1. Introduction

A primary oceanic transport pathway between the South Pacific Subtropical Gyre and the equator is the low-latitude western boundary current in the Solomon Sea known as the New Guinea Coastal Undercurrent (NGCUC). Tsuchiya et al. (1989) argued that the NGCUC is the main source for the Equatorial Undercurrent (EUC), a deduction corroborated by the CFC-11 maps of Fine et al. (1994). The EUC supplies much of the water that upwells along the equator and modulates equatorial sea surface temperature and air–sea interaction during ENSO (Lee and Fukumori 2003). Thus the NGCUC is a central link in the shallow meridional overturning circulation and potentially in interannual-to-decadal climate variability. Indeed, mechanisms have been proposed by which variability of the temperature (Gu and Philander 1997) and mass transport (Kleeman et al. 1999; McPhaden and Zhang 2002) in the low-latitude western boundary current modulates equatorial air–sea interaction on ENSO or decadal time scales. Despite their importance, neither the transport nor temperature on the isopycnals that intersect the EUC has been measured in the Solomon Sea, which remains a less-studied part of the Pacific.

Bounded by Papua New Guinea on the west and north and by the Solomon Islands on the east, the Solomon Sea has been sparsely observed because it is remote from research facilities. Hydrography has sketched the principal pathways that pass through the Solomon Sea (Tsuchiya 1968; Tsuchiya et al. 1989; Fine et al. 1994). At selected sites, short time series have been gathered of sea level (Ridgway et al. 1993; Melet et al. 2010) and currents (Lindstrom et al. 1987; Butt and Lindstrom 1994; Lindstrom et al. 1990; Murray et al. 1995) but none spans multiple ENSO events. Cravatte et al. (2011) reviewed observational and modeling work in the Solomon Sea and provide a detailed analysis of shipboard acoustic Doppler current profile data throughout the area, but studied only the region above 300-m depth. Because quantifying the mean and variability of mass and heat transport through the Solomon Sea is central to diagnosing the mechanism of seasonal-to-decadal climate variability, time series measurements of these transports is part of the Climate Variability (CLIVAR) Southwest Pacific circulation and Climate Experiment (SPICE; Ganachaud et al. 2008).

To gather time series of equatorward transport, a program using Spray underwater gliders to make repeated transects across the Solomon Sea was begun by the Consortium on the Ocean’s Role in Climate (CORC) program on boundary current measurement. The initial goal was to learn how to use gliders to measure transport well enough to detect the annual cycle and variability, first on ENSO time scales and eventually on decadal time scales. The feasibility of operating gliders from shore is important in this area where research vessels are largely unavailable for hydrographic sections or deploying moorings.

This paper reports initial results of glider sampling including descriptions of (i) the magnitude and structure of transports above 700-m depth, (ii) the substantial responses to two La Niñas and an El Niño, and (iii) of the scales of variability. Gathering these data has depended on generous and skillful support by the people of the Solomon Islands and Papua New Guinea and on the development of methods to measure absolute current velocities with gliders as they become fouled and damaged by fish strikes. The capabilities and limitations of gliders for this kind of long-term measurement is a focus of what follows.

2. The data

Operations began in August 2007 and the last transect reported here ended in January 2011. Most transects were between Gizo Island in the Solomon Islands and Rossel Island in Papua New Guinea, although three crossings were between Rossel and Honiara, Solomon Islands (see Figs. 1 and 6 for place names). The first glider was deployed from the R/V Alis off Rossel Island, but most were deployed and recovered from open boats close to shore, many by Gizo resident Mr. Danny Kennedy without CORC personnel present. A typical transect takes about 6 weeks, so the glider’s endurance allows round trips across the Solomon Sea, but because the currents averaged over the glider operating depths are often substantially faster than the glider’s motion through the water, some roundtrips took almost five months, nearly exhausting the glider batteries. Initial transects consisted of sequential dives to 500- and 600-m depth but most missions sampled to 700 m. Early sections were sometimes spaced by a few months as we learned how to operate effectively in this remote region. Later, successive transects usually overlapped.

Geographic features and nine typical glider transects. Honiara and Sudest, Rossel and Gizo Islands are glider landfalls. Grayscale bathymetric shading breaks at 0, 200, and 1000 m. Measured depth-average ocean velocities (usually to 700 m) are shown by vectors. Examples in the northern basin just south of New Britain all indicate flow eastward into Solomon Strait. The text discusses 18 transects through 2010 in the southern basin crossing between Rossel or Sudest Islands and either Honiara or Gizo Island in the Solomons. Transects from the Solomons pass south of Pocklington Reef to approach Sudest/Rossel from the south while eastward transects from Papua New Guinea (PNG) are typically swept north by the NGCUC before crossing to Gizo. The western boundary current is weak during La Niñas (blue and magenta) and strong in the 2009 El Niño (yellow and red).

