Regional circulation around New Caledonia from two decades of observations

Journal of Marine Systems 148 (2015) 249–271

Contents lists available at ScienceDirect

Journal of Marine Systems journal homepage: www.elsevier.com/locate/jmarsys

Regional circulation around New Caledonia from two decades of observations Sophie Cravatte a,b,⁎, Elodie Kestenare b,c, Gérard Eldin b,c, Alexandre Ganachaud b,c, Jérôme Lefèvre a,b, Frédéric Marin a,b, Christophe Menkes a,d, Jérôme Aucan a,b a

Centre IRD (Institut de Recherche pour le Développement), Nouméa, New Caledonia Université de Toulouse, UPS (OMP-PCA), LEGOS, 14 Avenue Edouard Belin, F-31400 Toulouse, France IRD, LEGOS, F-31400 Toulouse, France d Sorbonne Universités (UPMC, Univ Paris 06)-CNRS-IRD-MNHN, LOCEAN Laboratory, IPSL, Paris, France b c

a r t i c l e

i n f o

Article history: Received 25 November 2014 Received in revised form 6 March 2015 Accepted 7 March 2015 Available online 1 April 2015 Keywords: New Caledonia Oceanic circulation Regional and near-coastal currents Southwest tropical Pacific SADCP

a b s t r a c t The regional and near-coastal circulation around New Caledonia is investigated using a compilation of more than 20 years of observations. Velocity profiles acquired by Shipboard Acoustic Doppler Current Profiler (SADCP) during 109 research cruises and ship transits since 1991 are analyzed and compared with absolute geostrophic currents inferred from hydrographic profiles and Argo floats drifts. In addition, altimetric surface currents are used to explore the variability of the circulation at various timescales. By making the best use of the strength of these various observations, this study provides an unprecedented detailed picture of the mean circulation around New Caledonia and of its variability in the upper layers. New Caledonia, together with the Vanuatu Archipelago and the Fiji Islands, acts as a 750-km long obstacle to the westward South Equatorial Current (SEC) entering the Coral Sea. On average, the SEC bifurcates against New Caledonia's east coast into a northwestward boundary current, the East Caledonian Current, beginning east of the Loyalty Islands and extending to at least 1000 m depth, and into a weak southeastward current. The latter, the Vauban Current, flows into the Loyalty channel against the mean trade winds where it extends to at least 500 m depth. It is highly variable at intraseasonal timescales; it often reverses and its variability is mainly driven by incoming mesoscale eddies east and south of New Caledonia. West of the Island, the southeastward Alis Current of New Caledonia (ACNC) flows along the reef slope in the 0– 150 m layer. It overlays a weaker northwestward current, creating an unusual coastal circulation reminiscent of the current system along the Australian west coast. The ACNC is a persistent feature of the observations, even if its transport is also strongly modulated by the presence of offshore eddies. This study highlights the fact, if needed, that a snapshot view of the currents provided by a single transect can be strongly impacted by mesoscale eddies, and should be put into context, e.g. by using simultaneous altimetric data. © 2015 Elsevier B.V. All rights reserved.

1. Introduction New Caledonia (165°E, 19°S–23°S) is an archipelago located in the Southwest Pacific, at the entrance of the Coral Sea (Fig. 1). It is mainly composed of the Loyalty Islands and of the main island “Grande Terre”, surrounded by a 750 km long barrier reef enclosing a shallow lagoon. Together with the Fiji Islands (178°E, 18°S) and the Vanuatu Archipelago (169°E, 14°S–20°S), New Caledonia acts as an obstacle to the broadscale oceanic circulation, mainly composed of the South Equatorial Current (SEC) advecting subtropical thermocline waters westward from the Southeast Pacific to the Coral Sea (e.g. Qu and Lindstrom, 2002; Ridgway and Dunn, 2003: Kessler and Cravatte, 2013a). When encountering these islands, the broad SEC forms boundary currents and westward ⁎ Corresponding author at: Centre IRD de Nouméa, BPA5, 98848 Nouméa Cedex, New Calédonie. E-mail address: [email protected] (S. Cravatte).

http://dx.doi.org/10.1016/j.jmarsys.2015.03.004 0924-7963/© 2015 Elsevier B.V. All rights reserved.

zonal jets at their northern and southern tips (Couvelard et al., 2008; Webb, 2000). In the lee of the islands, recirculations and eastward countercurrents are also produced (Couvelard et al., 2008; Qiu et al., 2009). These interactions between the mean flow and the topography thus render the oceanic circulation quite complex in the Southwest Pacific: the westward SEC cannot be simply described as a broad westward flow, but should rather be considered as a combination of intricate currents. Describing them and understanding their connections is a necessary condition to understand if (and how) anomalies in water masses hydrological properties or transports are transmitted from the subtropical South Pacific to the Coral Sea, and downstream both to the equatorial band and to the Tasman Sea. Topography around islands creates local circulation features (including coastal currents, upwelling or downwelling conditions) that depend on the island orientation with respect to the main oceanic currents and atmospheric winds. These coastal circulations modify the thermohaline structure and nutrients' supply, and thus the local

250

S. Cravatte et al. / Journal of Marine Systems 148 (2015) 249–271

PNG

GPC

SOLOMON SEA

NVJ SANTO ISL.

NCJ

CHESTERFIELD ISLANDS

EAC

d’Entrecastaux Reefs

U UAT VAN

CORAL SEA

SOLOMON ISLANDS

LOYALTY ISL.

NEW SCJ CALEDONIA

SCCT STCC AUSTRALIA

SEC

Grand Passage

FIJI

Ouvéa

GR

AN

SEC

AC N

EC

C

Va

Lifou DE ub Maré an TE R

C

RE

C

TASMAN SEA

Fig. 1. Reference maps of the southwest Pacific (left) and New Caledonia (right), with names of the main Islands and countries, and schematic of the thermocline circulation (blue lines) and surface circulation, shown when different from the thermocline (red lines). The green dashed lines box in the left panel delineates the area shown in the right panel. Black and gray shading shows bathymetry at depths of 0, 100 and 300 m. «PNG» abbreviate Papua New Guinea. Indicated are the SEC (South Equatorial Current), NVJ (North Vanuatu Jet), NCJ and SCJ (North and South Caledonian Jet), STCC (SubTropical CounterCurrent), GPC (Gulf of Papua Current), EAC (East Australian Current) and ECC (East Caledonian Current). The dashed black lines on the right panel represent the sections used for Figs. 9, 13 and 15, showing the Vauban, ECC and Alis current of New Caledonia (ACNC) structures. Ouvéa, Lifou and Maré are the three main Loyalty Islands. The green stars show the positions of the tide gauges used in this study.

physical and biogeochemical oceanic properties, which impact the ecosystems up to higher trophic levels (e.g.: Ganachaud et al., 2010; Menkes et al., 2014). Yet, these local circulations are poorly known around most of the Pacific Islands. Describing them is an important task, firstly to better understand how the large-scale oceanic circulation and its variations at seasonal, interannual or decadal timescales affect local ecosystems and their connectivity, and secondly to infer with more confidence how climate change predicted at large scale in lowresolution climate models will influence these systems. These downscaling studies are an essential component of the CLIVAR/SPICE project (Climate Variability and Predictability/Southwest PacIfic ocean circulation and Climate Experiment) (Ganachaud et al., 2008b, 2014). As a first step toward this goal, we investigate here the mean observed circulation and its variability around New Caledonia, with a focus on the coastal currents. We are far from starting from scratch, and it is fair to pay tribute to the work historically done on the circulation in the region, thanks to organized surveys conducted as early as 1956 by ORSTOM (Office de la Recherche Scientifique et Technique Outre-Mer), which became IRD (Institut de Recherche pour le Développement) (Donguy et al., 1970; Henin et al., 1984; Rotschi and Lemasson, 1967). These pioneer cruises did allow describing the surface regional circulation, but did not allow an accurate description of the small-scale circulation around New Caledonia. The authors pointed out that the observed currents were quite different from one cruise to another, but they were not able to understand this variability. More recently, efforts have been made to better understand the large-scale circulation in the Southwest Pacific, and the jets feeding the Coral Sea. In that context, some aspects of the local circulation have been described from synoptic observations (Ganachaud et al., 2008a, 2010; Gasparin et al., 2011; Gourdeau et al., 2008; Maes et al., 2007). Numerical simulations have also been performed to investigate physical processes (Couvelard et al., 2008; Marchesiello et al., 2010). The knowledge gained from these studies is briefly summarized here. The southeast–northwest orientation of New Caledonia's main Island is at a small angle to the direction of the mean trade winds, resulting in interesting dynamical structures (Fig. 1). On the west coast, wind driven coastal upwelling events cool the sea surface temperature on the southern half of the barrier reef in austral summer (Alory et al., 2006; Ganachaud et al., 2010; Hénin and Cresswell, 2005; Marchesiello et al., 2010). A southeastward current flowing against the mean winds,

named “Alis Current of New Caledonia (ACNC)” after the R/V Alis, a Nouméa-based French research vessel (Ganachaud et al., 2010; Marchesiello et al., 2010), has been observed along that reef (Fig. 1). On the east coast, another southeastward boundary current flowing between the main island and the Loyalty Islands, in a 100 km wide channel, has also been episodically observed. It was named “Vauban Current” by Henin et al. (1984), after the R/V Vauban, another French research vessel formerly based in New Caledonia. Along the southeast coast, downwelling conditions, and southeastward advection of tropical warm waters result in sea surface temperature several tens of degrees warmer than on the west coast (Henin et al., 1984; Marchesiello et al., 2010). Yet, a comprehensive study relying on observations and describing the coastal currents' structure, extension, persistence and variability is still missing. Furthermore, the dynamics and variability of the currents still remain to be understood. Up to date, the most comprehensive analysis of the mean regional circulation around New Caledonia is provided in the modeling study of Marchesiello et al. (2010). Though very informative, this study has two major drawbacks. First, the model simulation they used is validated against observed surface temperature, but not against observed currents, questioning the validity of the simulated circulation. Secondly, the variability of the coastal currents was not studied. This paper aims at providing a useful reference database for process studies and numerical model validation, by taking advantage of the large amount of previously accumulated data. In particular, it relies on historical oceanic currents measurements from Shipboard Acoustic Doppler Current Profilers (SADCP) mounted on research vessels cruising in the vicinity of New Caledonia. These particularly valuable historical in situ data (e.g. Cravatte et al., 2011) are available from near surface to 200, 300 or 500 m depth, depending on the frequency of the mounted instrument. Some research cruises recorded currents deeper, but they are too few to be considered for this study. When analyzed together with hydrological and altimetric data, these SADCP data provide a description of the circulation around New Caledonia at relatively high resolution. This paper has the following structure. Section 2 describes the data and methods used. Section 3 summarizes our knowledge on the largescale circulation in the Southwest Pacific, and describes the details of the mean regional circulation around New Caledonia. Section 4 discusses the variability of this circulation, while Section 5 focuses on the Vauban Current, the East Caledonian Current and ACNC, which are

