The path of the shelf edge flow off southwestern Australia is documented using results from satellite altimetry and sea surface temperature (SST) and a climatogical in situ analysis. During Austral winter a continuous current is shown to extend from its origin at North West Cape to the southern tip of Tasmania, a distance of 5500 km. Satellite SST observations and surface buoy tracks confirm the location and continuity of the current trajectory. While the Leeuwin Current is forced by the strong alongshore pressure gradient associated with the meridional portion of the western Australian coastal boundary, our results suggest that the essentially zonal shelf edge flow along the southern Australian coast arises from the setup of coastal sea level by onshore Ekman flow driven by the winter westerly wind and possibly a further alongshore pressure gradient. The timing of these two different forcing mechanisms means that the west coast pressure gradient delivers the Leeuwin Current to the south coast just as the winds reverse and are thus able to maintain the eastward passage of the current. The shelf edge flow consists of two main water masses. A low-salinity, warm water type of tropical origin associated with the Leeuwin Current and a high-salinity, warm water inflow formed on the western end of the Great Australian Bight continental shelf. A naming convention is proposed, the Leeuwin Current representing flow from North West Cape to the Great Australian Bight (GAB); the South Australian Current, between the eastern GAB and western Bass Strait; and the Zeehan Current off western Tasmania.
 The unique character of the boundary flow off the southwestern coastal region of Australia has long been recognized. This is mainly due to unusual aspects of the regional density structure combined with an extensive continental margin which acts as a wave guide for the alongshore propagation of shelf water. The presence of the Indonesian Throughflow creates a warm water pool near the equatorial eastern boundary, which in turn establishes a meridional oceanic heat flux toward the pole [Godfrey et al., 1995]. A byproduct of this process is that a strong alongshore pressure gradient is maintained off the western Australian coast. This gradient is far stronger than on other eastern boundaries and provides the driving mechanism for an anomalous poleward boundary current [Godfrey and Ridgway, 1985].
 Considerable early evidence for a southward flowing warm current off western Australia has been available from fauna distributions [Saville-Kent, 1897], temperature patterns [Halligan, 1921; Gentilli, 1972], ships drift observations [Schott, 1935] and water properties [Rochford, 1969]. However, it was only within the past twenty years when these early data were reinforced with more contemporary observations [Cresswell and Golding, 1980] that the current was given a name. The Leeuwin Current (LC) is now recognized as an anomalous eastern boundary current that flows poleward against the prevailing winds along the meridional segment of the western coast of Australia [Godfrey and Ridgway, 1985]. It transports warm, low-salinity tropical waters from northwestern Australia, southward along the western Australian coast and then eastward across the Great Australian Bight (GAB). The current is driven by the strong alongshore pressure gradient, which is sufficient to overwhelm the effects of coastal wind-forced upwelling [Thompson, 1984; Godfrey and Ridgway, 1985]. It has a strong seasonal cycle with greatest strength in late autumn early winter which is possibly associated with a weakening of the alongshore winds [Smith et al., 1991].
 As for the western coast, it has long been known that a southeastward current exists off the southern Australian coast. In an early atlas, Black  showed a broad eastward flow from the eastern Indian Ocean to Tasmania that was named the South Australian Current. Halligan  presented a similar oriented flow, originating from an early representation of the southward flow of the LC rather than directly from the Indian Ocean. In his compilation of ships drift data, Schott  provided both greater spatial resolution and also indicated that the boundary flow reverses in summer. The Australia Pilot described an easterly coastal current that reaches its maximum strength between May and July [British Admiralty, 1937]. The more contemporary study of Cresswell and Golding  showed that the LC turns eastward at Cape Leeuwin (115°E) and penetrates eastward to 120°E, albeit with temporally variable offshoots out to sea. Studies based on satellite imagery and the dispersal of marine fauna extended this eastward limit to the eastern side of the GAB [Legeckis and Cresswell, 1981; Maxwell and Cresswell, 1981]. On the GAB shelf, the flow structure was found to be more complex, with the identification of a separate source of high-salinity, warm water that spreads eastward [Rochford, 1986; Bye, 1986; Herzfeld, 1997].
 Meanwhile, at the eastern end of the domain, Baines et al.  presented evidence of the Zeehan Current, a southward flow off the west coast of Tasmania. Inevitably this prompted such questions as, does a continuous shelf edge current along the south coast exist and if so is the Zeehan Current simply a continuation of the LC? Baines et al.  discounted the latter idea on the grounds of water type differences, although no evidence was offered. Rochford  argued that the LC terminates within the GAB and that Bight waters supply high-temperature, high-salinity input to a new shelf edge surface flow moving southeastward. He suggested that this flow is easily confused in satellite imagery with the LC. In fact, Godfrey et al.  found only a very weak current at the shelf edge across the GAB and suggested that it may actually lie further inshore on the wide shallow shelf region.
 The missing link between the flow exiting the eastern GAB and the Zeehan Current has always been the poor understanding of the circulation along the southeastern segment of the southern shelf. Rochford  proposed that highly saline outflow from the Spencer Gulf penetrated eastward along the coastal boundary to provide the source waters of a high-salinity northern Bass Strait water type. Limited measurements of the mean shelf edge currents from direct current observations (between 40 and 100 days duration, deployed between early April to early August) made along the southern coast, east of the GAB [Provis and Lennon, 1981; Hahn, 1986; Schahinger, 1987] (see also Cirano and Middleton  and Middleton and Platov  for a summary) show a shelf break intensified southeast flow of order 12–30 cm s−1 in non El Nino years. Godfrey et al.  carried out a major survey along the south coast of Australia in winter, 1982. Using a combined data set of buoy tracks, ship's drift vectors and SST they observed that a narrow shelf edge current continued at least as far eastward as Portland (140°E), their cruise destination.
 Although many aspects of the southwestern flow have been resolved, such as the LC characteristics off western Australia, the overall picture is still far from complete. The boundary current structure along the southern coast, particularly in the eastern sector, remains poorly understood. Recent results from a high-resolution model simulation of the southern Australian coast by Cirano and Middleton  have challenged the limited observational record. Downwelling favorable, westerly winds produce a continuous eastward coastal current in the model extending from Cape Leeuwin to the eastern coast of Tasmania. We might expect that the detail contained within satellite SST imagery should be capable of resolving this observational uncertainty, but in fact, the very complexity of the SST structure has often hampered efforts to determine the underlying temporal or spatial patterns of the flow [Cresswell and Peterson, 1993; Griffin et al., 1997; Herzfeld, 1997].
