The Southern Caribbean upwelling system off Colombia: Water masses and mixing processes

Deep–Sea Research I 155 (2020) 103145

Contents lists available at ScienceDirect

Deep-Sea Research Part I journal homepage: http://www.elsevier.com/locate/dsri

The Southern Caribbean upwelling system off Colombia: Water masses and mixing processes � Marco Correa-Ramirez a, *, Angel Rodriguez-Santana b, Constanza Ricaurte-Villota a, c Jorge Paramo a b c

Program of Marine and Coastal Geosciences, Institute of Marine and Coastal Research (INVEMAR), Santa Marta, Colombia Departamento Fisica, Universidad de Las Palmas de Gran Canaria, Spain Grupo de Investigaci� on en Ciencia y Tecnología Pesquera Tropical (CITEPT), Universidad Del Magdalena, Santa Marta, Colombia

A R T I C L E I N F O

A B S T R A C T

Keywords: Coastal upwelling Water masses Ocean mixing processes Caribbean sea

The upwelling system off the southern Caribbean coast is probably the main source of the nutrients that support biological productivity in the oligotrophic Caribbean Sea. Subtropical underwater (SUW) that forms the sub­ surface salinity maximum in the Caribbean Sea is the main source of upwelled waters in this system. Profiles of salinity and temperature with depth derived from four oceanographic cruises and Argo floats showed that upwelled waters have a salinity that is ~0.11 g kg 1 lower than the SUW in the central Caribbean Sea and a seasonal variation of approximately 0.09 g kg 1 that reflects the rainy/dry seasons. In addition, the SUW is ~50 m shallower on the continental shelf slope (~100 m) compared to the depth of the SUW in the central Caribbean Sea. The origin of these modified SUW was analyzed using the Mercator numerical model, which reproduces the main vertical characteristics of the subsurface salinity maximum. The modeled data showed that SUW upwelling off of the La Guajira Peninsula and Venezuela arrive into the system via an intense Caribbean Coastal Undercurrent (CaCU, mean speed ~0.28 m s 1). This current is formed in front of the Nicaragua platform from the divergence of subsurface water flow at the salinity maximum depth. The lower salinity observed in the upwelled waters may be the result of intense vertical mixing processes that could occur when the SUW are transported by the CaCU below the Panama-Colombia Gyre (PCG) region before reaching the upwelling zones. The mixing processes—involving double diffusion and mechanical turbulence driven by vertical shear of hori­ zontal currents—were analyzed using the Turner Angle and the Thorpe scale, respectively. Below a depth of 200 m, double diffusion by salt fingers (diffusivities > 5 x 10 5 m2 s 1) was the main process of salt diffusion, generating a downward salt flux of >2 x 10 2 g kg 1 m d 1 between the SUW and the North Atlantic Central Waters (NACW). Above a depth of 100 m, mechanical turbulent diffusion generates a salt flux towards the surface ranging 0.5–4 x 10 2 g kg 1 m d 1, where double diffusion by salt finger is not possible. The diluted SUW is subsequently transported by the CaCU, connecting—at the subsurface level—the PCG region with the upwelling zones off of Colombia and Venezuela. As well as modifying the salt content of the coastal SUW, these mixing processes may also alter the nutrient content of upwelling waters, the ecosystem effects of which are still unknown.

1. Introduction Trade winds over the Caribbean Sea generate northward Ekman transport and the upwelling of subsurface waters on the southern coast of the basin, off Colombia, Venezuela, and Trinidad (Andrade and Bar­ ton, 2005; Gordon, 1967; Rueda-Roa and Muller-Karger, 2013). This South Caribbean coastal Upwelling System (SCUS) is a low latitude

tropical upwelling system (~10 � N), the timing and spatial variability of which is determined by a combination of coastline orientation, the intensification of the trade winds in the Caribbean low-level jet (Ama­ ~ oz et al., 2008; Wang, 2007), and dor, 2008; Hidalgo et al., 2015; Mun the zonal differences in the stratification of the water column (Rue­ da-Roa and Muller-Karger, 2013). In the SCUS, about 21 upwelling foci have been identified (Castellanos et al., 2002; Paramo et al., 2011) that

* Corresponding author. E-mail address: [email protected] (M. Correa-Ramirez). https://doi.org/10.1016/j.dsr.2019.103145 Received 8 August 2019; Received in revised form 11 October 2019; Accepted 24 October 2019 Available online 4 November 2019 0967-0637/© 2019 The Author(s). Published by Elsevier Ltd. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/).

M. Correa-Ramirez et al.

Deep-Sea Research Part I 155 (2020) 103145

are distributed in the following two zones of different upwelling in­ tensity: (1) a Western Upwelling Zone (WUZ; 74–71� W) characterized by intense winds that generate an intense seasonal upwelling and high offshore transport of upwelled waters (Andrade and Barton, 2005); and (2) an Eastern Upwelling Zone (EUZ; 71–60� W) of less intense but upwelling-favorable conditions across the year that drive less offshore transport, where stratification favors the pumping of colder and deeper waters than those upwelled in the WUZ (Rueda-Roa and Muller-Karger, 2013). In addition to along shore winds, the Ekman pumping forced by the wind curl also produces a minor upwelling, which is only significant for the EUZ as it cannot overcome the high thermal stratification present in the WUZ (Rueda-Roa and Muller-Karger, 2013). The intraseasonal variability of the SCUS has been associated to short-term atmospheric perturbations that interrupt the upwelling process and induce down­ �nchez et al., 2018). In the last welling in the upper layer (Montoya-Sa three decades the upwelling has shown a trend to intensify, producing a moderate cooling of the sea surface temperature in the WUZ, which contrasts with the warming observed in most of the Caribbean Sea (Santos et al., 2016). The nutrient concentrations of upwelled waters in the SCUS increase phytoplankton productivity (Corredor, 1979; Muller-Karger et al., 2001, 1989), stimulates growth, and changes the ecological structure of coral reefs and macroalgal meadows (Diaz-Pulido and Garzon-Ferreira, 2002; Eidens et al., 2014). Estimations suggest that approximately 95% of the small pelagic biomass in the Southern Caribbean sea is sustained by the increase in biological productivity stimulated by the upwelled waters of the SCUS (Rueda-Roa and Muller-Karger, 2013). Upwelled waters in the SCUS derive from subtropical underwater (SUW) and share the geochemical composition of this water mass (Muller-Karger et al., 2001). It has been suggested that the low nutrient concentration of SUW is responsible for the lower phytoplankton growth observed in the SCUS compared to that typically observed in the large eastern boundary cur­ rents systems (EBCS, i.e., the Humboldt, California, Benguela, and Canarias Currents) (Corredor, 1979). The SUW originates in the central tropical Atlantic and freely enters into the Caribbean basin through the numerous passages between the Greater and Lesser Antilles (with a mean sill depth of approximately 1200 m) (Gordon, 1967). This forms a subsurface salinity maximum (SSM) at 100–200 m depth, which extends �ndez-Guerra and Joyce, 2000). How­ throughout the Caribbean (Herna ever, early hydrographic profiles developed in the Caribbean Sea have shown that the southern end of the SSM lacks spatial continuity off the Colombian and Venezuelan coasts where it is interrupted by the east­ ward flow of the Caribbean Coastal Undercurrent (CaCU). This provides a possible return flow for Atlantic waters that have entered the Carib­ �ndez-Guerra and Joyce, 2000). The waters trans­ bean basin (Herna ported by the CaCU may have a more recent origin than the SUW; it has been suggested that the CaCU originates at the surface, in the region of the Panama-Colombia gyre (PCG), and progressively deepens as it moves eastwards at a rate of approximately 0.1 m s 1, reaching a depth of ~100 m in the Colombian basin and ~200 m in the Cariaco basin (Andrade et al., 2003). It is still unknown whether the CaCU consists of a permanent flow or if it is spatially continuous along the southern Caribbean coast. This is largely because the available evidence for this flow comes from isolated hydrographic campaigns performed in different years (Andrade et al., 2003). The relationship between the CaCU, the SUW, and the coastal SSM with upwelled waters in the SCUS remains unclear. In the large EBCS, coastal undercurrents transport aged waters of equatorial origin, with low oxygen and high nutrient concentrations (Chavez and Messi� e, 2009). Throughout its trajectory beneath the up­ welling zones, a portion of the nutrients transported by the undercurrent waters can be upwelled directly or mixed with shallower waters that are eventually pumped by the upwelling (Glessmer et al., 2009; Messi�e et al., 2009). Therefore, the total amount of nutrients supplied by up­ welling depends on the original nutrient concentration of subsurface waters, the intensity of the vertical transport generated by winds

