a. Linear dynamics

The lowest-order estimate of the dynamics governing the flow around New Zealand is a linear model (labeled RG1, Table 1), the dynamics of which are essentially those of a Sverdrup (1947) interior with Munk (1950) western boundary layers and globally applied horizontal friction. The flow and sea surface height are calculated using a global numerical model with realistic coastline geometry and monthly climatological wind forcing. The model has one vertical mode and a bottom layer that is infinitely deep and at rest. Since the linear simulation reveals the lowest-order response to external forcing, it can be used as a baseline for the more complicated simulations that incorporate nonlinearities, bottom topography, multiple vertical modes, flow instabilities, and isopycnal outcropping.

The mean transport streamfunctions of the 1.5-layer linear 1/8° reduced-gravity simulation driven by the HR monthly wind stress climatology (Fig. 4a) reveal a system of nested anticyclonic gyres encompassing the Southern Hemisphere. Nested gyres were also identified numerically by De Szoeke (1987) and Godfrey (1989) and observationally by Davis (1998), using neutral buoyancy floats. A large supergyre extends through all three oceans following two different paths connecting the Pacific and Indian Oceans: one path extends through the IPT (purple, dark blue, and blue contours), while the other travels south of Australia (light blue and blue-green contours). Two smaller gyres compose an inner nesting: one gyre occupies the South Atlantic and Indian Oceans (dark green to maroon contours) and the other occupies the South Pacific Ocean and extends south of Australia (dark green contour bounded on west by Australia). Three gyres, one bounded on the west by Australia (yellow, orange, and red contours), the other two bounded by New Zealand (light green contours), compose the innermost nesting in our region of interest. Closer inspection of the region east of Australia (Fig. 4b) reveals how the currents of interest are connected to the larger flow field. The EAC and ECC are western boundary currents of the innermost nested gyres. The Tasman Front is the southern edge of the innermost gyre bounded by Australia. The northern anticyclonic gyre east of New Zealand is bounded on the west by the ECC, but the southern one is unobserved. The southward western boundary current associated with this gyre disagrees with the observed northward Southland Current and its subsequent eastward flow at the Chatham Rise as described by Chiswell (1996), Stramma et al. (1995), and Uddstrom and Oien (1999). Likewise, the simulated EAUC, the western boundary current connecting the two northern inner gyres, flows opposite to the observed direction (northwestward in the linear simulation versus southeastward observed). The simulated EAC and ECC both separate from the western boundaries at the observed locations (centered near 32°S on the east coast of Australia and near 42°S on the east coast of New Zealand, respectively), while the simulated Tasman Front appears in the correct geographic location.

Although observational values of the transports are sparse, a rudimentary model-data comparison can be made. The simulated linear EAC is 25.9 Sv, which agrees well with observational values. Ridgway and Godfrey (1997) estimated the flow of the EAC relative to 2000 db, using steric heights derived from historical hydrology and expendable bathythermograph (XBT) data collected in the region, and found that the EAC varies between 27.4 and 36.3 Sv seasonally. Chiswell et al. (1997), using CTD data gathered over five cruises between 1990 and 1994, estimated the geostrophic transport of the EAC relative to 2000 db to be between 22.2 and 42.2 Sv. The simulated linear transport associated with the Tasman Front, 10.6 Sv, also agrees well with observational values. Stanton (1979), using bathythermograph (BT) data, calculated geostrophic volume transport relative to 1000 db to be 7.6–8.5 Sv, which was slightly lower than his previous observations of 9–12 Sv (Stanton 1976). Andrews et al. (1980), using XBT data, estimated the net zonal transport to be approximately 15 Sv relative to 1300 db. More recent observations yielded integrated volume transports measured at 173°E ranging from −7.1 to 16.5 Sv over a 4-yr time frame (Chiswell et al. 1997). However, the authors hypothesized that the negative transport value in the Tasman Front transport was due to an error in measurement and should not be considered evidence of a westward flow into the Tasman Sea (Chiswell et al. 1997). The simulated transports from RG1 for selected sections along with those from the subsequent simulations and observations are summarized in Table 3. The transport sections surrounding New Zealand are shown as gray lines in Fig. 1.