 In this study we document the evolution of the shelf edge current system around the western and southern coastal boundaries of Australia using seasonally averaged SST, satellite altimeter and inferred water parcel trajectories. We seek answers to the following questions. With the latest generation of satellite data products, are we now able to determine the real nature of the boundary current system on the southwestern Australian coast? What are the forcing mechanisms driving the current system? How do the water mass characteristics vary along the current path? Is there a continuous current along the entire path and if so what is its relationship to the Leeuwin Current? Our overall aim is to treat the southwestern boundary as a coherent structure and hence to document the circulation over the whole system, hopefully replacing the somewhat piecemeal picture presently available.
 The paper is constructed as follows. The data and methods are described in section 2, and the underlying annual mean shelf edge circulation is described in section 3. Sections 4 and 5 present the evolution of the eastward flow of the shelf edge current system in winter, determined from a variety of data. In section 6, a contrast is drawn between the seasonal patterns of the alongshore forcing on the western and southern coasts. The water mass characteristics of the current system, as determined from temperature and salinity patterns are given in section 7, followed by a summary of biological evidence in section 8, and finally a summary discussion is given in section 9.
2. Study Domain, Data, and Methods
 Our focus is on the current system at the coastal margin stretching from northwestern Australia (114°E, 22°S; North West Cape) to the southeastern coast of Tasmania (Figure 1a). Note that this choice of domain excludes the North West Shelf, where shelf break currents have been observed in autumn which are likely to be intimately related to the LC flow [Holloway and Nye, 1985], and follows Godfrey and Ridgway  in reserving the term LC to flow beyond North West Cape. Off the western Australian coast the boundary is essentially meridional and stretches southward some 1900 km to Cape Leeuwin. Beyond Cape Leeuwin the southern coast is aligned in a predominately zonal orientation for more than 3000 km before reverting to a more meridional alignment for a further 600 km off western Tasmania. This is one of the very few extended zonal coastal boundaries found anywhere in the world and is possibly the longest. Over most of the domain the continental shelf width varies between 50 and 100 km, apart from the GAB (124°–135°E) where the shelf broadens to form a smooth plain (90–220 km wide) and a very gentle slope. Further important features that are likely to have a strong influence on the circulation are the gulf system east of the GAB (sometimes known as the South Australian Sea) [Hemer and Bye, 1999] and Bass Strait at the eastern end of the domain. Note that in this study we do not explicitly consider the flow within the confines of Bass Strait.
2.1. In Situ Climatology
 Vertical profiles of temperature and salinity were obtained from the World Ocean Database (WOD98) [Conkright et al., 1998], and CSIRO Marine Research archives for the region (110°–152°E; 20°–45°S). The temperature data set has been augmented by the expendable bathythermograph (XBT) casts deployed in the study region. All observed level casts were interpolated onto a set of 56 standard levels [Ridgway et al., 2002]. The spatial distribution of the combined temperature data set is presented in Figure 1b. Much of the data has been collected on the shelf and slope regions and the spatial sampling in deep water is adequate only in the region west of Australia. The data density deteriorates further when we examine the data distribution of deep casts and those collected over winter. Complete details of the data processing and quality control procedures are given by Ridgway et al. .
 Our interpolation tool is locally weighted least squares [Cleveland and Devlin, 1988]. The data are smoothed in space by projecting onto spatial quadratic functions, simultaneously fitted by annual and semiannual harmonic components and the influence of both variable bathymetry and coastal barriers is included. Fitting the spatial and temporal components in a single step minimizes the temporal bias in the mean. Uniform fields of T and S on a 0.5° grid covering the waters around Australia, form the CSIRO Atlas of Regional Seas (CARS). A detailed description of the analysis methods is presented by Ridgway et al.  and Dunn and Ridgway .
Figure 2 shows the mean pattern of surface steric height (relative to 2000 db) for the study domain determined from the CARS T and S fields. We have chosen to use this reference level as something of a compromise between a level which is deep enough to capture most of the baroclinic flow [Bye, 1972; Rintoul and Sokolov, 2001; Rintoul et al., 2002] and minimizing the regions where the water depth is less than the level. Since without ancillary alongshore current observations we cannot directly calculate the surface height in the shallower shelf and slope waters of particular interest here, we adopt the following procedure [from Godfrey and Ridgway, 1985]. In the deep water region the surface steric height field relative to 2000 db, h = 0/2000, is given by the standard expression;
where (x, y) represent the (cross-shelf, along-shelf (poleward)) directions, δ, the specific volume anomaly, is computed directly from the T and S fields in CARS, p is the pressure, g is the gravitational acceleration, and H is the water depth (Figure 2a). On the continental slope we exploit the fact that the flow of the LC in the upper portion of the water column is well represented by a reference level of 300 db [Thompson, 1984; Smith et al., 1991]. We combine the steric height referenced to 300 db with a deep component projected up the slope from the 2000-m isobath. On the slope this becomes,
where yA defines the alongshore 2000-m isobath (Figure 2a). No attempt is made to estimate the height pattern inshore from the 300-m isobath.
 We use altimeter data for the period October 1992 to December 2001 from both the TOPEX/Poseidon (T/P) mission and the ERS-1/2 satellites. Standard environmental corrections were applied to the T/P data [Benada, 1997] and ocean tides and tidal loading were obtained from the CSR 3.0 model and removed from the data. No orbit corrections were performed but the data were adjusted for the inverse barometer effect [Fu and Pihos, 1994]. When necessary, ERS-1/2 altimetric corrections were updated to make T/P and ERS-1/2 corrections consistent [Ducet et al., 2000].
 Monthly fields of sea level anomaly (SLA) were generated from annual and semiannual components derived from the time series of altimetric SLA. The components were determined at each point along individual satellite tracks. The track values were then interpolated onto a regular grid (0.25°) using a locally weighted least squares method with a minimum radius of 200 km, resulting in uniform fields of amplitude and phase of the two seasonal components.
