1654 JOURNAL OF PHYSICAL OCEANOGRAPHY VOLUME 44

Residual Flows in , a Large Tidally Dominated Strait

CRAIG STEVENS Marine Physics Group, National Institute for Water and Atmospheric Research, , and Department Physics, University of Auckland, Auckland,

(Manuscript received 20 February 2013, in final form 19 February 2014)

ABSTRACT

Flows at subtidal frequencies are resolved from 20 months of current observations at multiple sites across 2 the narrows of Cook Strait, New Zealand. The substantial tidal flows (spring .3ms 1) are challenging to 2 measure with moored ADCPs. Along-strait residual flows are on the order of 0.05–0.1 m s 1, but there were differences across the strait and also through the annual cycle. The calculated net residual transport is 20.25 2 Sverdrups (Sv; 1 Sv [ 106 m3 s 1) (i.e., negative sign implies to the south, from to Pacific Ocean). This is smaller (by a factor of 2–3) than that estimated from previous sparse current measurements. Crucially, however, these new data also suggest that, over the austral summer–autumn period, there is persistent re- sidual flow northward on the western side of the strait so that the summer–autumn net 20.15-Sv flow can be separated into 10.065-Sv westside and 20. 22-Sv eastside flow. This western northward flow had a seasonal bias, appearing only in the summer–autumn period. The observations also show persistent temperature stratification on the order of 28C over the full depth of the water column in this late summer–autumn period. An implication is that there is vertical shear influencing the residual flow estimates. A range of exchange flow scenarios are considered with the most extreme indicating that the flow on the western side of the strait in the January–June period might be as much as 10.08 Sv northward. This has substantial implications for regional circulation as well as for the nutrient supply for the nearby .

1. Introduction the eastward transport around the north [9 Sverdrups 2 (Sv; 1 Sv [ 106 m3 s 1)] and south (8 Sv) bounds of the Subtidal flows dominate the transport of materials main islands of New Zealand have been estimated that influence large-scale processes like ecosystem pro- (Roemmich and Sutton 1998; Sutton 2003), there is duction and climate variability. The challenge is to re- a third pathway for exchange. Cook Strait, the channel solve these modest residual flow speeds from often much separating New Zealand’s North and South Islands, larger tidal flows. This can be especially difficult when connects the eastern Tasman Sea (TS) to the western coastal topography introduces variability in flow struc- Pacific at 428S. At its narrowest point it is 22 km across, ture that can be asymmetric, causing rectification of tidal with 210- and 350-m average and maximum depths, re- transport (Loder 1980). In addition, wind forcing pro- spectively (Fig. 1). Its fast-flowing tidal currents have vides an episodic, and likely seasonal, driver of exchange been the focus of a number of studies, amongst which (Knutsen et al. 2005; Rossby et al. 2005). Finally, baro- the most notable conclusion was that the dominating clinicity and rotation can lead to complex and variable semidiurnal is around 1408 out of phase at the op- exchange flows (e.g., Umlauf and Arneborg 2009; Ott posite ends (north vs south) of the strait (Heath 1978; and Garrett 1998). Vennell 1994; Stanton et al. 2001). This phase difference New Zealand bisects the subtropical convergence 2 drives substantial flows, reaching as high as 3.4 m s 1 zone with warm waters arriving from the north via the during spring tides (Vennell and Collins 1991; Stevens East Australia Current and cooler water from the south et al. 2012). separating from the Antarctic Circumpolar Current. While The strait is nominally around 40 km ‘‘long’’ (Vennell 1998) so that even at the maximum flow speed described above, a parcel of fluid travels around 44 km in half Corresponding author address: Craig Stevens, Marine Physics Group, National Institute for Water and Atmospheric Research, a semidiurnal tidal cycle and so is only just exported Greta Point, 301 Evans Bay, Kilbirnie, Wellington 6021, New Zealand. through the strait in a single tidal period. Despite this, E-mail: [email protected] there has been little focus on residual currents and net

DOI: 10.1175/JPO-D-13-041.1

Ó 2014 American Meteorological Society Unauthenticated | Downloaded 09/27/21 02:11 AM UTC JUNE 2014 S T E V E N S 1655

FIG. 1. (a) New Zealand and the TS and Southern Ocean (SO) and regional circulation including the East Auckland Current (EAucC) , DC, SC, as well as the WE. (b) Cook Strait narrows is bounded by (CT) to the east and the headlands of the (shaded) MS to the west. The 200- and 400-m depth contours are marked, as well as the shoal at FR. Meteorological data come from the BI and MI. The locations of the four moorings (E, W, FR, and C) are marked. The dashed line is the power cable exclusion zone (CEZ). (c) The depth profile across the narrows with measurement locations marked and the proportional cross-sectional area as a function of depth shown as an inset. The cross section is divided into areas as discussed in the text. transport in the strait, an exception being the study of Sea is relatively quiescent in terms of circulation currents Heath (1986) using rotor-based current meters. (Sutton et al. 2005) that form the oceanic boundary for Transport through Cook Strait plays a role in influ- the shelf seas of the west coast of the . By encing upper water column ocean circulation in the most accounts (Heath 1971; Chiswell and Stevens 2010), central New Zealand region. The eastern central Tasman a northward coastal flow along the west coast of the

Unauthenticated | Downloaded 09/27/21 02:11 AM UTC 1656 JOURNAL OF PHYSICAL OCEANOGRAPHY VOLUME 44

TABLE 1. Residual speed and volume transport estimates from previous work where negative implies southward transport. The volumetric transport estimates were made here and not by the original authors (except where noted by an asterisk).