Citation: Journal of Physical Oceanography 42, 11; 10.1175/JPO-D-12-022.1

• Download Figure
• Download figure as PowerPoint slide

View Full Size

Geographic features and nine typical glider transects. Honiara and Sudest, Rossel and Gizo Islands are glider landfalls. Grayscale bathymetric shading breaks at 0, 200, and 1000 m. Measured depth-average ocean velocities (usually to 700 m) are shown by vectors. Examples in the northern basin just south of New Britain all indicate flow eastward into Solomon Strait. The text discusses 18 transects through 2010 in the southern basin crossing between Rossel or Sudest Islands and either Honiara or Gizo Island in the Solomons. Transects from the Solomons pass south of Pocklington Reef to approach Sudest/Rossel from the south while eastward transects from Papua New Guinea (PNG) are typically swept north by the NGCUC before crossing to Gizo. The western boundary current is weak during La Niñas (blue and magenta) and strong in the 2009 El Niño (yellow and red).

Citation: Journal of Physical Oceanography 42, 11; 10.1175/JPO-D-12-022.1

• Download Figure
• Download figure as PowerPoint slide

Geographic features and nine typical glider transects. Honiara and Sudest, Rossel and Gizo Islands are glider landfalls. Grayscale bathymetric shading breaks at 0, 200, and 1000 m. Measured depth-average ocean velocities (usually to 700 m) are shown by vectors. Examples in the northern basin just south of New Britain all indicate flow eastward into Solomon Strait. The text discusses 18 transects through 2010 in the southern basin crossing between Rossel or Sudest Islands and either Honiara or Gizo Island in the Solomons. Transects from the Solomons pass south of Pocklington Reef to approach Sudest/Rossel from the south while eastward transects from Papua New Guinea (PNG) are typically swept north by the NGCUC before crossing to Gizo. The western boundary current is weak during La Niñas (blue and magenta) and strong in the 2009 El Niño (yellow and red).

Citation: Journal of Physical Oceanography 42, 11; 10.1175/JPO-D-12-022.1

• Download Figure
• Download figure as PowerPoint slide

Spray gliders (Sherman et al. 2001) gathered temperature and salinity profiles on every ascent (typically spaced about 3 km) using a continuously operated pumped Seabird CP41 CTD; although the energy costs of running the CTD pump are significant, it is necessary for accurate salinity measurements. The CTDs were typically recalibrated every couple of years and spot checked (accuracy ±0.05) before and after each cruise. Factory recalibration showed small changes (±0.01) and Sea Bird recommended against more frequent recalibration because the changes associated with removing and replacing the CTD would be larger than the drifts seen.

Depth-average currents are deduced from navigation and the vehicle’s through-water velocity. Because the descent and ascent speeds vary little, the difference between glider movement at the surface and the distance traveled relative to the water during the intervening depth cycle measures the depth-average current velocity. Through-water speed is computed from vertical velocity (from pressure change), measured vehicle pitch, and a simple model of vehicle dynamics. The lynchpin of this calculation is the angle of attack between the relative flow past the vehicle and the plane of the vehicle’s wings and hull. This small angle determines the glide path and the lift produced by the wings. An approximate is adequate for general velocity estimates, but accurate measurement of transport requires knowing how vehicle drag, which affects the velocity calculation, varies as the vehicle becomes biofouled. It also requires knowing how vehicle roll causes the glider to slip sideways, causing the direction of vehicle motion to differ from its heading. The corrections for these factors are discussed in the appendix; typically these corrections change measured basin-wide transport several Sverdrups.

Geostrophic shears from the temperature/salinity data are referenced by depth-averaged absolute currents. This is imprecise because cross-track Ekman transport introduces a difference between the depth averages of velocity and of geostrophic velocity. Fortunately, prevailing winds in the Solomon Sea are along its axis, so the Ekman transport is across the sea and its impact on our velocity measurements is negligible.

Figure 1 shows the depth-average velocities measured along 9 selected cross-basin transects from the 24 available through 2010. Some of these explore the northern Solomon Sea with an emphasis on flow through Solomon Strait into the equatorial zone. There are too few of these for meaningful analysis and we note only that eastward transport above 700 m through the Strait is common. Other tracks cross the southern Solomon Sea between Honiara or Gizo Island in the Solomon Islands and the tip of the Louisiades Archipelago, which is the easternmost landmass of Papua New Guinea. Specifically, transects from the Solomon Islands generally reach land on Sudest or Rossel Island from the south while crossing the western boundary current. Transects returning to Gizo Island depart from Rossel Island to the northeast and are often carried across the mouth of Milne Bay by the strong western boundary current. Thus the westbound transects pass well south of Pocklington Reef while eastward tracks take a much more northerly route, sometimes coming close the Laughlin Island. The transports across these two paths should be approximately equal but the internal structures differ, as discussed below.

Glider paths in the eastern basin, where currents are weak, are fairly well repeated, allowing description of cruise-to-cruise variability. Typical fluctuations of depth-average velocity are O(10 cm s⁻¹) with scales of 100 to 200 km. The western-basin currents are stronger and more variable and, consequently, it is not possible to repeat tracks, particularly to the northeast of Rossel Island. These western basin segments were also intentionally varied to explore different techniques for sampling the western boundary current. The absence of repeated tracks in the western basin is a fundamental limitation to exploring the structure of western boundary flows.