 Interpolated fields of sea surface height (altimetry SLA plus 0/2000) have also been combined with modeled wind fields to generate estimates of currents and water parcel trajectories, using the methodology described by Griffin et al. . The wind-forced components of the currents were calculated over the region using surface Ekman layer dynamics [Pollard and Millard, 1970] and winds interpolated in space and time from the NCEP-NCAR 40-year reanalysis data set [Kalnay et al., 1996]. This formulation ignores the effects of bottom friction and is therefore considered most reliable in water depths greater than 50 m. The geostrophic component was neglected in waters shallower than 100 m, where locally wind-driven flows tend to dominate. Comparisons with currents estimated from satellite tracked ocean drifters indicated that this model could explain around half the total variance (r2 ∼ 0.5). The limited resolution of the wind and sea level fields also lead to systematic underestimation of velocities (slope ∼ 0.5), although peak velocities were well represented [Griffin et al., 2001].
 Water parcel trajectories were measured by individually tracking large numbers of particles seeded throughout the region of interest at a depth of 20 m. At 8-hour intervals they were moved individually according to the local interpolated current velocity. Particles moving onto land were held stationary in the water until the current patterns changed. The trajectories of all particles were recorded at 2-day intervals. The model was reseeded with particles every 3 months to counter losses through the outer edges of the domain.
 The particles followed complex paths, which were sensitive to their initial location. The results were therefore integrated into a statistical description providing the average distribution of particles from a specified source region after a specified dispersion time Td (S. A. Condie et al., Marine connectivity patterns around the Australian continent, submitted to Environmental Modelling and Software, 2004). Distributions were calculated from day 1 of the calendar quarter to day Td, then from day 2 to day Td + 1, until reaching the last day of the quarter. The probabilities were then averaged to give a probability distribution representative of that quarter on a 0.5° grid. All of the outputs shown below used a relatively long dispersion time of Td = 80 days, so as to allow mean transport patterns to emerge from the background eddy-driven dispersion.
2.3. Sea Surface Temperature
 SST satellite imagery was obtained from the advanced very high resolution radiometers (AVHRR) aboard the polar orbiting NOAA satellites using data received at Perth and Hobart. The images were first processed to give SST estimates, using the method of McMillin and Crosby  and remapped to a standard cartographic projection [Nilsson and Tildesley, 1986]. In order to obtain a consistent set of cloud-free images, sequences of individual images were then filtered and composited using a statistical median approach (C. Rathbone, personal communication, 2004). The method applies a simple median-like filter to clusters of pixels in space and time neighborhoods and then uses a sliding time window to generate a time series of uniformly spaced images. In this study we use a time series data set that covers the period 1989–2002, which has been generated using a 15-day compositing period.
3. Mean Circulation at the Coastal Boundary
 We now describe the circulation along the wave guide of the southwestern coastal boundary. The merged height field, which includes the estimates of the slope and shelf flow (equations (1) and (2)), is given in Figure 2a. We note that away from the shelf boundary the water depth is greater than 2000 m so that the map provides a reasonably accurate picture of the circulation. The errors in this region (Figure 2) are associated with sampling and interpolation uncertainties and with the reference level assumption. On the continental slope where the error is unknown, by inspection, our procedures appear to have captured the mean expression of the intense southward flow of the LC. Note that the majority of the current transport is located on or adjacent to the shelf edge.
 This flow is driven by the strong meridional gradient of steric height, which is represented by the southeastward oriented contours projecting seaward from the western Australian coastal boundary [Hamon, 1965; Godfrey and Ridgway, 1985]. The slope aligned flow penetrates eastward around Cape Leeuwin to 121°E and broadens onto the continental shelf [Cresswell and Golding, 1980; Cresswell and Peterson, 1993]. The along-shelf structure is maintained east to the GAB but it then breaks up into a series of smaller-scale eddy-like features. Along the southern coast mean field is far less prominent; its main features are a trough on the eastern side, west of Tasmania, and a ridge structure running parallel and some 200 km south of the western Australian coast. Off western Tasmania the alongshore structure is reintensified [Baines et al., 1983]. While there is a suggestion of the mean flow continuing eastward toward Tasmania, there is a clear need to resolve seasonal differences if we are to gain some understanding of the circulation in this region.
4. Seasonal Evolution of the Southeastward Flow
 The seasonal evolution of the boundary flow is illustrated by the sequence of monthly maps of the sea level anomaly (SLA) in Figure 3. There is very good agreement between these maps and equivalent monthly anomaly fields obtained from in situ data alone off western Australia. Off the southern shelf the in situ coverage is relatively poor (Figure 1b) and provides only a limited resolution of the spatial distribution of the seasonal cycle and hence we have used the altimeter data fields for our analysis. We also note that although there is some reduction in quality of the altimetry data at the coastal boundary [Ducet et al., 2000], a comparison of on-shelf estimates of the sea level seasonal cycle derived from altimetry, with coastal tide gauge values along the southern shelf shows excellent agreement (results not shown).
 The most energetic region with the largest SLA gradients is located off western Australia. There is clear evidence of a robust seasonal cycle - a band of negative SLA in summer is transformed into an equally strong positive feature between autumn and winter. Note that this seasonal change is superimposed on an annual mean of comparable strength: at the boundary the height fields in Figure 2 show a poleward flowing LC adjacent to the continental slope throughout the year. In fact, the flow patterns inferred from Figure 3 are a complicated sum of a poleward component and several eastward jets located between 20°–32°S. Of most interest in this study is that the seasonal poleward flow increases in autumn and is strongest in the period May–July. Associated with this intensification is the development of large-scale westward meanders.
 Along the southern coast the picture is rather different. The strong seasonal signal (the January and July fields represent the seasonal extremities) dominates the rather weak mean field seen in Figure 2. The coastal sea level is depressed by some 15 cm along the entire southern coast in summer and raised by the same amount in winter. Pariwono et al.  extracted a similar pattern from coastal tide gauges and observed that the transition from summer to winter states occurred almost simultaneously (within several days) over the entire coast. We observe that in January a coastal drop in sea level drives a meandering westward flow located some 150–200 km from the southern coastal boundary. This represents something of a surface expression of the Flinders Current, a year-around upwelling favorable subsurface boundary current, flowing from east to west along Australia's southern shelves [Bye, 1983; Middleton and Cirano, 2002].