(relative to the depth of the Ekman layer), and the mixing processes of nutrients that can occur between water masses in the water column. Some estimations suggest that turbulent mixing processes transport approximately 25% of the upwelled nutrients to the surface in the EBCS—transport that is not reversible during periods of upwelling relaxation when isopycnals descend to their original depth (Hales, 2005). In addition, the higher salt concentrations in the subsurface waters in comparison to the waters below them makes these waters prone to double diffusion processes, which occur alongside turbulent mixing processes in these upwelling systems (Arcos-Pulido et al., 2014). In the mid-latitudes, instabilities generated by double diffusion pro­ cesses can generate nutrient fluxes that are similar in magnitude to those induced by mechanical turbulence or mesoscale eddies (Oschlies et al., 2003). The magnitudes of these turbulent mixing and double diffusion processes have not previously been characterized for the SCUS. This information is required, however, to understand the levels of biological productivity observed here. Using observational and modeled data, this work aims to analyze the origin and vertical structure of the subsurface salinity maximum and its relationship with upwelled waters in the western zone of the SCUS. A description of the diffusive processes occurring in the water column is made first, and the influence of these processes on nutrient transport in this upwelling system is then discussed. 2. Methods 2.1. Observational data The vertical distribution and the characteristics of the water masses observed in the SCUS were analyzed using conductivity, temperature, and depth (CTD) data from four different survey cruises. Two of these surveys were made on the continental shelf slope off Colombia by the National Hydrocarbons Agency (ANH) at the beginning of the 2008 rainy season (ANH-I, May–June), and after the seasonal maximum of the 2009 rainy season (ANH-II, November–December) (Fig. 1a) that co­ incides with the seasonal maximum of river discharge into the Colom­ bian Basin of the Caribbean Sea (Beier et al., 2017). A total of 36 stations were surveyed on the ANH cruises at depths greater than 200 m using a self-contained Ocean Seven 316Plus profiler equipped with IDRONAUT full ocean depth, pump-free, and long-term stability sensors. The other two surveys were carried out as part of the “Marine Pro­ tected Areas (MPAs): a management tool for demersal fisheries in the northern zone of the Colombian Caribbean” project (Universidad del Magdalena - COLCIENCIAS code 020309-16652). Eighty-six coastal CTD profiles were surveyed during the MPAs cruises within the areas directly affected by coastal upwelling on the continental shelf off La Guajira Peninsula (Fig. 1b) using a Sea Bird SB-19 CTD profiler. MPA cruises were made at the beginning of the upwelling season in December 2005 (MPAs-I) and during the seasonal peak of upwelling in February 2006 (MPAs-II). Data for 13 (2008) and 28 (2009) CTD profiles were obtained using free-drifting Argo floats that crossed through an area of the central Caribbean Sea (between 75 and 68� W and 13.5–18� N) from the Copernicus Marine Environment Monitoring Service (http://marine. copernicus.eu). These profiles were compared with the ANH profiles to estimate differences in the subsurface salinity maxima between the coast and the central Caribbean Sea. Nineteen high-resolution profiles �-Colombia gyre region (~2 m) from a Argo float surveying the Panama between December 2018 and February 2019, were also obtained to es­ timate the offshore mixing parameters in this region. Instrument noise and error associated with the determination of salinity from the CTD measurements are generally propagated into the derived density profiles as noise and spike-like anomalies, which bias further mixing estimations (Gargett and Garner, 2008). These errors are mostly associated with pressure reversals due to ship roll and the salinity spiking caused by the differing time responses of the temperature and 2

M. Correa-Ramirez et al.

Deep-Sea Research Part I 155 (2020) 103145

Fig. 1. Locations of the CTD hydrographic stations. The upper image show the position of the Argo floats profiles surveyed in the central Caribbean sea during August to December 2008 (Gary diamonds) and January to November 2009 (black diamonds). Black triangles show the position of the profiles surveyed by an Argo float in the Panam� a-Colombia gyre region during December 2018 to January 2019. a) Stations from the National Hydrocarbons Agency of Colombia (ANH) were surveyed at the beginning of the 2008 rainy season (ANH-I, magenta dots) and after the seasonal peak of the 2009 rainy season (ANH-II, blue dots); b) The Marine Protected Areas project stations (MPAs) were surveyed at the beginning (MPAs-I, magenta dots) and during the seasonal maximum (MPA-II, blue dots) of the up­ welling season in 2005–2006. The continuous lines between the CTD stations represent the position of the hydrographic sections shown in Figs. 4–7 and in Fig. 12. The edge of the continental shelf is indicated as the 200-m isobath (gray line). (For interpretation of the references to color in this figure legend, the reader is referred to the Web version of this article.)