2 Study Speed (m s 1) Flow (Sv ) Notes Heath (1971) From 20.12 to 20.25 From 20.6 to 21.2 Speed identified using geostrophic balance Bowman et al. (1980) 20.3 21.2 Residual from numerical tidal model Heath (1986) From 20.01 to 20.03 From 20.04 to 20.14 60-day rotor-based current-meter deployment Vennell (1994) From 20.25 (west) to ;0 (east) 20.6 12-h ADCP survey Chiswell (2000) 20.14 20.8* Inferred from conservation of mass Walters et al. (2010) n/a 20.61* Baroclinic numerical model northern half of the South Island (Stanton and Moore effects. Crucially, from the present paper’s perspective, 1992) forms the upstream supply for Cook Strait. This Heath (1986) noted an east–west difference in direction, flow evolves into the d’Urville Current (DC) that moves whereby he found the water in the western zone of the 2 northeast around Farewell Spit and then across to the narrows moving southward (3 cm s 1), but the current west coast of the southern tip of the where it on the eastern side of the strait was found to be moving 2 turns southward (Bowman et al. 1983a; Chiswell and northward (5–15 cm s 1)—as will be seen, this is the Stevens 2010). On the central east coast of New Zealand, opposite to that found in the present work. transport is mainly associated with Eddy Vennell and Collins (1991) deployed month-long (WE) currents containing mainly subtropical water, plus acoustic Doppler current profiler (ADCP) moorings, some sporadic influence from the Southland Current (SC; concentrated to the east of the strait, but their analysis Chiswell 2000). Exchange between the three systems high-pass filtered the currents to focus on tides. Sub- (west coast, east coast–south, and east coast–north, where sequently, Vennell (1994) used vessel-based ADCP data 2 north and south refer to a dividing line at the strait latitude) to estimate residual flows ranging up to 25 cm s 1 to the affects a number of ecosystems and industries including the south and with an east–west bias (larger southward flow economically important Marlborough Sounds (MS; aqua- on the western side of the strait). Chiswell (2000) used culture, forestry, and tourism) and the canyon a mass balance approach to suggest the net residual 2 region (fisheries and tourism). flow (to the south) through the strait is 14 cm s 1 Olsson (1955) and Gilmour (1960) presented esti- (20.8 Sv). This latter value is comparable to the south- mates of flow in the strait using a combination of ap- ward 20.61 Sv estimated for zero wind conditions by proaches including cross-strait cable electrical potential, Walters et al. (2010) in a modeling study that included aerial tracking of surface dye, ship measurements, and Tasman–Pacific sea level differences and regional baro- bottom-mounted current meters. Unfortunately, the clinic gradients. 5-month monitoring of cable electrical potential differ- The wide range of conclusions based on often dis- ence indicated that the technique, while able to capture parate short-duration observations motivated a ;2-yr the tides, was not sensitive enough to measure flow re- ADCP deployment, providing the longest (by a factor of sidual. ;4) flow record in the strait. The objective here is to Previous estimates of residual flow in the strait are examine residual flows through the strait using this re- summarized in Table 1. Heath (1971) used geostrophy to cently acquired current data. This suggests several argue that the strait sustained a southward residual flow questions: (i) what is the depth-averaged residual flow 2 speed somewhere between 12.5 and 25 cm s 1. With through the strait and (ii) is there evidence of exchange a cross-sectional area at the narrows of ;4.7 3 106 m2 flow? Also what are the (iii) drivers of the subtidal re- (see next section), this implies (ignoring boundary layer sidual flows and (iv) the wider implications for regional effects) a volume flow in the range from 20.6 to 21.2 Sv. ecosystem behavior of this transport? Here, the convention used is to denote a southward oceanic flow with a negative sign. Bowman et al. (1980) used a tidal model to infer residuals and estimated 2. Methods 2 ;30 cm s 1 to the south (21.2 Sv), whereas Heath (1986) a. Site used analytical arguments to suggest it was more like 2 1cms 1 (20.04 Sv) but still to the south. Heath (1986) also The strait is described as having a length of 40 km presented a current-meter time series over periods of 30– (Vennell 1998). Water depths are around 250–300 m in 82 days and highlighted the importance of the averaging the main part of the channel, with shoals to the south and time scale. As will be seen, these early rotor current-meter the submerged Fishermans Rock (FR; a pinnacle) rising moorings likely suffered from current-induced sensor tilt to within 10 m of the surface to the northwest of the

Unauthenticated | Downloaded 09/27/21 02:11 AM UTC JUNE 2014 S T E V E N S 1657

TABLE 2. Mooring details where ‘‘C’’ and ‘‘T’’ indicate conductivity and temperature instruments, respectively.

Mooring Instruments Water depth (m) Lat Lon Start date Duration days Center SeaGuard, C and T 290 41815.0570S 174827.6250E 11 Dec 2008 50 East-1 75 kHz, C and T 204 41814.6340S 174833.6390E 25 Aug 2010 148 West-1 75 kHz, C and T 185 41812.5150S 174825.6100E 25 Aug 2010 148 East-2 75 kHz, C and T 210 41814.6180S 174833.6420E 12 Mar 2011 486 West-2 75 kHz, C and T 185 41812.5130S 174825.6020E 12 Mar 2011 486 FRock-2 75 kHz, C and T 276 41805.7810S 174835.0440E 12 Mar 2011 486

strait narrows (Fig. 1). The average depth of the marked eastern station (i.e., faster), the tilt exceeded 108 in fewer transect (Fig. 1b) is 219 m with a total cross-sectional than 8% of the samples and exceeded 158 in fewer than 2 area of 4.7 3 106 m2 (Fig. 1c). The cross section was di- 1% of the samples. A slight current (0.3 m s 1)wasre- vided into quadrants with separation based on the deepest quired to get zero tilt; however, the accuracy of the tilt point for horizontal and around midmaximum depth for sensor is 0.58, comparable to the zero flow offset. The the vertical (Fig. 1c). Quadrant designations are: west, manufacturer algorithm cannot remove bias from tilts upper (wu); west, lower (wl); east, upper (eu); and east, greater than 158, and so these samples have been removed lower (el). These quadrants have areas of (Awu, Aeu, Awl, from the dataset resulting in a slight bias to the data (i.e., 6 2 Ael) 5 (1.95, 1.5, 0.7, 0.5) 3 10 m . the fastest 1% of data are not included, which makes around a 5% reduction in the absolute flow estimate). If b. Instrumented moorings there is only a small bias in the directionality, this influence The primary data come from two stations with instru- may be further reduced in the time-average calculations. mented moorings deployed at locations one-third and The tidal signal dominates the flow most of the time, two-thirds of the way across the narrows [east (E), west creating challenges in terms of quantifying the residual. (W)] for a period spanning 2 yr, in two deployments, The temporal coverage of the data is good but there are starting in August 2010, although there was a 3-month gap after 5 months (Table 2). Coincidentally, the second deployment commenced around 4 h after the local arrival time of the 2011 Japanese Tsunami—so we saw no clear evidence of its passage in the data. In addition, this second phase included a third mooring station located to the north of the narrows for the last 15 months of the period (FR). The currents at this site are heavily biased by to- pography for the present purposes, and only the collo- cated scalar data will be described. Each mooring contained an upward-looking Teledyne RD Instruments 75-kHz ADCP mounted in a Flotation Technologies syntactic foam float and moored with 600 kg of iron and 10 m of chain. While a frame-based bed mounting would have given better coverage, this was not possible because of difficulties posed by small-scale topo- graphic variations. The ADCPs logged at 10-min intervals, sampling into 8-m depth bins. The presence of suspended sediment and strong signals meant that there were very few velocity data lost to quality control issues except for tilt effects described in the next section. Each float con- tained a Sea-Bird Electronics (SBE) MicroCAT (SBE 39) conductivity–temperature–depth sensor placed beneath the ADCPs. The MicroCATs sampled at 5-min intervals. Moored instruments can suffer ‘‘knock down’’ effects FIG. 2. ADCP tilt from the eastern ADCP as a function of the flow magnitude in the lowest bin (i.e., ;20 m above the seabed). whereby the instruments tilt in strong currents. The ve- The inset shows the log scale of the normalized cumulative distri- locity magnitude in the deepest ADCP bin reached nearly bution of tilt samples as a function of tilt, so that less than 1% of 2 2ms 1 (Fig. 2). In this example from the mooring at the samples exceed the 15% tilt.