 As the LC off the west coast strengthens in autumn the southern height regime begins a transition to its winter state. By May a positive sea level anomaly has been established along the entire southern shelf and an associated eastward current is located at the shelf edge. For the following 4 months there is a continuous warm current on the continental slope that flows from Northwest Cape, turns to the southeast around Cape Leeuwin, continues across the GAB and finally flows southward down the west coast of Tasmania. This represents possibly the first direct indication of a continuous flow between the eastern GAB and western Tasmania. While Cirano and Middleton  have presented similar results for a model forced by typical winter winds our results provide the observational linking element between surveys performed in the western [Godfrey et al., 1986], central [Hahn, 1986] and eastern sectors [Baines et al., 1983] of this region.
5. Flow Trajectories
 Many studies of ocean currents have demonstrated the value of the Lagrangian perspective of drifter tracks to describe flow trajectories. Unfortunately, there is no single period in which drifters were deployed over the entire current path; in fact, for much of the southern coast, only a very limited number of drifters of any type have ever been deployed. For example, Malcolm  reported on a single drift bottle that was recovered in the vicinity of Kangaroo Island in December after being released at Albany in May. A more systematic 2-year drift bottle study implied winter southeastward flow from the eastern portion of the southern coast to Tasmania with a summer reversal [Vaux and Olsen, 1961].
 To compensate for this comparative dearth of observations, we present results in the form of a composite autumn-winter drifter track in which we have patched together 4 tracks from 3 separate years (Figure 4). Note that for the first and last segment of the composite path off the western Australian coast and off western Tasmania, the displayed tracks are chosen from sets of many other similar examples, while the two central tracks appear to be the only buoys deployed in this region. The composite track traces out a predominantly poleward shelf edge path which matches that implied for the winter period in the height fields in Figure 3. The drifters show that the current reaches speeds of more than 1.5 m s−1 around Cape Leeuwin with more moderate eastward velocities across the GAB of order 0.2 m s−1.
 While very few actual drifter tracks have been obtained in the central portion of the southern shelf region, we are able to utilize a much larger set of pseudo-tracks determined from altimetry height fields. Figure 5 shows probability distributions of pseudo floats released at the indicated locations along the west and south coast determined from the procedures described in section 2. In each case the release location consists of a 2.0° × 0.5° band running across the outer shelf and slope. In the short term (<1 month), trajectories are dominated by eddy motions and probability distributions tend to form a localized cloud around the release locations (not shown). However, Figure 5 indicates that in the longer-term (>2 months) distributions are aligned along the shelf edge, suggesting that the float trajectories tend to be controlled by the coastal current system rather than offshore transitions. This is most obvious during the winter period, when the patterns reflect a strong tendency for floats to move south along the west coast, east across the GAB, and then poleward down the west coast of Tasmania. These results are entirely consistent with the current patterns described above. However, the advection timescales further suggest that it would be extremely rare for fluid parcels to travel from the west coast to southern Tasmania in a single year, which is consistent with observed drifter tracks (Figure 4).
6. Forcing Along the Current Path
 In our study domain, the dynamics associated with the boundary current on the western and southern coasts may differ but there are some common features. Following onshore transport there is a subsequent buildup of sea level along the entire southwestern coast and mass conservation induces coastal downwelling (Figure 3). The resultant onshore sea level slope drives a geostrophic flow poleward on the west coast and eastward on the southern coast.
where (x, y) represent the (cross-shelf, along-shelf (poleward)) directions, (u, v) are the corresponding velocity components, p is the depth-integrated steric height, τy is the alongshore wind stress, U, V are the depth integrals of (u, v), D/Dt is the Lagrangrian derivative, and Y is the depth integral of all the Reynolds stress terms.
6.1. Western Coast
 On the western coast the onshore flow is driven by the large-scale meridional pressure gradient, which is very clearly evident in Figure 2. This generates onshore geostrophic transport that is sufficient to exceed the offshore Ekman transport induced by the equatorward wind stress [Thompson, 1984; Godfrey and Ridgway, 1985]. Smith et al.  suggested that this onshore flow is balanced by offshore Ekman transport near the bottom, over the upper slope and outer shelf, under the LC. The induced poleward flow at the coastal boundary overcomes the equatorward forcing of the alongshore wind stress and is in turn balanced by bottom friction [Thompson, 1987].
 Off the western coast the dominant term on the rhs of equation (3) is the alongshore pressure gradient (Figure 6a) which drives the LC [Thompson, 1984; Godfrey and Ridgway, 1985]. This term reaches its maximum during autumn-winter at the same time as the northward winds weaken. This appears to confirm the contention of Godfrey and Ridgway  that the seasonality in the strength of the LC is due to seasonal variations in both terms rather than simply the slackening of the autumn-winter winds as suggested by Smith et al. . Their results showed little seasonal dependence in the alongshore gradient but being based on a single anomalous year (low Southern Oscillation Index) may not have been representative of ‘average’ conditions. We also note that the secondary peak in summer for g∂p/∂y is in agreement with the observations of the alongshore gradient inferred from altimetry [Morrow and Birol, 1998].
6.2. Southern Coast
 On the southern coast, beyond Albany, determining the influence of the alongshore pressure gradient is far less straightforward. From Figure 2 we have noted that the mean height field displays both nonuniform alongshore scale structure and offshore variations at relatively short scales. However, in this region mariners have long believed that the observed currents result from wind forcing [British Admiralty, 1937]. Observations provided by Newell  indicated that the currents along the eastern portion of the southern coast were correlated with the wind direction. Godfrey et al.  suggested that the prevailing westerlies and possibly independent thermohaline effects drive the eastward currents. In Figure 6b we observe that the alongshore wind stress does appear to be an important if not the main forcing term. It has a seasonal cycle which is in phase with the pattern of coastal sea level evident in Figure 3.