conductivity sensors. To reduce these errors, measurements located in pressure reversal loops were excluded from all of the CTD profiles, conserving only the consecutive salinity and temperature measurements associated to pressure increases in the downward direction along the profile. Most of the spikes in the salinity profiles (S) were removed using a salinity reference profile (Sref) that was obtained from seven-point median filtering of the S profiles. The CTD measurements at depths where ðSðiÞ SrefðiÞ Þ > 2 *stdS were excluded, where stdS is calculated as follows: sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi n �2 1X stdS ¼ SðiÞ Sref ðiÞ (1) n i¼1

mean square (rms) of the detrended density anomalies over successive 10 m segments from a selected “well-mixed” layer of each profile. It was assumed that the minimum rms value of each profile must come from the most mixed layer, and was considered as a multiple of 4 as a threshold value. The obtained threshold value was approximately 5.8 x 10 3 kg m 3. Based on this, two new density profiles were created by convolution in the top–bottom and bottom–top directions, where each density value was maintained as equal to the previous value if the ab­ solute difference between them did not exceed the threshold value; otherwise, the current density value was conserved. The intermediate density profile results from the average of these two profiles (Gargett and Garner, 2008). Water mass analysis was undertaken using graphs of potential tem­ perature against absolute salinity (Θ-SA plots). Some CTD profiles were selected to produce depth sections along the coast, which were inter­ polated with Data Interpolating Variational Analysis (DIVA; Troupin et al., 2012) to produce continuous fields. Satellite information of sea surface temperature and chlorophyll from the MODIS-Aqua mission (htt ps://oceancolor.gsfc.nasa.gov), and winds from the Blended Sea Winds combined product (https://www.ncdc.noaa.gov/data-access/marine ocean-data/blended-global/blended-sea-winds), were also used to establish the start and intensity of the 2005–2006 upwelling season.

With using this procedure, approximately 5% of the salinity data were identified as anomalous and were excluded. The remaining CTD data were averaged at regular depth intervals of 0.5 m. On average, approximately two to three scans fell within each 50 cm averaged in­ terval, which is consistent with a mean fall rate of ~0.25 m s 1 and the 1 Hz sampling frequency of the CTD measurements. The remaining noise in the density profiles was eliminated by constructing an intermediate density profile (Fig. 2a) as explained by Gargett and Garner (2008). This method considers only significant differences in the density profiles that are larger than a threshold noise level, which is calculated as the root 3

M. Correa-Ramirez et al.

Deep-Sea Research Part I 155 (2020) 103145

Fig. 2. Sample illustration showing the estimation of the Thorpe-scale-derived diffusivities at stations 267 (ANH-I, ae) 277 (ANH-II, f-j): (a,f) Intermediate density profile (black line). The sub­ panel is an enlargement where the blue and red lines shown the calculated pro­ files in the top-bottom and bottom-top directions, respectively, which were averaged to create the intermediate profile. The SA profile is depicted as a gray line; (b,g) the Brünt-Vaisala fre­ quency; (c,h) Thorpe displacements (L, gray line) and the Thorpe scale (LT, black line). The red line shows the excluded LT values corresponding to non-symmetric overturns where Ro < 0.25; (d,i) Diffusivity profile calculated with the Osborn parameteri­ zation (Eq. (7)); (e,j) Salt flux calculated using Fick’s law (Eq. (8)). (For inter­ pretation of the references to color in this figure legend, the reader is referred to the Web version of this article.)

2.2. Modeled data

where α is the coefficient of thermal expansion; β is the saline contrac­ tion coefficient; Θ is the conservative temperature; and SA is absolute salinity. The Turner Angle (Tu) is an estimate that allows to determine the relative influence of temperature and salinity in the stratification of the water column and the probability of double diffusion processes. The first of the two arguments of the arctangent function is the “y”-argument and the second one the “x”-argument in a Cartesian plane. The Turner angle Tu is quoted in degrees of rotation. If 45�
In order to analyze the subsurface circulation of the SUW, numerical outputs from the Operational Mercator ocean model (Global Ocean Analysis and Forecast System) were obtained from the Copernicus Ma­ rine Environment Monitoring Service (http://marine.copernicus.eu). This model has a spatial resolution of 1/12 of a degree, starts on December 27, 2006, and produces daily and monthly mean fields for temperature, salinity, currents, sea level, mixed layer depth, and ice parameters across the ocean. It also includes hourly mean surface fields for sea level height, temperature, and currents. Using the monthly Mercator outputs, the vertical sections in the same location in the ANH-I section were reproduced to compare modeled and surveyed data. Hor­ izontal maps at the depth of the subsurface salinity maximum were produced to describe the SUW circulation in the Caribbean Sea.

Ksf ¼ 2.3. Diapycnal mixing parameters

and the Turner angle (Ruddick, 1983; Turner, 1973): � � ∂Θ ∂SA ∂Θ ∂SA Tu ¼ tan 1 αΘ þ βΘ ; αΘ βΘ ∂z ∂z ∂z ∂z

(4)

where K ¼ 10 4 m2 s 1 is the maximum diapycnal diffusivity associated with salt fingers; Rc ¼ 1:6 is the threshold of Rc ¼ 1:6 density ratio where mixing due to salt fingers falls sharply due to the absence of staircases; and n ¼ 6 is an index to control the Ksf decay rate with the increase of Rρ (Zhang et al., 1998). The background coefficient K∞ ¼ 3x10 5 m2 s 1 is the diapycnal diffusion constant due to processes unrelated to double diffusion, such as internal wave breaking, that should be considered when estimating the total diffusivity of the salt. This coefficient was not considered in the Ksf estimates to guarantee a direct comparison between only the double diffusion process and me­ chanical diffusion occurring due to shear instabilities. The CTD measurements can also provide valuable information on the mixing induced by vertical shear of the horizontal flow throughout the Thorpe scale, which is an estimate of the vertical overturning scale

The diffusion of properties generated by turbulent mixing is several orders of magnitude more efficient than molecular diffusion (Thorpe, 2005). Double diffusion by salt fingers is one of the most important processes of turbulent mixing in the central waters of the tropical and subtropical oceans (Schmitt, 1981). The susceptibility of the water col­ umn to vertical convection through double diffusion processes can be evaluated with the Stability Ratio Rρ (Thorpe, 2005a): Rρ ¼ α∂Θ=β∂SA

K þ K∞ 1 þ ðRρ=RcÞn

(2)

(3)

4

M. Correa-Ramirez et al.

Deep-Sea Research Part I 155 (2020) 103145

associated with large turbulence eddies in an otherwise stably stratified fluid (Thorpe, 2005). To quantify the Thorpe scale, density profiles are sorted to produce an ideal stable state whilst keeping track of the min­ imum vertical distance each parcel of water has to move to establish this stable condition. The Thorpe displacement assigned to depth di is: Li ðdi Þ ¼ df

di ;

were calculated following Fick’s law: Fx ¼ Kx *∂SA =∂z

where x could take the form of sf or T depending on the specific case. 3. Results