As noted, there are some discrepancies between the linear simulation and observations. The simulated linear EAUC (3.1 Sv to the west-northwest) is opposite in direction to observations, which show part of the flow along the Tasman Front continuing along the northeast coast of New Zealand. Heath (1985b) describes an east-southeast flow of 2–10 Sv, while other observations (Stanton et al. 1997) describe larger transports of 11–34 Sv. The simulated ECC (2.1 Sv southward) is much less than the observed 22 Sv (Stanton et al. 1997). The simulated net southward transport through the Tasman Sea (12.1 Sv) is higher than observed values of 7.1 Sv at 44°S (Ridgway and Godfrey 1997) and 6.3–9.2 Sv at 43°S (Chiswell et al. 1997). Also, the southward-flowing boundary current along South Island disagrees with the observed northward-flowing Southland Current. These discrepancies suggest that linear dynamics are unable to explain some aspects of the flow field, and more complicated processes are necessary to fully describe this region.

Since the linear model agrees with observed locations of EAC separation, we are able to use linear theory to investigate the baseline dynamics behind this phenomenon. The unique position of New Zealand (directly in the path of a separating western boundary current) has fueled speculation concerning the role of the island in the separation of the EAC. Two main theories concerning its role focus on two different aspects of the flow. The first states that New Zealand blocks the westward-propagating Rossby waves south of 34°S, causing the EAC to separate at this latitude instead of much farther south (Warren 1970). The second states that the coastline geometry of Australia (i.e., the bend in the coastline at Sugarloaf Point) causes the separation (Godfrey et al. 1980; Ou and De Ruijter 1986), and the presence of New Zealand is unimportant.

Transport streamfunctions from a simulation (Fig. 5) where New Zealand has been removed (RG2) demonstrate that the removal of New Zealand does not alter the separation latitude of the EAC. In a linear simulation, complete separation of a western boundary current from the coast occurs at the latitude of the zonally integrated zero wind stress curl line integrated from the eastern boundary to the western boundary segment. This separation can be complicated by land masses that partially block gyres and by gradients in the wind stress that give zonal currents but do not act as outer-gyre boundaries. There is no zero wind stress curl line (Fig. 6) that would cause the EAC to completely separate from the coast of Australia. A gradient, however, exists in the zonally integrated wind stress curl (shown as a gray line in Fig. 6) that causes a zonal flow eastward at 34°S, providing a southern boundary to a nested gyre within a larger gyre. The separation points of the EAUC (37°S) and the ECC (42°S) also correspond to gradients in the zonally integrated wind stress curl. Zonal currents at these latitudes are seen at the same latitudes, even after New Zealand has been removed (Fig. 5). The partial separation of the EAC can be explained by the response of the ocean to the wind field with New Zealand present or removed, suggesting that New Zealand does not affect the separation point, aside from the island's effect on the wind field itself. Since a reduced gravity model produces the correct EAC separation point, bottom topography and coastline details of Australia must also have little effect on the separation latitude.

The reversed direction of the EAUC in the linear simulation RG1 can be linked to higher than observed southward transport through the Tasman Sea. The simulated transport is slightly less than that of the linear model of Godfrey (1989), discussed in Ridgway and Godfrey (1994) (12.1 versus 15 Sv) but still larger than observations. Analysis of the mass balance of this region shows four pathways available for water to enter or exit the region (Fig. 7): the Tasman Sea, the basinwide transport at 32°S, the transport of the South Pacific subtropical gyre east of New Zealand at 39°S, and the western boundary current along New Zealand (i.e., the EAUC). Since net vertical mixing is small compared to the horizontal transports in this latitude band (32° to 39°S, marked in Fig. 7), the residual of the mass balance of the transports along the three other pathways determines the direction of the EAUC. If the non-EAUC transport into the region (equatorward flow of the South Pacific subtropical gyre across 39°S) is larger than that out of the region (transport across 32°S + Tasman Sea) then the EAUC must flow southeastward (out of the latitude band) to achieve mass balance. If the transport into the region is less than the transport out, the EAUC must be northwestward (into the latitude band). From Table 3, we can see that, if the EAUC is to flow southeastward in RG1, the southward flow through the Tasman Sea must be less than 9.0 Sv [25.5 (transport across 39°S) −16.5 (transport across 32°S) = 9.0 Sv]. Since the simulated flow through the Tasman Sea in RG1 is larger than this value and the observed value, the simulated EAUC flows in the incorrect direction.

