South Equatorial Current Entering the Coral Sea

1

ENSO and short-term variability of the

South Equatorial Current entering the Coral Sea

To be submitted to the Journal of Physical Oceanography

William S. Kessler

NOAA Pacific Marine Environmental Laboratory, Seattle WA USA

Sophie Cravatte

Institut de Recherche pour le Développement, Nouméa, New Caledonia

Abstract

Historical section data extending to 1985 are used to estimate the interannual variability of transport entering the Coral Sea between New Caledonia and the Solomon Islands. Typical magnitudes of this variability are ±5-8Sv on a mean of about 30Sv. Transport increases a few months after an El Niño event, and decreases following a La Niña. Most of the interannual transport variability is well-simulated by a reduced-gravity long Rossby wave model. Vigorous westward-propagating mesoscale eddies can yield substantial aliasing (under- or over-estimates comparable to the mean transport) on individual ship or glider surveys. Since the transport variability is surface-intensified and highly correlated with satellite-derived surface geostrophic currents, a simple index of South Equatorial Current transport based on satellite altimetry is developed.

1. Introduction

The South Equatorial Current carries the westward limb of the South Pacific subtropical gyre to the Coral Sea where it bifurcates into poleward and equatorward branches (REFs). It has been long known, and is expected from simple dynamical ideas (REFs), that the incoming transport varies annually (Kessler and Gourdeau 20NN) and with the ENSO cycle (REFs), but developing time series of this transport variability has been difficult becausemuch of the historical data consists of occasional ship surveys that are hard to put in context. With the advent of the Argo program (REF), this problem will be rectified for the future, but Argo sampling of the Coral Sea became sufficiently dense only in 2005 or 2006. The purpose of this paper is to make use of older data to extend the transport time series back before the Argo era.We would also like to develop indices that could allow satellite altimetry to be used as a proxy for interannual transport variations; the [rapid and regular] satellite sampling gives the opportunity to evaluate high-frequency signals that could alias ship surveys.

Several estimatesof the transport entering the Coral Sea based on ship surveys have been published, and these vary by about a factor of two (REFs), confusing the interpretation of these data. There are several factors that cause such estimates to differ: sampling locations (especially obtaining adequate resolution of the narrow zonal jets and boundary currents), sampling depths, choices of reference level, timing relative to the seasonal or ENSO cycles, and short-term eddy aliasing. By extending the existing time series back before Argo, and making use of weekly satellite altimetry, we hope to evaluate these estimates.

2. Data and methods

Three sources of geostrophic current information provide time series of the transport entering the Coral Sea: First, the SIO/ORSTOM Volunteer Observing Ship XBT program (REF Donguy-Meyers) deployed XBTs along a regularly-repeated merchant ship track from Auckland, New Zealand to Solomon Strait, passing just west of New Caledonia (Fig.AAA), from 1985 through 2001. 3917 temperature profiles, usually to 400m, were made on this track and quality controlled by the Coriolis Data Centre; only profiles with quality flags indicating good data were considered. 1346 of these profiles were within the 20°S-11°S region considered here, for an average of about 1.5 profiles per degree latitude per month. For each XBT temperature value, a mean T-S relation from the CARS gridded compilation (Dunn and Ridgway 2002; was used to associate a salinity. The resulting temperatures and salinities were interpolated to regular 5m depth intervals, then gridded in (y,t) using a Gaussian mapping procedureto a 0.5° latitude by3-month grid, using mapping scales 0.75° latitude and 3 months. Dynamic height relative to 400m was found from these, and crosstrack (perpendicular, approximately southwestward) geostrophic velocity estimated by centered differencing.

Second, the Argo Atlas (Roemmich and Gilson 2009) provides gridded temperature and salinity profilesto 2000m on a global 1°by1° monthly grid. The Argo Atlas fields were sampled along the XBT track at 1° latitude intervals to produce a time series comparable to the XBT data from 2004 through 2011. Crosstrack and zonal geostrophic currents were found relative to 2000m or the deepest common level.

Third, weekly AVISO satellite altimetric sea surface height (SSH) data (REF) on a roughly 1/3° grid from August 1992 through 2011, was sampled along the same track and used to find surface geostrophic velocity anomalies.