 In summer a high-pressure ridge is maintained over the South Australian Basin which induces a consistent pattern of southeasterly wind. In contrast in winter the anticyclone moves to the north over central Australia resulting in a predominantly westerly wind regime. The transition from an easterly to a westerly alongshore pattern occurs from March to May and this is also the period that coastal sea level changes from a negative to a positive anomaly. The westerly wind stress at the coast drives an onshore Ekman mass transport which leads to coastal downwelling and an eastward shelf edge flow [Middleton and Cirano, 1999; Cirano and Middleton, 2004] representing a rare example of downwelling currents driven by seasonally reversing winds (Neshyba et al.  describes upwelling favorable cases). The mechanism described above thus is established in less than a month in April after the coastal winds switch over to their winter pattern. We note that in summer when the winds reverse a westward (upwelling) current is formed when similar dynamical processes lead to a drop in coastal sea level.
 We still need to resolve the importance of the pressure gradient forcing term. In Figure 6b the pressure gradient is determined over the entire southern coast from Albany to Portland. However, calculating the pressure term over this distance may ‘average’ out the effects of more localized gradients. For example, at the western end of the southern coast the current dynamics are somewhat more complicated. Figure 3 suggests that a substantial alongshore steric height gradient is continued eastward around Cape Leeuwin - it is certainly capable of driving the current around the Cape where it flows eastward along the southern shelf [Godfrey et al., 1986]. In fact, it is in this region, just beyond Cape Leeuwin and before the GAB, that the LC reaches its maximum speeds (Figure 4) [Cresswell and Golding, 1980; Cresswell and Peterson, 1993]. This is confirmed by a number of models with extended domains, which have successfully reproduced this behavior [Weaver and Middleton, 1989; Batteen and Butler, 1998; Batteen and Huang, 1998]. In all these cases the models were driven by thermohaline forcing alone (a smoothed climatological density gradient was imposed) and then were allowed to geostrophically adjust in the absence of external forcing. Cirano and Middleton  have examined the mix of forcing terms in this region. They found that the influence of the LC on the total flow steadily diminishes as it moves toward the east until off the eastern GAB (135°E) it drives only some 15% of the total flow. Of the remainder, wind forcing accounted for 47% and a pressure gradient term made up the final 38%. This latter contribution is certainly much higher than that implied by the results in Figure 6b which may indicate that our pressure gradient estimate may not be representative of the ‘true’ forcing term at the shelf edge. Either because of sampling or reference level shortcomings at the shelf edge, we are left with some uncertainty associated with our estimate of the strength of the pressure term.
 We are able to conclude that quite fortuitously, the west coast pressure gradient delivers the LC to the south coast at just the right time for the reversed wind distribution to maintain the currents passage eastward.
6.3. Western Tasmanian Coast
 The easternmost sector of the wave guide stretches from the western Victoria coast, where the shelf break changes direction to the southern tip of Tasmania. The shelf here has a much more meridional orientation and this means that the zonal winds are no longer parallel to the isobaths hence the alongshore forcing will be reduced. Middleton and Cirano  suggested that the alongshore flow in this region (Zeehan Current) may be an outcome of a forcing regime located to the northwest, the local density structure being set up by this remote forcing. Early drift bottle observations appeared to show that the current is seasonally reversing [Newell, 1961; Vaux and Olsen, 1961]; however, Baines et al.  suggested that the flow was permanently southward. Even the direct current observations have displayed some ambiguity in the seasonal patterns [Lyne and Thresher, 1994; Cresswell, 2000]. The seasonal map of steric height (Figures 3) clearly shows that there is a much stronger southward flow off western Tasmania in winter which tends to confirm Cresswell's  findings. We have been able to identify the alongshore wind stress as providing the extra forcing in this period. Figure 6c shows that the climatological winds have strong southward wind stress from May to September which corresponds to the period of high coastal sea level and induced southward currents (see Figure 3). The 1997 winds were very close to the climatological average (Figure 6c) and hence the current meter records by Cresswell  showed the expected seasonal pattern. Lyne and Thresher  found strong southward flow from June to November 1988, however, Figure 6c shows that by chance their deployment coincided with anomalously high southwest wind stress over spring. Recent seasonal model runs have shown that the Zeehan Current is clearly far stronger in winter [Bruce et al., 2001; Middleton and Platov, 2003; Cirano and Middleton, 2004].
 Despite the strong seasonality, a range of evidence strongly suggests that there is southward flow throughout the year [Baines et al., 1983; Cresswell, 2000]. For example, drifters follow southward trajectories in summer [Cresswell, 2000] and the mean surface pattern illustrates that there is a mean southward Zeehan Current (Figure 2). We are not aware of the mechanism that underlies this result as the alongshore pressure gradient appears to be weak (Figure 2).
7. Water Mass Characteristics
 We now examine the water mass characteristics of the boundary flow and thus to identify where there is entrainment of neighboring waters or modification through air-sea interactions. To assist this process we draw from the results presented in Figures 7–10. These include, a sequence of SST anomalies, seasonal salinity maps and related properties along the current path.
7.1. Sea Surface Temperature Fields
 SST imagery from satellites provides a very comprehensive picture of the sources and modification of water masses associated with the current system as well as supplementary information on the current strength and direction. The SST anomaly for the same sequence of months as the height fields in Figure 3 is presented in Figure 7. The anomalies are constructed by removing the spatial mean and to improve the visibility of the small-scale boundary structure we also remove the domain seasonal signal (the average monthly anomaly over the whole domain) from all of the maps. These SST patterns show a remarkably similar picture to the height fields of the spatial distribution of the boundary flow, including intensification at the shelf edge and the transformation that occurs from the summer to the winter flow regimes. However, the much greater spatial resolution of the SST imagery also provides a wealth of detail which is not contained in the altimeter fields. What is most striking is how the temperature structure of the shelf edge region stands out from the surrounding waters throughout the year as a narrow feature following the 200-m isobath. This attribute is also very obvious in other representations of the SST data such as the annual mean, and the annual components of amplitude and phase. It is also far more distinct than the coastal currents in other regions such as the East Australian Current.