(5)

where df is the depth to which the point originally at di has been moved. Thus, positive and negative Thorpe displacements correspond to downward and upward relocation of a water parcel by overturning, respectively. The Thorpe length scale (LT) is the rms of all Thorpe dis­ placements within a complete overturn cell, defined as a vertical dis­ tance over which Thorpe displacements sum to zero. In practice, the bounds of overturn cells are determined by those locations where the cumulative sum of the Thorpe displacements drops back to zero (Fig. 2c). Here, questionable instabilities were detected and removed from LT profiles using an overturn verification through the overturn ratio: � Ro ¼ min Lþ =L; L =L (6)

3.1. Water masses and mixing processes in the upwelling system The Θ-SA plot of CTD data from the ANH survey cruises (Fig. 3a) shows the presence of four water masses off the Colombian coast. At the surface (above the first 50 m), most of the SA and Θ values are in the range of typical values for Caribbean Surface Water (CSW, SA ~35 g kg 1, Θ ~29 ΊC), which is a water mass formed in the Caribbean Sea by a mixture of North Atlantic Surface Waters (NASWs) and riverine waters from the Orinoco and Amazon Rivers. Direct contributions from rain and freshwater from the South American rivers discharging within the basin are also a component of NASWs (Cherubin and Richardson, 2007; Wόst, 1964). Because the ANH-II survey was carried out during the rainy season, in 2009, the surface waters were more diluted than is typical for CSW west of 76.5ΊW, with salinity values < 33 g kg 1 (gray dots in Fig. 3). This observed low salinity is the result of a greater dilution in the PCG region caused by large rivers discharging into the Gulf of Darien (Beier et al., 2017). Under the CSW, higher salinity waters are observed with values close to those typical for Subtropical Under Water (SUW, SA ~37, Θ ~22 � C, Fig. 3a), forming a subsurface salinity maximum (SSM) in the water column. The observed salinity in the SSM during the ANH-I (37.00 � 0.04 g kg 1) and ANH-II (37.09 � 0.02 g kg 1) surveys was 0.05–0.14 g kg 1 lower than that observed in the central Caribbean (37.14 � 0.06 g kg 1) by Argo floats during the same years of the ANH cruises. In addition, there is a significant difference of 0.09 g kg 1 be­ tween the SSM salinity of the ANH profiles measured at the beginning of the rainy season (ANH-I, May–June 2008) with respect to those measured after the peak of the rainy season (ANH-II, November–De­ cember 2009). This difference could suggest a seasonal intensification in mixing processes, with a greater dilution of the SUW waters with the less-saline surface waters during the rainy season. In the center of the Caribbean Basin, the SSM formed with SUW is �ndez-Guerra and located at a depth of approximately 150 m (Herna Joyce, 2000). The ANH hydrographic sections at the edge of the conti­ nental shelf off Colombia showed the SSM at a depth of approximately 100 m depth, which is shallower than the SSM in the central Caribbean Sea (Fig. 4c and d). At the beginning of the rainy season (ANH-I) the SSM reaches the surface east of 73� W in front of La Guajira Peninsula

where L is the total vertical extent of an overturn cell, and Lþ and L are the cumulative extents occupied by positive and negative Thorpe dis­ placements within the cell, respectively. Since values of Ro < 0.25 sug­ gest non-symmetric density overturns that could be caused by residual salinity spiking (Park et al., 2014), the LT associated with these Ro values are excluded. Salt diffusivity associated with shear turbulence (KT) is calculated using the Osborn parameterization: KT ¼ 0:128L2T N

(8)

(7)

where N is the smoothed Brünt-Vaisala frequency with a 10 m mobile mean. The regions where no overturns are detected do not necessarily mean there is no vertical mixing. The resolution of CTD sensors and the amplitude of the depth intervals in the density profiles imposes a basic constraint in overturn detection, since small overturns could exist whilst not being detected. Considering the smallest detectable overturn should be approximately 1 m (i.e., twice the vertical resolution), a conservative value of 1 x 10 6 m2 s 1 was set for the regions where no overturning was detected, as suggested by Zhu and Zhang (2018). Diffusivities below 300 m, where the respective N values were lower than 3 x 10 5 s 2, were excluded since low N values could produce high erroneous diffusivities. The resulting diffusivities were averaged over regular vertical intervals of 10 m (Fig. 2). Salt diffusive fluxes caused by both salt fingers and shear turbulence

Fig. 3. a) The conservative temper­ ature–absolute salinity (Θ-SA) plot of the CTD data from ANH-I (black dots) and ANH-II (gray dots) survey cruises; b) Θ-SA plot of the CTD profiles from the free-drifting Argo floats crossing an area of the central Caribbean Sea located between 75 and 68� W and 13.5–18� N in 2008 (black dots) and 2009 (gray dots). In both panels, the red squares represent the typical values for Caribbean Surface Water (CSW), Sub­ tropical Under Water (SUW), western North Atlantic Central Water (wNACW), and Antarctic Intermediate Water (AAIW). (For interpretation of the references to color in this figure legend, the reader is referred to the Web version of this article.)

5

M. Correa-Ramirez et al.

Deep-Sea Research Part I 155 (2020) 103145

(Fig. 3c). This upwelling of saltier and colder SUW can be a remnant effect of the seasonal intensification of upwelling-favorable winds that generally occurs during the previous dry season (Andrade and Barton, 2005). After the peak of the rainy season (ANH-II), the SSM is slightly deeper than observed before, although the progressive lifting of the SSM towards the east is still observed. Unfortunately, the ANH-II section (Fig. 3d) did not have hydrographic profiles off of La Guajira, meaning it was not possible to observe the depth of the SSM during this period of

upwelling relaxation. The high salt concentration of the SUW at the SSM depth creates favorable conditions for the double diffusion processes and the forma­ tion of salt fingers with less saline and colder waters located below the SSM (Schmitt, 2005; Schmitt et al., 1987). The vertical mixing of salt by double diffusion could be one of the processes determining the lowest salinities observed in the SUW off of Colombia. Below the SSM, the SA and Θ of the water decrease to typical values for the wNACW (i.e.,

Fig. 4. Hydrographic sections along the coastal shelf off of Colombia based on selected CTD stations of survey cruises ANH-I (left panels) and ANH-II (right panels): (a–b) Conservative temperature (Θ); (c–d) Absolute Salinity (SA); (e–f) Salt diffusivity by salt fingers (Ksf, colored tones) and Turner angle (black contours); (g–h) salt flux by salt fingers (Fsf). The dashed lines in panels c–h are the contour for SA ¼ 36.8 g kg 1, showing the depth of the subsurface salinity maximum zone. 6