In all three cases, anomalies were found as differences from the average annual cycle over the record length of each time series, then filtered with a 7-month triangle. This filter has a half-power point of about 8 months, and passes more than 80% of the variability with timescales longer than 13 months; the result is here referred to as "interannual".

The South Equatorial Current (SEC) can have ambiguous and fluctuating boundaries (REFs). In this work, however, we are interested in anomalies to the flow entering the Coral Sea, not to a specifically-defined SEC. Thus we have chosen a simple metric: transportperpendicular to the XBT line between 20°S and 11°S, which approximates the transport entering the Coral Sea between New Caledonia and the Solomon Islands (Fig.1). The anomalous transports are not sensitive to the actual boundaries of the current at a given time because interannual dynamic height anomalies are highly correlated over latitude in this range.

3. Results

3.1 Mean transports and representativeness of shallow transport anomalies

Although here we are concerned with anomalies, especially since we consider disparate measuresof flow, one motivation is to place historical cruise data into its ENSO context, so total transport valuesare relevant. Additional benefits of doing this are to show what is and is not sampled by the 400m XBT profiles, to compare perpendicular transport acrossthe slanted XBT track (Fig.1) to meridional transport sections, and to evaluate the magnitudes of ENSO and other anomalies relative to mean transport values.

Crosstrack speed across the XBT track (Fig.BBBa) shows two well-known westward branches of the SEC: the narrow North Caledonian Jet (NCJ;Gourdeau et al. 2007; REFs) at 19°S, with a mean speed relative to 400m of about 1cms–1 and transport of about 0.5Sv, and the North Vanuatu Jet (NVJ; REFs) from about 15°S to 9.5°Swith speeds to about 6cms–1 and transport of about 7Sv. Between them the shallow eastward Coral Sea Countercurrent (CSCC; Qiu et al. 2010) iscentered at 15.5°S.Comparable mean crosstrack speed relative to 400m from the Argo Atlas is very similar, with slightly weaker velocities (Fig.BBBb).

Relative to the 2000m reference level available from the Argo Atlas data, the NVJ is found to be stronger,with a similar structure, but the NCJ extends much deeper and its speeds are at least five times larger (Fig.BBBc), as has been previously described (Gourdeau etal. 2007). The CSCC is weaker and narrower. NCJ transport relative to 2000m from the Argo Atlas is about 11Sv, and NVJ transport about 15Sv. Thus the relative-to-400m transports omit most of the mean transport of the NCJ, and about half the mean transport of the NVJ.

The Argo Atlas provides a further check on the representativeness of transport variability relative to 400m. Time series of interannual SEC transport relative to 400m and 1000m are very highly correlated with that relative to 2000m (r=0.98 and 0.93respectively) . The magnitude of transport variance relative to 1000m is about 46% larger than to 400m, and that relative to 2000m is about 54% larger. Regression of the 0/400m onto 0/2000m transport gives a slope of 0.61 and intercept of 0.02Sv (essentially zero). The high correlationsare consistent with the idea that ENSO variability is focused in the thermocline, and suggests that the XBT 400m transport gives a good representation of the temporal variance pattern, but the interannual magnitude of 20°S-10°S transport variability is larger by about 50-60% than measured by the XBT line.

The representativeness of perpendicular velocities across the slanted section can be checked using a meridional section at 160°E from the Argo Atlas. Relative to 2000m, the three currents on the purely meridional section have an almost identical structure to that on the track-sampled field, with slightly stronger velocities and total (20°S-10°S) zonal transport of 28Sv (Fig.BBBd). This correspondence is expected since the vector flow is nearly zonal so the crosstrack velocity component gives the same integral. A similar calculation at 160°E from the CARS climatology relative to 2000m gives a transport of just under 30Sv.