 As with the height fields the autumn-winter growth of the LC poleward flow of warm tropical water at the western Australian boundary is striking. The initial small positive anomaly off North West Cape in March rapidly develops into the full strength LC which floods southward in May and June. The warm water pulse first penetrates around Cape Leeuwin and onto the southern shelf in May and maintains a continuous presence for the next 6 months (there is a weak remnant still visible in October; not shown). We note that the full temperature of this pulse steadily decreases along the path, from 26.3°C at the northern origin, to 21.5°C at Cape Leeuwin reducing to ∼18oC on reaching the GAB (Figure 8a).
 A second source of warm water forms on the western edge of the GAB (∼126°E) in February (not shown). A warm pulse is evident in March which spreads eastward across the shelf in a broad tongue through April and May [Rochford, 1986]. Herzfeld  has shown that it is formed in the region of shallow water in the western end of the GAB (124°–129°E) by a combination of a positive heat flux and the passage of anticyclonic weather systems. A positive heat flux exists during the spring-autumn period, which is sufficient to both heat the water some 2°–3°C above the surrounding waters and to increase the salinity by evaporation. In addition, Herzfeld and Tomczak  showed that the eastward flowing plume is actually part of an anticyclonic gyre on the eastern GAB shelf which includes a pool of cooler upwelled water in the east. This cool water flow is evident in Figure 7 on the eastern GAB from March onward. It actually first develops in January as an upwelling shelf edge flow east of the GAB at the Bonney coast of South Australia (139°E) which spreads northwestward onto the shelf by March.
 In the transition from March to April the warm tongue moves beyond the confines of the broad shelf region onto the slope, as a narrow tendril of warm water that spreads southeastward along the shelf edge beyond Kangaroo Island. From Figure 3 we note that this warm feature is forced by the alongshore steric height gradient which is now favorable for eastward flow. In fact, April is the month of transition between the summer easterlies and the winter westerlies. Once the full winter state has been established with positive sea level anomaly along the entire southern shelf and the associated eastward current, the seasonal jet of warm water moves progressively eastward until it reaches the southeast of Tasmania in July. In tandem with this process, the flow of LC water from the western region continues and its warm tongue penetrates onto the western flank of the GAB from May to August. We observe that a continuous jet of warm water stretches from North West Cape to Tasmania in July corresponding to the maximum cross shelf slope of the steric height fields in Figure 3. However, it is also apparent that the along-shelf height gradient is set up in April–May at the time of the wind reversal well prior to the June–July period when the warm tongue reaches Tasmania. In fact, while the maximum temperature along the boundary occurs in July, Figure 3 shows that the cross-shelf structure of the sea level gradient has already begun to weaken at this time. Figure 8a indicates that the temperature decreases from 18.3° to 14°C along the path from the GAB to Tasmania.
 One further aspect is that the jet appears much cooler across the central GAB region and the temperature contrast with the surrounding waters is much weaker in this region although it is still quite distinct. The continuous shelf edge flow appears to be formed from two separate water masses located either side of the GAB. From these maps we are not able to determine an unambiguous penetration distance for the LC across the GAB.
7.2. Salinity Patterns
 Salinity is often a better tracer for the influence of interacting water masses as it is not nearly as susceptible as SST, to contamination by a range of forcing factors with similar amplitude. Bye  showed that in this region temperature tends to be influenced by local forcing whereas salinity is controlled by advection and hence is a far better indicator of dynamic processes. However, without satellite measurements we must depend on the much poorer coverage of in situ data. In Figure 9 we present the January and July patterns of salinity (50-m depth) from the CARS climatology supported by the individual cast values for October–March and April–September. In addition the salinity along the shelf edge path (200-m isobath) is shown for the two seasonal extremes in Figure 8b. As predicted the salinity climatology does not resolve the shelf and slope structure in anywhere near the detail provided by the satellite SST patterns in Figure 7. There is no sign of a salinity analogue of the narrow shelf-edge temperature jet. However, several general observations may still be made from these results. The winter pulse of low-salinity tropical water associated with the LC off the western Australian coast is evident. In fact, as the LC tongue propagates southward it passes over the high-salinity subsurface core of South Indian Central water (SICW) and associated mixing processes make it progressively more saline and cooler [Webster et al., 1979]. We also note the year-round high-salinity core of water on the GAB shelf which marches eastward in winter [Rochford, 1986], which is clearly the same water mass indicated by the warm water pulse identified previously. A third observation is that east of the GAB as far as Tasmania, the shelf edge salinity is consistently higher in winter (Figure 8b).
7.3. Time Series of T and S Along Current Axis
 We now seek to obtain a clearer representation of the 2 water masses already identified. Figure 10 presents plots of T and S seasonal anomalies along 2 tracks shown in Figures 10a and 10b over a year. The first track (red line) shows the path traced out by the warm water filament in July (Figure 7) along the shelf edge between north West Cape and Tasmania: it is very close to the 200-m isobath. Figures 10c and 10d show the T and S structure, respectively, of the waters along the path off western Australia, southern Australia and western Tasmania. The second track begins in the western GAB and follows more closely the expected path of the GAB water mass (blue line in Figures 10a and 10b) before rejoining the shelf edge track at 132°E. The progression of the two major water masses and of other modified minor species is apparent, albeit with somewhat more resolution for temperature than salinity. First, the warm, low-salinity LC water spreads southward and then eastward to the western edge of the GAB in autumn-winter. The temperature pattern indicates that the pulse traveled 3000 km in 2.5 months, an along-track speed of ∼0.4 m s−1. Note that Figure 10d shows that, at least in late autumn, there is a seasonal increase in the salinity of the water mass as it approaches the Cape Leeuwin region and in the southern shelf waters to the east. This is associated with the drawing up of higher-salinity SICW by the southward current. Having rounded Cape Leeuwin, the current is now more saline than the surrounding waters. Cresswell and Peterson  observed that in this fresher regime the current carried a sheath of the high-salinity SICW, which had been entrained upstream. The sheath was slowly lost downstream through energetic mixing with the fresher offshore waters.