M. Correa-Ramirez et al.

Deep-Sea Research Part I 155 (2020) 103145

SA ¼ 35.85 g kg 1, Θ ¼ 13 � C, Fig. 3) at a depth of approximately 300 m. At depth of 600 m, these values fall to SA ¼ 35 g kg 1 and Θ ¼ 6 � C, approaching the typical values of the AAIW (i.e., SA ¼ 34.25 g kg 1, Θ ¼ 4 � C). The Turner angles below the SSM depth range between 45� and 90� , which shows a high probability of salt finger formation in the water column below a depth of 100 m. However, the higher salt diffu­ sivities for the salt fingers (Ksf > 4 x 10 5 m2 s 1) are observed between 300 and 500 m depths where Tu > 75� (Fig. 4e and f). This suggests that the double diffusion processes are faster below wNACW waters. Despite this, the vertical gradient of salt below the SSM drives a negative (downward) salt flux via the salt fingers, which is more intense between depths of 250–350 m (Fsf > 2 x10 2 g kg 1 m d 1). This indicates that the largest double diffusive salt transport occurs between the SUW and the WNACW waters (Fig. 4g and h). The mechanical turbulence produced by the vertical shear of the horizontal currents is another process that can potentially contribute to the mixing of the SUW with the adjacent water masses above and below the SSM. Since the stratification in the water column generally opposites the occurrence of turbulence by shear, stratification also provides

information on the susceptibility to mechanical mixing. The BrüntVaisala frequency (N2) shows the Caribbean Sea stratification off of Colombia is dominated by salinity, which makes the water column highly stable to the depth of the SSM (Fig. 5a and b). The entry of lower salinity waters in the PCG region during the rainy season (ANH-II) further increases the surface stratification in the first 50 m mainly to the west of 75� W (Fig. 4b). This makes the mechanical turbulence diffusivity (KT) higher below the depth of the SSM (~3 x 10 5 m2 s 1) than in the surface layer, generating a salt flux of approximately 2 x 10 2 g kg 1 m d 1 from the SSM towards deep waters below ~300 m (Fig. 4e and f). This is lower than the flux generated by the salt fingers at the same depth. Although the KT values at the surface (<2 x 10 5 m2 s 1) are lower than the deeper ones, the intense salt gradient imposed by the entry of low salinity waters into the surface layer during the rainy season also generates a positive salt flux from the SSM to the surface. This can reach comparable rates to those observed at depth, mainly at 75.3� W near the mouth of the Magdalena River (Fig. 5f). Due to the fact that double diffusive processes are not possible above the SSM depth, me­ chanical mixing is the dominant process of salt diffusion from the SUW

Fig. 5. Hydrographic sections along the coastal shelf off of Colombia based on selected CTD stations of survey cruises ANH-I (left panels) and ANH-II (right panels): (a–b) Brünt-Vaisala frequency (N2); (c–d) Salt diffusivity associated with shear turbulence (KT); (e–f) Mechanical salt flux (FsT). The dashed lines in all panels are the contour for SA ¼ 36.8 g kg 1, showing the depth of the subsurface salinity maximum zone. 7

M. Correa-Ramirez et al.

Deep-Sea Research Part I 155 (2020) 103145

m d 1 (Fig. 6c). Below the SSM depth (<100 m), no overturns larger than the Argo profile resolution were observed and, therefore, the Kt estimates were low. Although overturns smaller than 2 m may occur and not be detected (because the low resolution of the Argos profiles), the absence of large overturns at depth could also be related to the offshore location of these profiles (~50 km north of the continental shelf). The intense subsurface coastal currents could contribute to creating the shear instabilities, thus forcing the large overturns observed in ANH

towards the surface. Mixing estimates derived from Argo float profiles in the PCG region (Fig. 6) show similar mechanical diffusion and salt fluxes as those observed at surface in the ANH profiles off of Colombia. Despite their lower vertical resolution (~2 m), large overturns of 11–15 m were observed in the first 50 m in 70% of the Argo profiles. The diffusivities derived from the Thorpe scale in the surface layer (Kt > 0.2 x 10 5 m2 s 1) gives a salt flux towards the surface of >0.2 x 10 2 g kg 1

Fig. 6. Sections from Argo floats in a transect located ~10.5� N in the Panam� a-Colombia gyre region: (a) Absolute salinity (SA); (b) Brünt-Vaisala frequency (N2); (c) Salt diffusivity associated with shear turbulence (KT); and (d) Mechanical salt flux (FT). The dashed lines in right panels show the contour for SA ¼ 36.8 g kg-1, showing the depth of the subsurface salinity maximum zone. The left panels show the mean (black line) and the standard deviation (gray area) of each section. 8

M. Correa-Ramirez et al.

Deep-Sea Research Part I 155 (2020) 103145

sections. The results above show that both the mechanical turbulence and the salt fingers can contribute to the diffusive loss of salt from the SSM and could explain the lower salinity of the SUW observed off of Colombia. However, the net loss of salt is also a function of the exposure time to these mixing processes during the movement of SUW before reaching the upwelling zones. Since no measurements of currents were made during the ANH surveys, the circulation pattern of SUW in the Caribbean was analyzed through the numerical outputs of the Mercator model.

would result in the observed salt and temperature deficits below the MSE in the modeled section. The parameterization of double diffusion is currently one of the major challenges for ocean circulation models as it significantly affects the estimates of the speed of large-scale thermoha­ line circulation (Zhang et al., 1998) and the efficiency of vertical nutrient transport (Dietze et al., 2004). Fig. 8 shows the monthly variability of salinity and currents modeled by Mercator for a section 74� W from Colombia (14.5� N, Tayrona Na­ tional Park, Santa Marta) to Haiti (18� N, Port Salut). Throughout the year, the offshore currents (north of 12� N) are observed flowing pre­ dominantly towards the west (shown by the negative values in blue tones in the left-hand panels in Fig. 8). Mesoscale variability in the currents is also observed, with narrow sectors (~1� amplitude) of strong currents flowing to the east, which are limited by sectors of similar size flowing to the west. This is a characteristic circulation pattern of mesoscale eddies often observed in transit throughout the Caribbean Sea, which represent the main variability source of currents in the basin (Centurioni and Niiler, 2003; Jouanno et al., 2008; Richardson, 2005). On the slope of the Colombian continental shelf (south of 12� N), an intense eastward flow is observed during most of the year forming a core at 100 m depth with maximum velocities of approximately 0.4 m s 1. The position of this subsurface flow corresponds to the location of the Caribbean Coastal Undercurrent (CaCU) reported by Andrade et al. (2003). However, the intensity of the CaCU reproduced by the Mercator model is higher than that reported by Andrade et al. (2003). The modeled currents also suggest that the flow of the CaCU is not perma­ nent throughout the year; the CaCU experiences its greatest intensifi­ cation between the months of January to April, weakening slightly during May to June. In July and August, the CaCU intensifies again but disappears almost completely during October. Finally, between November and December, the beginning of a new pulse of intensification is observed.