Decreasing the flow through the Tasman Sea in the linear simulation, to agree with observed values by increasing the local horizontal friction within the Tasman Sea (RG3) (Fig. 8a), produces an EAUC that flows southeastward, connecting the two northern inner gyres seen in Fig. 4b. The EAUC can be divided into two flow regimes: 1) before separation and 2) after separation. Before separation corresponds to the alongshore transport passing between the northern tip of New Zealand and 33.6°S, 176.5°E. After separation corresponds to the alongshore transport passing between the eastern coast of New Zealand and 36.0°S, 178.9°E, after the majority of the EAUC has turned offshore, leaving a much weaker flow attached to the coast (see Fig. 1). The linear EAUC from RG3 has a magnitude of 6.5 Sv before separation and 0.9 Sv after separation, in agreement with the lower bound of observations. The weaker attached flow continues around the East Cape and joins with a westward onshore flow to produce an ECC of 6.5 Sv. The separation point of the EAUC agrees well with observations and is situated at a sharp gradient in the zonally integrated wind stress curl (Fig. 6).

Since the model domain has no transport through the Bering Sea, all of the northward transport across the basin at 32°S must leave the basin through the IPT. Since estimates of the transport through the IPT vary from <5 to 18.6 ± 7 Sv (Godfrey 1996), an incorrect choice of a simulated IPT might conceivably affect the mass balance of our region. However, varying the magnitude of the IPT, and therefore the transport through 32°S, has no effect on the EAUC or ECC. Shutting off the flow through the IPT (RG4) does not change the total flow out of the region (Fig. 8b); it merely diverts the flow through the Tasman Sea via an increased EAC (42.4 versus 25.9 Sv) with no change in the EAUC or the ECC.

The failure of the linear model to produce the correct direction of the EAUC may be in part due to its inability to simulate isopycnal outcropping, which is discussed in the next section. Although the lowest-order response of the ocean to the wind stress can explain the direction and separation of most of the major currents in our region of interest, it fails to reproduce the complicated flow field to the north of New Zealand and the meanders in the Tasman Front.

b. Nonlinear dynamics: Flat-bottom case

A 6-layer nonlinear 1/8° simulation with a flat bottom (FB) driven by HR winds is employed to examine vertical structure and nonlinear dynamics. Mean surface currents calculated from the last four years of the simulation (Fig. 9a) reveal flow meanders in the Tasman Front and the relatively persistent eddies observed by Roemmich and Sutton (1998) north and east of New Zealand. The simulated surface currents reveal an elongated eddy corresponding to the observed NCE and eddies to the north and southeast of East Cape corresponding to the ECE and WE. The simulated meanders in the Tasman Front have wavelengths (∼410 km) that agree with the observational values of 300–500 km (Stanton 1979; Andrews et al. 1980; and Mulhearn 1987) although their amplitude and phase do not.

Stanton (1976) postulates that the wavelengths of the meanders over the Norfolk Ridge might correspond to those of a westward-propagating Rossby wave, although there was not sufficient data at the time for a definite conclusion. Andrews et al. (1980) found that the wavelengths required for a Rossby wave with the observed propagation speed of the meanders were much larger than those observed and suggested that nonlinearities were affecting the flow. Both Stanton (1979, 1981) and Mulhearn (1987) mention the possibility of topographic influences on the meanders, although Mulhearn (1987) notes that he was unable to find a clear indication of the mechanism.

Several other authors have examined the dynamics behind stationary meanders associated with a separating western boundary current. Ou and De Ruijter (1986) argued that the exact separation point of a western boundary current was a function of the transport within the current and the shape of the coastline. When the coastline slopes away from the separation point (similar to the orientation of the Australian coastline), the boundary current continues southward after separation and then retroflects to return to the latitude of initial separation, forming meanders that are then governed by conservation of absolute vorticity. Campos and Olson (1991) and Da Silveira et al. (1999) were both able to produce mean meanders in a zonal separation jet in both barotropic and baroclinic nonlinear simulations. Both sets of authors also found that the separation point and amplitude of the meanders were greatly influenced by the shape of the coastline. Holland (1978), using a quasigeostrophic 2-layer model with a straight western boundary did not reproduce meanders, demonstrating that some sort of asymmetry such as a sloped coastline is required to produce meanders.