 From Figures 10c and 10d it is clear that there is a major change of regime at the GAB (125°–130°E). Both the T and S patterns show discontinuities. These changes imply that the high-salinity, warm water mass formed on the shelf region of the GAB is entrained into the eastward flow of the shelf edge current driven by the sea level gradient shown in Figure 3. The propagation pattern for track 2 (Figure 10e) illustrates the passage of this GAB water more precisely (the salinity pattern is very similar along both tracks) - both the temperature and salinity now show a smooth and consistent evolution from source to destination. In fact, the implied propagation speed of the water mass further to the east (130°–140°E) is of similar order to that shown by the LC waters. Overall the T and S results in Figures 10c–10e show quite a consistent propagation structure given the respective resolution capacity of each data type. The S pattern displays more small-scale structure, which may be due to shortcomings in the data distribution.
 East of the GAB an eastward flow at the shelf edge of high-salinity water was proposed in several early studies of the water mass properties in the region [Rochford, 1957; Newell, 1961, 1974]. Rochford suggested that the presence of a high-salinity water mass in the northwestern approaches of Bass Strait was directly attributable to the influence of this postulated eastward flow. Newell  made the further connection between this eastward flow and the high-salinity water found off western Tasmania in winter. While Rochford  surmised that the source of the saline water mass was the Spencer and St. Vincent Gulfs, Newell  expressed more uncertainty and could not distinguish between the Gulfs, the GAB or indeed the South Indian Gyre as possible origins for the water. Our results (Figure 10) show quite unequivocally that this saline flow arises from the eastern GAB [Herzfeld, 1997] and is introduced into the shelf edge current at the western GAB as suggested by Bye . Furthermore, it can be argued that there is little augmentation of this current from outflow from the Gulfs. A winter density outflow of cool and very saline water from Spencer Gulf intrudes onto the shelf and slope region [Lennon et al., 1987] before finding its own density level at ∼250–300 m. However, there is little opportunity for this bottom flow to influence the surface layer occupied by the GAB outflow [Hahn, 1986].
 We note that prior to the appearance of the GAB water mass, the shelf edge temperature in Figure 10c steadily decreases (between 120° and 125°E). Along this portion of the southern coastal boundary the current undergoes a very vigorous interchange with the open ocean [Griffiths and Pearce, 1985: Godfrey et al., 1986; Cresswell and Peterson, 1993]. The actual role of eddies breaking away from the shelf west of 124°E in removing warm water from the slope/shelf region is yet to be determined [Griffiths and Pearce, 1985; Godfrey et al., 1986]. However, instability processes are likely to be involved in causing a horizontal diffusion of heat by decelerating the alongshore flow and reducing the cross-shelf temperature gradient. There is also strong evidence of instability processes occurring east of the GAB which lead to exchange with the deep ocean (Figures 3, 5, and 7; see the July patterns) - these features have also been observed in previous studies [Bye, 1983; Godfrey et al., 1986]. We plan to explore this exchange process in detail in a future study.
 What then is the fate of the Leeuwin Current? If we follow the seasonal propagation of the temperature front of the current from Cape Leeuwin, within the GAB the location of the front becomes more difficult to determine because of the intrusion of GAB shelf water onto the slope. However, there is an implied eastward penetration limit from the temperature pattern (see the red dashed line) which is essentially confirmed by the sharp transition from fresh to saline water in Figure 10d. The LC appears to reach a longitude of ∼130°E before its TS signature is completely swamped by the shelf outflow of high-salinity GAB water. This result is in general agreement with the conclusions of Rochford  and Herzfeld . However, it is interesting to note that Cirano and Middleton  found in their model that although the influence of the LC does steadily decrease in this region it is still detectable at 135°E, accounting for 15% of the total flow.
8. Biological Evidence
 There is a large body of evidence that demonstrates the important biological impact of the southwestern coastal current system [Edyvane, 2000]. In particular, it has been shown to be a vital factor in a range of ecological mechanisms. We are interested here in organisms carried great distances along its path and so acting as ongoing biological tracers illustrating the complete path of the shelf edge current. These may be compared with the trajectories indicated by both real and pseudo drifters in section 5. Some of the earliest recorded observations of the LC came from inferences drawn from biological data. Saville-Kent  reported that tropical marine species such as coelentrerates and holothurians were found in the Houtman Abrolhos at 29°S, far south of their expected range. In fact, the poleward penetration of the warm low-salinity waters of the LC is the underlying mechanism for the migration and subsequent distribution of many marine pelagic fauna, from the tropical northern waters to the temperate southern shelf [Maxwell and Cresswell, 1981; Edyvane, 2000]. Maxwell and Cresswell  have described a range of observations which demonstrate that the GAB has a distinctive ‘Indo-Pacific’ character.
 While there is an abundant data set describing the penetration of the LC as far as the GAB, there are somewhat fewer biological studies which make the further link to the southern Tasmania. Wood  observed a distribution of dinoflagellates from southwest Australia to Tasmania and noted that they are excellent indicators of water masses. He surmised that an eastward subtropical current, flowing from Cape Leeuwin to western Bass Strait was responsible. He further suggested that the existence of such a warm current was confirmed by the discovery of warm water turtles on the west coast of Tasmania. In back to back cruises in 1982, a tropical coccolithoprid (Scyphosphaera apsteinii) was present in a continuous sample from Cape Leeuwin to western Tasmania [Godfrey et al., 1986]. The coastal circulation along the southwestern shelf appears to be intimately related to the spawning behavior of the Australian salmon species Arripis trutta [Malcolm, 1960]. The most intense spawning occurs in April–May near Cape Naturaliste: the eggs and larvae are introduced into the system just at the time of the development of the eastward flow. Thus salmon eggs produced off western Australia are transported along the shelf edge current and appear to successfully bypass the unproductive waters of the GAB shelf [Malcolm, 1973].
9. Discussion and Conclusion
 Using a wide range of evidence from climatological fields, satellite altimetry, SST imagery, drifters, and biological observations we have demonstrated that during winter there is a continuous southeastward flow at the shelf edge from the North West Shelf to South East Cape; a distance of 5500 km. The system is driven by the combined effects of the alongshore pressure gradient and the local wind forcing. On the west coast the pressure gradient provides the primary forcing; it is strong enough to overwhelm the equatorward alongshore winds. On the southern shelf our results suggest that the pressure gradient is much weaker and the eastward current is mainly forced by the seasonal reversing alongshore winds.