3.2. Subsurface circulation at the salinity maximum depth in the Caribbean Sea Fig. 7 reproduces the hydrographic section of the ANH-I cruise using the numerical outputs of the Mercator model for the month of June, which corresponds to the beginning of the rainy season. This section shows that the model is capable of reproducing the main hydrographic characteristics observed in the ANH section, specifically a surface layer (0–50 m) with low salinity to the west of 75� W in the PCG region; a subsurface salinity maximum located at a depth of ~100 m; and waters with characteristics similar to wNACW below a depth of 250 m. Despite this, the modeled profiles underestimate Θ (in ~2 � C) and SA (in ~0.4 g kg 1) between 150 and 250 m depths, where the main thermo­ cline and halocline are located. This causes the modeled halocline (corresponding to the 36.2 g kg 1 isohaline) to be narrower and 50 m shallower (located ~200 m deep) than the halocline observed in the ANH profiles at a depth of approximately 250 m. This underestimation also results in the SSM layer having a lower vertical depth amplitude in the Mercator sections. The underestimation of the model outputs at depth could be linked to inadequate modeling of the diffusive processes in the water column, particularly those associated with double diffusion by salt fingers between the SUW and the wNACW water masses. This

Fig. 7. Sections showing: (a) Conservative temperature (Θ); and (d) Absolute salinity (SA) based on numerical outputs of the Mercator operational model for June 2008 for the same location of the ANH-I section along the coastal shelf off of Colombia. b)The average Θ profile of the ANH-I (black solid line) and Mercator (gray solid line) sections are shown along with their respective standard deviations (dashed lines); c) Difference between the average Θ profiles of the ANH-I and Mercator sections; e) and f) show the averaged SA profiles and their differences between the ANH-I and Mercator sections in a similar way to b) and c) for the Θ profiles. 9

M. Correa-Ramirez et al.

Deep-Sea Research Part I 155 (2020) 103145

Fig. 8. Three-year climatological average (2016–2018) of Mercator model sections for a 74� W transect from Tayrona National Park, Colombia, to Port Salut, Haití. In the left-hand panels, monthly means of the zonal currents are shown in blue and red tones for westward and eastward velocities, respectively. In the right-hand panels, the climatological monthly means of salinity are shown. (For interpretation of the references to color in this figure legend, the reader is referred to the Web version of this article.)

The periods of CaCU intensification shown in Fig. 8 correspond to the seasonal intensification of coastal upwelling in the western zone of the SCUS. Off the Colombian coast, the upwelling intensifies mainly in two periods; first, between December to March and, second, during July

(Andrade and Barton, 2005) in response to the intensification of the trade winds in the Caribbean low-level jet (Hidalgo et al., 2015). During the periods of intense upwelling (January–April and July–August), the modeled salinity sections (right-hand panels in Fig. 8) show that the 10

M. Correa-Ramirez et al.

Deep-Sea Research Part I 155 (2020) 103145

waters of the SSM rise to the surface near to the Colombian coast to be subsequently advected northwards beyond 13� N. The model thus re­ produces the upwelling of subsurface waters from the SSM—as is similar to the ANH profiles—and suggests that the intensification of the CaCU could have a dynamical link with the intensification of the upwelling. Since upwelled waters come from the same depth of the CaCU, most must be replaced by waters transported by the CaCU. This could induce the observed acceleration of the CaCU, supplying new subsurface waters to the upwelling system. This also suggests that a significant portion of the SUW upwelled in the system do not come directly from the center of the Caribbean Sea but, instead, upwelled waters might come from coastal subsurface waters transported from the west of the Caribbean Sea by the CaCU. Fig. 9 shows how the depth and salt concentrations of the modeled MSE vary spatially in the Caribbean Sea. In the north of the basin, the SSM lies between 120 and 140 m and has a higher salt concentration than in the south. Here, the SA is between 37.0 and 37.1 g kg 1 (Fig. 9b and c), which is close to the original salinity of the SUW in the North Atlantic. In the southwest sector of the Caribbean Sea, the MSE de­ creases its SA to 36.9 g kg 1 and becomes shallower. During the season of intense upwelling (from January to March), the cyclonic circulation in the PCG raises the SSM, forming two domes approximately 100 m deep between 84� W and 75� W (Fig. 9d). In front of the coasts affected by the upwelling (i.e., La Guajira Peninsula and the Venezuelan coast), the SSM is at is shallowest for the whole basin at approximately 70 m deep. When the upwelling weakens during October–December, the SSM deepens again in the southern Caribbean to a depth of approximately 100 m (Fig. 9e). The currents at the depth of the SSM presented in Fig. 10 show the circulation of the SUW in the Caribbean basin. This figure shows that the SUW enter the basin through passages between the Greater and Lesser

Antilles, which agrees with Gordon (1967). After entering, SUW flow predominantly to the west, from the Venezuela basin to the continental shelf of Nicaragua and the Jamaican ridge. In some sectors, the currents of the SSM are inverted to the east due to the presence of mesoscale eddies—structures that can modify the kinetic energy in the water col­ umn up to a depth of 800 m (Jouanno et al., 2008). The SUW that flow into the Nicaragua shelf divide at around 83� W and 12� N, generating two branches that flow along the platform, one flowing to the northeast and the other flowing to the southwest. The northeast branch surrounds the continental shelf of Nicaragua and then exits towards the Cayman basin through the passages of the Jamaica ridge. The southwest branch changes direction when it reaches the continental shelf off of Costa Rica and continues eastward along the shelves of Panama, Colombia, and Venezuela, forming the CaCU. Therefore, the currents simulated by the Mercator model suggest that the CaCU is formed in front of the Nicaragua platform from the divergence of subsurface water flowing at depth of the SSM. This differs from what was suggested by Andrade et al. (2003), who propose that the CaCU has a more recent formation from the deepening of surface waters in the PCG region. The transport of SUW by the CaCU from Nicaragua could be connecting—at the subsurface level—the highly diluted region of the PCG (Beier et al., 2017), with the upwelling zones off of Colombia and Venezuela. This could explain the lower salinities observed in the upwelled waters. The simulated currents show that the CaCU has an average speed 0.28 m s 1 and displays significant seasonal variation (Fig. 9a), losing its spatial continuity throughout the southern Caribbean Sea during some periods of the year. During the intense upwelling season, cyclonic cir­ culation intensifies in the PCG region (Fig. 10a), which may contribute to a decrease in the SSM depth, as shown in Fig. 9d. During this period, the CaCU is more intense in the PCG region and in La Guajira Peninsula (70–73� W) with speeds of 0.3–0.4 m s 1 (Fig. 10a), but it is less intense

Fig. 9. a) Three-year climatological average (2016–2018) of the current velocities in the core of the CaCU located at a depth of 50–150 m off the coast of the Tayrona National Park (74� W). The dashed lines represent the standard deviation and the horizontal dashed-dotte line shows the CaCU mean speed of 2.8 m s-1. Salinity (b, c) and depth (d, e) of the subsurface salinity maximum (SSM) in the Caribbean Sea as simulated by the Mercator model are also shown for the periods of upwelling intensification (January–March) and relaxation (October–December). 11