Unauthenticated | Downloaded 09/27/21 02:11 AM UTC 1658 JOURNAL OF PHYSICAL OCEANOGRAPHY VOLUME 44

FIG. 3. (a) A 24-h section of north–south aligned velocity structure at the eastern mooring and (b) example ‘‘flood’’ velocity profile from time (DOY) 279.275 [vertical dashed line in (a)] with extrapolated regions included. The lower horizontal dashed line shows the limit of data, and the horizontal blue bar shows the upper averaging region used. The dashed–dotted profiles show the nearest before and after profiles. gaps in the vertical coverage. The benthic boundary layer east, respectively, Fig. 1; data available at http://cliflo. region is partly lost due to the instrument blanking dis- niwa.co.nz). Convention is followed whereby wind di- tance and the mooring arrangement whereby the bottom rections are ‘‘from’’ and ocean current directions are ;20 m are not sampled. This can be replaced with ‘‘to.’’ Two different sea surface temperature (SST) prod- a boundary layer model assuming a logarithmic distri- ucts are used. The National Oceanic and Atmospheric bution and a drag coefficient (Fig. 3). The details of this Administration (NOAA) 1/48 product (Reynolds et al. extrapolation do not particularly influence the transport 2007) is used for long time series, and 1-km imagery is used results, as they are well constrained by no slip at the bed to provide spatial snapshots (Uddstrom and Oien 1999). and have generally lower magnitudes. More importantly, the top 20% of the water column is lost due to side-lobe 3. Results interference exacerbated by the relatively large tilts. Two assumptions are possible: (i) constant flow in the upper a. Wind region using an average of the uppermost few reliable At least locally to the Cook Strait narrows, the wind is data bins or (ii) extrapolate based on slope between the considered an important driver of transport (e.g., Zeldis uppermost few data bins. The former, more conservative et al. 2013). The wind speeds on either side of the strait are approach was used here as the slope approach proved comparable during northwesterly (NW) winds. However, prone to generating unrealistic surface currents (Fig. 3). the Brothers (western side) wind speeds are often double Another issue with the estimation of the residual is the that of Mana Island during southerly (i.e., from the south) lack of coverage in the center of the strait, especially at winds. This occurs because this latter station is not located depths beneath the sampling stations (i.e., .200 m). within the narrowest region and is also set back from the A 50-day deployment of an Aanderaa SeaGuard single- likely fastest wind flow topographically steered through point current meter (‘‘center’’; Table 2) from December the strait. Consequently, the Mana Island station to the 2008 to February 2009, sampled at 4-min intervals, is east is dominated by NW winds, whereas the western examined to augment this. Brothers station has an even mix of southerlies and northerlies while the directions are spread over 458 (Fig. 4). c. Meteorological and SST data Winds can rapidly switch directions at both stations. For Meteorological data from either side of the strait example, in Fig. 4c, at around day of year (DOY) 2011 63.5 2 come from the Brothers Islands (BI) and Mana Island the wind switches from a 10 m s 1 northerly (i.e., from the 2 (MI) automatic weather station (AWS) (in the west and north) to 10–20 m s 1 southerly in a matter of an hour.

Unauthenticated | Downloaded 09/27/21 02:11 AM UTC JUNE 2014 S T E V E N S 1659

21 FIG. 4. Wind rose for (a) BI and (b) MI AWS, where color is the wind speed (m s ) and the circles represent percentage (of the direction of wind origin). (c) The wind component along the strait axis (3458 true north), where positive is from the north. (d) Events are derived from this using the low pass–filtered (1-day filter constant) north–south wind 2 component was searched for events contiguously above a threshold (here 5 m s 1) for a minimum period (here 2 days).

To place the wind forcing for the present dataset the number of events fell within the standard deviation. into some long-term context, seasonal averages of wind In no season did the events from both sampled years fall events were collated over a 27-yr period. The wind outside the standard deviation. events were determined by looking in the low pass– b. ADCP data and flow residuals filtered north–south component of wind for events above a given threshold that lasted for at least a selected Scatter diagrams for the vertically averaged in- 2 period (Fig. 4d). The events were separated into north- stantaneous currents show flows extend to over 2 m s 1 erly and southerly wind events. With the subjective pa- with a strong north–south bias in orientation (Fig. 5). 2 rameters used (1-day low-pass filter; 5 m s 1 speed It also shows a slight bias in the second deployment on threshold; 2-day minimum event period) there was an the western side, presumably due to a slightly different average of five events per season (Table 3). In 12 of the mooring location. Considering the average of the in- total of 16 seasons experienced by the pair of moorings, stantaneous vertical shear at the two stations, separated

Unauthenticated | Downloaded 09/27/21 02:11 AM UTC 1660 JOURNAL OF PHYSICAL OCEANOGRAPHY VOLUME 44

TABLE 3. Number of wind events (northerly 5 southward wind) (hence the mainly positive values in Fig. 6b). The two collated into 3-month seasonal bins based on BI wind data from eastern side records are very similar. The same cannot 1987 to the end of 2012. Here, an ‘‘event’’ is defined as wind con- 2 be said for the western side. While the slope is similar, it stantly exceeding 5 m s 1 for 2 days where the wind speed has a 1-day low-pass filter applied. The rows include the numbers for appears there can be differences in the depth-averaged the 2 yr of observation 2011 and 2012 and the average (std dev in flow. A summer-averaged profile illustrates the north- parentheses) for that season for the entire period. ward current structure that is faster near the bed. Somewhere in the missing section of the profile there JFM AMJ JAS OND must be a turning point to be consistent with a boundary Northerly 2011 5 4 3 2 layer. These velocity profiles do support the notion that, Northerly 2012 6 8 5 2 Northerly 1987–2012 4.8 (2.3) 4.5 (2.1) 4.4 (2.2) 4.4 (2.5) apart from the along-strait eastern side, the missing Southerly 2011 4 3 2 4 profile end zones are at least not the major components Southerly 2012 6 5 6 6 of the residual flow. Southerly 1987–2012 5.0 (1.9) 3.6 (1.8) 3.8 (2.0) 5.0 (2.4) Considering the across-strait flows (Fig. 7b), the maximum speed (westward) lies at around 120–130-m depth (although the peak across-strait speed for the based on depth-averaged direction (Fig. 6), the western western side in the second deployment is at 100 m, with station is almost symmetric with the southward flows the speed distribution being broader than the other av- having a slight bias to a positive shear (i.e., faster at the eraged profiles). Also, a switch is seen from the first surface). The eastern station (Fig. 6b) is quite different deployment where the eastern station speed is around with clear directional asymmetry. The distribution of double that of the western side to the second deploy- the slope of southward flows has a higher kurtosis and ment where there is only around 15% difference in the mostly positive slopes. The northward slope distribution peak speeds. Given the dominant flow in the strait is on the other hand is much broader in spread with a near- southward, the westward-biased cross-strait flow sug- zero average. gests rotation is not a controlling factor in this case. The The time-averaged along-strait velocity profiles for summer profile across-strait flow is consistent with the each deployment and location (Fig. 7a) are modestly other periods. consistent in form, with a southward bias increasing with It is useful to compare the various low pass–filtered height above the bed. The eastern side is dominated by time series (filter constants ;48 h), including the wind, southward flow with the uppermost measured speed currents, tidal phase, and scalar quantities (temperature around 4 times that of the lowermost measured speed and salinity). The north–south components of wind and