3. Transport and its variability

To measure transport, we treat the velocity field as frozen and integrate along the track the flow through the track toward the equator. Volume transport above D, the transect’s depth, with the date t and total track-length L is then

View Expanded

where is the equatorward water velocity normal to the glider track (to the left proceeding from west to east), s is along-track distance, and z is depth. The laterally integrated transport per unit depth, V(z), allows comparison of the vertical structure of transport between our quite different transect paths; unmeasured s, z are excluded from (1). The gliders are typically launched less than 5 km from the shore on the Solomon Islands side and approach within the same distance to the Papua New Guinea side, so the entire transport is sampled. This is especially important near Sudest and Rossel Islands where the western boundary current is often strong close to the coast.

In a frozen and nondivergent velocity field, Q would not depend on the glider’s path, but in a time varying field transient velocities add sampling noise to the transport measurement. Individual tracks (see Fig. 1) show ample evidence of strong transient cross-track flow, perhaps from waves or migrating eddies, which can induce transport errors. Many of these errors are random and average out. But because glider paths are greatly affected by the currents they are measuring, eddies and current fluctuations could produce a nonlinear bias of measured transport.

Our primary objective was to determine the equatorward transport of the NGCUC and other currents in the Solomon Sea with an eye to how these might affect the equatorial zone. Energetic transients seen in the data suggest it will require many years to accurately measure the mean transport, annual cycle, and ENSO response of transport; we have only begun the needed measurements. Figure 2 shows the time series through 2010 of transport through the southern Solomon Sea above the deepest measurement, usually 700 m, integrated from coast to coast. Two equatorward volume transports are shown: (i) the depth-average absolute flow (Q_A) measured directly from the glider motion and (ii) the vertical integral (Q_G) of geostrophic velocity relative to the glider’s deepest depth. Here, Q_A is typically 10 Sv greater than Q_G, showing the depth extent of the currents and the importance of measuring absolute velocity or geostrophic shear below 700 m.

Time series of equatorward volume transport (Sverdrups) through the Solomon Sea. Red is absolute transport above 700 m from measured depth-average velocity. Blue is transport above 700 m from geostrophic shear referenced to 700 m. Black symbols are T_3.4, the anomaly of monthly NCEP Niño-3.4 SST relative to the 30-yr average for that month.

Citation: Journal of Physical Oceanography 42, 11; 10.1175/JPO-D-12-022.1

• Download Figure
• Download figure as PowerPoint slide

View Full Size

Time series of equatorward volume transport (Sverdrups) through the Solomon Sea. Red is absolute transport above 700 m from measured depth-average velocity. Blue is transport above 700 m from geostrophic shear referenced to 700 m. Black symbols are T_3.4, the anomaly of monthly NCEP Niño-3.4 SST relative to the 30-yr average for that month.

Citation: Journal of Physical Oceanography 42, 11; 10.1175/JPO-D-12-022.1

• Download Figure
• Download figure as PowerPoint slide

Time series of equatorward volume transport (Sverdrups) through the Solomon Sea. Red is absolute transport above 700 m from measured depth-average velocity. Blue is transport above 700 m from geostrophic shear referenced to 700 m. Black symbols are T_3.4, the anomaly of monthly NCEP Niño-3.4 SST relative to the 30-yr average for that month.

Citation: Journal of Physical Oceanography 42, 11; 10.1175/JPO-D-12-022.1

• Download Figure
• Download figure as PowerPoint slide

Both series exhibit variability of O(15–25 Sv) over time scales of 6 months to a year with additional changes of 5 to 10 Sv over a few months. In the second half of 2009, two gliders were started from Gizo about 10 days apart, producing a pair of transects in which the rms instantaneous glider separations were between 40 and 250 km. The Q_A from the two nearly simultaneous transects differ by about 5 Sv while the Q_G differ somewhat less. From this and the apparent cruise-to-cruise noise, we believe that the time series reflect seasonal-interannual changes of transport with O(3 Sv) noise from transient eddies superposed.

Both Q_A and Q_G in Fig. 2 exhibit minima in late 2007, early 2009, and late 2010 separated by maxima in late 2008 and late 2009. The fluctuations in the two curves differ from each other by more than the apparent eddy noise level. Also shown in Fig. 2 is T_3.4, the monthly anomaly of the National Centers for Environmental Prediction (NCEP) Niño-3.4 SST relative to the average for that month. The ratio of month-to-month variability to slower variability in T_3.4 is similar to that in Q_A and Q_G. All three series exhibit the same maxima and minima on seasonal-interannual scales. The shape of Q_G is closer to T_3.4 (correlation r = 0.94) than is Q_A (r = 0.53) suggesting that the response to ENSO is strongest shallow where Q_G is focused. From T_3.4 it is evident that the Q minima in late 2007 and late 2010 are associated with La Niñas and the maximum of Q in late 2009 reflects an El Niño. The transport maximum in mid-2008, most notable in Q_G, is stronger than the corresponding maximum of T_3.4, which is well below El Niño levels. Any annual cycle in Q is completely masked by the ENSO modulation. A time lag between Q and T_3.4 or Southern Oscillation index is not clear, but a lag of 3–4 months might not be detected amid the differences in the temporal shapes of the transport and ENSO indicators.