M. Correa-Ramirez et al.

Deep-Sea Research Part I 155 (2020) 103145

Fig. 10. Subsurface currents at the depth of the SSM in the Caribbean Sea for periods of upwelling intensification (January–March) and relaxation (October–De­ cember). The blue and red tones represent westward and eastward velocities, respectively. Streamlines (with directional arrows) are depicted in yellow. (For interpretation of the references to color in this figure legend, the reader is referred to the Web version of this article.)

in Venezuela (~0.2 m s 1) and completely absent between Cabo de la Vela (71.6� W) and Cabo San Roman (70� W). This suggests that the CaCU does not have a continuous flow from Colombia to Venezuela during this time. During the upwelling relaxation season (Fig. 10b), the CaCU weakens around the La Guajira Peninsula, flows continuously between Colombia and Venezuela, and intensifies slightly around Venezuela, reaching speeds of up to 0.25 m s 1.

and minimum surface temperature. Therefore, this campaign corre­ sponded to the seasonal maximum of upwelling. The satellite images for sea surface temperature show how, in the initial stage of the upwelling season (December), low-temperature waters were restricted 100 km from the coast (Fig. 11b) while later, during the seasonal maximum (February), the upwelled waters were advected offshore beyond 14� N, thus forming an extensive tongue of cold water (Fig. 11d). The surface extension of the upwelled waters is closely related to the intensity of the upwelling-favorable winds, with a lag of up to ~4 months (Alonso et al., 2015). In the MPAs-I section at the beginning of the upwelling season, the SSM was observed between 30 m and 70 m depths (Fig. 12c). This is shallower than that observed during the ANH survey (~100 m depth). Three projections (at approximately 73.8� W, 72.9� W, and 72.2� W) of high-salinity and low-temperature SUW from the depth of the SSM to­ wards surface show the locations where upwelling was active during this time (Fig. 12c). In contrast, at the surface to the east of 71.8� W, the hightemperature and low-salinity waters produced a highly stratified water column (Fig. 12e) where upwelling had not yet started to the north of La Guajira Peninsula. During the upwelling seasonal maximum (MPAs-II), the SSM layer expanded vertically and occupied almost the entire water column from a depth of ~120 m to the surface (Fig. 12d). The Θ-SA diagram shows that most of the surface and subsurface waters corresponded to the diluted SUW (having lower salinities by approximately 0.05 g kg 1, Fig. 13) observed in the ANH sections. These waters may have been pumped from the SSM and pulled over the coastal shelf by the upwelling. The highest temperature observed in these SUW could be obtained when

3.3. The 2005–2006 upwelling season off the Colombian coast Since the ANH sections only provided information for the water column at the edge of the continental shelf, the variability of the water column above the continental shelf was also analyzed using the coastal CTD profiles made during the MPAs survey cruises (Fig. 1b). These profiles were made off La Guajira Peninsula at two different times during the 2005–2006 upwelling season (Fig. 11). During this period, Ekman offshore transport calculated from the winds along the coast began to increase in mid-October, from 1 m2 s 2 (not favorable for upwelling) to 7 m2 s 2 (favorable for upwelling). The MPAs-I survey took place in mid-December, close to the onset of the season (Fig. 11a). The upwelling of the cold subsurface waters generated by the enhanced Ekman trans­ port gradually decreased the surface temperature during this initial stage from 30 � C to 26 � C. Between December 2005 and February 2006, Ekman transport remained favorable for upwelling (~4 m2 s 2), showing several intensification phases with a synoptic periodicity of 5–7 days. During this period, the surface water temperature fell to 24 � C, which was the lowest of the season. The MPAs-II survey was carried out in mid-February 2006 in the days before the maximum Ekman transport 12

M. Correa-Ramirez et al.

Deep-Sea Research Part I 155 (2020) 103145

Fig. 11. a) Ekman transport variability based on satellite-derived alongshore winds (blue line) and sea surface tem­ peratures off La Guajira Peninsula dur­ ing the upwelling season in 2005/2006. The horizontal black bars show the dates of the MPAs-I and MPAs-II survey cruises; b) and c) shown the mean sea surface temperature and the mean winds during the MPAs-I survey (December 2005); d) and e) show the same variables during MPAs-II survey. The black dots and lines in b) and d) show the location of the CTD stations and the hydrographic sections. (For interpretation of the references to color in this figure legend, the reader is referred to the Web version of this article.)

reaching the surface, and by mixing with warmer surface waters. In fact, the Brünt-Vaisala frequency during this time showed a lower stratifi­ cation compared to the start of the upwelling season, which could favor an intense vertical mixing in the water column.

Turbulent diffusion processes are more intense by several orders of magnitude compared to molecular diffusion processes (Thorpe, 2005b). In the central waters of the oceans, the most important turbulent diffusion processes are the mechanical mixing generated by the hori­ zontal shear of currents and salt fingers (McDougall and Ruddick, 1992). These two processes have been observed to coexist with similar mag­ nitudes (KT � Ksf � 10 5 m2 s 2) in the Canary upwelling system (off the east coast of the North Atlantic) (Arcos-Pulido et al., 2014). Our observations also show that the presence of these two diffusive pro­ cesses, and at similar magnitudes. However, in the SCUS, double diffu­ sion by salt fingers was found to be the dominant process below the SSM at the depth of the NACW (~300 m), with mechanical mixing being more important above the SSM, where salt finger is not possible. The mechanical diffusivities reported in this paper were indirectly estimated from the overturns in the density profile (Thorpe scale) because there are no direct measurements available for turbulence in the region. Although it has been reported that indirect estimates are comparable with direct turbulence measurements (Park et al., 2014), additional direct measurements are necessary to adjust current estimates of

4. Discussion Compared to the mid-latitude upwelling systems in the EBCS, the SCUS has been described as a tropical upwelling system with low bio­ logical productivity due to the low nutrient content of the SUW that feed into it (Corredor, 1979). SUW form in the middle of the tropical Atlantic and enter the Caribbean at a depth of approximately 100 m (Gordon, 1967). These waters are, therefore, not subject to significant biological processes such as photosynthesis that could otherwise consume or in­ crease their nutrient contents. Therefore, before reaching the upwelling system, variations in the concentrations of nutrients in the SUW are almost exclusively due to the physical processes of mixing with adjacent water masses. These processes may explain the observed decrease in the salinity of the SUW that arise in the SCUS. 13

M. Correa-Ramirez et al.

Deep-Sea Research Part I 155 (2020) 103145

Fig. 12. Hydrographic sections on the continental shelf off La Guajira Peninsula based on selected CTD stations from the MPAs-I (December 2005, right panels) and MPAs-II (February 2006, left panels) survey cruises: (a–b) Conservative temperature (Θ); (c–d) Absolute salinity (SA); and (e–f) Brünt-Vaisala frequency (N2).