S. Cravatte et al. / Journal of Marine Systems 148 (2015) 249–271

New Caledonia's coastal currents, and discusses the causes of their variability. Discussion and conclusions are provided in Section 6. 2. In situ and remote data, and methods 2.1. SADCP data sources and processing For the past decades, research vessels have been commonly equipped with SADCP mounted on the hull of vessels. When navigation and heading data are reliable, SADCP provide precise estimates of the absolute horizontal velocity components as a function of depth recorded during either transits or research cruises. Following the methodology used by Cravatte et al. (2011), we gathered historical SADCP data from 109 cruises that sampled around New Caledonia (hereafter NC), in the 156°E–174°E, 28°S–11°S region, from 1990 to 2014 (see Table 1 and Fig. 2). These data were extracted from various data centers: 73 cruises came from IRD database, and especially from R/V Alis (55 cruises). These data were processed using the freely available CODAS software (http://currents.soest.hawaii.edu/). The processing procedure is described in Hummon and Firing (2003). Other data were provided by the Joint Archive for Shipboard ADCP at the University of Hawaii (11 cruises), the MarLIN database of the CSIRO (Commonwealth Scientific and Industrial Research Organization, Australia), mainly consisting of SADCP measurements from the R/V Franklin and R/V Southern Surveyor (23 cruises). Two cruises came from the Japanese JAMSTEC data center. Most of the data come from 150 kHz SADCP instruments, configured to produce profiles from about 25 to 300 m depth, with a typical 8-meter vertical resolution and 5 minute ensemble averages. Note that some ships have several SADCPs mounted on the hull at different frequencies; in that case, we chose data from the instrument giving the best resolution in the 20–300 m layer. We further filtered the data for each cruise according to the following procedure: (a) hourly averaging when not already done; (b) elimination of values inside the lagoon and where the bathymetry is shallower than 300 m depth; (c) elimination of profiles where measurements are not available deeper than 100 m, since they reveal a probable failure of the instrument; and (d) detection of outliers when velocities are larger than 3.5 standard deviations in time at each depth. Ideally, we would have liked to remove the tidal currents from the SADCP data. As observed during cruises that held a fixed position for several days (see Section 5.1.b), diurnal and semidiurnal tidal amplitudes may reach 10–15 cm s − 1 (not shown), and may cause aliasing in SADCP data. Unfortunately, these observed signals are mostly associated with baroclinic internal waves originating from interactions between barotropic tides and bathymetry (e.g. Gourdeau, 1998). The available barotropic tidal models, even those developed at regional scale (J. Lefevre, unpublished results) are thus not useful to filter the tides from SADCP data. The optimal interpolation method used to grid the SADCP data is able to filter the tides in well-sampled regions. In some locations, and for synoptic transects, they may cause aliasing. This will be highlighted and discussed when necessary. The seasonal distribution of SADCP profiles is shown in Fig. 2, together with the number of observations in 0.5° per 0.5° boxes. It indicates good sampling all along the western coast of NC and on the eastern coast between the main land and the Loyalty Islands. It also indicates good sampling along 165°E, where repeated hydrographic cruises took place (Delcroix and Eldin, 1995), and on the northern and southern tips of the New Caledonian reef. At these locations, the seasonal distribution is homogeneous, and currents have been sampled throughout the year. Sampling is sparser east of 170°E, especially in austral summer, and large regions remain unsampled there. 2.2. SADCP data horizontal gridding SADCP data profiles were first vertically averaged from 25 to 100 m (the surface layer), and from 100 to 300 m (referred to as the thermocline

251

layer), and then mapped on a regular 0.25 by 0.25° longitude–latitude seasonal quarterly grid using the three-dimensional (longitude, latitude and time) optimal interpolation (OI) method of De Mey and Ménard (1989). This vertical averaging within two depth ranges is necessary before applying the horizontal mapping to filter out small-scale vertical noise. It is worth mentioning that the 25–100 m mean is reliable, since all profiles extend to at least 100 m depth, whereas the 100–300 m vertical mean is sometimes biased toward shallow depths, when not all profiles extend to 300 m. We separately mapped zonal and meridional components. The OI method provides an optimal value of the components of the velocity at each grid point (x, y, t) using data included in predefined influence radii. A first guess field is obtained from Argo– CARS (CSIRO Atlas of Regional Seas) merged velocities (Kessler and Cravatte, 2013a, Section 2.5). The differences between this first guess and the data are then mapped using the OI method. In fact, whether or not this first guess is used makes little difference; it only slightly smoothes the velocity patterns. The shape of the correlation function used for the mapping is a spatial exponential function multiplied by a Gaussian time decay. C ðr; tÞ ¼


 t2 1 2 ð−rÞ 1þrþ r e  e− T 2 : 3

In the expression above, r is a nondimensional relative space increment; t is the relative time. r ¼ 2:1038 

r ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi x 2  y 2 þ : RCX RCY

To avoid making pre-assumptions on the spatial scales of the currents, spatial and time correlation radii (RCX = RCY = 50 km and T = 91 days) were chosen empirically to avoid excessive smoothing of the small spatial structures, while resolving seasonal variability. We chose a large noise/signal ratio of 10 for the OI. This parameter is an estimate of the ratio of the variance of the noise (all variability which is not seasonal) to the variance of the process of interest (seasonal variability). In our case, the “noise” is primarily due to intraseasonal (or tidal) variability (see Section 4). In addition to the interpolated current, the OI method provides an error estimate at each grid point (comprised between 0 and 1), which is given as a percentage of the variance of the signal represented. It mostly depends on the number of available surrounding data. In our analysis, only interpolated fields with an error estimate of less than 85% of the variance of the signal represented will be considered. This threshold is rather qualitative; choosing 70 or 90% instead does not significantly change the results. 2.3. SADCP data vertical sections in coastal currents In specific locations where many profiles are available, a description of the vertical structure of the currents from SADCP data is possible. This is the case along the eastern and western coasts of NC, and east of Lifou Island (see Section 5). In these locations, a different gridding was performed, and velocity sections were calculated in the following way. In the core of the coastal currents, reference sections perpendicular to the coast (hereafter Vauban, East Caledonian Current (ECC) and Alis Sections) were defined (dashed lines in Figs. 1, 9, 13 and 15), where numerous observations are available (Fig. 2). Around these sections, 5 minute velocities (when available) from each cruise were first rotated by 52° anticlockwise to obtain the cross-shore and alongshore velocity components. Velocities from the SADCP profiles selected within a rectangular area whose length in the cross section direction was chosen as a function of data distribution and size of the current (220 km for the Vauban current, 200 km for the Alis current and 350 km for the ECC) were then averaged in 10 km bins for the Vauban and ECC sections and 15 km bins for the Alis Section, referenced from the Grande Terre Coast. In total, 33

252

S. Cravatte et al. / Journal of Marine Systems 148 (2015) 249–271

Table 1 Cruise list with dates, name of the ship, mounted instrument and data sources. JASADCP = University of Hawaii Joint Archive for Shipboard ADCP (http://ilikai.soest.hawaii.edu/sadcp/ index.html); MARLIN = CSIRO Marine Laboratories Information Network (http://www.marine.csiro.au/maru/marlin_admin.survey_list); IRD-SADCP = ADCP-Institut de Recherche pour le Développement (http://www.ird.nc/ECOP/sadcp_inventory.html); JAMSTEC: (http://www.godac.jamstec.go.jp/darwin/e/). The last column indicates if the cruise data are used to compute vertical velocity sections in the core of the Alis, Vauban and ECC coastal currents (see Section 2.3). Numbers in parentheses for the Vauban current refer to the abscissa axis in Fig. 9a. Cruise

Start

End

Ship

Instrument

Source

FR02/90–FR03/90 Alize2 Surtropac14 Surtropac15 Coare1 FR07/91 FR10/91 Surtropac16 Coare2 Noutah92 Surtropac17 Coare3 FR08/92 Poi FR02/93 Zoneco1 tn026/1 p10 P21w Flupac Nouzu Kaonou75 Zoneco3 Zoneco4 Ebene FR02/99 FR03/99 Wespalis1 Wespalis2 Frontalis1 FR06/01 FR07/01 FR08/01 Diapalis1 FR09/01 Diapalis2 Diapalis3 FR03/02 Diapalis4 Diapalis5 Trar Diapalis7 Diapalis8 Secalis1 Pil2 Snm3 Bula4a diapalis9 Nor1 Teralis1 Pil3 Pvs1 Frontalis2 SS200405 SS200406 BULA5A BULA5R Ss200410 Salomon Km0418 Pvs3 Secalis2 Km0420 SS200502 Km0505

26-Feb-1990 03-Jan-1991 12-Mar-1991 18-Jul-1991 20-Aug-1991 31-Aug-1991 15-Nov-1991 21-Jan-1992 21-Feb-1992 05-May-1992 06-Aug-1992 05-Sep-1992 07-Oct-1992 02-Dec-1992 10-Feb-1993 27-Jun-1993 05-Oct-1993 20-May-1994 23-Sep-1994 24-Feb-1996 23-Jun-1996 30-Aug-1996 21-Sep-1996 21-Oct-1996 12-Mar-1999 10-Apr-1999 14-Oct-1999 13-Apr-2000 29-Mar-2001 08-Jul-2001 29-Aug-2001 04-Sep-2001 25-Oct-2001 13-Nov-2001 10-Dec-2001 18-Jan-2002 26-Mar-2002 04-Apr-2002 20-May-2002 11-Dec-2002 03-Feb-2003 10-Jun-2003 04-Jul-2003 28-Jul-2003 06-Aug-2003 25-Aug-2003 08-Oct-2003 20-Oct-2003 08-Dec-2003 28-Jan-2004 11-Feb-2004 02-Apr-2004 03-May-2004 02-Jun-2004 31-May-2004 14-Jun-2004 02-Oct-2004 14-Oct-2004 21-Oct-2004 26-Nov-2004 08-Dec-2004 19-Dec-2004 24-Jan-2005 24-Mar-2005

06-Apr-1990 05-Mar-1991 06-Apr-1991 15-Aug-1991 15-Sep-1991 23-Sep-1991 14-Dec-1991 13-Feb-1992 17-Mar-1992 27-Jun-1992 31-Aug-1992 01-Oct-1992 20-Oct-1992 28-Feb-1993 26-Feb-1993 15-Jul-1993 09-Nov-1993 26-Jun-1994 29-Oct-1994 09-Mar-1996 09-Jul-1996 19-Sep-1996 12-Oct-1996 20-Nov-1996 08-Apr-1999 04-May-1999 09-Nov-1999 10-May-2000 25-Apr-2001 22-Jul-2001 04-Sep-2001 24-Sep-2001 30-Oct-2001 05-Dec-2001 21-Dec-2001 22-Jan-2002 23-Apr-2002 08-Apr-2002 26-May-2002 23-Dec-2002 12-Feb-2003 13-Jun-2003 17-Jul-2003 31-Jul-2003 19-Aug-2003 30-Aug-2003 15-Oct-2003 06-Nov-2003 21-Dec-2003 01-Feb-2004 19-Feb-2004 29-Apr-2004 27-May-2004 26-Jun-2004 05-Jun-2004 18-Jun-2004 26-Oct-2004 22-Nov-2004 10-Nov-2004 02-Dec-2004 21-Dec-2004 27-Dec-2004 19-Feb-2005 01-Apr-2005

Franklin Le Noroît Le Noroît Le Noroît Le Noroît Franklin Franklin Le Noroît Le Noroît Le Noroît Le Noroît Le Noroît Franklin Le Noroît Franklin Atalante Thomson T. Melville Atalante Atalante Atalante Atalante Atalante Atalante Franklin Franklin Alis Alis Alis Franklin Franklin Franklin Alis Franklin Alis Alis Franklin Alis Alis Alis Alis Alis Alis Alis Alis Alis Alis Alis Alis Alis Alis Alis S. Surveyor S. Surveyor Alis Alis S. Surveyor Alis Kilo Moana Alis Alis Kilo Moana S. Surveyor Kilo Moana

NB150 NB150 NB150 NB150 NB150

JASADCP IRD-SADCP IRD-SADCP IRD-SADCP IRD-SADCP MARLIN JASADCP IRD-SADCP IRD-SADCP IRD-SADCP IRD-SADCP IRD-SADCP MARLIN IRD-SADCP MARLIN IRD-SADCP JASADCP JASADCP IRD-SADCP IRD-SADCP IRD-SADCP IRD-SADCP IRD-SADCP IRD-SADCP MARLIN MARLIN IRD-SADCP IRD-SADCP IRD-SADCP MARLIN MARLIN MARLIN IRD-SADCP MARLIN IRD-SADCP IRD-SADCP MARLIN IRD-SADCP IRD-SADCP IRD-SADCP IRD-SADCP IRD-SADCP IRD-SADCP IRD-SADCP IRD-SADCP IRD-SADCP IRD-SADCP IRD-SADCP IRD-SADCP IRD-SADCP IRD-SADCP IRD-SADCP MARLIN MARLIN IRD-SADCP IRD-SADCP MARLIN IRD-SADCP JASADCP IRD-SADCP IRD-SADCP JASADCP MARLIN JASADCP

Frontalis3 Secalis3 Emerlis Mgln06mv SS200608 ST200602 Secalis4

24-Apr-2005 11-Jul-2005 14-Dec-2005 21-Jul-2006 20-Aug-2006 11-Sept-2006 08-Nov-2006

19-May-2005 25-Jul-2005 17-Dec-2005 01-Sep-2006 11-Sep-2006 17-Sep-2006 21-Nov-2006

Alis Alis Alis Melville S. Surveyor S. Surveyor Alis

NB150 NB150 NB150 NB150 NB150 NB150 no doc NB150 no doc NB75 NB150 NB150 NB75–NB300 NB75–NB300 NB75 NB75 NB75 NB75-NB300

