gyre links up with the shoreward arm of the Adelaide trough to drive water along the shelf break and from Esperance in the west to Robe in the east. This eastward current flows in the opposite direction to the Flinders Current and the coastal currents that are driven by the upwelling-favorable winds. The cause and nature of this current and associated downwelling will be shown below to result from the interaction of the gyre, trough, and Flinders current. Support for the existence of the above features of circulation is given by the plot (Fig. 6) of summertime dynamic height relative to the 2000-dbar level, a product of the CARS dataset. A general north and then westward flow (nearer the shelf) is indicated while in the west, a large-scale ridge provides evidence for the Albany high. In the east the low found off western Tasmania is possibly a reflection of that found in the model (Fig. 5a). Immediately south of the bight, a low is found in dynamic height to provide evidence for the existence of the Adelaide trough. The dynamic height contours presented for depths less than 2000 dbar were determined by horizontal extrapolation of seaward data into the shelf (CARS), and, while unrealistic, an anticyclonic circulation is indicated for both the bight and the Coorong (Fig. 5b). The pattern of alternating highs and lows shown in FIG. 7. Results obtained at day 4 for the case of quasigeostrophic adjustment to the initial density field. The wind stress and boundary transports are set to zero. (a) Sea level (interval 2 cm). The (unlabeled) bold contour indicates the 200-m isobath. (b) The depth-averaged velocity (a vector of length 2 cm s 1 is indicated). dynamic height was originally noted by Bye (1972) and suggested to be a form of frictional coastal wave driven by variations in the alongshore component of the wind stress, linear friction, and the beta effect (Bye 1983). Here, the explanation for these high and lows will be shown to involve the surface Ekman transport and circulation induced by density field anomalies related to wintertime processes and the Flinders Current. b. The Adelaide trough In order to determine the origin of the trough, solutions were obtained with the wind stress and boundary forcing transports set to zero so that the model can only geostrophically adjust to the initial stratification. (The nonlinear and bottom stress terms were retained but the associated effects on the circulation were small.) As before, the pressure gradients are ramped in from zero over 3 days and results for sea level and depth-averaged velocity at day 4 are presented in Fig. 7. The results in

7 2276 JOURNA OF PHYSICA OCEANOGRAPHY VOUME 33 Fig. 7a correspond to the adjustment of sea level to water that is either relatively light or dense. ighter (denser) water sits higher (lower) in the water column. Thus, the colder and denser water found to the east and south (the Victorian Tasmanian region) leads to a low in sea level while the warmer water in the northwest leads to the high. Water near the coast and within the top 400 m is also generally lighter than that farther offshore so that sea level is higher over the shelf. The associated currents (Fig. 7b) over the shelf are to the east (10 cm s 1 ) and oppose the westward, wind-driven currents. The density anomaly that leads to these results is shown in Fig. 3a and results from deep wintertime mixing and downwelling. Analytical results for the geostrophic adjustment of such a light water anomaly (appendix A) show that it will lead to a rise in sea level of 3 cm that decays exponentially away from the shelf break over a scale of 200 km. Near the surface, the geostrophic currents are directed to the east and of order 3cms 1, while below the anomaly, the currents are directed to the west and will reinforce the Flinders Current. However, the exponential decay of sea level due to the light water anomaly fails to explain the trough evident in Fig. 7a and its absence off Tasmania. Moreover, the width of the trough would seem to be set by the width of the shelf slope and is narrowest off Kangaroo Island where the slope is narrowest. The explanation for these features is that the sea level is also affected by the deep upwelled dense water anomaly (between 500 and 800 m) that is shown in Fig. 3 and that results from the Flinders Current. This dense water anomaly is weakest off Tasmania so that sea level there simply decays away from the shelf break. Off Kangaroo Island, the upwelled dense water reduces sea level over the slope region and this perturbation leads to the trough shown in Figs. 5a and 7a. Further confirmation that the trough results from the density field is given by barotropic results obtained with density gradients set to zero. All other forcing terms were included. In this case, no trough is found, and the flow is generally directed along isobaths and to the west. c. The anticyclonic gyre One feature not found in the adjustment solutions is the anticyclonic gyre within the bight. As noted, Hertzfeld and Tomczak (1999) have suggested that such a gyre