3.2 Internannual transport anomalies

Unfortunately the XBT and Argo time series do not overlap so no direct comparison can be made. We note that the 1985-2001 period covered by the XBT data included the strong El Niño events of 1986-87, 1991-92, 1994-95 and 1997-98, as well as the strong La Niña of 1988, while the Argo time series has been made during a period of generally weaker El Niño events. Two other ENSO-relevant time series span these in situ data sources: The Southern Oscillation Index (SOI; and crosstrack surface geostrophic currents derived from AVISO satellite SSH (Fig.CCC). All theestimatesof the SEC are well-correlated with the SOI with a few-month lag consistent with Rossby wave propagation from the mid-basin focus of ENSO wind variability (REFs). The 16-year XBT transport correlates with the SOI at 0.88 with a 4-month lag, the 7-year Argo transport at 0.67 with a 2-month lag, and the 19-year AVISO surface velocity at 0.75 with a 3-month lag. The shorter recent lag times could reflect the change from the east-Pacific-centered El Niños during the 1980s and 1990s sampled by the XBT and AVISO data, to the late 2000s sampled by Argo when central Pacific “Modoki” El Niños (Ashok et al. 2007) produced wind anomalies further west.

The ENSO signal in crosstrack velocity is clearly seen in examples of El Niños and La Niñas on the track.Anomalies of the zonal flow were strongest at the surface and toward the north end of the New Caledonia-Solomons gap (Fig.DDD). Near-surface NVJ speeds increased by about 50% following the El Niño of 2009-10, and decreased about the same following the La Niña of 2008-09, while NCJ changes were much smaller (Fig.DDD). Very similar anomalies were seen during the other ENSO events, both in the XBT and Argo data. This weakening of interannual variancewith depth is similar to that seen in glider data across the mouth of the Solomon Sea (Davis etal. 2012).

Since ENSO anomalies are surface-intensified, it is not surprising that AVISO surface geostrophic velocities are highly correlated with transport anomalies: the correlation of AVISO speeds (averaged along the XBT line from 20°S to 11°S) with XBT transport is 0.93, and with Argo transport is 0.94, as is obvious from Fig.CCC. Thus AVISO could be used as a proxy for interannual transport anomalies, with the factor of about 1.3Sv(transport relative to 400m) per cm/s (surface speed) given by the regression slope. To approximateArgo transport variability relative to 2000m, the factor would be about 2.1.

Published estimates of the transport at or near 160°E from synoptic sampling were coincidentally taken at times when the ENSO cycle was nearly neutral, and thus do not provide a useful test of the SEC response to ENSO. Scully-Power (1973) found 30Sv from a cruise at 162°E in Aug-Sep 1970, while Gourdeau etal. (2008) found 32Sv absolute transport above 600m from glider data and 28Sv from cruise data based on geostrophy relative to 2000m, both in Aug-Oct 2005. These values are all indistinguishable from the mean transport from the Argo Atlas and CARS climatological means (28 and 30Sv respectively). A somewhat larger estimate, again from Scully-Power (1973) based on data taken during mid-1968 (very weak La Niña conditions), was 37Sv.

There is one outlier to these values: Sokolov and Rintoul’s (2000) SEC transport estimate of 54Sv westward, relative to 2000db,from the WOCE P11S cruise data taken during June-July 1993, a value well above any others noted before or since. This cruise took place following the relatively weak El Niño of 1993, and the AVISO proxy index suggests that westward transport would have therefore been about6-7Sv greater than normalduring this period; not nearly enough to account for the large P11S value. Since thisapparently cannot be explained based on ENSO conditions, we use the weekly AVISO data to ask if higher-frequency variability could have aliased the sampling.

3.3 An example of eddy aliasing?

The standard deviation of high-passed AVISO crosstrack velocity averaged along the XBT section is about 1.4cm/s, which is about 60% the size of the interannual RMS, and about 3 times larger than the annual cycle RMS. Inspection of weekly maps shows that much of this variability is due to westward-movingmesoscale eddies with typical diameters of 100-300km and anomalous surface geostrophic velocities of 20cm/s or more (e.g. Fig.EEE). These features translate west across the Coral Sea with speeds decreasing from about 17cm/s at 12°S to 12cm/s at 20°S; that is, with somewhat smaller meridional speed gradientthan would be expected from linear Rossby wave propagation (Chelton; Maharaj etal. 2009; REF).At these speeds, an individual eddy affects a section line for 15-25 days. However, most of the time such an eddy will produce oppositely-directed velocity anomalies on its flanks that do not result in a correspondingly-large transport anomaly when integrated over a section as we consider here.