BB150 BB150 BB150

BB150 BB150 BB150 BB150 BB150 BB150 BB150 BB150 BB150 BB150 BB150 BB150 BB150 BB150 BB150 BB150 BB150 BB150

BB150 BB150 BB150 OS38BB/NB BB150 BB150 OS38BB/NB OS38BB/NB WH300 BB150 BB150 BB150 OS75NB

BB150

IRD-SADCP IRD-SADCP IRD-SADCP JASADCP MARLIN MARLIN IRD-SADCP

Currents sampled

Vauban (12), ECC

Alis Vauban (8), ECC Alis Vauban (26), ECC

Vauban (28) Vauban (23) Alis Vauban (41), ECC

Alis Vauban (37/39) Vauban (48/51) Vauban (1) Vauban (11/13) Vauban (17/19) Vauban (4/5), Alis

Vauban (21/22)

Vauban (29/30), Alis Vauban (46/47), ECC Vauban (3/6) Vauban (14)

Vauban (42), ECC Alis Vauban (45), Alis, ECC

Vauban (20), ECC Alis

Vauban (40), ECC

S. Cravatte et al. / Journal of Marine Systems 148 (2015) 249–271

253

Table 1 (continued) Cruise

Start

End

Ship

Instrument

Source

Currents sampled

Km0701

04-Jan-2007

11-Feb-2007

Kilo Moana

JASADCP

Alis

Bsmf Flusec MR07-07 Zonalis Valhybio SS200806 Concalis SS200807 Gyrafort SS200802 49mr0901 Secargo Biopapua Calcot Nectalis1 Sprayalis2 Nectalis2 SS2012_T02 Essai-glider Geodeva5 SS2012_V02 Momalis1 SS2012_T03 Pandora IPOD Bifurcation Sprayalis3 SPOT1 SS2012_V06 MR12-05 SPOT2 Polynésie2013 SPOT3 SPOT4 LOSS SPOT5 SPOT6

03-May-2007 12-Aug-2007 27-Dec-2007 01-Mar-2008 27-Mar-2008 16-Apr-2008 27-Apr-2008 30-Apr-2008 01-Jun-2008 09-Jun-2008 14-Apr-2009 07-May-2010 13-Aug-2010 18-Jul-2011 30-Jul-2011 16-Oct-2011 26-Nov-2011 02-May-2012 6-May-2012 10-May-2012 12-May-2012 23-May-2012 07-Jun-2012 28-Jun-2012 05-Aug-2012 01-Sep-2012 01-Oct-2012 07-Oct-2012 26-Oct-2012 05-Nov-2012 10-Feb-2013 17-Feb-2013 26-Jul-2013 09-Oct-2013 29-Oct-2013 04-Dec-2013 28-Feb-2014

31-May-2007 28-Aug-2007 25-Jan-2008 14-Mar-2008 04-Apr-2008 29-Apr-2008 11-May-2008 7-Jun-2008 22-Jun-2008 18-Jun-2008 19-Jun-2009 19-May-2010 26-Oct-2010 23-Jul-2011 15-Aug-2011 20-Oct-2011 14-Dec-2011 10-May-2012 8-May-2012 20-May-2012 05-Jun-2012 07-Jun-2012 17-Jun-2012 06-Aug-2012 11-Aug-2012 15-Sep-2012 3-Oct-2012 10-Oct-2012 20-Nov-2012 25-Nov-2012 12-Feb-2013 20-Jul-2013 29-Jul-2013 11-Oct-2013 04-Nov-2013 06-Dec-2013 05-Mar-2014

Alis Alis Mirai Alis Alis S. Surveyor Alis S. Surveyor Alis S. Surveyor Mirai Alis Alis Alis Alis Alis Alis S. Surveyor Alis Alis S. Surveyor Alis S. Surveyor Atalante Alis Alis Alis Alis S. Surveyor Mirai Alis Alis Alis Alis Alis Alis Alis

OS38BB/NB WH300 BB150 BB150 OS75 BB150 BB150

cruises sampled the Loyalty channel, making 51 crossing transects across the Vauban current (with 22 transects extending below 300 m depth) (see Table 1). 13 cruises sampled across the ACNC, making 20 transects. 20 cruises sampled the ECC, making 29 transects. All SADCP profiles included within 100 km from the eastern coast (corresponding to the width of the Vauban current) and within 70 km from the western coast (corresponding to the width of the ACNC) were averaged in the direction perpendicular to the coast every 5 m depth (Figs. 8 and 14). They were then gridded both seasonally and in the alongshore direction, to produce monthly vertical sections of the coastal currents as a function of latitude. The latitude varies along the section perpendicular to the coast; the latitude in abscissa in Figs. 8 and 14 is the latitude at the coast of the “Grande Terre”. At each grid point, the number of SADCP observations and their standard deviation were computed, and used to estimate the 90% confidence interval for the mean current. We consider the estimation of the mean current reliable if it is different from zero at 90% confidence level, or if the error on the mean is smaller than 2 cm/s. In Figs. 8, 9, 14 and 15, the grid points where the estimation of the mean is not reliable are hatched in black. 2.4. Sea surface height and surface currents from satellites 2.4.1. Data We also used the gridded sea level anomalies and corresponding geostrophic velocity anomalies from the Archiving, Validation, and Interpretation of Satellite Oceanographic (AVISO) SSALTO/DUACS2014 delayed time product. In this new version, the reference period of the sea level anomalies is based on the 1993–2012 period. The sea level

BB150 BB150 OS75BB BB150 BB150 BB150 BB150 BB150 BB150 BB150 BB150 BB150 OS150/OS38 OS75 OS75 OS75 OS75 OS75 OS75 OS75 OS75 OS75 OS75 OS75 OS75

IRD-SADCP IRD-SADCP JAMSTEC IRD-SADCP IRD-SADCP MARLIN IRD-SADCP MARLIN IRD-SADCP MARLIN JASADCP IRD-SADCP IRD-SADCP IRD-SADCP IRD-SADCP IRD-SADCP IRD-SADCP MARLIN IRD-SADCP IRD-SADCP MARLIN IRD-SADCP MARLIN IRD-SADCP IRD-SADCP IRD-SADCP IRD-SADCP IRD-SADCP MARLIN JAMSTEC IRD-SADCP IRD-SADCP IRD-SADCP IRD-SADCP IRD-SADCP IRD-SADCP IRD-SADCP

Alis

Vauban (15), ECC

Vauban (27), ECC Vauban (33/34), ECC Alis

Vauban (16/18)

Vauban (31/32), ECC Vauban (38), ECC

Vauban (2/7), ECC Vauban (24/25), ECC Vauban (35/36), ECC Vauban (43/44), Alis, ECC Vauban (49/50), ECC Vauban (9/10), ECC

anomalies are based on up to 4 satellites available at a given time (T/P and ERS followed by Jason-1 or OSTM/Jason-2 and Envisat); this “allsat” product quality is improved, thanks to a better sampling, but not homogeneous. The distributed product is a daily 1/4° resolution gridded field of surface geostrophic currents. We found that daily outputs are not necessary for our purpose, and we regridded the currents on a weekly time basis. The AVISO altimetric product was suspected to not adequately resolve currents close to the coasts, which may be an issue when studying the coastal currents structures along NC. Alternative along-track sea level anomaly products dedicated to the coastal domain are developed by the CTOH/LEGOS (http://ctoh.legos.obs-mip.fr/) for T/P, Jason and ENVISAT tracks, with adapted geophysical and atmospheric corrections and reprocessing algorithms. Once adequately filtered, they allow deriving a more reliable coastal geostrophic surface current (see for example Durand et al., 2008). This product was tested for the Vauban current, but in our case it did not provide a significant added value above the gridded AVISO product while being more complicated to use. The coastal currents derived from the DUACS gridded product finally exhibit a realistic variability; as will be shown later (Section 5), they compare reasonably well with in situ tide gauge observations. In addition, we use the OSCAR (Ocean Surface Current Analysis Realtime) near-surface ocean current estimates (Bonjean and Lagerloef, 2002). The horizontal currents are directly estimated from sea surface height, surface vector wind and sea surface temperature. Currents are calculated using a quasi-steady geostrophic model together with an eddy viscosity based wind-driven ageostrophic component and a thermal wind adjustment. The model calculates a surface current averaged over the top 30 m of the upper ocean. Data are on a 1/3° grid with a 5 day resolution.

254

S. Cravatte et al. / Journal of Marine Systems 148 (2015) 249–271

D N O S A J J M A M F J

Fig. 2. Sampling of SADCP data. Upper panel: tracks of the cruises included in the database. Points represent the positions of the hourly profiles, and colors the month during which the profiles were measured. Bottom panel: number of SADCP profiles in 0.5° × 0.5° boxes. Black and gray shading shows bathymetry at depths of 0, 100 and 300 m.

2.4.2. Filtering To distinguish between the seasonal, the low-frequency (interannual and decadal) and the high-frequency (intraseasonal) variability, we filtered the weekly currents with the following steps: a—we computed the mean monthly climatology (Useas); b—we filtered the non-seasonal current (U-Useas) with a 198-day half-power parzen filter to obtain high-pass Uhf and low-pass Ulf currents. With this filtering, the high-pass currents contain all the signal at periods of 88 days (and below), knowing that the dominant period of the mesoscale eddy variability in the region is 70 days (Qiu et al., 2009); c—we computed the kinetic energy for the seasonal, high-pass and low-pass sig!  0:5 2013  1 nals as KEi ¼ nt ∑ Ui ðt Þ2 þ Vi ðt Þ2 where i can stand for seas, t¼1993

lf or hf, U and V are respectively the zonal and meridional currents, and nt is the number of observations.

2.5. Other in situ data 2.5.1. Tide gauges Tide gauges data from three stations were used to estimate the Vauban coastal current transport and compare it with the estimation derived from satellite altimetry data. Their positions are indicated in Fig. 11. Daily sea level data from the Maré tide gauge was obtained by applying a Demerliac filter to the quality-controlled hourly data obtained from the Refmar website (http://refmar.shom.fr/maregraphes_ french-tide-gauges-data) for this station. Data are available from this site from 2012 to 2014. No quality-controlled hourly data was available at this time for the Lifou and Ouinné tide gauges (available from 2011 to 2014, with gaps). We thus used the raw 1-minute data to calculate hourly data if more than 50% of the data was available at each hour, then applied a Demerliac filter to obtain the daily sea level timeseries at these stations.

S. Cravatte et al. / Journal of Marine Systems 148 (2015) 249–271

2.5.2. Argo–CARS merged velocity product (ACMerged) This product constructed by Kessler and Cravatte (2013a) estimates the mean absolute geostrophic currents from the surface to 2000 m with climatological hydrographic data referenced to a 1000 m-velocity field. The absolute velocity at 1000 m is estimated from drift trajectories of Argo floats, and is used as a level of known motion to reference geostrophic shear from the CARS Atlas [Ridgway et al., 2002; Condie and Dunn, 2006; http://www.cmar.csiro.au/cars], giving a field of threedimensional mean absolute geostrophic velocity in the Coral Sea. Description of data, sampling and mapping procedures, can be found in Kessler and Cravatte (2013a). It should be noted that this product partly relies on the sampling of Argo floats, which are sparse west and southeast of NC, and might be too smooth for some local details of the circulation. 2.5.3. Hydrological observations The CARS atlas of temperature and salinity is provided on a 0.5° by 0.5° grid. To resolve smaller scales features, we constructed our own gridded product. We gathered CTD (Conductivity, Temperature, and Depth) casts and Argo float temperature and salinity profiles from the World Ocean DataBase 2013 (http://www.nodc.noaa.gov/OC5/ WOD13/), from 1991 to 2013, in our region of interest. Salinity on isopycnals was computed for individual profiles. On selected isopycnals, corresponding to different water masses, we gridded the salinity data using the same OI method and the same correlation scales as for the SADCP data, using the CARS salinity climatology as a guess field. 3. Mean circulation around New Caledonia 3.1. Background: large-scale circulation and water masses in the Southwest Pacific To set this study around NC into a larger context, we briefly summarize here important elements of the large-scale circulation in the southwestern tropical Pacific, which has been largely described (e.g. Ceccarelli et al., 2013; Ganachaud et al., 2014; Kessler and Cravatte, 2013a; Qu and Lindstrom, 2002) and which is schematically shown in Fig. 1. NC is located in the western branch of the subtropical gyre, roughly at the latitude of the bifurcation of the SEC against the western boundary of the basin, as predicted by the Island Rule (Couvelard et al., 2008; Kessler and Cravatte, 2013a). It means that on the vertical average, waters flowing north of NC mainly bifurcate equatorward at the Australian coast, and waters flowing south of NC bifurcate poleward into the East Australian Current. The subtropical gyre at the entrance of the Coral Sea and the bifurcation at the Australian coast are in fact vertically tilted (Huang and Qiu, 1998; Kessler and Cravatte, 2013a; Qu and Lindstrom, 2002), and the SEC inflow deepens with latitude, with upper eastward shear overlying deeper westward flow. As a result, north of NC, the mean flow is roughly westward at all depths; south of NC, the mean subsurface flow is still westward, whereas the surface current is eastward. This mean schematic view of the oceanic circulation is complicated by its interaction with topography. As stated in the introduction, the westward SEC is not a continuous current, but splits into several zonal jets when encountering the islands of Fiji, Vanuatu and New Caledonia: the north and south Fiji jets, the North Vanuatu Jet (NVJ), and the North Caledonian Jet (NCJ) (Couvelard et al., 2008; Gourdeau et al., 2008). All have a strong surface signature as revealed by drifter observations (Choukroun et al., 2010). South of NC, limited observations suggest the presence, at least episodically, of a subsurface South Caledonian Jet (SCJ) (e.g. Ganachaud et al., 2008a). The eastward surface current flowing south of NC is called the SubTropical CounterCurrent (STCC); Ridgway and Dunn (2003) showed that this surface current is not continuous but occurs as a series of branches that emanate from the East Australian Current recirculation and flow toward or south of the main land of NC (see their Fig. 7; see also Fig. 2 from Marchesiello et al., 2010).