FIG. 5. Scatter diagrams for vertically averaged instantaneous flows from the (a) western and (b) eastern moorings, where the lighter symbols are for the second deployment. Angled reference lines are included for along-strait de- composition.

Unauthenticated | Downloaded 09/27/21 02:11 AM UTC JUNE 2014 S T E V E N S 1661

FIG. 6. Distributions of velocity shear for (a) western and (b) eastern stations, separated based on instantaneous averaged direction, where a positive slope is speed increasing with height above bed.

currents, in combination with the Brothers (western) The difference in the behavior on either side of the wind data show that, despite the filtering, wind speeds strait is made clear in the comparison of the two moor- 2 still exceed 20 m s 1 with the majority of the stronger ings made using 2-day aggregates of the 48-h filtered events being from the south (Fig. 8). The low pass– results (Fig. 10a). Few of the data points fall on or near 2 filtered currents reach nearly 0.6 m s 1, around 20% of the 1:1 line. The residual component on the eastern side the maximum tidal speed. Broadly, the current time se- is rarely greater than zero, whereas a reasonable pro- ries is dominated by the southerly storms and the western portion of the western mooring samples (33%) flow side response. The effect of the wind field (Fig. 8a) northward. Furthermore, the eastern residual is typi- manifests itself more clearly at the western station, al- cally larger in magnitude than the western side. Aggre- though this may be influenced by using wind data from gating the wind into the same time bins as the flow that side of the strait. This is especially noticeable in the residuals makes it clear that southerly (i.e., negative) large event starting around day 283 and persisting for winds drive the majority of the northward flows on 2 3–4 days where winds reached 25 m s 1 (Fig. 9). This the western side of the strait (Fig. 10b). It also dem- generated a substantial northward flow on the western onstrates how the currents are aligned with the to- side and was even sufficient to counter the large depth- pography. Because of the contraction, it is likely that averaged southward flow at the eastern station. This sit- wind forcing generates a barotropic pressure gradient at uation rapidly changed however, as the southerly wind a scale greater than the narrows and that this dominates event was followed by a northerly wind that generated any wind-driven surface boundary layer. Furthermore, 2 a southward flow residual of 0.5 m s 1. In the cases of the this surface boundary layer may be difficult to observe in southerly wind (day ;283) at the western station and the data collected here because of the lack of near- the northerly wind (day ;288) at the eastern station, the surface data. wind-driven flow was sufficient to counter the tide so that It has been hypothesized that tidal rectification plays it would have appeared to have not ‘‘turned.’’ a role in driving the flow residual (Walters et al. 2010). While the currents on the western side provide the The expanded time series (Fig. 9) are visually suggestive strongest transient peaks, in the record average the of some connection between austral spring–neap tidal eastern mooring is the stronger of the two with an av- variations and residual flow. However, again using 2-day 2 erage over the entire time series of 10 cm s 1 flowing to aggregates, comparing the residuals with the spring– the south, whereas the western station data average was neap envelope did not resolve any trend (Fig. 10c), so 2 close to zero with 0.2 cm s 1 flowing to the south (Figs. 7, that the rectification process does not appear to manifest 8). This transverse variability is the reverse of that sug- itself in a way that is captured by the moorings. This is gested by Heath (1986). location dependent because the region of rectified flow

Unauthenticated | Downloaded 09/27/21 02:11 AM UTC 1662 JOURNAL OF PHYSICAL OCEANOGRAPHY VOLUME 44

2 these averages are 211.0 and 12.4 cm s 1 for eastern and western stations. The standard deviations shown in Fig. 11 are associated with a 30-day low-pass filtering, and while their distribution is helpful as a guide, the de- viation magnitude is largely a function of the filtering time constant. Figure 11 also includes the individual quar- terly residual speed estimates for each of the two sam- pled years. Two of the western side estimates pairs fall on top of each other [April–June (AMJ) and October– December (OND)], as do the OND estimates for the eastern side. Importantly, this also means the northward summer–autumn trend seen in the west is seen in both sampling years. The switch in net direction on the western side in winter–spring (JAS–OND) is balanced by a reduction in the southward residual on the eastern side. The north- ward flow at the western station is substantial during the summer (JFM), reaching a magnitude of nearly half that of the southward flow at the eastern station. While only 50 days in duration, the point-measurement SeaGuard current meter for 2008/09 at 284-m depth in the center of the strait (Fig. 12) provides a complemen- tary story to that observed at the main sampling sites. Although near the bed and thus well into the boundary layer at most times, it captured instantaneous flow speeds 2 almost reaching 1 m s 1. Filtering the signal in the same way as for Fig. 8 results in a consistent northward flow of 2 around 0.14 m s 1 (Fig. 12a) that is in keeping with the western station flow direction but not magnitude. The flow at this station did not appear to be as closely related to the local wind driver as seen at the other stations (Fig. 9). This could suggest wind-driven shear or some baroclinic driver plays a role in the deeper region of the strait. The frequency structure of the vertically averaged current data (i.e., no low-pass filter; Fig. 13) yields a key difference between the sides of the strait. At diurnal

frequencies (K1; Stanton et al. 2001) and higher, the FIG. 7. Time-averaged (a) along- and (b) across-strait velocity western side of the strait contains moderately weaker profiles for each of the deployments showing the measured sections flows compared to the east. For example, the semi- only. The horizontal bars are the local bed depths at the mooring locations. A western side summer (JFM) profile is also included for diurnal peak is larger to the east, as are its harmonics comparison. (Fig. 13b). However, at subtidal frequencies there is considerably more energy in the west. The variability is typically constrained to an eddy scale related to the associated with the peak in the western station residual tidal excursion. at around 0.1 cycle per day (cpd) can be readily seen in Seasonal (quarterly) averages of the east and west the expanded residual time series (Fig. 9). This supports residuals (Fig. 11) show that the northward flow on the the observation that the synoptic 5–10-day period is the western side is concentrated in the summer–autumn dominant source of variability in the current records. period (January–June inclusive). It must be noted that The wind forcing has a broad peak in the weather band— there are biases in these data due to the sampling gap in nominally marked here at 5–10 days. The coherence 2 the January–March (JFM) and July–September (JAS) spectrum is given jSxyj /(SxxSyy), where Sxx, Syy,andSxy aggregates. The complete time series averages are 210.0 are the spectra for wind, north–south velocity, and the 2 and 20.2 cm s 1 for eastern and western stations, respec- cross spectrum, respectively (Emery and Thomson 2001). tively. During the combined summer and autumn periods, This shows a degree of covariance at low frequencies but