Zilberman et al. (2012, manuscript submitted to J. Climate) have examined the shallow overturning circulation in the South Pacific as observed in Argo hydrography. Using the available 7-year Argo record from 2004 to 2010, they estimated geostrophic meridional flow relative to 1000 m through 7.5°S, including a “Solomon Sea transport index” that averages velocity west of 170°E. Since few Argo floats have drifted into the Solomon Sea, the objective analysis procedure combined data from both sides of the Solomon Islands. The resultant time series of 12-month smoothed transport shows show the Solomon Sea index to be well correlated with, and in phase with, Niño-3.4 temperature. There is little detectable phase lag between Argo transport and Niño-3.4. The sensitivity of 4 Sv (°C)⁻¹ is about half what the gliders show in Fig. 2 in part because of the smoothing inherent in the Argo mapping and the 12-month running mean used.

Analysis of the vertical structure of transport, as indicated by V(z) of (1), helps explain the transport changes seen in Fig. 2. Profiles of the transport per unit depth V(z) and cross-basin-average potential density are shown in Fig. 3 for each cruise. (Here, V has units of and can be converted to the cross-basin average velocity by dividing by the basin width.) The salient features are as follows: mean isopycnals heave in phase and by amounts that decrease slowly with depth; variability of V, which depends on tilt of these isopycnals, increases sharply above 250 m; and the range reaches over 100 Sv per kilometer depth near the surface. Below 300 m, V fluctuation are of O(20 Sv km⁻¹) and the shear fluctuations are reduced even more. The flow (Fig. 3, left) at 700 m shows the error made by referencing geostrophic shear at 700 m and explains the typical O(10 Sv) difference between Q_A and Q_G in Fig. 2.

Profiles of (left) transport per unit depth and (right) . Shears are greatest in midthermocline and decrease below 300 m. Shore-to-shore average isopycnals heave in phase with amplitudes that decrease slightly with depth.

Citation: Journal of Physical Oceanography 42, 11; 10.1175/JPO-D-12-022.1

• Download Figure
• Download figure as PowerPoint slide

View Full Size

Profiles of (left) transport per unit depth and (right) . Shears are greatest in midthermocline and decrease below 300 m. Shore-to-shore average isopycnals heave in phase with amplitudes that decrease slightly with depth.

Citation: Journal of Physical Oceanography 42, 11; 10.1175/JPO-D-12-022.1

• Download Figure
• Download figure as PowerPoint slide

Profiles of (left) transport per unit depth and (right) . Shears are greatest in midthermocline and decrease below 300 m. Shore-to-shore average isopycnals heave in phase with amplitudes that decrease slightly with depth.

Citation: Journal of Physical Oceanography 42, 11; 10.1175/JPO-D-12-022.1

• Download Figure
• Download figure as PowerPoint slide

Variability of V(z) is well described using empirical orthogonal functions (EOFs) of depth and associated time-dependent amplitudes a(t) as . Figure 4 shows the mean of V(z), the two leading , and their amplitudes . The mean has an undercurrent profile with an equatorward maximum near 250 m of about 50 Sv per kilometer depth. The two remarkably describe a two-layer flow separated by a shear zone between 200 and 300 m with uncorrelated motion in each layer; they explain 80% and 18% of the depth integrated variance. The (dominant) upper layer mode’s amplitude tracks ENSO indices like Niño 3.4 SST (r = 0.94) or the Southern Oscillation index (r = −0.65). The lower-layer mode time series is more erratic and is poorly correlated with ENSO.

(left) Mean profile of transport per unit depth (black) and the first two EOFs of (blue and red) explaining 80% and 18% of variance. (right) Amplitudes of the EOFs (same color code). Amplitudes have unit mean squares so the magnitude of the spatial EOFs represents rms transport per unit depth contributed by that mode.

Citation: Journal of Physical Oceanography 42, 11; 10.1175/JPO-D-12-022.1

• Download Figure
• Download figure as PowerPoint slide

View Full Size

(left) Mean profile of transport per unit depth (black) and the first two EOFs of (blue and red) explaining 80% and 18% of variance. (right) Amplitudes of the EOFs (same color code). Amplitudes have unit mean squares so the magnitude of the spatial EOFs represents rms transport per unit depth contributed by that mode.

Citation: Journal of Physical Oceanography 42, 11; 10.1175/JPO-D-12-022.1

• Download Figure
• Download figure as PowerPoint slide

(left) Mean profile of transport per unit depth (black) and the first two EOFs of (blue and red) explaining 80% and 18% of variance. (right) Amplitudes of the EOFs (same color code). Amplitudes have unit mean squares so the magnitude of the spatial EOFs represents rms transport per unit depth contributed by that mode.

Citation: Journal of Physical Oceanography 42, 11; 10.1175/JPO-D-12-022.1

• Download Figure
• Download figure as PowerPoint slide

The “undercurrent” structure of cannot be approximated by a linear combination of the two EOFs. It is perhaps surprising that transport variability has a vertical profile so completely different from the “undercurrent” structure of the mean. The dominant variability is in the “upper layer,” which is responsive to ENSO and approximately in phase with T_3.4. Geostrophic shear referenced to 700 m is better correlated with Niño-3.4 SST than is absolute transport above 700 m (Fig. 2) because flow variation at the reference level is dominated by the lower-layer EOF and is largely unrelated to ENSO. Density stratification (Fig. 3) does not explain why the separation between the two “layers” is as sharp and shallow as it is.