The simulated meanders of the flat-bottom model (Fig. 9a) are approximately stationary, with the first meander extending southward and retroflecting northward, consistent with Ou and De Ruijter (1986), Campos and Olson (1991), and Da Silveira et al. (1999). The wavelengths of the meanders correspond to the most probable path discussed by Mulhearn (1987) and Stanton (1979) but the amplitudes or phase do not. The amplitudes of the simulated meanders decrease as the flow proceeds away from the western boundary as would be expected in an inertial flow, but in direct disagreement with observations. The wavelengths of both simulated (∼410 km) and observed meanders (300–500 km) are less than those of a stationary Rossby wave pattern or constant absolute vorticity (CAV) trajectory (Haltiner and Martin 1957). For small amplitude trajectories, stationary wavelengths in CAV trajectories are related to the characteristic velocity of the flow by

View Expanded

where υ_c is the characteristic velocity at the core of the current and λ is the stationary wavelength.

The observed and simulated stationary wavelengths are less than the stationary wavelength (600 km) predicted from the depth-averaged velocity of the top five layers of FB (0.17 m s⁻¹). According to CAV theory, the simulated meanders should propagate eastward, not remain stationary as observed. When the bottom layer is constrained to be at rest, creating a 5.5-layer reduced-gravity simulation (Fig. 9b), the simulated meanders have longer than predicted stationary wavelengths and propagate westward, consistent with CAV theory. A long-term mean averages out these time-dependent features, resulting in a mean flow field with little or no meanders (Fig. 9b). Interestingly, the simulated EAC in the reduced gravity simulation separates farther south and retroflects northward to a larger degree than FB. The propagation in the reduced-gravity model and the stationarity in the flat-bottom model are clearly seen in a longitude–time plot of sea surface height deviation (Fig. 10). These differences suggest that, while the meanders within the reduced gravity model can be explained by CAV theory, other dynamics must be present in FB. The inclusion of the bottom boundary and, therefore the barotropic mode, allows the formation of baroclinic instabilities involving the barotropic mode within FB. The appearance in FB of meanders with wavelengths shorter than those predicted by CAV theory indicates that flow instabilities are shortening the wavelength of the mean meanders via vortex stretching. Evidence of the presence of baroclinic instabilities can be obtained by inspection of eddies formed from the meanders within the Tasman Front. The beta Rossby number, R_b = υ/βr², is the ratio of relative to planetary vorticity advection, where υ = maximum swirl velocity within the eddy and r = mean radius of the eddy. An R_b of order 1 suggests that the eddy was formed by barotropic instabilities involving the first baroclinic mode, while an R_b of order 10 suggests that it was formed by baroclinic instabilities involving the barotropic mode (McWilliams and Flierl 1979; Hurlburt and Thompson 1982, 1984; Murphy et al. 1999). Eddies formed during the last two years of the simulation (not shown) were characterized by υ ≈ 35 cm s⁻¹ and r ≈ 47 km, resulting in R_b ≈ 7.9, suggesting that the eddies and, therefore, the meanders were formed by baroclinic instabilities.

Baroclinic instabilities may also be identified by the wavelength of the meanders within the flow field, since the growth rates of disturbances due to baroclinic instabilities are functions of the disturbance wavelength. Although the flow field is constantly exposed to disturbances of different wavelengths, it is predisposed to the propagation and growth of a limited range of wavelengths. Within a long-term mean of the flow field, stationary disturbances with large growth rates will dominate. Mean meanders produced by baroclinic instabilities will most likely exhibit the wavelength of these disturbances. Tilburg (2000), using an approach similar to Phillips (1954) and Haidvogel and Holland (1978), analyzed the susceptibility of the mean flow within FB to baroclinic instabilities by introducing to the mean layered vorticity equations a disturbance and solving for the growth rate. He found that the average wavelength within the simulated Tasman Front corresponded to a stationary, unstable disturbance, providing further evidence that the meanders are due to the presence of baroclinic instabilities.