Unauthenticated | Downloaded 09/27/21 02:11 AM UTC JUNE 2014 S T E V E N S 1663

FIG. 8. Mooring deployment 1: (a) NS wind and (b) corrected depth-averaged north–south velocity from the western and eastern moorings. (c) Temperature time series that include data from FR and SST [(b) and (c) include spring–neap height envelope (arbitrary scale) from the center of the narrows at bottom]. (d) Salinity time series including FR data. nothing substantial around the 5-day band (Fig. 13c), al- value by May 2012. There is no evidence of this effect though there is a peak at 0.07 cpd. a year earlier, although this is not certain because of the gap in the mooring deployments. The Fishermans Rock c. Temperature and salinity time series data to the north of the central strait region tracks the Temperature and salinity from the near-bed moorings western side data both for temperature and salinity. range over 108–158C and 34–35 psu, respectively (Figs. Fouling of the eastern side conductivity sensor cannot be 8c,d). Temperature is consistent across the strait (Fig. 8c). ruled out, but it is unlikely as it returns to a value con- However, the cross-strait variation in salinity is not as sistent with the other side of the strait and the departure consistent. During the period from Nov 2011 (DOY 675) roughly coincides with a change in the difference be- to May 2012 (DOY 840), a ;0.5 psu drop on the eastern tween surface and near-bed temperatures (DOY 675; side grows and decays, ultimately returning to the western Figs. 8c,d).

Unauthenticated | Downloaded 09/27/21 02:11 AM UTC 1664 JOURNAL OF PHYSICAL OCEANOGRAPHY VOLUME 44

FIG. 9. Comparison of wind and residual flow for a 100-day section of the record showing the along-strait wind (note shift in directionality from previous figure) and the corrected depth- averaged filtered north–south velocity from the western and eastern moorings. The unscaled spring–neap height envelope is included at bottom. d. SST imagery The satellite-derived SST data track the bottom temperature well during April–November, but the sur- face data rise to nearly 178C in the summer of both years of sampling representing around 28C of thermal strati- fication (Fig. 8c). A sequence of SST images from March to April 2009 (Fig. 14) captures many of the facets of the horizontal structure seen in the mooring data. Images from 3 to 9 April (Figs. 14d,e) probably best encapsulate the mean structure suggested by the residual quantities. The other three images, 23–27 March (Fig. 14a), 27–30 March (Fig. 14b), and 10–14 April (Fig. 14f), give the impression that the exchange through the strait is oc- curring mainly through the central portion of the nar- rows, with reduced flows in the inshore waters. It is not particularly apparent that there is a strong relationship between the structure in the images and the low pass– filtered wind forcing. This is, in part, because the SST distributions are influenced by tidal excursions. These snapshots, in conjunction with the SST–temperature bifurcation seen in Fig. 8c and the data from Bowman et al. (1983a,b), indicates that Cook Strait is stratified for the majority of the austral summer, contrary to Walters et al. (2010). As with the wind time series (Fig. 4), in order to fur- ther compare the period of the observations with the FIG. 10. The 2-day aggregated results showing (a) a comparison long-term average, the SSTs from the NOAA 1/48 dataset of the along-strait components from the west and east moorings (1982–2012) were collated into an annual average and (diagonal line is 1:1 comparison), (b) along-strait flow speed re- standard deviation (Fig. 15). The data from the sampling siduals for both moorings as a function of wind, and (c) flow speed residuals as a function of the spring–neap cycle. The spring–neap period fall mostly within plus or minus one standard cycle here is represented by the normalized 2-day average of the deviation. There does appear to be a small bias to cooler absolute tidal max amplitude.

Unauthenticated | Downloaded 09/27/21 02:11 AM UTC JUNE 2014 S T E V E N S 1665

FIG. 12. Wind and near-bed (6 m above bed in 290 m of water) currents at Cook Strait center for 2008/09 mooring.

The lack of coverage near the surface is likely an issue FIG. 11. Seasonal residual averages (thick lines). East and west mooring residuals are averaged over quarterly periods with in the summer months, when the persistent stratification northward flow being positive. The short bars within each quarter identified here potentially could enable enhanced ve- show the estimates for each of the 2 yr of the sampling. The shaded locity shear (Fig. 7). The stratification has similarities to 6 regions show the std dev of 30-day low pass–filtered signal. data collected in January (of 1980 and 1981) by Bowman et al. (1983a). The block-average approach, whereby the cross-sectional areas described in Fig. 1c are used conditions in the summer period (January–March) and to estimate the flow residual, would be challenged by early spring (October–December) relative to average highly localized shifts in the exchange flow, for example, temperatures. The autumn (AMJ) period was the only where there is a consistent tidal front. The east–west season that came close to departing substantially from 2-day comparison (Fig. 10a) suggests that there is this average picture with temperatures nearly 28C above modest correlation between sides (i.e., both sides going the mean. Based on this and the wind analysis, there in the same direction at the same time). is reasonable confidence that the sampled years were While the time series–averaged residual speeds in within the range of typical behavior as characterized by 2 the east of 10 cm s 1 flowing to the south, whereas the the 30-yr SST time series. 2 western mooring observed 0.2 cm s 1,itisusefultoalso consider total flow rates. However, quantifying the net 4. Discussion flow through the strait averages out any exchange as- pect of the flow. The various exchange scenarios can be a. The residual and its uncertainty examined by compartmentalizing the flow using the This sampling represents a substantial increase in the zones defined in Fig. 1c and considering a number of flow data available for Cook Strait. While there are flow regimes (Table 4). These include 1) separating uncertainties in the estimates, the measurements upon east and west flows and their sum, 2) the same but just which they are based are also the first to cover the an- for the January–June period, 3) assuming the flow nual cycle with simultaneous measurements across the through Ael is as per the western mooring, and 4) again narrows and also to have reasonable vertical coverage assuming the flow through Ael is as per the western without the effect of mooring knockdown. The lack of mooring but for the January–June period. The net coverage in the coastal inshore regions, the upper wa- volume flow then is 20.25 Sv as per the total for sce- ter column, and the deep central zone are the primary nario 1. Assuming the western and eastern halves of the contributors to uncertainty. Furthermore, interannual strait are independent, then this comprises two flows, variability may limit the generality of the conclusions. western side southward 20.05 and eastern side south- With regard to inshore regions this may be a possible ward 20.2 Sv, respectively (scenario 1). The eastern zone of enhanced flow, but the steep-sided coastal to- flow measurement is smaller than existing estimates of pography and comparison with previous inshore mea- transport in the d’Urville Current that is thought to be surements (Stevens et al. 2012; Heath 1986) do not from around 20.6 to 20.8 Sv (Chiswell 2000; Walters suggest any radical departure from the measurements in et al. 2010). Furthermore, given that there is an apparent the main moorings described here. seasonal variation, repeating this for January–June gives