255

Salinity is a useful parameter to trace water masses, and is used as a complementary tool to follow current pathways. The water masses' properties entering the Coral Sea have been described in detail (see for example Tomczak and Hao, 1989; Qu and Lindstrom, 2002, 2004; Qu et al., 2009; Grenier et al., 2013, 2014; Kessler and Cravatte, 2013a; Ganachaud et al., 2014; Gasparin et al., 2014), and for the purpose of this study, only a few points of their findings need to be reminded. Two distinct thermocline water masses originating from different subduction zones enter the Coral Sea at different densities. At sigma 24.5 (at typical depths of 110 m in the region, see Fig. 3a), high salinity water (above 35.9 pss) transported by the NVJ is found north of Vanuatu, and was formerly subducted in the high evaporation region around 20°S, 120°W. At sigma 26 (at typical depths of 300 m, see Fig. 3b), high salinity water (~35.4 pss) originating from north of New Zealand is found south of Vanuatu. Both water masses exhibit a front near 18°S east of Vanuatu and near 16°S west of Vanuatu. Below the thermocline, at sigma 27.2 (around 830 m depth, see Fig. 3c), low salinity Antarctic Intermediate Water with salinity ~ 34.4 pss is advected from the southeastern region, although its pathways through the Coral Sea are not precisely known (Kessler and Cravatte, 2013a). 3.2. Regional circulation around New Caledonia Mean circulation deduced from SADCP data gridding and hydrological data is shown in Figs. 4 and 5 for the surface and thermocline layers around New Caledonia. The mean OSCAR surface currents are also shown in Fig. 6. In the surface layer, the three independent products show similar features. The SADCP and ACMerged products also show similar features in the thermocline layer, with noticeable differences from the surface layer. For clarity, we describe the surface circulation, and point out when it is different in the thermocline. The SEC branch flowing north of 16°S between 168°E and 174°E and originating from north of Fiji is mostly diverted northward when encountering the Vanuatu Archipelago, and feeds the westward NVJ north of Santo Island. The NVJ is stronger in the SADCP data than in the other products, with westward velocities above 25 cm/s (Fig. 4b). A small part of this SEC branch also appears to flow through the permeable Vanuatu Archipelago and to directly feed the North Caledonian Jet near 17°S or recirculate eastward toward Fiji (Fig. 4a and b), as also suggested by the high salinities on isopycnal 24.5 near 17.5°S both west and east of Vanuatu islands (Fig. 3a). The SEC branch flowing between 18°S and 22°S (the South Fiji Jet) is mostly diverted southward when encountering the Vanuatu Archipelago, and feeds the northwestward East Caledonian Current flowing east of the Loyalty Islands (Ganachaud et al., 2008a; Gasparin et al., 2011). This connection, clearly visible in the ACMerged product and in the hydrological data (Figs. 4a, 3a), is not as evident in the SADCP product. Part of the South Fiji Jet also appears to recirculate northeastward in the lee of Fiji in the surface layer. The East Caledonian Current further feeds both the narrow North Caledonian Jet flowing westward along 17°S–18°S, and potentially the “Grand Passage”, a 500 m deep, 52 km wide passage between the main Island reef and d'Entrecasteaux reef (Fig. 1), although the resolution of our products is not sufficient to resolve it. This current has also been hypothesized to feed the southeastward Vauban Current along the eastern coast of NC's main land. However, this connection is not clear in Fig. 4, and the mean Vauban current itself is weak (a few cm/s) in the surface layer, and it does not emerge as a mean in the thermocline layer. Its horizontal and vertical structures will be studied in more detail in Section 5.2. When exiting the Loyalty channel to the south, the weak Vauban current appears to feed partly the eastward current flowing along 24°S–22°S (Fig. 4a), which is one branch of the STCC (Marchesiello et al., 2010; Ridgway and Dunn, 2003). South of NC, the circulation appears more convoluted. The three products suggest the existence of a meandering South Caledonian Jet in the thermocline (Fig. 5), with a signature at the surface (Fig. 4), possibly surrounded by branches of the eastward STCC. Its origin is not

256

S. Cravatte et al. / Journal of Marine Systems 148 (2015) 249–271

S on sigma 24.5

35 +

a)

z~110 m

S on sigma 26

35 +

b)

z~300 m

S on sigma 27.2

z~830 m

34 +

c)

Fig. 3. Maps of salinity (pss) on the isopycnal 24.5 (a), 26 (b) and 27.2 (c) from in situ observations (see section 2.4). Overlaid are the ACMerged geostrophic velocity vectors on the same isopycnal surfaces. Each map has its own color scale, given as hundreds of pss plus the value indicated above the color bar. Mean depths of each isopycnal in the region are indicated at bottom left, and values on the isopycnal 24.5 are only considered when this isopycnal is deeper than 80 m.

S. Cravatte et al. / Journal of Marine Systems 148 (2015) 249–271

257

ACMerged

cm/s

SADCP

a)

b)

Fig. 4. (a) Mean 25–100 m Argo–Cars merged geostrophic currents. (b) Mean SADCP currents from the optimal interpolation analysis averaged in the 25–100 m layer. Colors represent the mean zonal velocity (red: eastward and blue: westward, in cm/s). Black and gray shading shows bathymetry at depths of 0, 100 and 300 m. Grid points with no data are blanked.

clear: the ACMerged product suggests that it is in part fed by the Vauban current as found in numerical simulations results (e.g. Ganachaud et al., 2008a), whereas such a connection does not clearly appear in the SADCP product. At thermocline level, the existence of a branch supplying the SCJ from the north is strongly supported by a low salinity signature on isopycnal 26 compared to the neighboring salinity (Fig. 3b), although it is rapidly damped. The South Caledonian Jet is not continuous in the SADCP product, especially in the surface layer, and seems dislocated by strong local eastward currents. As will be shown in Section 5, high mesoscale eddy variability is observed in this latitude band, from both in situ observations (Ganachaud et al., 2008a), and satellite altimetry (Qiu et al., 2008, 2009). This high variability renders the

observation of a mean current by sparse observations difficult, and it is very likely that the structures shown are in fact artifacts from unresolved eddies (see Section 3.3 and Fig. 6). Finally, in the lee of the islands, surface intensified eastward countercurrents are observed, in opposition to the surface and thermocline layers' circulation (compare Figs. 4 and 5). The Coral Sea Countercurrent along 15°–16°S, in the lee of Vanuatu, and the Fiji countercurrent along 17°–18°S, in the lee of Fiji islands, stand out in the two products. It was suggested that they are generated by local wind curl dipoles produced by Islands topography resulting in local eastward currents (Qiu et al., 2009). Alternatively, Couvelard et al. (2008) suggested with numerical modeling sensitivity studies that they result from non-linear oceanic

258

S. Cravatte et al. / Journal of Marine Systems 148 (2015) 249–271

ACMerged

cm/s

SADCP

a)

b)

Fig. 5. Same as Fig. 4 for the 100–300 m currents.

advection. In the lee of the southernmost Vanuatu Islands (~ 19°S 169°E), there is also indication of an undocumented shallow eastward surface current in all three products. Interestingly, another eastward current, not predicted by the Island Rule calculation as for the aforementioned countercurrents, is observed in the lee of NC (compare Fig. 4a and Fig. 4c in Kessler and Cravatte, 2013a). Its exact pathways are product-dependent. In the SADCP product (Fig. 4b), it appears to be meandering around the Chesterfield Islands. This current is one of the branches of the STCC, as described in Ridgway and Dunn (2003) and Marchesiello et al. (2010). It feeds the northern part of the coastal ACNC at 21°S. A second branch is observed further south in the two products, feeding the ACNC around 23°S. In the SADCP product, it appears to be strong and part of a cyclonic eddy, whose origin and connection with the northern part of the ACNC are

not clear; this feature is probably biased by insufficient sampling (Section 3.3). Both the Vauban Current and ACNC structures and variability will be discussed in greater detail in Section 5. 3.3. Variability and sampling issues The optimal interpolation method we used provides a map of the mean currents from irregularly sampled SADCP profiles. At grid points in which many profiles are averaged, the estimated mean can be considered as a representative mean. However, in other places with few profiles and high variability, the mean could be seriously biased. There are several possible sources of aliasing; among them the tidal currents, and the intraseasonal variability. Several important questions thus arise: which spatial structures seen in the mean (Figs. 4b and 5b) are

S. Cravatte et al. / Journal of Marine Systems 148 (2015) 249–271

259

cm/s

a)

b)

c)

Fig. 6. (a) Mean OSCAR currents. (b) Mean OSCAR currents subsampled at the same dates and positions as the SADCP data, and gridded with the same method. (c) b minus a. Colors represent the mean zonal velocity (red: eastward and blue: westward, in cm/s). Black and gray shading shows bathymetry at depths of 0, 100 and 300 m. Grid points with no data are blanked.

260

S. Cravatte et al. / Journal of Marine Systems 148 (2015) 249–271

biased by insufficient sampling? How do current snapshots, as gathered during an individual observation (cruise or glider survey), compare to the mean currents? How persistent are the currents shown in the 1991–2013 mean? To evaluate the possible impact of irregular and insufficient sampling on the mean circulation inferred from SADCP data, we subsampled the OSCAR product at the same grid points, and same days as the SADCP database, and regridded the subsampled currents with the same technique and parameters as the SADCP data. The results are thus compared to the OSCAR 1993–2014 mean circulation (Fig. 6a and b), and their difference gives an idea of the aliasing, and a map of surface current errors (Fig. 6c). Aliasing due to tidal currents cannot be evaluated with this method. Interestingly, it appears clearly that some of the features shown in Fig. 6c shows up in Fig. 4b (such as the strong eastward current at 162°E, 22°S, the eastward current at 163°E, 25–27°S; 156°E, 18–19°S). These features in Fig. 4b are thus not representative of the true mean, but are the result of undersampled eddies. Obviously, in this region of high variability and sparse sampling, our OI method does not allow obtaining a reliable mean velocity field.

4. Variability of the circulation around New Caledonia SADCP data in fact reveal an important variability with respect to the mean. The irregular SADCP data sampling does not allow estimating the frequency (tidal, intraseasonal, seasonal or interannual) of the variability by its own. The only observations available to infer the currents variability at the regional scale are the altimetric geostrophic surface currents, and the OSCAR currents. We used the altimetric data and assume that thermocline currents and surface geostrophic currents are well correlated. This has indeed been shown to be true north of NC (Kessler and Cravatte, 2013b; Melet et al., 2010; Ridgway et al., 1993) where surface currents were successively used as a proxy to estimate thermocline transports

variability. To distinguish between the low-frequency (seasonal and interannual variability) and the high-frequency (intraseasonal) variability, weekly altimetric currents were filtered as described in Section 2.4.