Tentatively, we take the upper-layer EOF as representing the largely horizontal response of the Solomon Sea to varying tropical winds. During La Niña, equatorward flow through the Sea is retarded or reversed; when tropical winds relax in El Niño, equatorward Solomon Sea flow reaches its maximum. Argo (Zilberman et al. 2012, manuscript submitted to J. Climate) and our data show that Solomon Sea transport responds to ENSO mainly above 250 m and, schematically, with the same timing and direction as the central Pacific Ekman flow. This is consistent with models like that of Lee and Fukumori (2003), but observations do not show the transport clearly lagging by several months as such models predict.

Are the glider-observed transport fluctuations typical of other time periods and is their relation to ENSO stable? This was addressed using the Aviso assimilation of satellite-altimeter sea surface height data to represent shallow Solomon Sea transport from 1993 through 2010. Figure 5 shows time series of the following: (i) the cross-basin equatorward geostrophic surface transport per unit depth from Aviso averaged between 11.3°–8.0°S, (ii) the Niño-3.4 surface temperature anomaly, and (iii) the equatorward surface transport per unit depth of the dominant “upper layer” EOF depicted in Fig. 4. Each series is of the perturbation from the average annual cycle and is smoothed with a 15-week triangular filter and normalized by the standard deviation of the series over the plotted period. Reasonable agreement between Aviso currents and the surface current of the “upper layer” EOF might be expected because (i) the EOF is weakly sheared above 100 m, and (ii), as is shown below, shallow transport variability is concentrated in the interior of the Sea where altimeters are relatively unaffected by bathymetry and land.

Time series of the smoothed equatorward geostrophic surface transport per unit depth from Aviso (red), the Niño-3.4 Sea Surface Temperature anomaly (blue), and the equatorward transport per depth by the dominant “upper layer” EOF depicted in Fig. 4. See text for details.

Citation: Journal of Physical Oceanography 42, 11; 10.1175/JPO-D-12-022.1

• Download Figure
• Download figure as PowerPoint slide

View Full Size

Time series of the smoothed equatorward geostrophic surface transport per unit depth from Aviso (red), the Niño-3.4 Sea Surface Temperature anomaly (blue), and the equatorward transport per depth by the dominant “upper layer” EOF depicted in Fig. 4. See text for details.

Citation: Journal of Physical Oceanography 42, 11; 10.1175/JPO-D-12-022.1

• Download Figure
• Download figure as PowerPoint slide

Time series of the smoothed equatorward geostrophic surface transport per unit depth from Aviso (red), the Niño-3.4 Sea Surface Temperature anomaly (blue), and the equatorward transport per depth by the dominant “upper layer” EOF depicted in Fig. 4. See text for details.

Citation: Journal of Physical Oceanography 42, 11; 10.1175/JPO-D-12-022.1

• Download Figure
• Download figure as PowerPoint slide

The Aviso surface current and Niño-3.4 temperature are reasonably well correlated (r = 0.67) and approximately in phase during 1993–2010. The glider EOF’s amplitude is slightly better correlated with Niño-3.4 in 2007–10 (r = 0.79). The fluctuations of both velocity series from the shorter period are clearly of the same character, and similarly related to Niño-3.4, as in the longer time series. The surface-velocity standard deviations over the shorter period are 41 Sv per kilometer depth for the glider EOF and about 27% smaller at 28 Sv km⁻¹ for Aviso. Considering the differences in sampling and the problems extracting near-coast sea surface height from satellite altimeters (Melet et al. 2010), the agreement shown in Fig. 5 is good and suggests that the fluctuations seen in glider data from 2007–2010, and their phase relation with ENSO, are typical of the last 17 years.

4. Scales of transport and variability

It is evident in Fig. 1 that flow in the Solomon Sea is highly variable and no single transect describes it. Particularly in the western basin, individual glider paths vary substantially, making description of variability elusive. Initial efforts to compare data using distance from the Rossel Island were confused by large downstream perturbations to glider tracks. To deal with this, we introduce the cross-basin coordinate depicted in Fig. 6. It obeys and is made to follow coastlines by boundary conditions on Papua New Guinea, on the Solomon Islands, and on New Britain, all applied on the 200-m isobath. An adjustment erased Laughlin Island to make better follow velocities in that region. Erasing other small islands produced a smooth with a relatively constant gradient across the Sea. Analyses based on are, of course, most descriptive if properties change slowly along ; this seems reasonable for averages and the best assumption among those that could be used to compare the ocean measured along different tracks.

The field used as the “cross-basin” coordinate in statistical descriptions of variability and structure.

Citation: Journal of Physical Oceanography 42, 11; 10.1175/JPO-D-12-022.1

• Download Figure
• Download figure as PowerPoint slide

View Full Size

The field used as the “cross-basin” coordinate in statistical descriptions of variability and structure.

Citation: Journal of Physical Oceanography 42, 11; 10.1175/JPO-D-12-022.1

• Download Figure
• Download figure as PowerPoint slide

The field used as the “cross-basin” coordinate in statistical descriptions of variability and structure.