Unauthenticated | Downloaded 09/27/21 02:11 AM UTC 1666 JOURNAL OF PHYSICAL OCEANOGRAPHY VOLUME 44

ADCP records, but still comparable or longer in duration than previous records. Consideration of the 2008/09 mooring data suggests it is correlated with the western mooring, in which case the west side flow is 20.062 Sv and the east side flow is 20.15 Sv (scenario 3). Repeating this again only for summer results in a 1 0.08- and 20.17-Sv west–east split. Other than the numerical analysis of Walters et al. (2010), most of the estimates of volume flow described in the introduction were based on published velocity esti- mates, and so the volume flows are not claims of the original authors. In many cases, the previous data were recorded at locations some moderate distance away from the narrows where the flow speeds are generally larger, and so the estimates should be considered lower bounds. This only serves to enhance the point that the present estimate of 20.25 Sv, along with the scenario estimates, are substantially smaller than previous esti- mates. Furthermore, the present results suggest in sum- mer the southward-flowing paradigm is reversed with as much as 10.09-Sv net flow northward being possible for a time. The 10-day peak in the current spectral energy on the western side (Fig. 13) is not apparent in the data of Chiswell and Schiel (2001), who examined coherence between wind and near-surface temperatures offshore of Kaikoura, around 150 km to the south of Cook Strait on the east coast of the South Island. They found no FIG. 13. (a) Variance-preserved power spectral densities for the coherence at the 10-day band. Although they do not north–south component of BI wind and the vertically padded and show the actual driving spectra, consideration of their averaged north–south ADCP data for the second deployment of time series data suggests that there is some energy at the east and west moorings (includes 95% confidence intervals). around 10–15 days in the thermal data. (b) The ratio of west to east power spectral density. Vertical lines mark various key frequencies including nominal 5–10-day weather b. Wider implications band (WB) and buoyancy frequency (cpd). (c) Spectral coherence between north–south components of wind and vertically integrated Rectification a priori seemed a strong candidate for along-strait velocities (includes 95% confidence level). driving flow through the strait, and it is certainly ap- parent in sufficiently resolved modeling approaches western side northward 10.065-Sv and eastern side (Popinet and Rickard 2007; Walters et al. 2010; Zeldis southward 20.22-Sv flows (scenario 2). This suggests et al. 2013). It may be responsible for the differences, a strengthening of exchange in the observed summer based on direction, seen in the slope of the instanta- periods, but a reduction to small (10.15 Sv) net flow. The neous velocity profiles (Fig. 6). However, based on cor- time series suggest that this northward bias at the western relation with the tidal spring–neap variations, there station might be related to strong southerly wind events was no apparent effect in the observations (Fig. 10c). (Fig. 9). As rectification is highly spatially variable, potential Further residual flow scenarios are possible by con- observations at a different site would exhibit rectifi- sidering the flow of the deeper water in the center of the cation effects. As it stands, the present observations strait that is poorly constrained by present data. The suggest a combination of local wind, and the drivers above estimates (scenarios 1 and 2) assume the deeper behind the d’Urville Current control the residual flows water follows that observed on the eastern and western in the system. Local winds will drive a Coriolis-influenced moorings. However, Fig. 12 shows that the deeper flow flow structure, whereby southerly winds will tend to en- does not follow the wind as seen with the shallower side hance flow along the western side of the strait. The op- stations (e.g., Fig. 9) The data record from the deep posite is true on the eastern side, and this flow is then central mooring is much shorter in duration than the incorporated into the regional-scale d’Urville Current.

Unauthenticated | Downloaded 09/27/21 02:11 AM UTC JUNE 2014 S T E V E N S 1667

FIG. 14. SST image sequence from 23 Mar to 14 Apr 2009: (a) 23–27 Mar, (b) 27–30 Mar, (c) 30 Mar–2 Apr, (d) 3–6 Apr, (e) 6–9 Apr, and (f) 10–14 Apr. (g) The north–south wind record from the BI station low-pass filtered with a 2-day filter window.

Locally, the modest northward flow for January–June main components of the Marlborough Sounds, so that on the western side of the strait is quite important. If, as the northward flow identified here may not penetrate as the present analysis suggests, there is a northward flow far as Pelorus Sound (the northwestern, and larger, of on the western side of the strait for reasonable periods, the two sounds). Using volumes based on estimates of nutrients driving ecological conditions in the Marl- area and average depth given by Heath (1974) and using borough Sounds might be more readily associated with the range of volumetric flows on the western side from the Wairarapa Eddy or even residual Southland Current above ;10.06, then filling time estimates for the Marl- waters (Bowman et al. 1983b) rather than with flow from borough Sounds are on the order of 2–4 days. It is ap- the west coast of the South Island, as is presently thought parent from Fig. 9 that this duration of flow event is (e.g., Harris 1990; Zeldis et al. 2013). Zeldis et al. (2013) common. observe that there appears to be a separation point be- The net flow estimate here of 20.25 Sv with the po- tween Pelorus and Queen Charlotte Sound, the two tential for a 10.065:20.15-Sv split between west and

Unauthenticated | Downloaded 09/27/21 02:11 AM UTC 1668 JOURNAL OF PHYSICAL OCEANOGRAPHY VOLUME 44

TABLE 4. Residual scenarios.