 The current system is also made up of contributions from two major water mass inputs although there is a continuous modification of the water properties along the current trajectory. The LC provides an inflow of warm, low-salinity water of tropical origin that is modified significantly by contributions from subtropical higher-salinity SICW originating further to the west. This LC water mass may be discerned as far as the central-eastern GAB in midwinter. A fresh input to the boundary flow comes from the eastern GAB after being formed in the shallow shelf waters to the west of the GAB. A gravity outflow of a warm, high-salinity water, formed in the shallow waters on the western GAB shelf, is injected into the eastward shelf edge current stream. The higher-salinity structure of this intrusion is observed eastward to western Tasmania.
 How then should the southwestern boundary current system be described encompassing as it does the complexity of multiple forcing mechanisms and distinct water mass inputs? Should the flow we identified by the unifying factor, its continuity, or by its distinguishing features of forcing and composition? Adopting the first approach the system may be defined as a quasi-continuous flow with the potential to move water from the source to the sink, albeit with much mixing and gradual alteration of the water properties along the way. However, since the idea of individual components such as the Leeuwin and Zeehan Currents has become entrenched in the literature, a more pragmatic approach would be to adopt the following current definitions.
 The southward, shelf edge flow off western Australia that turns around Cape Leeuwin and penetrates eastward as far as the central GAB should clearly continue to be referred to as the Leeuwin Current. This includes the autumn-winter low-salinity warm water flow of predominantly tropical origin as well as a summer flow of mainly more saline SICW. This essentially corresponds to the original definition presented by Cresswell and Golding . Note that while this current is forced by the alongshore pressure gradient off the western coast, on the southern shelf there is a major alongshore wind component. This definition differs somewhat from that employed by Cirano and Middleton . Along the southern coast they defined the LC as the eastward inflow at 120°E (their boundary) and distinguished between this flow and other components arising from separate forcing mechanisms. While this approach may seem useful in a model study when it is possible to diagnose the individual forcing terms, it becomes impractical in any other setting.
 The section of the shelf edge flow from the eastern end of the GAB to the western edge of Bass Strait could be referred to as the South Australian Current (SAC). This reclaims the name found in an early atlas (Black , earliest available version) to describe a broad eastward stream south of Australia, emerging from the south Indian gyre circulation: our usage is rather different. Hahn  initially proposed that the term SAC be applied to the entire southern shelf edge flow but clearly with the presence of the LC west of the GAB this becomes problematic. The SAC is thus the mainly winter shelf edge flow largely driven by seasonally reversing winds which is identified by a warm, saline core that originates from a gravity outflow from the eastern GAB and spreads eastward as far as the western edge of Bass Strait.
 The final component is the Zeehan Current a southward flow of saline, relatively warm water off western Tasmania that spreads beyond the southern tip of Tasmania and ultimately reaches the southern portion of the east coast [Cresswell, 2000; Bruce et al., 2001]. Apart from the changed orientation of the coastal boundary there seems to be no real distinction between the Zeehan Current and the SAC. There is no clear cut transition in either the forcing or water mass composition along the current pathway from the eastern GAB to Tasmania. However, for mainly historical reasons it is unlikely that this name will die out.
 The intriguing aspect of the current system is that the two independent forcing mechanisms act in such a synergistic manner. While the two mechanisms are very different they are each influenced by changes in the regional wind regime. The southern coast component is directly forced by the alongshore wind system whereas the western coast region is effected by the wind in a far more indirect manner. The fact is that in winter the weakening in the alongshore southerly winds off the western Australian coast is in phase with the reversal of the southern coast system. Both of these changes are local expressions of basin-scale systems that are unlikely to be influenced by the coastal geometry. There appears to be no underlying reason why the wind regime operates in such a way to reinforce the shelf edge current flow over its entire trajectory.
 The significant regional climatic implications of the Leeuwin Current flow in autumn−winter off the western Australian region have been explored in several studies [Gentilli, 1991]. Similarly we would expect a strong regional response off the southern coast following the passage of a warm eastward flow in winter. This leads us to question the stability of the ‘seasonal’ pattern presented here. It has been well known that the LC off the western Australian coast exhibits considerable interannual variability [Griffin et al., 2001] with consequent climatic influences [Pattiaratchi and Buchan, 1989]. Smith et al.  and Morrow and Birol  have shown that the alongshore gradient, g∂p/∂y, has clear differences from year to year but limited observations indicate that the seasonal variation of the alongshore wind remained fairly constant [Morrow and Birol, 1998]. The strength of the LC on the southern coast is very sensitive to the coastal winds - how stable is the mean seasonal cycle? Harris et al.  showed a clear cyclical pattern in the zonal westerly winds over Tasmania and related this to interannual variations in the regional fisheries. The continental high-pressure cell over Australia, which drives the westerly wind belt, is known to have interannual variations [Thresher, 2002]. In fact from 1940 to 1990, the subtropical ridge exhibited fluctuation of ±2° of latitude in winter. There is likely to have been a corresponding effect on the frequency of downwelling favorable winds along the southern coast in winter. For example, pseudo-drifter results (as shown in Figure 5) display observable year to year variations. Floats released off the western GAB in different years, generally have an eastward, shelf edge preferred trajectory (as in Figure 5 for 1996). However, in 1998 the floats have a strong tendency to move offshore, with no obvious preference for eastward or westward movement indicating that the westerly wind forcing may have been anomalously weak in that year.
 The aspects of the southwestern boundary flow presented in this paper provide further evidence of its unique character. We have noted that because of a combination of certain quirks in the large-scale wind systems and the coastal geometry of the Australian continent it has a combined path length of over 5500 km. This may well be the longest continuous coastal current system in the world. Underpinning this behavior is that there is also no other comparable continental system as Australia. The particular attributes in question are the open passage between the Pacific and Indian Oceans in the north, which maintains the alongshore gradient off the western coast of Australia and the unusual length of the zonally oriented coast in the south.
 This project was carried out with the support of the Strategic Research Fund for the Marine Environment (SRFME). We wish to thank George Cresswell, David Griffin, and John Middleton for their careful reading of earlier versions of this paper and for their helpful comments. We acknowledge Jeff Dunn, Madeline Cahill, Jim Mansbridge and Jason Waring for assistance with the data analysis, and Chris Rathbone and Neil White for providing the satellite altimeter and SST data sets.