4.1. Seasonal and interannual variability In the altimetric data, the surface current variability is strongly dominated by its high-frequency component (Fig. 7c). The seasonal variability appears to be weak except north of 12°S, and negligible compared to the intraseasonal variability. Kessler and Gourdeau (2007) analyzed the annual variability of the vertically integrated circulation in the Southwestern Pacific in an oceanic model. They showed that except north of 13°S, and in the western boundary flows, there is very little transport variability at an annual timescale. They demonstrated that between 10°S and 25°S, in almost all our region of interest, the thermocline depth oscillates annually in phase, and transport variability is generated only on the gyre edges, consistently with our findings here. The low frequency (mostly interannual) variability of the currents also appears weaker than intraseasonal variability. Its origins and dynamics are not well documented. Kessler and Cravatte (2013b), using XBT, Argo floats and altimetric data, showed that the transport along a track between NC and the Solomon Islands is correlated with ENSO (El Nino Southern Oscillation) with a few months lag. More precisely, the near-surface North Vanuatu Jet is stronger following an El Nino event, whereas the North Caledonian Jet is only weakly impacted. These interannual transport anomalies can be explained by ENSO-related wind curl changes in the southwestern Pacific. To complement these findings, we computed the lagged regression between the low frequency altimetric current anomalies and the Nino3.4 index. North of 15°S, the SEC and NVJ velocities are significantly increased three months after an El Nino peak (not shown). The branch of the SEC flowing south of Fiji, the ECC and the NCJ also appear to increase with 5 to 7 months lag after an El

cm/s A B a) Seasonal

b) LF

C

c) HF

d) eddies

Fig. 7. Kinetic energy of (a) seasonal (b) low-pass filtered and (c) high-pass filtered velocities from AVISO surface geostrophic data (in cm/s). A, B and C indicate the three regions of high intraseasonal variability discussed in the text. The filter has a half-power at 198 days. (d) Mean rotational speeds of eddies detected by Chelton's method (in cm/s).

S. Cravatte et al. / Journal of Marine Systems 148 (2015) 249–271

Nino peak. Southward of 20°S, correlations are not significant, and only a small fraction of the variance is explained by ENSO. It is possible that the low frequency variability of Fig. 7b is related to low frequency modulation of mesoscale eddies that rectify into low frequency flow variability. Investigating this possible rectification is an interesting task that will be undertaken in future studies. 4.2. Intraseasonal variability and mesoscale eddies As shown in Fig. 7c, three regions of particularly high intraseasonal variability are observed with altimetry: one corresponding to the East Australian Current system west of 158°E (C), one to the STCC band between 21°S and 28°S (B), and one along the Coral Sea countercurrent at 16°S (A). With a lesser amplitude, another band of high intraseasonal variability is observed along 18°S, east of Vanuatu, corresponding to the position of the eastward Fiji countercurrent. The characteristics of the mesoscale eddy variability in the STCC band and in the Coral Sea have been studied in detail by Qiu and Chen (2004) and Qiu et al. (2008, 2009). In the Coral Sea, the eddy signals have a dominant period of 70 days. They are generated by barotropic instability of the horizontally sheared zonal currents (Qiu et al., 2009). In the STCC band, eddies are generated by baroclinic instabilities in the vertically sheared SEC/SCJSTCC currents (Qiu and Chen, 2004). Another band of relatively high eddy variability also emerges: along and east of the East Caledonian Current. To further document the eddy variability, we use the eddy dataset from Chelton et al. (2011), from October 1992 to April 2012. The eddy identification and tracking procedure applied on sea surface height fields confirms that mesoscale eddies with typical radii of 80–150 km are found all around NC, with greater amplitudes and lifetimes south of about 22°S (not shown). The mean eddies' rotational speeds (defined as the maximum of the average geostrophic speeds around all of the

a)

b)

261

closed contours of SSH inside the eddy, see Chelton et al., 2011) are shown in Fig. 7d. They compare quite well, in amplitude and location, to the velocity RMS (root-mean square) shown in Fig. 7c, reinforcing the underlying assumption made earlier that the intraseasonal variability in the region is mostly linked to mesoscale eddy activity. According to the Chelton et al. (2011) method, eddies form throughout most of the region, but some are also generated to the east of the dateline and enter in the region as they propagate westward; their detection is stopped when arriving at the eastern coast of the islands, east of Vanuatu and east of the Loyalty Islands: they die or may be too distorted by the interaction with the topography to be detected by the procedure. For example, south of NC, half of the eddies detected have lifetimes longer than 4 months. West of the islands, new eddies are detected: whether they are indeed formed in the lee of the islands, or whether they are just the reappearance of eddies that were temporarily lost to the tracking procedure is not known. As will be shown later, these eddies are important drivers of the coastal currents' variability. To conclude, around NC, the intraseasonal velocity variability is similar to the typical rotational speeds of eddies and is comprised between 15 and 25 cm s−1 rms. West and South of NC, the mean surface flow is around 10 cm s−1 rms (Fig. 4). The currents' variability is thus higher than the mean, with instantaneous currents changing direction. 5. Coastal currents of New Caledonia 5.1. On the eastern coast of the “Grande Terre”: the Vauban current The Vauban current has been described in previous studies as a coastal southeastward current flowing between NC's east coast and the Loyalty Islands (Henin et al., 1984; Marchesiello et al., 2010, see Fig. 1). It was first discovered and described thanks to a series of 20 cruises carried out from 1978 to 1980. Based on surface-only direct

c)

d)

Fig. 8. Extraction of the mean alongshore current on the NC eastern coast. (a) SADCP profiles selected; (b) number of SADCP data available between 0 to 100 km from the coast in 0.25° latitude boxes, as a function of depth; (c) mean alongshore current (in cm/s), averaged between 0 to 100 km from the coast, as a function of the latitude at the Grande Terre coast, from ACMerged geostrophic velocities; (d) same as (c) from SADCP data. Positive currents are northwestward. The black line on panel (c) shows the 480 m depth, corresponding to the vertical extend of the bottom panels. Locations where the mean current is not reliable (see text) are hatched in black. The dashed black line on the four panels shows the location of the crosssections shown in Fig. 9.

S. Cravatte et al. / Journal of Marine Systems 148 (2015) 249–271

extending below sigma 26, and transporting 6 ± 2 Sv southeastward between Lifou Island and the “Grande Terre”. Marchesiello et al. (2010) also simulated a strong southeastward Vauban current (with mean velocity reaching 20 cm/s) in a numerical simulation. From these papers, one may conclude that the Vauban current is a persistent feature of the circulation. However, as will be discussed below, SADCP

measurements, Henin et al. (1984) pointed out the existence of a southeastward current at the surface, with a mean velocity of 30 cm/s within 65 km from the coast, relatively stable in direction, whose amplitude and direction showed little correlation with the local wind. During the SECALIS cruise in 2004, Ganachaud et al. (2008a) used an inverse method to estimate the geostrophic velocity; they found a deep current

Grande Terre

b) STD

Lifou

Grande Terre

a) Alongshore SADCP current

Lifou

262

cm/s c) Cross-shore SADCP current

d) STD

Lifou

Grande Terre

Lifou

Grande Terre

Lifou

Fig. 9. (a) Mean SADCP current parallel to the coast (in cm/s) as a function of the distance from the NC eastern coast (in km), along the section perpendicular to the NC coast and shown in insert. Positive currents are northwestward. (b) Standard deviation of this alongshore current (in cm/s). (c, d) Same as (a, b), for the SADCP current perpendicular to the coast. Positive values mean offshore. Locations where the mean current is not reliable (see text) are hatched in black.

S. Cravatte et al. / Journal of Marine Systems 148 (2015) 249–271

observations contradict this description, and reveal highly variable currents in the Loyalty channel, whose direction and intensity vary from cruise to cruise.

5.1.1. Structure of the Vauban current The large number of velocity profiles within the Loyalty channel (Fig. 2) allows studying the vertical and horizontal structures and variability of the Vauban Current. The mean alongshore current inferred from the ACMerged product and the SADCP profiles is shown as a function of latitude (Fig. 8c and d), together with the number of SADCP observations used for this average (Fig. 8a and b, see Section 2.3 for details about the method). In the ACMerged product, there is a clear bifurcation at the eastern coast around 21.4°S, roughly the latitude of the northernmost Loyalty Island, Ouvéa. Mean southeastward flow is observed south of this latitude, extending to 500 m depth (Fig. 8c). Mean northwestward flow is observed north of this latitude: this is the East Caledonian Current (Gasparin et al., 2011), extending from the surface to at least 1000 m depth, and beginning as a subsurface current below the Vauban current. Below 600 m depth, the current is northwestward all along the coast, as confirmed by the 1000 m mean Argo float drifts (Fig. 3c; Kessler and Cravatte, 2013a). This deep current advects low-salinity and highoxygen Antarctic Intermediate waters from the southeastern NC region west and east of the Loyalty Islands, and downstream in the NCJ (Fig. 3c). The SADCP data averages (Fig. 8d) reveal similar currents, but stronger and with a more complex structure. The Vauban current has a mean southeastward surface component in the first 100 m depth from 20.2°S to the southern horn, in a direction opposite to the trade winds; this surface current can thus not be explained by Ekman dynamics. However, this mean southeastward current is significantly different from zero

263

only in the Loyalty channel, between 21°S and 22.5°S. There, it extends down to 500 m depth. In this latitude range, a 10 km-resolution cross-channel section perpendicular to the coast reveals that the mean alongshore current is southeastward across most of the channel (Fig. 9a) although its vertical extent cannot be inferred with great confidence where the variability is high (Fig. 9b). As explained in Section 2.3, these 10-km sections are obtained with SADCP data from 51 crossing transects (with 22 transects extending below 300 m depth). The current is maximum toward the surface and the coast, with typical mean speeds of 7–8 cm/s, weaker than that inferred by Henin et al. (1984) at the surface. It also has an offshore component (Fig. 9c), consistently with the simulation by Marchesiello et al. (2010). As for the western boundary current system along Australia's eastern coast, a deep northward current is found below the southward current (Fig. 8c, 3c); whether the dynamics of these tilted bifurcations are similar remains to be studied. Finally, tight to the reef slope, an opposite mean northwestward current is observed in the thermocline layer (Fig. 9a). This alongshore current is observed during most of the cruises, especially south of 22°S, where it is stronger. As stated above, a striking feature of the Vauban current is its high variability. The alongshore current's variability is surface-intensified, and maximum toward the coast of NC. Its standard deviation reaches 26 cm/s (20 cm/s for its cross-shore component, Figs. 9b and d), which is much larger than the mean current. As a result, the transport of the current in the Loyalty channel is highly variable; during the 51 transects, it varies between +/− 7 Sv in the first 500 m (its mean and standard deviation are 0.5 and 2.3 Sv respectively), and varies between +/− 4 Sv in the first 200 m (see Fig. 10). Similarly, some cruises crossed the channel twice, and measured currents in opposite direction a few days apart. Such a short-time variability can be attributed to tidal, inertial (the inertial oscillations period is around 32 h at this latitude), or

northwestward

a)

b)

southeastward

c)

Fig. 10. a) Vauban current alongshore 0–200 m transport anomalies computed from the 51 SADCP transects, in Sverdrup (black, and dashed when the transit lasted less than 6 h). The numbers in abscissa axis correspond to the cruise numbers given in Table 1. Positive values represent northwestward flow, and negative values southeastward flow, and the dashed line represents the mean transport from the 51 transects. The red (resp. green) bars show the transports at the same dates, estimated from altimetry (resp. OSCAR currents), with a 140 m equivalent depth. (b and c) Example of sea level (colors, in cm) and geostrophic surface currents anomalies (black vectors) during two cruises, Diapalis2 (cruise 48) and Spot2 (cruise 7). The vectors overlaid in colors are the 25–200 m SADCP currents anomalies from these cruises.