Citation: Journal of Physical Oceanography 42, 11; 10.1175/JPO-D-12-022.1

• Download Figure
• Download figure as PowerPoint slide

For scalars like temperature or salinity, is simply used as the cross-basin coordinate when mapping or averaging. Describing velocity is more complex. Although our gliders measure both components of depth-averaged velocity, only the cross-track component of shear is known. While defines transport, it is a poor velocity descriptor because it depends as much on the direction of glider-track to which it is normal as on the ocean. More useful descriptors are obtained by rewriting (1) in terms of or in terms of r, the distance across the basin moving normal to :

View Expanded

From this we can define three metrics of equatorward flow that preserve transport:

View Expanded

Here, is the transport through the area perpendicular to lines of . A more physically descriptive measure is provided by , which converts to velocity units. We refer to as a pseudo velocity because it is not the actual component of velocity parallel to , but it is rather the velocity parallel to that would produce the measured equatorward transport Q. The depth-integrated transport through an increment is .

Figure 7 shows statistics of , the density of transport in . The mean is concentrated near the western boundary but extends out to midbasin near , showing that significant transport is carried well outside the western boundary current. Variance of is also stronger in the west, but is less concentrated than the mean. Addressing the nature of eddy variability and changes in the circulation, Fig. 7 also shows the first two EOFs of . The spectrum is not steep, suggesting there is a wide variety of eddies and basin-scale variability. Remarkably, the first EOF is focused in midbasin reiterating the importance of transport processes away from the western boundary current. The second EOF has rapid and complex variation in the boundary current, probably representing both meanders of the boundary current and less organized eddy motions. The shallow spectrum and shape of the EOFs beyond the first EOF suggests there is little organization to the variability.

Properties of , the depth-integrated transport per unit in (3). (left) Mean (black) and standard deviation (red). (right) The two leading EOFs explaining 25% (black) and 18% (red) of variance. Note that the largest variability is in midbasin.

Citation: Journal of Physical Oceanography 42, 11; 10.1175/JPO-D-12-022.1

• Download Figure
• Download figure as PowerPoint slide

View Full Size

Properties of , the depth-integrated transport per unit in (3). (left) Mean (black) and standard deviation (red). (right) The two leading EOFs explaining 25% (black) and 18% (red) of variance. Note that the largest variability is in midbasin.

Citation: Journal of Physical Oceanography 42, 11; 10.1175/JPO-D-12-022.1

• Download Figure
• Download figure as PowerPoint slide

Properties of , the depth-integrated transport per unit in (3). (left) Mean (black) and standard deviation (red). (right) The two leading EOFs explaining 25% (black) and 18% (red) of variance. Note that the largest variability is in midbasin.

Citation: Journal of Physical Oceanography 42, 11; 10.1175/JPO-D-12-022.1

• Download Figure
• Download figure as PowerPoint slide

The pseudo velocity is more descriptive of the actual flow along contours than one might at first guess. This is so because along the western boundary strong currents are fairly well directed along contours, so that the relation defining is correct for the true velocity along . Away from the boundary, weak flow allows gliders to cross at near-normal angles so that is relatively insensitive to the crossing angle. The relation between and true flow along is illustrated in Fig. 8, where the true depth averages parallel to are plotted along with their differences from the depth-averaged pseudo velocity ; only cases where the track crosses at more than 15° are plotted. This comparison is possible because, unlike vertical shear, both components of depth-average velocity are measured. We do not suggest that the flow parallel to has been measured, but Fig. 8 suggests that, with moderate averaging to minimize the impact of outliers, is a reasonable estimate of that velocity.

The true depth-average flow along contours (black) and its difference from the depth-average of the pseudo velocity (blue). While not a measurement, evidently provides a reasonable estimate of equatorward flow parallel to .

Citation: Journal of Physical Oceanography 42, 11; 10.1175/JPO-D-12-022.1

• Download Figure
• Download figure as PowerPoint slide

View Full Size

The true depth-average flow along contours (black) and its difference from the depth-average of the pseudo velocity (blue). While not a measurement, evidently provides a reasonable estimate of equatorward flow parallel to .

Citation: Journal of Physical Oceanography 42, 11; 10.1175/JPO-D-12-022.1

• Download Figure
• Download figure as PowerPoint slide

The true depth-average flow along contours (black) and its difference from the depth-average of the pseudo velocity (blue). While not a measurement, evidently provides a reasonable estimate of equatorward flow parallel to .

Citation: Journal of Physical Oceanography 42, 11; 10.1175/JPO-D-12-022.1

• Download Figure
• Download figure as PowerPoint slide

To examine the structure and variability of the flows producing equatorward transport, for each transect was converted to and averaged in bins of and z using only measurements with tracks crossing at 15° or more. Initial analysis of these sections showed the dominant variability to be associated with the direction that the glider crossed the Solomon Sea. This happens because, as seen in Fig. 1, east-to-west tracks take a southerly route to approach Sudest from the south while west-to-east transects leave Rossel on a more northerly path back to Gizo. The discovery of fundamental differences in the flow across these two groups of paths describe how flow enters the Solomon Sea.