will result from the wind stress curl and the sloping shelf topography. These effects combine to produce a southward topographic Sverdrup transport within the bight that is returned as an eastern boundary current along the west coast of the Eyre Peninsula. Their conclusions were however based upon results from an idealized numerical model and are examined next by estimating the sources of seaward transport from the vorticity equation [ ] f ( W B ) 1 H H H J, k J, other, TF/H WSC BSC JEBAR OTHER where 0 x 0 H H udz and dz, y W and B denote the surface and bottom stress, and g 0 H z dz. The term TF/H on the left of (1) denotes the production of vorticity V x U y of the depth-averaged flow U (U, V), which accompanies flow across f/h contours (Mertz and Wright 1992). The first two terms on the right of (1) denote the production (or loss) of vorticity through Ekman pumping by the wind stress curl (WSC) and spindown by bottom stress curl (BSC). The third term, JEBAR, denotes the joint effect of baroclinicity and topographic relief (see below). The final term, OTH- ER, includes t, the nonlinear terms and horizontal diffusion. These will be neglected as they were found to be generally an order of magnitude smaller than the other terms. Now, the wind stress curl WSC is positive for the bight region and can lead to the seaward transport of fluid columns to regions where the depth H is larger and f/h smaller (less cyclonic). This is simply illustrated by the following model for the shelf in the midbight. We assume that H H(y) and here take y to be directed to the north and x to the east. For illustrative purposes, we first take the transport across the f/h contour term (TF/H), to balance that due to the wind stress curl ( T )V WSC 0, (2) (1) where HV is the Sverdrup transport and T fh y /H denotes the topographic vorticity gradient. For the region bounded by the 70- and 100-m isobaths (Fig. 8a), T is of order 10 9 (ms) 1, positive and much larger that the planetary gradient (ms) 1. Thus, V is negative, and the transport is directed into deeper water where f/h is smaller. Now consider the JEBAR term. As Mertz and Wright (1992) point out, it represents a correction term to the topographic stretching term (TF/H) that uses the depthaveraged velocity U rather than the bottom velocity u b. A two-layer model is developed in appendix B and shows that the JEBAR term is only affected by horizontal density gradients. For the case here, lighter water is found to the west and JEBAR is shown to lead to a shoreward transport (opposite to the topographic Sverdrup transport). The topographic Sverdrup transport

8 NOVEMBER 2003 MIDDETON AND PATOV 2277 FIG. 8. (a) The region for the topographic Sverdrup, JEBAR, and vorticity calculations is shaded gray. The relevant northern and southern f /H contours (dark lines) are defined by the H 70 m and H 100 m isobaths (light lines). (b) The sea level results obtained at day 87.5 (cm). The boldface contour indicates the 200-m isobath, and the location of the j 26 midbight cross-shelf transect is shown. The location of the ridge in sea level is indicated. should therefore exceed that estimated directly from the model. In order to evaluate the contribution of each term in (1) to the total transport across f/h contours, we adopt the method outlined by Thompson et al. (1986). The transort term in (1) is integrated over the gray region shown in Fig. 8a that is bounded by the f/h contours to the north ( N ) and south ( S ) and the Stokes theorem is invoked to yield R k ( ) da N N E W C S S dr d] d], (3) where the subscripts E and W denote path integration at the eastern and western ends of the domain: the anticlockwise integral along the f/h contours vanishes. Now provided is approximately constant at the eastern and western ends, the right side of (3) can be rewritten as an average over distance to yield the transport across f/h contours TABE 1. Onshore transport within the bight. The transport estimates for the bottom stress curl (BSC), JEBAR, wind stress curl (WSC), and their sum e in the vorticity equation (4) (Sv). The total transport as estimated directly from the model is also given (). Positive transports indicate flow towards shallow water ( north). The two cases shown correspond to the central case discussed here and the case in which enhanced bottom friction is adopted. Case BSC JEBAR WSC e Central Enhanced friction N N 1 E W N S S S e d] d]. The transport for each term in (1) may then be estimated from W B 1 ( ) [ ] e dr JEBAR da H N S C R (4) and compared with that obtained directly from the model () and across the 85-m isobath. The results at day 87.5 are presented in Table 1 under the heading central case. The bottom friction curl of the anticyclonic gyre is negative (clockwise) and so from (4) leads to a positive onshore transport: Similarly, the JEBAR term is also positive as expected and also larger than that due to friction. The wind stress curl contribution is negative as indicated above and larger in magnitude (0.33 Sv, where Sv 10 6 m 3 s 1 ) than the total (0.19 