264

S. Cravatte et al. / Journal of Marine Systems 148 (2015) 249–271

intraseasonal variability, either due to local forcing or related to largerscale oceanic current instabilities. 5.1.2. Sources of variability for the Vauban current What can be inferred about this variability from the SADCP data alone? Seasonally, the Vauban current is slightly weaker in May– June–July (not shown), but does not reverse. Between October 2001 and October 2003, 9 cruises sampled the Loyalty channel (the “Diapalis” cruises, see Table 1), and held a fixed position for several hours (up to 78 h) in the middle of the channel, around 167°E, 21.5°S. The SADCP time series obtained during these fixed stations allow us to roughly estimate the amplitude of the tidal current variability. A semi-diurnal signal of variable amplitude (typically from 13 cm/s to 24 cm/s peak to peak for the 20–250 m vertically-averaged current, not shown) is observed during each fixed station. This observed signal must be associated at least in part with baroclinic internal waves; indeed, simulated barotropic tidal velocities from a regional tidal model reach a maximum of 5 cm/s in the channel during spring tide. This strong semi-diurnal variability, though possibly biasing currents measured during a cross section lasting typically a few hours, is not the sole source of variability explaining the reversals of the Vauban current. Indeed, during the “diapalis1” fixed station, the vertically-averaged current oscillated at semidiurnal period from 9 cm/s northwestward to 3 cm/s southeastward around a mean 4 cm/s northwestward current. During the “diapalis2” fixed station two months later, the vertically averaged current oscillated at semidiurnal period from 17 cm/s to 31 cm/s southeastward around a mean 24 cm/s southeastward current. We therefore suspect another forcing for the Vauban current variability. To further investigate the source of this variability, the Vauban current transport anomalies measured during the 51 SADCP transects (the mean of the 51 cruises was removed) are compared to the Vauban transport anomalies estimated with altimetric geostrophic velocity anomalies, and with OSCAR velocity anomalies, integrated along the same perpendicular section. A 140 m equivalent depth is chosen to estimate the 0–200 m

transport anomalies from surface current anomalies only; this corresponds to the depth over which vertically uniform surface velocities would best approximate the 0–200-m transport. This equivalent depth was estimated by the regression of the 25 m SADCP velocities on the 0– 200 m transport anomalies. Still, these comparisons remain qualitative. SADCP transports and concomitant DUACS and OSCAR transports are shown in Fig. 10a; when the SADCP transect lasted less than 6 h, and therefore could potentially be strongly biased by tidal variability, the transport is represented by dashed lines. The correspondence between SADCP transports and transports estimated from the DUACS or OSCAR products is not perfect, as can be seen by anomalous transports during cruises 3, 21, 26 and 48. However, this correspondence can hardly be perfect, because what are compared are transports from instantaneous currents, measured typically during a few hours and resolving non-geostrophic scales, to geostrophic weekly estimations of the transports. Some of the temporal and spatial variability included in the SADCP data (such as tides, high-frequency variability, or submesoscale filaments), are not resolved in the DUACS or OSCAR surface currents. Yet, during most of the cruises, altimetric large-scale anomalous signals explain the SADCP anomalous transports. Examples of large-scale conditions during two cruises, the “diapalis2” and “SPOT2” cruises, are shown in Fig. 10b, c. Typically, when cruises observed a strong Vauban current (negative anomalous southeastward transports), an anticyclonic eddy was centered on the Loyalty Islands, and/or a cyclonic eddy was located southeast of the mainland of NC (Fig. 10b). On the contrary, when cruises observed a reversed Vauban current (positive northwestward transports), a cyclonic eddy was present north of the Loyalty Islands, and/or an anticyclonic eddy was located southeast of NC mainland (Fig. 10c). These dipoles of anticyclonic and cyclonic eddies apparently contributed to generate current anomalies in the Loyalty channel. Our hypothesis is therefore that the intraseasonal variability of the Vauban current is mainly controlled by large-scale mesoscale activity. Obviously, the local wind forcing is not a major factor of the Vauban current's variability, as seen from the good correspondence between the OSCAR

Transport (Sv)

northwestward

a)

southeastward

LIFOU DUACS2014 TIDE GAUGES Ouinne-Lifou MARE

OUINNE

b)

c)

Fig. 11. (a) Vauban current alongshore transport anomaly (in Sv) estimated from altimetry with a 140 m equivalent depth. Positive values represent anomalous northwestward flow; negative values southeastward flow. (b) Topography of the southeastern NC coast; the positions of the tide gauges are shown by red crosses, together with the red line along which the altimetric current is estimated. (c) Vauban current alongshore transport anomaly estimated from altimetry (black) and tide gauges sea level anomalies (red), in Sv.

S. Cravatte et al. / Journal of Marine Systems 148 (2015) 249–271

anomalies (containing the geostrophic and ageostrophic Ekman components) and DUACS anomalies (containing only the geostrophic component) in Fig. 10a. To check the robustness of this hypothesis, the timeseries of the 1993– 2013 altimetric Vauban transport anomalies were computed from DUACS (Fig. 11a). Using this dataset to compute a transport in a 100 km large channel could be questionable, as discussed in Section 2.4; however, we validated it against the transport estimated with weekly averages of tide gauges sea level data differences, located in Ouinné and Lifou, on both sides of the channel (Fig. 11b, c). During their common periods of time, the transport estimates match well in phase and amplitude, giving us confidence in the altimetric Vauban transport. We also computed the sea level difference between tide gauges in Ouinné and Maré. The altimetric Vauban transport compares better with the Ouinné-Lifou transport than with the Ouinné-Maré one, suggesting that the transport between Lifou and Maré Islands is not negligible. We then computed composites based on this altimetric transport. Maps of sea level and surface geostrophic current anomalies were extracted for the weeks of Vauban current reversals, when anomalous transports were greater than 1 Sv to the northwest (Fig. 12d). A dipole pattern of cyclonic and anticyclonic eddies emerges, and appears to be significant. Similar composites computed 4 weeks, 8 weeks and 12 weeks in advance (Fig. 12c, b, a) indicate that these eddies propagate from the east, with propagation speeds of about 7 cm/s, somewhat faster than the long baroclinic Rossby wave phase speed at these latitudes (4–5 cm/s). Eddies at lower latitudes propagate slightly faster than eddies at 23°–25°S, as expected from theory and observations (Chelton et al., 2011). Composites are also computed when the Vauban current is enhanced, with negative transports anomalies greater than 1 Sv. The opposite dipole pattern is found, with an anticyclonic eddy

265

north of the Loyalty Islands, and a cyclonic eddy south of NC, propagating from the east (not shown). Vauban current variability thus appears to be related to the arrival on the NC east coast of mesoscale eddies generated further east. 5.2. East of the Loyalty Islands: the East Caledonian current East of the Loyalty Islands, a westward/northwestward current is seen in Figs. 4 and 5. This current, named the “East Caledonian Current”, was first described by Gasparin et al. (2011). It transports waters from the SEC, flows along the Loyalty Islands and downstream along the northeastern coast of NC, and ultimately feeds the North Caledonian Jet, possibly partly through the “Grand Passage”; part of the ECC also feeds the Vauban current (Figs. 1b, 4 and 5). Thanks to individual cruises, Gasparin et al. (2011) showed that the ECC extends about 100 km horizontally, and transports around 14.5 Sv in the 0–1000 m layer off the northeast coast of NC, and 8 Sv upstream east of the Loyalty Islands. In our dataset, 29 SADCP transects crossing the ECC east of Lifou Island (Table 1) are available and allow us to estimate the mean current and its variability with statistical confidence. As opposed to the coastal Vauban current, the ECC is far from being parallel to the coast, and its direction varies with depth. Thus, instead of showing the alongshore and cross-shore velocity components, Fig. 13 shows the amplitude of the current, and its angle relative to the northwestward alongshore direction. Since this angle is computed relative to the 308° azimuthal direction, the current is northwestward if the angle is lower than 38°; it is southwestward if the angle is greater than 38°. The ECC extends deep. Its core is located in the 150–400 m layer, at 45 km from Lifou's coast (170 km from the “Grande Terre” eastern coast), with speeds greater

a) 12 weeks in advance

b) 8 weeks in advance

cm

c) 4 weeks in advance

d) No lag

cm

20 cm/s

Fig. 12. Composites of altimetric sea level (colors) and surface geostrophic current (vectors) anomalies for weeks when the Vauban current is reversed (anomalous transport greater than 1 Sv), with no time lag (d). Similar composites computed 4 weeks, 8 weeks and 12 weeks in advance (c,b,a); only the significant composites are shown.

S. Cravatte et al. / Journal of Marine Systems 148 (2015) 249–271

Lifou

a) amplitude of the current

b) alpha (degrees)

Lifou

266

c) STD of the alongshore current d) SADCP section

alpha

Lifou

Lifou

Fig. 13. (a) Mean SADCP current's amplitude (in cm/s) as a function of the distance from the Grande Terre eastern coast (in km). (b) Angle relative to the alongshore direction (defined by azimuth 308°), positive anticlockwise (in degrees), as shown in (d). The current is northwestward if the angle is lower than 38°; it is southwestward if the angle is greater than 38°. (c) Standard deviation of the alongshore current.

than 12 cm/s (Fig. 13a). Argo float drifts reveal that this current is much deeper, and that its magnitude is still higher than 4 cm/s at 1000 m depth (Fig. 3c). In the surface layer, away from the coast, the ECC flows southwestward (with alpha around 55/60°, Fig. 13b), and turns westward when approaching the coast (with alpha close to 38°). At subsurface, the ECC progressively rotates to almost align alongshore (alpha around 15°). The alongshore current's standard deviation is shown in Fig. 13c (note that it is similar to the standard deviation of the current projected along its mean direction, but less noisy). It is surface-intensified, and similar in amplitude to that of the Vauban current. It reaches 24 cm/s in the near-surface, which is higher than the mean current. Deeper, in the core of the current, the standard deviation is around 8–10 cm/s, which is smaller than the mean current. The ECC is thus persistent at subsurface, but highly variable and often reversed in the surface layer. As for the Vauban current, the ECC variability is mainly controlled by large-scale mesoscale activity.

5.3. On the western coast: the Alis current of New Caledonia The ACNC was only recently observed and described by Ganachaud et al. (2010). Data from two cruises that took place one year apart close to NC southern west coast revealed the existence of a prominent coastal southeastward current in the first 150 m, reaching 30–40 cm/s. This current, flowing against the mean trade winds, was found to be in geostrophic balance. Comparing data from these two cruises, Ganachaud et al. (2010) suggested that the location and strength of this current was strongly impacted by the local winds. Especially, during upwelling favorable winds, the ACNC weakened at the surface and was displaced offshore from the reef slope. Numerical simulations (Marchesiello et al., 2010) further suggested that the ACNC is not confined to the southern west coast, but extends all along the coast, with a core at around 50 m depth. Advecting warm waters southeastward and pushing down the thermocline toward the reef, it attenuates the cooling due to the upwelling.

S. Cravatte et al. / Journal of Marine Systems 148 (2015) 249–271

Numerous cruises sailed along NC's west coast, providing a very good sampling of the coastal current (Fig. 14a, b). Data from these cruises are combined to obtain the mean alongshore currents within 70 km off the western coast as a function of latitude (Fig. 14d). The mean alongshore current inferred from the ACMerged product is also shown (Fig. 14c). These observations reveal that the ACNC begins as a shallow current at the northern tip of NC (around 20°S, less than 100 m deep), and progressively reinforces and extends deeper to the southern tip (23°S, to more than 200 m depth) (Fig. 14c,d). It is maximum towards the surface, at the shallowest depth of the SADCP measurements (20–30 m depth). In the mean observed by the SADCP, it has two latitudinal maxima, corresponding roughly to the arrival of waters from the west, at 21.5°S and 23°S, clearly visible in Fig. 4. An opposite northwestward current of around 10 cm/s is observed below, beginning at 300 m depth at 23°S, and reaching the surface at 20°S. Whether this subsurface current is continuous or composed of distinct branches is not clear. These surface and subsurface currents appear to be in geostrophic balance, since they have similar structures in the SADCP and ACMerged product. In addition, 20 velocity transects crossed the ACNC perpendicularly to NC, around 22.4°S (Table 1, Section 2.3). They are combined to produce a 15 km-resolution cross-shore section (Fig. 15). In its southern portion (at 22.4°S), the mean ACNC flows along the reef slope and extends approximately 60 km offshore (Fig. 15a). It has a convergent component toward the coast (Fig. 15c), in agreement with the Marchesiello et al. (2010) modeling study. Marchesiello et al. (2010) indeed suggested that the Ekman offshore flow is very shallow, approximatively 10 m deep, and compensated below by geostrophic onshore flow. The alongshore ACNC current is highly persistent, and is observed along the reef during the twenty available transects. Its amplitude (from 9 to 60 cm/s), its depth (from 100 m to 350 m depth), and its width (from 15 to 75 km from the reef) are however cruise dependent. As a result, the alongshore transport, defined as the integral of

a)

b)

267

alongshore current from 0 to 150 m depth and to 70 km offshore varies from −1.5 Sv to 4 Sv southeastward. In this southern section, the ACNC is seasonally varying; it is weaker in austral winter, from May to July, and stronger in austral summer, from October to February (not shown). In its northern portion, the current is much more variable, often reverses, and there are no persistent features worth to be described (not shown). Thus, it appears that the southeastward current flowing along NC west coast is not continuous, and is rather composed of independent portions that exhibit different variability. We tried to investigate the source of variability for the ACNC. With 21 transects, we were not able to find any clear relationship between the intensity or the position of the ACNC, and the local winds. Alternatively, we found that the reversals of ACNC transport are linked to the occasional presence of an anticyclonic eddy offshore, and strengthening on the transport is linked to the presence of a cyclonic eddy offshore. The origin of these eddies could not be inferred; as opposed to the eddies modulating the Vauban transport that propagate from the east, eddies modulating the ACNC transport seem to grow locally.