Figure 9 shows the mean along the southern and northern tracks. The southern paths are dominated by two main equatorward currents, a subsurface western boundary current, and a shallow (0–250 m) jet in midbasin () that reflects shear from sloping thermocline isopycnals. By the time the flow crosses the northern paths these two currents join to make a western boundary current with much greater vertical extent. The two sections also show substantial differences in the western boundary current. The western boundary current off Sudest on the southern paths is weaker and deeper than that off Rossel. Both currents are centered near and extend to , but spacing of phi contours south of Sudest is nearly twice that off Rossel (Fig. 6), making the center(edge) of the current south of Sudest 40(80) km offshore while off Rossel these distances are nearer 20(40) km; the transports are similar. A small satellite undercurrent appears just offshore of the NGCUC north of Rossel. Both mean sections also show poleward shallow flow in the eastern basin and generally equatorward flow around 250-m depth

(top) Mean equatorward pseudo velocity along (left) southern east-to-west tracks and (right) northern west-to east tracks; potential-density depths are overlaid. Note the significant differences in the western boundary current between the tracks. (bottom) The two dominant EOFs of variability describe (left) 28% and (right) 14% of variance on all tracks.

Citation: Journal of Physical Oceanography 42, 11; 10.1175/JPO-D-12-022.1

• Download Figure
• Download figure as PowerPoint slide

View Full Size

(top) Mean equatorward pseudo velocity along (left) southern east-to-west tracks and (right) northern west-to east tracks; potential-density depths are overlaid. Note the significant differences in the western boundary current between the tracks. (bottom) The two dominant EOFs of variability describe (left) 28% and (right) 14% of variance on all tracks.

Citation: Journal of Physical Oceanography 42, 11; 10.1175/JPO-D-12-022.1

• Download Figure
• Download figure as PowerPoint slide

(top) Mean equatorward pseudo velocity along (left) southern east-to-west tracks and (right) northern west-to east tracks; potential-density depths are overlaid. Note the significant differences in the western boundary current between the tracks. (bottom) The two dominant EOFs of variability describe (left) 28% and (right) 14% of variance on all tracks.

Citation: Journal of Physical Oceanography 42, 11; 10.1175/JPO-D-12-022.1

• Download Figure
• Download figure as PowerPoint slide

Figure 9 also shows the two dominant EOFs computed from fluctuations around the mean of 18 transects. The main feature in the dominant EOF is a pair of oppositely directed flows in the upper 200 m, one centered in the shallow western boundary current and the other in a midbasin jet near . This suggests that some input to the Solomon Sea alternates between the subsurface western boundary, which comes from the Coral Sea, and the shallow midbasin current, which apparently is fed directly from the southeast by the shallow South Equatorial Current. As was true of the EOFs of in Fig. 7, the shallowness of the EOF spectrum and complexity of the EOFs beyond the first suggests a lack of organization of the higher modes. Examination of individual transects shows that, like the midbasin feature seen in the mean in Fig. 9, midbasin fluctuations are stronger and more common on southern east-to-west transects than the northern paths, and are associated with slopes in the thermocline rather than property distributions in the surface layer as might be associated with a front. Perhaps most importantly, the EOFs in Fig. 9 show that the two-layer pattern of the cross-basin integral seen in Fig. 4 is a consequence of integrating through complex patterns of variability.

5. Properties and isopycnals

The NGCUC carries water with potential densities between 25 and at least 27 from the SEC toward the equator. The upper half of the NGCUC has a density near that of the Equatorial Undercurrent core ( Johnson et al. 2002). Because the SEC in this density range is fed by subducted evaporative subtropical waters, it is marked by elevated salinity (Kessler 1999). Climatologies like the Pacific Marine Atlas (available online at sio-argo.ucsd.edu/Marine_Atlas.html; Roemmich and Gilson 2009) and the high-resolution analysis from the Commonwealth Scientific and Industrial Research Organisation (CSIRO) Atlas of Regional Seas (CARS; Ridgway Dunn and Wilkin 2002) show that for the SEC brings higher salinity water to the Australian coast and along the landward boundary of the Coral Sea and western Solomon Sea. Gu and Philander’s (1997) suggestion that changing T/S properties of these subducted waters may affect equatorial air-sea interaction motivates examination of temperature/salinity variability in the Solomon Sea.

Figure 10 shows the distribution with and of equatorward volume transport and of salinity anomalies on isopycnals. The salinity plotted is the time-average minus the cross-basin mean of salinity on the same isopycnal: . The plots of support the interpretation from Fig. 9 that the western boundary current on the northern glider lines is the combination of the deep NGCUC and a shallow jet that is found at midbasin on the southern paths. The salinity plots add to this interpretation by (i) marking the NGCUC on both paths with a higher salinity (and temperature) core for , (ii) showing that the shallow midbasin jet is also marked by the higher salinity and temperature of their subtropical origin, and (iii) confirming that the shallow high-salinity core on the southern lines merges with the shallow western boundary current on the northern line. The salty warm core of the NGCUC is denser than the core of the Equatorial Undercurrent, suggesting that the source for the EUC is the upper NGCUC that includes waters from the Coral Sea and from the shallow midbasin jet.