Sv). The sum of these terms does not equal that computed directly from the model (0.17 Sv) because of the approximate nature of (4) above and the neglected terms. The results do support the conclusion of Hertzfeld and Tomczak (1999) that topographic Sverdrup transport for the region is important for the formation of the anticyclonic high within the bight. Moreover, the results here also show that JEBAR, and to a lesser extent bottom friction, are important and act to oppose this transport. d. Downwelling and the eastward shelfbreak current From Figs. 5c and 5d, an eastward current is centered over the shelf break (200-m isobath) and flows across the entire bight. In the western bight, this eastward current lies on the seaward side of a ridge of high pressure (sea level) that is shown in Fig. 8b (see the 31-cm contour). As we show below, the ridge results from a convergence over the shelf break of the seaward topographic Sverdrup and Ekman transports with the equatorward deep ocean Sverdrup transport. This convergence leads to an elevation of sea level and downwelling over the shelf break. The ridge of sea level so formed (Fig. 8b)

9 2278 JOURNA OF PHYSICA OCEANOGRAPHY VOUME 33 FIG. 9. Density results obtained at the j 26 (midbight) cross-shelf transect (a) at day 7.5 and (b) at day 87.5: contour interval 0.05 kg m 3. Results obtained at day 87.5 and at the j 26 (midbight) cross-shelf transect for (c) cross-slope velocity (u, 1000w): a vector of magnitude 1 cm s 1 is indicated. (d) The alongslope velocity. The contour interval is 2 cm s 1 and positive currents to the east are shaded gray. combines with the anticyclonic high within the bight and the inshore arm of the Adelaide trough to drive an east to southeast current that flows from Esperance to Robe (Fig. 5b). Downwelling over the shelf break is FIG. 10. The seasonally averaged temperature field obtained from the CARS data at a 130E meridional transect and for (a) winter and (b) summer. The shelf topography is that adopted in the OCCAM model. also found in the eastern half of the bight because of the convergence of the cross-isobath component of this current (Figs. 5c,d) with the onshore Sverdrup transport. The shelfbreak convergence and associated downwelling is illustrated in the model results by the evolution of the density field between days 7.5 and 87.5 and at the midbight transect j 26 (Figs. 9a,b). Here, both the 26.1 and contours are downwelled near the shelf break by 50 and 100 m between days 7.5 and Upwelling is found at shallower regions farther inshore because of the offshore Ekman transport, while at depths of 300 m and away from the shelf, the isopycnals are moved onshore by the equatorward Sverdrup transport (see the 26.3 isopycnal). The directions of isopycnal displacements are also broadly consistent with the cross-shelf velocities shown on Fig. 9c. Far from the shelf, an onshore flow is indicated and an offshore flow over the shelf and near the surface is found. Downward motion is found over the shelf break. The eastward current that results has a maximum amplitude of6cms 1 and extends to depths of 250 m (Fig. 9d). Evidence for the downwelling over the shelf break and along the entire bight region is given by the CARS data. As an example, we consider the temperature fields for both a winter and summer average and for a midbight transect at 130E (Fig. 10). During winter, the isotherms

10 NOVEMBER 2003 MIDDETON AND PATOV 2279 FIG. 11. Sea surface temperature results for the South Australian region at day 87.5 (contour interval 0.4C). Water with temperatures greater than 17.4C is not contoured. The location of the Mount Hope transect off the Eyre Peninsula is indicated. Coorong, Robe, and Kangaroo Island (KI) are indicated. near the shelf break are downwelled by 200 m or so by the eastward winds and the intense cooling leads to a surface mixed layer with depths m (Fig. 10a). The water temperature over the shelf break is 15C. During the transition to summer, the surface heating leads to a seasonal thermocline at depth of m (Fig. 10b). This thermocline water has temperatures greater than 15C (absent during winter) and is upwelled near the coast, but notably downwelled above the shelf break. The downwelling cannot be a remnant of wintertime processes since water temperatures greater than 15C are not found during that season. A possible alternative to summertime downwelling is that the downward sloping isotherms result from enhanced vertical diffusion within the bottom boundary layer of the shelf break and slope. However, the downwelling of the 16C isotherm shown in Fig. 10 occurs over ½ of latitude or 55 km seaward of the shelf break. This would appear to be a large distance for boundaryinduced mixing to act. 4. Upwelling around the Gulfs and Robe region As noted, the Gulfs and Robe region is one of the few locations along the Australian coastline where a surface signature of upwelling is regularly found. In the following we examine the upwelling produced by the model and inquire into the importance of local winds, the role