6. Discussion The objective of this paper was to present a comprehensive view of the observed circulation around New Caledonia, to describe its variability in the upper layer, and to provide a reference for regional studies, ecosystem connectivity and numerical model validation. Our observed currents are derived from four independent data sources: from over 20 years of direct SADCP measurements onboard research vessels, from indirect estimations using hydrographic profiles, Lagrangian floats' drifts, and from in situ and satellite sea level measurements. In addition, salinity has been used to trace the currents' pathways. Each dataset has its own limitations in space, time or depth, being pointwise and depthlimited for ADCP values to 2D-only but continuous for satellite surface

c)

d)

Fig. 14. Extraction of the mean alongshore current on the NC west coast. (a) SADCP profiles selected; (b) number of SADCP data available between 0 to 70 km from the coast in 0.25° latitude boxes, as a function of depth; (c) mean alongshore current (in cm/s), averaged between 0 to 70 km from the coast, as a function of latitude, from ACMerged geostrophic velocities; (d) same as (c) from SADCP data. Positive currents are northwestward. The black line on panel (c) shows the 280 m depth, corresponding to the vertical extend of the bottom panels. Locations where the mean current is not reliable are hatched in black. The black dashed line on the four panels indicates the location of the cross-sections shown in Fig. 15.

268

S. Cravatte et al. / Journal of Marine Systems 148 (2015) 249–271

b) STD

d) STD

Grande Terre

Grande Terre

c) Crosshore SADCP current

Grande Terre

Grande Terre

a) Alongshore SADCP current

Fig. 15. (a) Mean SADCP current parallel to the coast (in cm/s) as a function of the distance from the NC western coast (in km). Positive currents are northwestward. (b) Standard deviation of this current (in cm/s). (c, d) Same as (a,b) for the SADCP current perpendicular to the coast. Positive values mean onshore. Locations where the mean current is not reliable (see text) are hatched in black.

maps. This paper takes advantage of all these various datasets to produce a “multi-data set” view of the circulation around New Caledonia. One of the outcomes of this study is to highlight the high variability of the circulation in the upper 300 m, questioning even the significance of a defined “mean current” for some of the regional circulation features previously described in the literature, at least in the surface layer. This is

not the case for the westward branches of the SEC observed north of 18°/16°S, that are the most persistent features of the regional circulation, and whose variability at seasonal and interannual timescales have been partially documented in previous papers. South of 18°S, however, around and south of New Caledonia, variability is strongly governed by mesoscale eddies that dominate instantaneous maps of

S. Cravatte et al. / Journal of Marine Systems 148 (2015) 249–271

of New Caledonia, of similar amplitude to that off the Australia's west coast (Fig. 16). The contrast between warm and fresh tropical waters north of New Caledonia advected from the northeast, and colder and saltier waters south of New Caledonia results in a density and dynamic height gradient reversing with depth. Associated with this meridional gradient, the eastward surface STCC (Fig. 4a) flows toward New Caledonia and feeds the ACNC. The arrival of its different branches, probably due to the presence of topographic obstacles or recirculations, produces a poleward strengthening of the ACNC, consistently with observations (Fig. 14d). Deeper, the meridional pressure gradient reverses (Fig. 16), and the subsurface current is equatorward. Fig. 15c also confirms that the mean flow has a component toward the shore, the geostrophic flow being stronger than the Ekman offshore flow, at least below 25 m. As suggested by the Lagrangian analysis of Marchesiello et al. (2010), the ACNC may also be partly fed by waters flowing north of New Caledonia through the Grand Passage or the NCJ, and retroflecting southward. A full analysis of the ACNC dynamics would also require dedicated model experiments. Two other important issues related to the variability of the currents are left unresolved. The first issue concerns the vertical extent of intraseasonal variability, which has been explored only at the surface or in the first 300 m. We concluded that this variability is governed by mesoscale eddies that dominate instantaneous maps of currents, and that mean currents would be masked by these eddies for any snapshot. However, the eddy vertical scales are largely unknown in the region. In New Caledonia, at least one cruise (ZONALIS) sampled an eddy in the SCJ down to at least to 700 m (Menkes, unpublished results). Several other studies suggested that eddies extend below the thermocline (e.g Chaigneau et al., 2011) and much deeper in some regions (e.g. Morrow et al., 2003). Determining the depth to which they affect

Meridional dynamic height gradient (dynamic cm per degrees)

Depth

currents. The coherent SCJ as seen in the long-term mean near-surface field (Fig. 4b) is most often masked by eddies aliasing synoptic surveys. As a result, it should be borne in mind that a snapshot view such as given by a one-shot cruise cannot represent the mean situation, which may prevent one to draw general conclusions on the area functioning (e.g. Menkes et al., 2014). A focus was made on the coastal currents flowing west and east of the main island of New Caledonia. As a mean, both coastal currents are flowing southeastward against the prevailing trade winds in the upper layer. Interestingly, they display numerous similarities with the coastal currents around Australia, despite the differences in Islands sizes. On NC's eastern coast, the westward SEC bifurcates into a northward current and a weak southward Vauban current (3–9 cm/s), overlying a deeper northward current of a few cm/s. This creates a vertical tilt in the SEC bifurcation, as on the Australian east coast (e.g. Kessler and Cravatte, 2013a): above 500 m, part of the SEC inflow bifurcates south, and part of it bifurcates north. Deeper, at 800–1000 m, the whole SEC bifurcates northward along the coast, west and east of the Loyalty islands, advecting Antarctic Intermediate Waters to the NCJ. A striking characteristic of the Vauban current is its high intraseasonal variability; it often reverses, and its residual mean amplitude is weak. It is shown that this variability is not locally forced, but is mainly driven by the occurrence of mesoscale eddies east and south of New Caledonia. The existence of the Vauban current between the mainland and the Loyalty Islands cannot be inferred from the Island Rule. The linear Island Rule simply states that there should be a bifurcation of the SEC east of New Caledonia and that a southeastward current should flow along New Caledonia eastern coast at some point. However, predicting if this current should flow east of the Loyalty Islands or in the channel between the “Grande Terre” and the Loyalty Islands is beyond the dynamics resolved by the Island Rule: the Loyalty Islands lie at least partially within the “Grande Terre” frictional western boundary. Moreover, the Island Rule dynamics relies on strong assumptions (especially on the assumption of a flat bottom), and non-linearities and topographic effects are certainly important factors affecting the coastal currents around the NC islands (Couvelard et al., 2008; Pedlosky et al., 2009). It is also interesting to note that the observed mean Vauban current is much weaker than suggested by modeling works (Marchesiello et al., 2010), and that this difference cannot solely be explained by insufficient sampling in this highly variable current. The influence of topography, the parameterization of the friction and the boundary conditions in the model are possibly important factors contributing to the presence and intensity of the Vauban current; their influence could be evaluated in dedicated model experiments. On the western coast, the near-surface ACNC is a persistent feature of the observations, despite the variation of its amplitude, width and depth. It flows poleward and overlies a weaker subsurface equatorward countercurrent. As pointed out in Marchesiello et al. (2010), this unusual eastern boundary current system is similar to the current system observed along the Australian west coast, where the surface Leeuwin Current, flowing southward against the prevailing winds, deepening and strengthening poleward, overlies the northward subsurface Leeuwin Undercurrent (e.g. Furue et al., 2013; Godfrey and Ridgway, 1985; Godfrey and Weaver, 1991; McCreary et al., 1986). Godfrey and Ridgway (1985) and McCreary et al. (1986) first suggested that the Leeuwin current is forced by a strong alongshore meridional dynamic height gradient. This largescale gradient drives a geostrophic eastward surface flow into the open Indian Ocean. As it approaches the coast, this incoming flow is strong enough to overwhelm the offshore Ekman transport (driven by the alongshore equatorward winds), and drives the coastal Leeuwin current, which is trapped over the slope and down the pressure gradient. Interestingly, Godfrey and Weaver (1991) suggested that the meridional dynamic height gradient is maintained by the transport of warm waters from the Pacific to the Indian Ocean through the Indonesian Throughflow. Going back to the ACNC along NC's western coast, it is found that a large surface meridional dynamic height gradient is also present west

269

Off West Australia Off West NC

Fig. 16. Meridional dynamic height gradient as a function of depth, referenced to 1500 m, computed from CARS2009 climatology. Thick line: off western coast of Australia (difference between 17.5°S and 37.5°S). Dashed line: off western coast of New Caledonia (difference between 19°S and 23°S).

270

S. Cravatte et al. / Journal of Marine Systems 148 (2015) 249–271

the currents is a key issue for oceanographic sampling, which would require mooring measurements over the water column. The second issue relates to tidal variability. We only briefly discussed it but could not assess its amplitude due to inadequate sampling. The few data available at fixed points indicate that internal waves dominate the signal. They may affect the currents and water masses in different ways, in addition to being a cause of bias in sampling during synoptic measurements. There is possible upscaling from submesoscale to mesoscale: tidal currents may generate coastal jets and eddies advected along the east coast, possibly interacting with the Vauban current. Barotropic tides and internal waves may also contribute significantly to the water mixing and change the density gradients. Similar effects are expected on the west coast. Investigating these interactions opens up promising avenues to better understand the local circulation features that are also known to impact the structures of New Caledonia's ecosystems and coral habitats. Acknowledgments This study has been possible thanks to the effort of many scientists, engineers and crew who carefully recorded, processed and made available SADCP data. The authors are very grateful to all who contributed, to the PIs of the cruises and in particular to the R/V Alis crew, who for years have been collecting data and complied with the demands of scientists. SADCP data were downloaded freely from various databases. Many of them come from the Joint Archive for Shipboard ADCP (JASADCP, http://ilikai.soest.hawaii.edu/sadcp/ main-inv.html). We greatly thank E. Firing, J. Hummon and P. Caldwell for maintaining this very useful resource. Other SADCP data come from CSIRO (http://www.marine.csiro. au/ maru/marlin_admin.survey_list), and from JAMSTEC, with the help of T. Hasegawa. The tide gauge observations of Ouinné, Lifou and Maré are the joint property of DéGéOM, of the Government of New Caledonia and SHOM, and are available on the REFMAR website (refmar.shom.fr). The authors also wish to acknowledge Ssalto/Duacs AVISO who produced the altimeter products, with support from CNES (http://www. aviso.altimetry.fr/duacs/), the OSCAR Project Office, the National Oceanographic Data Center, the International Argo Program and the national programs that contribute to it (http://www.argo.ucsd.edu), and the CARS compilation of hydrographic data (http://www.cmar/csiro.au/ cars). They also thank D. Chelton and M. Schlax for making their eddy dataset available. Comments from Lionel Gourdeau and an anonymous reviewer were helpful to improve an earlier version of this manuscript; the careful reading of Lydia Keppler is also appreciated. The authors finally wish to acknowledge the use of the Ferret program for analysis and graphics in this paper. Ferret is a product of NOAA's Pacific Marine Environmental Laboratory. (Information is available at http://ferret. pmel.noaa.gov/Ferret.). This work is co-funded by ANR project ANR09-BLAN-0233-01 and INSU/LEFE project Solwara; it is a contribution to the CLIVAR/SPICE International program (http://www.clivar.org). References Alory, G., Vega, A., Ganachaud, A., Despinoy, M., 2006. Influence of upwelling, subsurface stratification, and heat fluxes on coastal sea surface temperature off southwestern New Caledonia. J. Geophys. Res. Oceans 111. Bonjean, F., Lagerloef, G.S.E., 2002. Diagnostic model and analysis of the surface currents in the tropical Pacific Ocean. J. Phys. Oceanogr. 32, 2938–2954. Ceccarelli, et al., 2013. The Coral Sea: physical environment, ecosystem status and biodiversity assets. In: Lesser, Michael (Ed.), Advances in Marine Biology vol. 66. Academic Press, AMB, UK, pp. 213–290. Chaigneau, A., Le Texier, M., Eldin, G., Grados, C., Pizarro, O., 2011. Vertical structure of mesoscale eddies in the eastern South Pacific Ocean: a composite analysis from altimetry and Argo profiling floats. J. Geophys. Res. 116, C11025. http://dx.doi.org/10. 1029/2011JC007134. Chelton, D.B., Schlax, M.G., Samelson, R.M., 2011. Global observations of nonlinear mesoscale eddies. Prog. Oceanogr. (ISSN: 0079-6611) 91 (2), 167–216. http://dx.doi.org/ 10.1016/j.pocean.2011.01.002 (October 2011). Choukroun, S., Ridd, P.V., Brinkman, R., McKinna, L.I.W., 2010. Surface circulation in the western Coral Sea and residence times in the Great Barrier Reef. J. Geophys. Res. 115, C06013. http://dx.doi.org/10.1029/2009JC005761.