The simulated directions and separation points of the investigated currents in FB generally agree with observations. Notably, the EAUC flows in the correct direction, while the ECC separates at 42°S and forms a broad band of eastward flow. However, the flow through the Tasman Sea in FB is no longer southward but is mostly confined to the western boundary and travels northward with a transport of 5.3 Sv. The confinement of a northward flow to the western boundary agrees with observations. Huyer et al. (1988), using both current meters and CTD/XBT surveys along the continental margin, found predominantly northward flow along the Australian shelf but southward flow in the interior of the Tasman Sea. The simulated surface flow is northward along the coast and slightly southward throughout the rest of the Tasman Sea, although the combined flow is northward, not southward as observed by Huyer et al. (1988) and others (Chiswell et al. 1997 and Ridgway and Godfrey 1997). The northward flow through the Tasman Sea combines with the EAC as it separates from the coast of Australia and flows eastward, producing significantly larger transports along the Tasman Front (31.6 Sv) than the linear simulations or observations. Although some transports greater than 30 Sv have been observed along the Tasman Front due to recirculation of the flow by eddies, time-averaged transports are much smaller (Stanton 1981). The simulated Southland Current also does not agree with observations.

Two mechanisms may contribute to the simulated northward transport through the Tasman Sea: unrealistically high IPT transport and exaggeration of the effect of isopycnal outcropping in reducing eastward baroclinic transport south of New Zealand. As mentioned in section 3a, any change in the transport through the IPT directly affects the transport at 32°S and through the Tasman Sea. An increased IPT would reduce the southward (or increase the northward) transport through the Tasman Sea but have little effect on the EAUC or the ECC. The introduction of isopycnal outcropping, however, significantly alters the nested gyres seen in the linear simulations. Isopycnal outcropping in the Antarctic region in response to the large negative wind stress curl (Fig. 6) reduces the transport passing south of New Zealand. The depths of the mean interfaces south of Australia and New Zealand (Table 4) are at a minimum along the coast of Antarctica, indicating significant outcropping. Within the model, an outcropped layer does not result in a zero thickness, but instead a minimum thickness. When all five upper layers in the hydrodynamic model outcrop (as is the case for the majority of the Antarctic region), only the nonsteric contribution to the barotropic transport remains, resulting in drastically reduced transport in the upper five layers. The mean interface depths at the coast of New Zealand are much shallower than those at the coast of Australia (Table 4), allowing less transport south of New Zealand than south of Australia, forcing transport northward through the Tasman Sea.

Returning to the linear simulation of the triple nested subtropical gyre system of the South Pacific (Fig. 4a), we see the impact of the reduced transport south of New Zealand on other current systems. Focusing on the middle-nested gyre in the southeast Pacific (represented by the dark green contour in Fig. 4a) we see flow associated with this gyre passing south of New Zealand. As this flow is reduced, more of it must pass north of New Zealand, impacting the transport through the Tasman Sea and the current system that comprises the Tasman Front, the EAUC, and the ECC. Since in FB the net transport through the Tasman Sea is northward, a portion of the supergyre transport feeding the IPT no longer passes south of New Zealand but instead flows north through the Tasman Sea and along the Tasman Front. To achieve mass balance in the latitudinal band described in Fig. 7, the EAUC now flows southeastward along the coast of North Island. The EAUC has a mean magnitude of 24.6 Sv before separation and 13.5 Sv after separation, both of which are significantly larger than those produced in the modified linear case (RG3) but still within the range of observational values.

Clearly the model overestimates the effect of isopycnal outcropping in this region, but it does reproduce the correct direction of the EAUC, suggesting that the dynamics of isopycnal outcropping, if not the magnitude of its simulated effect, are correct. The exaggerated effect of this outcropping on the transport of the ACC has been noted earlier. Shriver and Hurlburt (1997), using a lower-resolution version of this model, found that the simulated transport of the ACC through the Drake Passage was much less than observed values and attributed this difference to the effect of the modeled isopycnal outcropping.

The linear model RG1 does not reproduce isopycnal outcropping and therefore must be modified to reproduce the reduced flow (RG3) through the Tasman Sea and the observed direction of the EAUC. A model that incorporates isopycnal outcropping (FB) does reproduce the observed direction of the EAUC. The flat-bottom simulation also produces meanders with wavelengths that agree with observations but phases and amplitudes that do not.