Nevertheless, during the WOCE P11S cruise in June-July 1993, such mesoscale eddies lined up so as to produce eastward surface velocity anomalies within 100km of the coast of Papua New Guinea, where the mean flow is the eastward boundary current (the Gulf of Papua Current; GPC), and westward anomalies across the background SEC from 13°S to 19°S, thus temporarily augmenting both mean currents. In the GPC, geostrophic crosstrack surface currents relative to 2000db measured by P11S were about 50cm/s, but the short-term eddy anomalies seen by AVISO at this time were about 20cm/s. Across the SEC, P11S surface currents averaged about 13cm/s (westward), while AVISO anomalies were about 8 cm/s, thus more than half the westward surface velocity measured during P11S was transient.We note that if the cruise had been conducted 2 months earlier, the AVISO anomalies along the P11S sectionwould have had the opposite sign, emphasizing the short-term nature of these eddy signals.

It is impossible to determine the depth extent of the eddy anomalies, but it is reasonable to think that they represent fluctuations of the upper thermocline, and an exponential decay scale of 400m would be a conservative estimate of the eddy vertical structure. If that is the case, the SEC eddy anomalies on the P11S section amount to about 23Sv of transient westward transport. Thus, the background SEC seen in the P11S cruise could then be estimated at about 31Sv, a value consistent with other samples and climatologies. This has an important implication for the redistribution of incoming flow from the SEC. Sokolov and Rintoul (2000) noted that the 54Sv SEC transport across P11S implied that about 29Sv of the arriving westward flow turns south into the East Australian Current. If the typical transport is instead onlyabout 30Sv, then the required southward flow would be small, and the bulk of the SEC would be found to turn north into the Solomon Sea, as most other studies have found (REFs).

4. A simple Rossby wave model

The simplest representation of low-frequency wind-driven dynamics is the linear, reduced-gravity long Rossby wave model as has been frequently usedto study tropical ocean variability (Meyers 1979; Kessler 1990; Chen and Qiu 2004). The model is:

(1)

where h is the pycnocline depth anomaly (positive down). The long Rossby speed is cr=c2/f2, (c is the internal long gravity wave speed, f is the Coriolis parameter and  its meridional derivative), R is a damping timescale and  is the wind stress (data sources are given below). The two parameters to be chosen are c and R, which we take to be c=2.5ms–1 and R=(24 months)–1. In this case, the results are only weakly sensitive to these choices because the largest interannual forcing is in the western half of the basin (Fig.FFF, bottom panels), so differences in propagation have little time to manifest themselves. We ignore signals that might emanate from the eastern boundary (Kessler 1990; Minobe and Takeuchi 1995; Vega etal. 2003; Fu and Qiu 2002), so solutions represent purely the linear response to interior-ocean wind stress curl.

The model was forced with a sequence of interannually-filteredmonthly wind stresses: fromJanuary 1981 to July 1991 with ship-derived winds from the Center for Ocean-Atmospheric Predictions Studies at Florida State University (Bourassa etal. 2005), then until July 1999 with the scatterometer winds from ERS 1 and2 (European Space Agency Remote Sensing) satellites, then until November 2009 with QuikSCAT winds [REF]. We ignored the effects of islands: the stresses  were linearly interpolated in x across all the islands east of Australia-New Guinea, then we did not consider any effects of island blocking (Island Rule; Firing etal. 1999).

Geostrophic zonal transports analogous to the XBT and Argo estimates (Fig.3) were found from the solution h(x,y,t) of (1), using U=(c2/f)dh/dy, then similarly integrated from 20°S to 11°S.

The Rossby transport timeseries is very clearly correlated with the 16-year XBT transport (r=0.86), but does less well during the 5-year period when the winds and Argo data overlap (r=0.53; Fig.FFFa). We note that Argo sampling of the southwest Pacific was sparse until at least mid-2005, when the Rossby and Argo transport time seriesappear dissimilar, which perhaps contributes to the weaker overall relation.The magnitude of the Rossby-predicted transport anomalies is comparable to the observed as well (Fig.FFFa, where the RMS of XBT anomalies is 4.2Sv, of Argo 3.1Sv, and of the Rossby model 5.1Sv). As with the XBT, Argo and Aviso timeseries, the Rossby solution has a strong ENSO signal, with a lag relation to the SOI similar to what was seen for the XBT and Argo transport (section 3.2; Fig.FFFa,b).