West East Total Scenario (Sv) (Sv) (Sv) 1 Split at center 20.052 20.2 20.25 2 Split at center 10.065 20.22 20.16 (JFM 1 AMJ) 3 All bottom water moving 20.062 20.15 20.21 as per west 4 All bottom water moving 10.07 20.17 10.09 as per west (JFM 1 AMJ)

between Tasmania and the Australian mainland, per- mits flows on the order of 0.5 Sv flowing eastward with much of the signal being associated with the arrival of atmospheric transients (Baines et al. 1991). Bass Strait then, in turn, drives coastal-trapped waves. Cahill et al. (1991) suggested similar connection for Cook Strait whereby coastal-trapped waves on the west coast of the South Island were the result of wind forcing within the strait itself. The present data support this in that the effect of the southerly wind events manifests itself in the northward (i.e., from east to west in a bulk sense) vol- FIG. 15. Annual cycle of SST near the strait center as a function ume transport. of day of year. All years 1982–2012 are shown in gray. Thick From an exchange perspective, Cook Strait contrasts line shows 5-day filtered daily average 6 one std dev. The in- dividual samples for the period of the moorings are included as dramatically from the baroclinically dominated Straits individual dots. of Gibraltar that experience an exchange flow (varying in the vertical) of around 0.8 Sv inwards into the Medi- east sides is, of course, substantially smaller than that terranean and around 25% smaller outflow (Tsimplis associated with the very large (in an instantaneous and Bryden 2000). The mixing that occurs in the hy- sense) tidal flow that is nearly two orders of magnitude draulically controlled exchange flow then feeds back and larger at around 66 Sv. At a regional scale, the present influences the overall hydrostatic balance (Wesson and residual flow estimate is dwarfed by East Auckland Gregg 1994). While Cook Strait’s more modest ex- (9 Sv; Roemich and Sutton 1998) and Southland (8 Sv; change flows also generate horizontal and vertical shear, Sutton 2003) Currents. It is also small compared to the instantaneous velocity structure will be dominated d’Urville Current flow estimates (e.g., 20.8 Sv; Chiswell by tides. 2000). This last point confounds the expectation that the c. Summary and future emphases Cook Strait residual flow should in fact comprise the d’Urville Current. Bowman et al. (1983a) note that, dur- These data, the most substantial of their kind yet ing southerly winds, the d’Urville Current may be de- recorded in Cook Strait, suggest that a number of issues flected northward out of the greater Cook Strait system need to be rethought with regard to circulation in the entirely. Part of the answer may lie in the present lack of central New Zealand zone. Primarily, these observa- sampling of the inshore region to the eastern side of the tions suggest that the net transport is less than supposed narrows. Potentially, there is enhanced residual flow in from previous analysis, by as much as a factor of 5. this region that might enhance overall transport through Furthermore, the suggestion of a reasonable northward the strait (M. Hadfield 2013, personal communication). current on the western side of the strait is new, although However, there is no consistent evidence of this in the SST consideration of bedform orientation and structure imagery (Fig. 14). Alternately, the present sampling might provides some supporting evidence (Black 1986). The miss a tidal rectification process by simply being too far northward current must influence the provenance of outside the zone of influence. However, as noted above, nutrient supply to the productive Marlborough Sounds. this does not manifest itself in the spring–neap analysis The data also suggest a number of areas for future (Fig. 10). development. If this exchange flow is common, what is These residual transport quantities can be placed in the horizontal structure of the residual? While the in- a global context by looking at other systems. Bass Strait, stantaneous velocity profile across the strait could be

Unauthenticated | Downloaded 09/27/21 02:11 AM UTC JUNE 2014 S T E V E N S 1669 determined from modeling, these models would need to Buijsman, M. C., and H. Ridderinkhof, 2007: Long-term - capture the present apparent exchange before the low ADCP observations of tidal currents in the Marsdiep inlet. pass–filtered version would be useful. Taking this point J. Sea Res., 57, 237–256, doi:10.1016/j.seares.2006.11.004. Cahill, M. L., J. H. Middleton, and B. R. Stanton, 1991: Coastal- further, the hypothesis that the d’Urville Current is trapped waves on the west coast of South Island, New compressed within the easternmost few kilometers of Zealand. J. Phys. Oceanogr., 21, 541–557, doi:10.1175/ the strait, not sampled here, needs to be tested. In this 1520-0485(1991)021,0541:CTWOTW.2.0.CO;2. regard, it would appear useful to place instrumentation Chiswell, S. M., 2000: The Wairarapa Coastal Current. N. Z. J. Mar. aboard one or more of the that regularly cross the Freshwater Res., 34, 303–315, doi:10.1080/00288330.2000.9516934. ——, and D. R. Schiel, 2001: Influence of along-shore advection strait (e.g., Buijsman and Ridderinkhof 2007) in order to and upwelling on coastal temperature at Kaikoura Peninsula, build up a picture of the transverse flow structure. In New Zealand. N. Z. J. Mar. Freshwater Res., 35, 307–318, addition, the transport in the upper 20–50 m is not ade- doi:10.1080/00288330.2001.9517000. quately sampled here and, while the assumption of ho- ——, and C. L. Stevens, 2010: Lagrangian and Eulerian estimates mogeneity in current vertical structure is reasonable, the of circulation in the lee of , New Zealand. Cont. Shelf Res., 30, 515–532, doi:10.1016/j.csr.2010.01.004. present data suggest the potential exists for it to improve Emery, W. J., and R. E. Thomson, 2001: Data Analysis Methods in the conclusions drawn here if it can be adequately Physical Oceanography. Elsevier, 638 pp. measured. Additionally, the seasonal cycle needs fur- Gilmour, A. E., 1960: Currents in Cook Strait. N. Z. J. Geol. Geo- ther exploration. This is especially true for the stratified phys., 3, 410–431. period from the late summer through early autumn Harris, T. F. W., 1990: Greater Cook Strait: Form and Flow. DSIR Marine and Freshwater Publishing, 212 pp. months. How does stratification persist in one of the Heath, R. A., 1971: Hydrology and circulation of central and strongest large-scale tidal flows on the planet, and does southern Cook Strait. N. Z. J. Mar. Freshwater Res., 5, 178– this influence exchange? 199, doi:10.1080/00288330.1971.9515375. ——, 1974: Physical oceanographic observations in Marlborough Acknowledgments. The author thanks Mark Hadfield, Sounds, New Zealand. N. Z. J. Mar. Freshwater Res., 8, 691– Philip Sutton, and John Zeldis for discussions and ideas. 708, doi:10.1080/00288330.1974.9515538. ——, 1978: Semidiurnal tides in Cook Strait. N. Z. J. Mar. Fresh- Craig Stewart and Brett Grant conducted the mooring water Res., 12, 87–97, doi:10.1080/00288330.1978.9515730. design, initial deployment, and the microstructure work. ——, 1986: In which direction is the mean flow through Cook The mooring maintenance and retrieval was achieved by Strait, New Zealand—Evidence of 1 to 4 week variability? Michael Brewer, Fiona Elliott, Tim Divett, and the N. Z. J. Mar. Freshwater Res., 20, 119–137, doi:10.1080/ skippers and crew of R/V Kaharoa and R/V Ikatere. 00288330.1986.9516136. Knutsen, Ø., H. Svendsen, S. Østerhus, T. Rossby, and B. Hansen, Graham Rickard, Melissa Bowen, Murray Smith, and 2005: Direct measurements of the mean flow and eddy kinetic Steve Chiswell are thanked for discussions around the energy structure of the upper ocean circulation in the NE science. Two anonymous reviewers are thanked for their Atlantic. Geophys. Res. Lett., 32, L14604, doi:10.1029/ constructive comments on an earlier version of the 2005GL023615. manuscript. Charlotte Severne and Murray Poulter are Loder, J. W., 1980: Topographic rectification of tidal currents on the sides of Georges Bank. J. Phys. Oceanogr., 10, 1399–1416, thanked for their support with securing the funding that doi:10.1175/1520-0485(1980)010,1399:TROTCO.2.0.CO;2. was provided by the New Zealand Ministry for Science Olsson, B. H., 1955: The electrical effects of tidal streams in Cook and Innovation and NIWA Core Funding. This paper is Strait, New Zealand. Deep-Sea Res., 2, 204–212, doi:10.1016/ dedicated to the memory of Dr. Fiona Proffitt. 0146-6313(55)90025-3. Ott, M., and C. Garrett, 1998: Frictional estuarine flow in Juan de Fuca Strait, with implications for secondary circulation. J. Geophys. Res., REFERENCES 103, 15 657–15 666, doi:10.1029/98JC00019. Popinet, S., and G. Rickard, 2007: A tree-based solver for adaptive Baines, P. G., G. Hubbert, and S. Power, 1991: Fluid transport ocean modelling. Ocean Modell., 16, 224–249, doi:10.1016/ through Bass Strait. Cont. Shelf Res., 11, 269–293, doi:10.1016/ j.ocemod.2006.10.002. 0278-4343(91)90069-I. Reynolds, R. W., T. M. Smith, C. Liu, D. B. Chelton, K. S. Casey, Black, K. P., 1986: Sediment transport rates in Cook Strait. NZOI and M. G. Schlax, 2007: Daily high-resolution-blended anal- Oceanographic Field Rep. 24, 14 pp. yses for sea surface temperature. J. Climate, 20, 5473–5496,