c. Nonlinear dynamics: Realistic bottom topography case

A 6-layer nonlinear 1/8° simulation with realistic bottom topography (RB8) is used to investigate the role of upper-ocean–topographic coupling in this region. Mean surface currents superimposed on bottom topography (Fig. 11) show that the simulation reproduces stationary meanders (Fig. 10) in the eastern portion of the Tasman Front (meanders E, F, and G, in Fig. 2), the phase of which more closely agrees with observations, revealing the importance of upper-ocean–topographic coupling. The simulated EAC separates slightly farther south than FB (35° versus 32°S) and creates a large retroflection similar to the reduced gravity simulation. RB8 also reproduces all three quasi-stationary eddies and the low observed by Roemmich and Sutton (1998) north and east of the North Island of New Zealand. In RB8, the simulated surface ECC separates from the coast of New Zealand and flows eastward south of the Chatham Rise rather than north of the rise as seen in observations and the other simulations. A summation of the five top layers (not shown) shows a zonal current at the observed latitude of 42°S, revealing the baroclinic nature of the flow and depth-integrated agreement with observations. The introduction of the bottom topography greatly reduces the northward transport through the Tasman Sea (0.1 versus 5.3 Sv), while increasing the IPT transport (13.7 versus 10.5 Sv). These changes combine to produce a reduced transport in the Tasman Front (23.1 versus 31.6 Sv). Although mass balance of the region still requires the EAUC and ECC to proceed southward, the decrease in the Tasman Sea transport gives a smaller transport in both the after separation EAUC (3.8 versus 13.5 Sv) and the ECC (11.7 versus 14.9 Sv). The simulated Southland Current flows in the observed direction, revealing that its location is determined by bottom topography.

The major differences between FB and RB8 are the incorporation of bottom topography, which introduces a bottom layer that is constrained to follow f/h contours, and the change in the initial depth of the mean interfaces. The introduction of bottom topography has two major ramifications: 1) reduction of baroclinic instabilities and 2) bottom steering. When the bottom layer is constrained by topography, it cannot freely adjust to perturbations and become out of phase with the surface layer. This restriction reduces the distribution of baroclinic instabilities and their potential for occurrence. Baroclinic instabilities convert potential energy of the mean flow to eddy potential energy, effectively flattening mean isopycnals. Since transport is related to the change in depth of isopycnals via the thermal wind relation, this flattening restricts the transport south of New Zealand. The greater depth of the original interfaces in RB8 (1500 m for interface 5) compared to FB (1000 m for interface 5) mitigates some of the effect of isopycnal outcropping and provides greater potential for transport south of New Zealand and Australia. The decrease of baroclinic instabilities and the increase in initial interface depth result in much deeper ispoycnals south of Australia and New Zealand in RB8 than FB (Table 4). The relative deepening of the isopycnals allows more transport south of New Zealand, and, consequently, decreases the northward flow through the Tasman Sea.

The inclusion of bottom topography can substantially affect surface currents that do not directly impinge on the topography. Hurlburt and Metzger (1998) found indications of the Shatsky Rise topography affecting the bifurcation of the Kuroshio. This steering is due to the abyssal currents advecting upper-layer thickness gradients and, therefore, surface currents.

As discussed in the previous section, the presence of baroclinic instabilities in FB regulates the wavelength of the meanders in the Tasman Front. The effect on the meander wavelength of the decrease of baroclinic instabilities due to the inclusion of bottom topography in RB8 is compensated by bottom steering, resulting in meanders that are similar in wavelength to those in FB but have a phase that now agrees more closely with observations. Inspection of the bottom topography (Fig. 2) beneath the Tasman Front reveals a series of ridges and valleys, which constrain the abyssal flow and affect the surface flow. Since the presence of bottom topography does not allow the abyssal layer to freely adjust and create meanders solely due to baroclinic instabilities, the wavelengths of the meanders are not directly determined by the presence of the instabilities themselves, although the instabilities may provide a length scale for the meanders. Instead, the flow instabilities contribute energy to the abyssal layer providing greater bottom steering, which determines the wavelengths.