of the bottom boundary layer and advection in the upwelling process. a. The Coorong and Robe The sea surface temperature plot (Fig. 11) shows the upwelling to be largest off the Coorong, the Gulfs, Kangaroo Island, and the Eyre Peninsula. For these regions, the surface Ekman transport is directed offshore and calculations show that this transport is approximately equal to the onshore interior transport within 20 km or so of the coast. That is, the upwelling in the near-shore resembles two-dimensional wind forced upwelling. It should also be noted that the transport within the bottom boundary layer for these and all other regions was found to be negligible in comparison with the interior transport. The reason for this is simply that the interior currents are generally weak (10 cm s 1 ). For the Coorong region, the near-surface currents (depth 5 m) shown in Fig. 12 are directed to the west and northwest and lead to the westward spread of cool ( C) water shown in Fig. 11. The current field at 35 m (Fig. 13) is similar to that near the bottom and, as indicated, most upwelling occurs off Robe and to depths of more than 100 m. b. Kangaroo Island A detail of the bottom velocity and temperature for the region from Kangaroo Island to the Eyre Peninsula is shown in Fig. 14. The deepest upwelling for the region occurs south of Kangaroo Island where 15.6C water is upwelled from depths of 150 m (the thick black con-

11 2280 JOURNA OF PHYSICA OCEANOGRAPHY VOUME 33 FIG. 12. The model horizontal velocity field on day 87.5 at a depth of5m(avector of magnitude 10 cm s 1 is indicated). The locations of the mooring sites S, G, E, F, and A are labeled along with the mean currents estimated from current-meter data (Table 2). The origin of the arrow indicates the location of the mooring site. The boldface contour indicates the 200-m isobath. tour). The currents here are up to 10 cm s 1 and drive a plume of 15.6C water toward the northwest and along the north coast of Kangaroo Island. The advection here results from the north, then eastward, current that arises from the upwelling-induced drop in sea level along the west and northern coast of Kangaroo Island (Fig. 8b). Cold water is also driven to the northwest with a magnitude of about 5 cm s 1 or 43 km over 10 days. This speed is more than enough to account for the evident km northwestward displacement of the 15.8C bottom isotherm apparent in Fig. 14. Some observational evidence for the upwelling described above is given by satellite images of sea surface temperature (e.g., Griffin et al. 1997) as well as by hydrographic transects made for the Kangaroo Island region (Hahn 1986) and Robe region (Schahinger 1987). In agreement with the model results, the latter indicate upwelling to be confined to depths of 150 m or less and within 15 km of the shelf break. c. The Eyre Peninsula Temperature results were also found for the Mount Hope transect shown in Fig. 11. These results (Fig. 15) show that, between days 7.5 and 87.5, 16.4C water is upwelled from depths of 50 m and then transported offshore in a 30-m-deep surface Ekman layer. On the shelf a plume of colder (16C) water is found at day 87.5 and from the plot of bottom velocities and temperature (Fig. 14), the source of this water lies to the south and southeast. Over the shelf break, the 16C isotherm is downwelled by 80 m between days 7.5 and 87.5: the downwelling here results from the cross-isobath excursion of the southeast shelfbreak current noted above. FIG. 13. The model horizontal velocity field on day 87.5 at a depth of 35 m (a vector of magnitude 10 cm s 1 is indicated). The locations of the mooring sites S, G, E, F, A, and B are labeled along with the mean currents estimated from current-meter data (Table 2). The origin of the arrow indicates the location of the mooring site. Note that for sites A and B both the 1983 and 1984 summer vector mean currents are shown. The boldface contour indicates the 0.2-km (200 m) isobath. While bottom boundary layer transport was also found to be insignificant for the Mount Hope transect, it may become important for episodic upwelling when the alongshore component of wind stress can reach extreme values of 0.35 Pa (Griffin et al. 1997) or when the wind stress curl is large. In the case of the latter, the anticyclonic gyre within the bight can be intensified, possibly leading to large-amplitude currents off the west coast of the Eyre Peninsula and significant bottom boundary layer upwelling. Indeed, observations show that anomalously cool water is occasionally observed off the west coast of the Eyre Peninsula even when the wind direction is not upwelling favorable. This scenario was examined by Hertzfeld and Tomczak (1999) using an idealized numerical model. They found that upwelling was dominated by that within the bottom boundary layer although a much stronger wind stress was adopted (up to 0.35 Pa). The larger stress led to much stronger coastal currents (up to 56 cm s 1 ) and bottom Ekman transport since the latter is proportional to the square of the current speed. [The wind stress adopted here is smaller (0.05 Pa) and the currents are smaller (5cm s 1 ).] 5. A comparison with current observations and the OCCAM results In the following the circulation results are compared with current-meter observations obtained off the Gulfs and Robe regions and provide support for the existence of both the nearshore northwestward coastal current and southeastward shelfbreak current. The regional model results above have also been obtained by initializing