Bowman, M. J., A. C. Kibblewhite, and D. E. Ash, 1980: M2 tidal doi:10.1175/2007JCLI1824.1. effects in greater Cook Strait, New Zealand. J. Geophys. Res., Roemmich, D., and P. J. H. Sutton, 1998: The mean and variability 85, 2728–2742, doi:10.1029/JC085iC05p02728. of ocean circulation past northern New Zealand: Determining ——, ——, S. M. Chiswell, and R. Murtagh, 1983a: Shelf fronts and the representativeness of hydrographic climatologies. J. Geo- tidal stirring in Greater Cook Strait, New Zealand. Oceanol. phys. Res., 103, 13 041–13 054, doi:10.1029/98JC00583. Acta, 6, 119–129. Rossby, T., C. N. Flagg, and K. Donohue, 2005: Interannual vari- ——, ——, R. Murtagh, S. M. Chiswell, and B. G. Sanderson, ations in upper-ocean transport by the Gulf Stream and ad- 1983b: Circulation and mixing in greater Cook Strait, New jacent waters between New Jersey and Bermuda. J. Mar. Res., Zealand. Oceanol. Acta, 6, 383–391. 63, 203–226, doi:10.1357/0022240053693851.

Unauthenticated | Downloaded 09/27/21 02:11 AM UTC 1670 JOURNAL OF PHYSICAL OCEANOGRAPHY VOLUME 44

Stanton, B. R., and M. I. Moore, 1992: Hydrographic observations Umlauf, L., and L. Arneborg, 2009: Dynamics of rotating shallow during the Tasman boundary experiment off the west coast of gravity currents passing through a channel. Part I: Observation South Island, New Zealand. N. Z. J. Mar. Freshwater Res., 26, of transverse structure. J. Phys. Oceanogr., 39, 2385–2401, 339–358, doi:10.1080/00288330.1992.9516528. doi:10.1175/2009JPO4159.1. ——, D. G. Goring, and R. G. Bell, 2001: Observed and modelled Vennell, R., 1994: Acoustic Doppler current profiler mea- tidal currents in the New Zealand region. N. Z. J. Mar. Freshwater surements of tidal phase and amplitude in Cook Strait, Res., 35, 397–415, doi:10.1080/00288330.2001.9517010. New Zealand. Cont. Shelf Res., 14, 353–364, doi:10.1016/ Stevens, C. L., M. J. Smith, B. Grant, C. L. Stewart, and T. Divett, 0278-4343(94)90023-X. 2012: Tidal stream energy extraction in a large deep strait: The ——, 1998: Observations of the phase of tidal currents along Karori Rip, Cook Strait. Cont. Shelf Res., 33, 100–109, astrait.J. Phys. Oceanogr., 28, 1570–1577, doi:10.1175/ doi:10.1016/j.csr.2011.11.012. 1520-0485(1998)028,1570:OOTPOT.2.0.CO;2. Sutton, P. J. H., 2003: The Southland Current: A subantarctic ——, and N. R. Collins, 1991: Acoustic Doppler current profiler current. N. Z. J. Mar. Freshwater Res., 37, 645–652, doi:10.1080/ measurements of tides in Cook Strait, NZ. Proc. 10th Aus- 00288330.2003.9517195. tralasian Conf. on Coastal and Ocean Engineering, Auckland, ——, M. Bowen, and D. Roemmich, 2005: Decadal temperature Australia, DSIR, 529–534. changes in the Tasman Sea. N. Z. J. Mar. Freshwater Res., 39, Walters, R. A., P. A. Gillibrand, R. Bell, and E. M. Lane, 2010: 1321–1329, doi:10.1080/00288330.2005.9517396. A study of tides and currents in Cook Strait, New Zealand. Tsimplis, M. N., and H. L. Bryden, 2000: Estimation of the trans- Ocean Dyn., 60, 1559–1580, doi:10.1007/s10236-010-0353-8. ports through the Strait of Gibraltar. Deep-Sea Res. I, 47, Wesson, J. C., and M. C. Gregg, 1994: Mixing at Camarinal Sill in 2219–2242, doi:10.1016/S0967-0637(00)00024-8. the Strait of Gibraltar. J. Geophys. Res., 99, 9847–9878, Uddstrom, M. J., and N. A. Oien, 1999: On the use of high- doi:10.1029/94JC00256. resolution satellite data to describe the spatial and temporal Zeldis, J. R., M. G. Hadfield, and D. J. Booker, 2013: Influence of variability of sea surface temperatures in the New Zealand climate on Pelorus Sound mussel aquaculture yields: Pre- region. J. Geophys. Res., 104, 20 729–20 751, doi:10.1029/ dictive models and underlying mechanisms. Aquacult. Envi- 1999JC900167. ron. Interact., 4, 1–15, doi:10.3354/aei00066.

Unauthenticated | Downloaded 09/27/21 02:11 AM UTC