Increasing the resolution of RB8 to 1/16° (RB16) (Fig. 12a) significantly increases the amplitude of the meanders in the entire Tasman Front and produces meanders in the western portion (meanders A, B, C, and D, in Fig. 2) that agree quite well in phase, amplitude, and wavelength with observations. Comparison of the mean sea surface deviations of the model and the most probable path from Fig. 2 (Fig. 12b) demonstrates that the simulation resolves each observed meander. The increased resolution in RB16 increases the northward transport through the Tasman Sea (6.4 versus 0.1 Sv) with little effect on the IPT and strengthens the EAUC (8.5 versus 3.8 Sv) and the ECC (15.9 versus 11.7 Sv). The quasi-stationary eddies corresponding to the Wairarapa Eddy and the NCE are evident, but in the mean the ECE and the low have merged. Uddstrom and Oien (1999) were also unable to detect a robust ECE in their SST fields, indicating a high amount of variability in this region. The closer agreement of the simulated meanders with observations suggests stronger upper-ocean–topographic coupling in RB16 than in RB8.

Preliminary results from a further increase in resolution to 1/32° (Fig. 13) shows very little difference between the 1/16° and 1/32° simulations in the vicinity of the Tasman Front. This simulation was initialized from simulation RB16 and extended 8 years at 1/32° resolution and, hence, is not completely equilibrated at this resolution but is sufficiently equilibrated to show the impact of the increase in resolution on flow instabilities and therefore the mean meanders within the Tasman Front. The agreement between the two simulations in the Tasman Front suggests that the simulation has reached a convergent solution, and the flow features are not functions of resolution. The disagreement between the two simulations in the region north of New Zealand (Figs. 12a and 13) should be expected since within this region both simulations exhibit a large degree of variability and should not be considered inconsistent with the convergence of the simulations.

The simulated sea surface height variability of RB16 qualitatively agrees with TOPEX/Poseidon altimeter data (Fig. 14). The peak values of variability in the EAC and New Zealand region (the separation point of the EAC, along the Tasman Front, and directly north and east of North Island) are colocated in the simulations and the observations and can be directly linked to physical phenomena. The large variability in the region of the EAC separation is due to both the seasonal migration of the separation point and eddy shedding, while the large amount of variability along the Tasman Front indicates significant movement of the meanders. The large variability north and east of North Island can be linked to the presence of the quasi-stationary eddies and meandering of the EAUC and ECC.

Since the model is hydrodynamic and forced by monthly a wind stress climatology and the actual ocean is subject to seasonal thermal forcing and high-frequency winds, the overall variability of the model is less than that measured by TOPEX/Poseidon. Lower modeled SSH variability in locations such as east of Tasmania and within the southern section of our region of interest is expected. However, while the simulated variability north of New Zealand agrees qualitatively with observations, it is significantly higher than observations. The disagreement between the model and observations can be linked to the model's inability to adequately resolve the transport south of New Zealand and Australia. As discussed in section 3b, the reduction of flow south of New Zealand due to the model's insufficient vertical resolution in that region forces northward flow through the Tasman Sea, resulting in an unrealistically large current associated with the Tasman Front and therefore the EAUC. Instabilities and meanders within these unrealistically large currents result in larger than observed SSH variability as seen in Fig. 14. Examination of the SSH variability of the lower-resolution RB8a (not shown) reveals less variability throughout our region of interest and much less variability north and east of New Zealand. The decreased variability north and east of New Zealand in RB8a is consistent with our explanation of the origin of the SSH variability, since RB8a contains much smaller transports within the Tasman Front, EAC, and ECC. The location of the observed ECE is a region of extremely high variability in the model and observations, providing a possible explanation for the mixing of the low and the ECE in RB16.

Maps of surface and abyssal eddy kinetic energy (EKE) from RB8 and RB16 (Fig. 15) demonstrate that the energy in both the surface and abyssal layers increases with this increase in grid resolution. The increased resolution in RB16 allows greater flow instabilities and an increased rate of energy transfer to the abyssal layers via these instabilities. Hogan and Hurlburt (2000) provide a clear discussion and illustration of the vertical transfer of energy. A mean value of R_b for eddies produced by RB16 is 6.9, which is slightly lower than FB, suggesting that the eddies' formation may be due to mixed barotropic–baroclinic instabilities. In fact, latitude–time plots of the meridional derivative of the barotropic absolute vorticity (dQ/dy) at 165°E (Fig. 16) show that the derivative changes sign frequently within our domain, which is a necessary (but not sufficient) condition for barotropic instability and suggests that barotropic instabilities are probable in this region.