Condie, S.A., Dunn, J.R., 2006. Seasonal characteristics of the surface mixed layer in the Australasian region: implications for primary production regimes and biogeography. Mar. Freshw. Res. 57 (6), 569–590. Couvelard, X., Marchesiello, P., Gourdeau, L., Lefèvre, J., 2008. Barotropic zonal jets induced by islands in the southwest Pacific. J. Phys. Oceanogr. 38, 2185–2204. Cravatte, S., Ganachaud, A., Duong, Q., Kessler, W.S., Eldin, G., Dutrieux, P., 2011. Observed circulation in the Solomon Sea from SADCP data. Prog. Oceanogr. 88 (1–4), 116–130. De Mey, P., Ménard, Y., 1989. Synoptic analysis and dynamical adjustment of GEOS 3 and Seasat altimeter eddy fields in the northwest Atlantic. J. Geophys. Res. 94, 6221–6231. Delcroix, T., Eldin, G., 1995. Observations Hydrologiques dans l'Océan Pacifique Tropical Ouest. Campagnes SURTROPAC 1 à 17, de janvier 1984 à août 1992, campagnes COARE156 1 à 3, d'août 1991 à octobre 1992. TDM 141. ORSTOM Editions, Paris, p. 78. Donguy, J.-R., Oudot, C., Rougerie, F., 1970. Circulation superficielle et subsuperficielle en mer du Corail et a 170°E, Cah. ORSTOM. Ser. Oceanogr. 8 (1), 3–20. Durand, F., Shankar, D., Birol, F., Shenoi, S.S.C., 2008. Estimating boundary currents from satellite altimetry: a case study for the east coast of India. J. Oceanogr. 64 (6), 831–845. Furue, R., McCreary, J.P., Benthuysen, J., Phillips, H.E., Bindoff, N.L., 2013. Dynamics of the Leeuwin current: part 1. Coastal flows in an inviscid, variable-density, layer model. Dyn. Atmos. Oceans 63, 24–59. Ganachaud, A., Gourdeau, L., Kessler, W., 2008a. Bifurcation of the subtropical south equatorial current against New Caledonia in December 2004 from a hydrographic inverse box model. J. Phys. Oceanogr. 38, 2072–2084. Ganachaud, A., Brassington, G., Kessler, W., Mechoso, C.R., Wijffels, S., Ridgway, K., Cai, W., Holbrook, N., Sutton, P., Bowen, M., Qiu, B., Timmermann, A., Roemmich, D., Sprintall, J., Neelin, D., Lintner, B., Diamond, H., Cravatte, S., Gourdeau, L., Eastwood, P., Aung, T., 2008b. Southwest Pacific Ocean circulation and climate experiment—part II. Implementation plan. CLIVAR Publication Series No. 133, NOAA OAR Special Report. International CLIVAR Project Office (www.clivar.org). Ganachaud, A., Vega, A., Rodier, M., Dupouy, C., Maes, C., Marchesiello, P., Eldin, G., Ridgway, K., Le Borgne, R., 2010. Observed impact of upwelling events on water properties and biological activity off the southwest coast of New Caledonia. Mar. Pollut. Bull. 61, 449–464. Ganachaud, A., Cravatte, S., Melet, A., Schiller, A., Holbrook, N., Sloyan, B., Widlansky, M., Bowen, M., Verron, J., Wiles, P., Ridgway, K., Sutton, P., Sprintall, J., Steinberg, C.R., Brassington, G., Cai, W., Davis, R., Gasparin, F., Gourdeau, L., Hasegawa, T., Kessler, W., Maes, C., Takahashi, K., Richards, K.J., Send, U., 2014. The southwest Pacific Ocean and climate experiment (SPICE). J. Geophys. Res. Oceans, Spec. Issue West. Pac. Ocean Circ. Clim. 119, 7660–7686. http://dx.doi.org/10.1002/2013JC009678. Gasparin, F., Ganachaud, A., Maes, C., 2011. A western boundary current east of New Caledonia: observed characteristics. Deep-Sea Res. I Oceanogr. Res. Pap. 58, 956–969. Gasparin, F., Maes, C., Sudre, J., Garcon, V., Ganachaud, A., 2014. Water mass analysis of the Coral Sea through an optimum multiparameter method. J. Geophys. Res. Oceans 119. http://dx.doi.org/10.1002/2014JC010246. Godfrey, J.S., Ridgway, K.R., 1985. The large-scale environment of the poleward-flowing Leeuwin Current, Western Australia: longshore steric height gradients, wind stresses and geostrophic flow. J. Phys. Oceanogr. 15, 481–495. Godfrey, J.S., Weaver, A.J., 1991. Is the Leeuwin Current driven by Pacific heating and winds? Prog. Oceanogr. 27, 225–272. Gourdeau, L., 1998. Internal tides observed at 2°S, 156°E by in situ and TOPEX/POSEIDON data during the Coupled Ocean-Atmosphere Response Experiment (COARE). J. Geophys. Res. 103 (C6), 12629. Gourdeau, L., Kessler, W.S., Davis, R.E., Sherman, J., Maes, C., Kestenare, E., 2008. Zonal jets entering the Coral Sea. J. Phys. Oceanogr. 38 (3), 715–725. Grenier, M., Jeandel, C., Lacan, F., Vance, D., Venchiarutti, C., Cros, A., Cravatte, S., 2013. From the subtropics to the central equatorial Pacific Ocean: neodymium isotropic composition and rare earth element concentration variations. J. Geophys. Res. 118, 592–618. http://dx.doi.org/10.1029/2012JC008239. Grenier, M., Jeandel, C., Cravatte, S., 2014. From the subtropics to the equator in the southwest Pacific: continental material fluxes quantified using neodymium parameters along modeled thermocline water pathways. J. Geophys. Res. 119 (6), 3948–3966. Hénin, C., Cresswell, G., 2005. Upwelling along the western barrier reef of New Caledonia. Mar. Freshw. Res. 56, 1005–1010. Henin, C., Guillerm, J.-M., Chabert, L., 1984. Circulation superficielle autour de la NouvelleCalédonie. Oceanogr. Trop. 19 (2), 113–126. Huang, R.X., Qiu, B., 1998. The structure of the wind-driven circulation in the subtropical Pacific Ocean. J. Phys. Oceanogr. 28, 1173–1186. Hummon, J., Firing, E., 2003. A direct comparison of two RDI shipboard ADCPs: a 75-kHz ocean surveyor and a 150-kHz narrow band. J. Atmos. Ocean. Technol. 20 (6), 872–888. Kessler, W., Cravatte, S., 2013a. Mean circulation of the Coral Sea. J. Geophys. Res. Oceans 118, 6385–6410. http://dx.doi.org/10.1002/2013JC009117. Kessler, W.S., Cravatte, S., 2013b. ENSO and short-term variability of the south equatorial current entering the Coral Sea. J. Phys. Oceanogr. http://dx.doi.org/10.1175/JPO-D-120113.1. Kessler, W.S., Gourdeau, L., 2007. The annual cycle of circulation of the southwest subtropical pacific, analyzed in an ocean GCM. J. Phys. Oceanogr. 37 (6), 1610–1627. http://dx.doi.org/10.1175/JPO3046.1. Maes, C., Gourdeau, L., Couvelard, X., Ganachaud, A., 2007. What are the origins of the Antarctic intermediate waters transported by the North Caledonian Jet? Geophys. Res. Lett. 34, L21608. http://dx.doi.org/10.1029/2007GL031546. Marchesiello, P., Lefevre, J., Vega, A., Couvelard, X., Menkes, C., 2010. Coastal upwelling, circulation and heat balance around New Caledonia's barrier reef. Mar. Pollut. Bull. 61, 432–448.

S. Cravatte et al. / Journal of Marine Systems 148 (2015) 249–271 McCreary, J.P., Shetye, S.R., Kundu, P.K., 1986. Thermohaline forcing of eastern boundary currents: with application to the circulation off the west coast of Australia. J. Mar. Res. 44, 71–92. Melet, A., Gourdeau, L., Verron, J., 2010. Variability in Solomon Sea circulation derived from altimeter sea level data. Ocean Dyn. 60, 883–900. Menkes, et al., 2014. Seasonal oceanography from physics to Micronekton in the southwest Pacific. Deep Sea Res. II http://dx.doi.org/10.1016/j.dsr2.2014.10.026. Morrow, R.A., Fang, F., Fieux, M., Molcard, R., 2003. Anatomy of three warm-core Leeuwin Current eddies. Deep Sea Res. II 50 (2229–2243), 2003. Pedlosky, J., Iacono, R., Napolitano, E., Helfrich, K., 2009. The skirted island: the effect of topography on the flow around planetary scale islands. J. Mar. Res. 67 (4), 435–478. Qiu, B., Chen, S., 2004. Seasonal modulations in the eddy field of the South Pacific Ocean. J. Phys. Oceanogr. 34, 1515–1527. Qiu, B., Scott, R.B., Chen, S., 2008. Length scales of eddy generation and nonlinear evolution of the seasonally modulated South Pacific subtropical countercurrent. J. Phys. Oceanogr. 38 (7), 1515–1528. Qiu, B., Chen, S., Kessler, W.S., 2009. Source of the 70-day mesoscale eddy variability in the Coral Sea and North Fiji Basin. J. Phys. Oceanogr. 39, 404–420. Qu, T., Lindstrom, E.J., 2002. A climatological interpretation of the circulation in the western South Pacific. J. Phys. Oceanogr. 32 (9), 2492–2508.

271

Qu, T., Lindstrom, E.J., 2004. Northward intrusion of Antarctic intermediate water in the western Pacific. J. Phys. Oceanogr. 34 (9), 2104–2118. Qu, T., Gao, S., Fukumori, I., Fine, R.A., Lindstrom, E.J., 2009. Origin and pathway of equatorial 13 °C water in the Pacific identified by a simulated passive tracer and its adjoint. J. Phys. Oceanogr. 39 (8), 1836–1853. Ridgway, K.R., Dunn, J.R., 2003. Mesoscale structure of the mean East Australian Current System and its relationship with topography. Prog. Oceanogr. 56 (2), 188–222. Ridgway, K.R., Godfrey, J.S., Meyers, G., Bailey, R., 1993. Sea level response to the 1986–1987 El-Nino southern oscillation event in the western Pacific in the vicinity of Papua New Guinea. J. Geophys. Res. Oceans 98 (C9), 16387–16395. Ridgway, K.R., Dunn, J.R., Wilkin, J.L., 2002. Ocean interpolation by four-dimensional weighted least squares — application to the waters around Australasia. J. Atmos. Ocean. Technol. 19 (9), 1357–1375. Rotschi, H., Lemasson, L., 1967. Oceanography of the Coral and Tasman Seas. Oceanogr. Mar. Biol. Annu. Rev. 5, 49–97. Tomczak, M., Hao, D., 1989. Water masses in the thermocline of the Coral Sea. Deep Sea Res. Part A 36 (10), 1503–1514. Webb, D., 2000. Evidence for shallow zonal jets in the South Equatorial Current region of the southwest Pacific. J. Phys. Oceanogr. 30, 706–720.