12 NOVEMBER 2003 MIDDETON AND PATOV 2281 FIG. 14. The horizontal component of bottom velocity at day 87.5 (a vector of magnitude 10 cm s 1 is indicated). The bottom temperature at day 87.5 is indicated by the shaded contours (interval 0.2C). Temperature contours less than 15.2C are not shown. The solid contour denotes the location of the 150-m isobath. The Eyre Peninsula and Kangaroo Island (KI) are indicated. The boldface contour indicates the 150-m isobath. with the OCCAM density field and by using the OC- CAM transports along the open boundaries. The question of interest then is just how the regional model solutions compare with OCCAM and what advances have been made. a. A comparison with current-meter observations Current-meter observations for the summer period and the Gulf and Robe regions were obtained at the sites shown in Fig. 12. The sites of the model results were obtained from available charts and by ensuring that the water and model depths were nearly equal. The latitudes and longitudes of the mooring sites are at best a guide to the model location because of the smoothing of topography necessary to minimize two-grid-point instabilities. Where appropriate, the data were low-pass filtered, and the magnitude of the vector mean current and direction (anticlockwise from east) are presented in the last column of Table 2 along with the model results. The vector means for currents meters in less than 22 m of water are also presented in Fig. 12 (the 5-m-depth model results), while data obtained at depths greater than 22 m are presented in Fig. 13 (the 35-m-depth model results). The currents at 35 m are indicative of those at greater depths. Site S lies nearest the gulf (Fig. 12), and near the surface quite good agreement is obtained between the model and data, with currents directed to the west with a magnitude of 7 cm s 1 or so (see also Table 2). At the lower 35-m level, both the model and observations indicate a northwest current of 2 4 cm s 1 (Fig. 13). Near the shelf break, the near-surface model currents are directed to the northwest and along the 100-m isobath (Fig. 12) with amplitude of 5 10 cm s 1. The upper mooring observations at E are broadly compatible with this although at G the observations indicate a much stronger current (27 cm s 1 ). At site F, a weaker southward current (3 cm s 1 ) is broadly consistent with the model results in Table 2 and Fig. 12. At the lower mooring depths (100 m) the model circulation is similar to that at 35 m (Fig. 13). At site G both the observed and the modeled currents are approximately to the northwest with amplitudes of 4 cm s 1 or so. At site E the model current is to the northwest and, while the direction of the observed vector mean is directed onshore, the amplitude is similar to the model results (8 cm s 1 ). At the deeper mooring, site F, (and with the exception of the 1981 results) the directions of the observations below 40 m are in good agreement with the model results, indicating the existence of the southeast shelfbreak current (Table 2, Fig. 13). The mooring data also indicate some interannual variability. The means for 1981 at site F indicate a northwestward current at depths of

13 2282 JOURNA OF PHYSICA OCEANOGRAPHY VOUME 33 FIG. 15. Results obtained along the Mount Hope cross-shelf transect for temperature at (a) day 7.5 and (b) day 87.5 (contour interval 0.4C). 46 and 115 m. Small variations in the position of the Adelaide trough could easily lead to changes in current direction at this site. The model results off the Robe region (Fig. 13) indicate considerable horizontal current shear with north to northwestward currents near the coast (site B) and southeastward currents farther offshore (near site A). Only the 1983 mean 18-m depth is presented in Fig. 12, while both 1983 and 1984 means are presented in Fig. 13. The 1984 observations differ somewhat from those of 1983, and such differences could arise from changes in mean winds and in the location of the current system shown in Fig. 13. The observations do indicate currents that are consistent with the nearby model results. b. A comparison with the OCCAM results To compare results, the depth-averaged summertime circulation from OCCAM was interpolated onto the regional model grid and is presented in Fig. 16. A northwest to westward Flinders Current is found in the OC- CAM results and is similar to that found in the regional model (Figs. 5c,d). Cross-shelf plots confirm this. Such similarity results from the comparable wind stress curl of the OCCAM winds (Seifridt and Barnier 1993) and those used to