Although mixed barotropic–baroclinic instabilities are very efficient mechanisms for transferring energy from the surface layers to the abyssal layers, other mechanisms may also drive abyssal EKE, such as vertical mixing or direct barotropic forcing by temporal variations in the wind stress. Very little vertical mixing between the surface and abyssal layers occurs in the area of the Tasman Front and New Zealand, although a significant amount does occur south of New Zealand. The collocation of EKE maxima in the surface and abyssal layers combined with the increase in variability with resolution is clear evidence that instabilities are present (Holland and Lin 1975) and that they, and not a barotropic response to the seasonal wind variations, are the major source of abyssal EKE (Hurlburt and Metzger 1998). The increase in baroclinic instability in RB16 over RB8 reduces the depth of the isopycnal trough north of the ACC as a result of the increased transfer between the potential energy of the mean flow to the eddy potential energy and consequent isopycnal flattening (reduced range of temporal mean isopycnal depths in the region). This flattening reduces the mean isopycnal depths at the southern tips of Australia and New Zealand (Table 4), more so at the southern tip of New Zealand, which extends farther south (Fig. 2). In all cases the isopycnals outcrop at Antarctica (reach the minimum depth allowed in the model). Since all other factors are invariant in RB8 and RB16, the relative magnitude of the baroclinic transport through sections is determined by the isopycnal depth ranges across the sections. Thus, the results in Table 4 show greater restriction on the baroclinic ACC transport south of New Zealand in RB16 and therefore increased northward transport through the Tasman Sea in accord with Table 2.

The link between abyssal energy and upper-ocean–topographic coupling can be clearly seen. Large amounts of abyssal energy are collocated with bottom steering of upper-ocean currents such as the meanders in the Tasman Front (Figs. 11, 12, and 15). RB16 with a larger amount of abyssal energy and, therefore, more energetic abyssal currents, produces more bottom steering and larger meanders than RB8. The increased variability and barotropic–baroclinic instabilities enable more upper-ocean–topographic coupling in the higher-resolution simulation. For this reason, we use these simulations to investigate in detail the role of several topographic features in steering the surface flow field. Two separate areas merit discussion: the meanders of the Tasman Front and the apparent bottom steering of the ECC over the Hikurangi Trench. As seen in Fig. 12, many surface features appear to be coupled with particular topographic features. Topographically driven abyssal currents intersecting with surface currents at large angles force the surface currents to be displaced in the direction of the abyssal flow where the flows overlay (Hurlburt and Thompson 1980, 1982, 1984). The surface currents tend to return to the same latitude after the perturbation by the abyssal currents, consistent with potential vorticity conservation. Abyssal currents superimposed onto the sea surface height contours (Fig. 17) show that all meanders, with the exception of the first, of the Tasman Front are associated with meridional abyssal flow on the side of a ridge or a trough. The first meander is consistent with the separation overshoot and retroflection described by Ou and De Ruijter (1987). Two meanders discussed by Stanton (1979), the large southward meander located at the New Caledonia Trough (166°E, meander F) and the northward meander over Norfolk Ridge (167°E, meander G) are both driven by meridional flows along ridges. Abyssal currents also drive meanders documented by Mulhearn (1987) that are associated with the Dampier Ridge (159°E, meander B) and Lord Howe Rise (161°E, meander E). While the directions of meanders A, C, E, F, and G might be expected from conservation of potential vorticity of a barotropic flow, Stanton (1979) clearly shows that the flow associated with the Tasman Front is highly baroclinic. Also, the southward meanders located directly east of Dampier Ridge (meander B) and on the western side of the Lord Howe Rise (meander D) are opposite in direction to those predicted by conservation of potential vorticity. However, the corresponding simulated meanders are directly above southward abyssal flow along the ridge, which forces the surface flow to form southward meanders.

At first glance, the flow of the ECC appears to be governed by the Hikurangi Trough. It separates very near to the location of the Chatham Rise and apparently follows this ridge line. In fact, the separation point in the realistic bottom topography simulations is the same as the flat-bottom model. As mentioned in section 3a, the separation is governed by the integrated wind stress curl and not bottom topography.

From examination of the simulations we are able to determine the governing dynamics behind several features of the flow field surrounding New Zealand. Upper-ocean–topographic coupling via baroclinic instabilities governs the phase and amplitude of the mean meanders in the Tasman Front. The flat-bottom nonlinear model and even the linear model with the added friction patch produce the correct directions and separation points for all examined currents. The shape of the wind stress curl and not bottom topography determines the separation points of these currents.