Mean of the transport density from (3) and salinity anomalies on (left) northern, eastbound paths and (right) southern, westbound paths. (top) 〈 〉 time- averaged on isopycnals; (bottom) mean salinity on isopycnals minus the cross-basin average of that salinity. The velocity plots give an alternative view of the mean currents in Fig. 9. Salinity plots show the NGCUC as a salty/warm core along the western boundary that is similar on the two sections. A salty/warm core also marks the midbasin shallow jet evident in both Figs. 9 and 10. The shallow fresh/cold pool along the eastern boundary in both sections, and the associated band of cold/fresh descending to the west, are unexplained.

Citation: Journal of Physical Oceanography 42, 11; 10.1175/JPO-D-12-022.1

• Download Figure
• Download figure as PowerPoint slide

View Full Size

Mean of the transport density from (3) and salinity anomalies on (left) northern, eastbound paths and (right) southern, westbound paths. (top) 〈 〉 time- averaged on isopycnals; (bottom) mean salinity on isopycnals minus the cross-basin average of that salinity. The velocity plots give an alternative view of the mean currents in Fig. 9. Salinity plots show the NGCUC as a salty/warm core along the western boundary that is similar on the two sections. A salty/warm core also marks the midbasin shallow jet evident in both Figs. 9 and 10. The shallow fresh/cold pool along the eastern boundary in both sections, and the associated band of cold/fresh descending to the west, are unexplained.

Citation: Journal of Physical Oceanography 42, 11; 10.1175/JPO-D-12-022.1

• Download Figure
• Download figure as PowerPoint slide

Mean of the transport density from (3) and salinity anomalies on (left) northern, eastbound paths and (right) southern, westbound paths. (top) 〈 〉 time- averaged on isopycnals; (bottom) mean salinity on isopycnals minus the cross-basin average of that salinity. The velocity plots give an alternative view of the mean currents in Fig. 9. Salinity plots show the NGCUC as a salty/warm core along the western boundary that is similar on the two sections. A salty/warm core also marks the midbasin shallow jet evident in both Figs. 9 and 10. The shallow fresh/cold pool along the eastern boundary in both sections, and the associated band of cold/fresh descending to the west, are unexplained.

Citation: Journal of Physical Oceanography 42, 11; 10.1175/JPO-D-12-022.1

• Download Figure
• Download figure as PowerPoint slide

Relatively fresh/cold water is found in a diagonal band across the Solomon Sea from a shallow cold area on the eastern boundary and extending to greater densities in the west. The range of temperature variation, reaching nearly 0.04 on , is surprising and unexplained. The cool freshwater along the eastern boundary is deeper and much closer to shore than the shallow poleward flow seen in Fig. 9, but both may be results of fresher water flowing southward into the Solomon Sea from the equatorial zone, perhaps through Solomon Strait.

In an attempt to detect variation in the properties or transport in the NGCUC, time series of the average temperature and equatorward transport in different regions of space around the core of the NGCUC were constructed. These show seemingly chaotic and rapid changes of O(0.1°C) and O(1 Sv). Unless these can be understood and accounted for, this may represent the detection limit for changes in the heat transport variations potentially caused by variations in subduction.

6. Conclusions

We have shown that gliders can be used to economically monitor subsurface flow, properties, and transport, even in remote regions with moderately strong boundary currents. The data quality is high enough to support detailed scientific studies. The main limitations to this type of sampling are as follows: (i) strong currents control the sampling paths and generally preclude repeating the same track and (ii) the absence of shear measurements in the along-track direction complicates describing the patterns of transport fluctuations. Nevertheless, credible analyses can be carried out using data gathered with modest effort in remote western boundary currents.

In 3.5 years, our time series shows transport through the Solomon Sea above 700 m to fluctuate around a value of 15 Sv with nearly 100% variation on approximately yearly time scales. The most significant driving force is ENSO with El Niño bringing strong equatorward currents without apparent time lag. Aviso surface currents suggest the current fluctuations seen in our short record are typical of the last 17 years, as is the relation with ENSO. Eddy noise permits accurate measurement of the transport averaged over the observed 3.5 years, but interannual variability prevents even approximating the long-term mean, annual cycle or decadal variability and makes inaccurate measurement of the response to ENSO.

Transport variability has the character of two independent layers separated by an overlap zone between 200 and 300 m, which corresponds to the maximum of the mean equatorward transport profile and the density of the Equatorial Undercurrent core. An analysis based on defining a cross-basin coordinate identifies energetic patterns of variability. This shows that the largest changes in depth-average flow are in midbasin, not near the western boundary (Fig. 7). One- and two-dimensional EOF analyses show that (i) most of the response to ENSO is found in the upper 250 m, (ii) much of the variability in this shallow layer occurs in midbasin, and (iii) the strongest pattern of variability involves a shift in the proportions of shallow input to the Solomon Sea from the Coral Sea or directly from the shallow SEC in midbasin.

The deep NGCUC is marked by strong equatorward velocity and high salinity that traces back along the southern flank of the Louisiades Archipelago into Coral Sea and the SEC. The deep NGCUC velocity structure varies somewhat between the region south of Sudest, where the current is weaker, deeper and wider (core about 40 km from shore), and northeast of Rossel, where the entire current is within 40 km of shore. Currents in the NGCUC core frequently exceed 60 cm s⁻¹ and the largest mean velocity, just east of Rossel Island, is more than 40 cm s⁻¹.