drive the regional model (Trenberth et al. 1989). The equatorward Sverdrup transport at 39S between 115 and 140E is 16 Sv for OCCAM and 19 Sv for the regional model. (As noted, the latter mean winds were determined using a longer dataset and should provide a better climatology.) Indeed, the spatial pattern of the two wind stress fields is very similar although the curl within the bight and the alongshore component of that used in OCCAM is about 20% smaller than those used here. The westward coastal circulation in OCCAM might therefore be slightly weaker than found here. However, the results in Figs. 16 and 5c,d show that the OCCAM coastal circulation and anticyclonic gyre are much weaker than in the regional model. The reason for this is related to the crude ETOPO5 topography adopted in OCCAM along with the coarse ¼ grid resolution. In the regional model, the 30 arc s AGSO topographic data have been interpolated onto the reasonably fine cross-shelf grid and the gentle slope of the shelf within the bight is reproduced (Fig. 9). The OCCAM topography for the midbight is illustrated in Fig. 10, and the slope of the shelf is confined to within about 55 km of the coast while the shelf farther offshore is flat. The implications for the circulation and anticyclonic gyre are twofold. First, the topographic Sverdrup transport is only generated within 55 km of the coast and leads to the weaker gyre shown in Fig. 16. Second, the topographic Sverdrup transport and gyre do not extend as far seaward as in the case of the regional model, and the resultant convergence with the onshore deep water Sverdrup transport is weaker. This leads to an eastward shelfbreak current at 130E that is much weaker (2 cms 1 ) and less well defined than found in the regional model (8 cms 1 ). Moreover, isopycnals (and isotherms) over the slope in the regional model are downwelled by an additional 100 m or so as compared with the OCCAM results: the latter are very similar to those presented in Figs. 9a and 15a since the OCCAM fields were used to initialize the regional model. A final deficiency of the OCCAM topography and circulation relates to the depth of the first oceanic cell, which varies markedly in the alongshore direction. At 130, 135.5, and 140E, the first cell has a depth of 60, 25, and 150 m, respectively. Such artificial variations in shelf depth will lead to vortex stretching and act to distort the direction and amplitudes of near-steady alongshore currents that would otherwise tend to follow isobaths. In the model here, the first sea cell is generally set equal to 40 m and the nearshore circulation between Robe and the bight is much stronger and well defined than in OCCAM. Other deficiencies of the OCCAM topography are discussed by CAM. 6. Summary and discussion A high-resolution numerical model has been used to study the important physics and circulation of Austra-

14 NOVEMBER 2003 MIDDETON AND PATOV 2283 TABE 2. Summertime current-meter observations. The following table presents summertime statistics obtained from available data and published data. Sites are shown in Fig. 12; U and represent the upper and lower current meters, respectively. The first record and days columns indicate the start time (day/month/year) and number of days used to compute the statistics. The latitude and longitude are in degrees south and east, respectively. In the final two columns, the magnitude of the vector mean U (cm s 1 ) is given along with its direction (degrees anticlockwise from east). These statistics may be compared with the model results that are presented immediately below. Site First record Days at on Water depth (m) Instrument depth (m) U S 89 S 89 S 88 E 89 G 89 E 89 G 89 F 83 F 82 F 82 F 82 F 82 F 83 F 82 F 81 F 81 B B A 83 A 83 A 84 U U U U U1 U2 Mid 1 2 U U 24 Jan Jan Feb Dec Dec Dec Dec Dec Nov Nov Nov Nov Dec Nov Dec Dec Feb Jan Feb Feb Jan Gulfs region Robe lia s southern shelves as forced by mean summertime winds. The positive curl of the mean winds results in an equatorward deep ocean Sverdrup transport, westward Flinders Current, and upwelling of the permanent thermocline (depth 600 m). The amplitude of the Flinders Current increases from 5 (140E) to 16 cm s 1 (120E) and has a maximum within the permanent thermocline and extends from the surface to 2000 m or so. The current also forms the nearshore arm of a large anticyclonic gyre south of Esperance (the Albany high) that is set up by the convergence of the Ekman transport of the anticyclonic winds that characterize the summer season. A map of dynamic height relative to 2000 dbar provides support for the existence of the large gyre and Flinders Current (see also Middleton and Cirano 2002). The wind stress curl also leads to an offshore topographic Sverdrup transport within the wide sloping shelves of the bight that returns along the west coast of the Eyre Peninsula, leading to an anticyclonic gyre within the bight (see also Herzfeld and Tomczak 1999). The joint effect of baroclinicity and topographic relief (JE- BAR) was investigated and shown to reduce the topo-