diffusion in this system. Currents simulated by the Mercator model show that the CaCU flows eastward along the continental shelf at the same depth as the SSM (50–150 m), transporting SUW at a rate of 0.28 m s 1. This is a higher velocity than previously reported by Andrade et al. (2003), possibly because their speed estimates were not based on direct measurements but rather geostrophic flow estimates from CTD data that might un­ derestimate the flow rate of the CaCU. Unfortunately, there is no way to determine the accuracy of the previous estimates or those reported here as there are no direct measurements available for the CaCU. The simu­ lated currents also suggest a remote origin of the CaCU, in front of the Nicaragua shelf, from a coastal branch of the western SUW, which return towards the east along the southern shelf of the Caribbean Basin. This contrasts with the hypothesis of CaCU formation from surface waters in the PCG region, previously proposed by Andrade et al. (2003). Indeed, our CTD observations show the high thermal and saline stratification in

the PCG region (Fig. 5b) makes the formation of the CaCU via the sub­ duction of surface waters unlikely. Our results show that the CaCU provides the diluted SUW waters that are upwelled in the SCUS and, therefore, most of the upwelled waters may not come directly from the central Caribbean Sea. Instead, the upwelled waters seem to derive from modified SUW, diluted in the PCG region located to the west of the SCUS. This region is a dilution sub-basin due to the large inflow of freshwater by rain and the discharge of large rivers into the Gulf of Darien (south-southwest Caribbean), which 3 together exceed 2800 m s 1 (Beier et al., 2017). The coefficients of mechanical diffusivity and the highly positive salt fluxes observed at the surface suggest that the SUW transported as subsurface flow by the CaCU are mixed with the low-salinity surface waters in this region. This mixing is favored by the predominant cyclonic circulation in the PCG region that generates an SSM elevation from 140 m to <100 m depths, thus increasing the contact between the SUW and the surface water. 14

M. Correa-Ramirez et al.

Deep-Sea Research Part I 155 (2020) 103145

transport of nutrients comparable to other turbulent processes of verti­ cal mixing (Arcos-Pulido et al., 2014). In the SCUS, double diffusion and mechanical mixing (mainly with surface waters) can modify nutrient concentrations in the SUW that eventually reach the surface in the up­ welling system. Despite this, it is still unknown if these processes contribute to higher or lower concentrations of nutrients in the SUW. Further studies are required to quantify the variability of nutrient diffusive transport over time and between different sectors of the SCUS. The relationship between nutrient concentrations and periods of inten­ sified or weakened CaCU, and its effect on biological productivity, also require further study in this upwelling system. 5. Conclusions Using CTD data from two different oceanographic campaigns (ANH and MPAs survey cruises) and Argo float profiles together with simu­ lated data from the Mercator oceanic numeric model, this study has shown that the upwelled waters in the Southern Caribbean Upwelling System are derived from SUW that have been slightly diluted—to a salt concentration of 0.05–0.14 g kg 1—in the PCG region. Simulated data suggest that part of the SUW entering through the passages between the Antilles and crossing the Caribbean Sea return to the east in front of the continental shelf of Nicaragua. This operated via a coastal undercurrent (CaCU) that flows along the edge of the continental shelf at a mean rate of 0.28 m s 1, passing below the main upwelling region off the coasts of Colombia and Venezuela. The CaCU is seasonally intensified by the intensification of coastal upwelling and provides most of the upwelled SUW. The transport by CaCU generates a longer retention time (between 3 and 6 months more) in the subsurface during which SUW undergoes intense turbulent mixing processes that alter their salt contents. The dilution of the coastal SUW is probably mainly the result of mechanical mixing with low-salinity surface waters in the PCG region, which occurs when these waters are transported under this region by the CaCU. In addition, SUW also have a permanent (downwards) loss of salt via double diffusion. These mixing processes may modify the nutrient content of upwelled waters, to an unknown extent, which may have important implications for the biological productivity of this system.

Fig. 13. Conservative temperature–absolute salinity (Θ-SA) plot for CTD data derived from the MPAs-I (black dots) and MPAs-II (gray dots) survey cruises. Red squares represent typical values for the CSW, SUW, and the wNACW. (For interpretation of the references to color in this figure legend, the reader is referred to the Web version of this article.)

Therefore, the CaCU acts as the main physical structure that connects, at the subsurface level, the PCG region with the upwelling areas of the SCUS. Considering that SUW in the central Caribbean Sea must travel to­ wards the Nicaragua shelf before returning back as the CaCU, the SUW flows an additional of 2600 km and 4400 km before reaching the up­ welling zones off the coast of Colombia and Venezuela, respectively. This implies that the SUWs are subjected to an additional subsurface retention time of 3.6 (6.1) months before being upwelled in front of Colombia (Venezuela) coast, during which they will be exposed to vertical salt diffusion process, assuming a continuous CaCU velocity of ~0.3 m s 1 as indicated by the Mercator model. This longer residence time may contributes to increase (or decrease) the concentration of nutrients in the upwelled SUW that support the higher levels of bio­ logical productivity observed in the upwelling system. There is extensive observational evidence of double diffusion pro­ cesses with the formation of salt fingers in the central waters of the tropical Atlantic Ocean and the Caribbean Sea (Schmitt, 2005; Schmitt et al., 1987). Salt finger formation generates turbulent salt transport approximately two orders of magnitude greater than molecular salt transport (Schmitt, 2005). This is comparable to fluxes induced by me­ chanical turbulence or mesoscale eddies (Oschlies et al., 2003). Since the main salt ions of seawater have molecular characteristics similar to the most of the nutrients, salt diffusivity is similar to nutrient diffusivity (Hamilton et al., 1989). In the subtropical North Atlantic, the transport of nutrients towards the surface layer by the salt fingers has been observed to be approximately five times greater than the transport due to other turbulent processes in the thermocline, such as that generated by internal waves (Dietze et al., 2004). Thus, the diffusion of nitrate by salt fingers could account for approximately 20% of the supply of new nitrogen to the tropical Atlantic (Fern� andez-Castro et al., 2015). Nutrient diffusion by salt fingers has not been analyzed in the SCUS, although this information could be important to fully understand the nutrient balance and the levels of biological productivity observed in this system. In the Canary upwelling system located on the eastern edge of the North Atlantic, salt fingers generate an important vertical

Declaration of competing interest The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper. Acknowledgments This work was funded by the Autonomous National Fund Patrimony “Francisco Jos� e de Caldas” of the Administrative Department of Science, Technology, and Innovation - COLCIENCIAS, through the grant No. FP44842-138-2016 COLCIENCIAS–INVEMAR, project code 210571451272. A. Rodriguez-Santana was funded by the FLUXES project (CTM2015-69392-C3-3-R) of the Spanish National Research Program and the European Regional Development Fund. J. Paramo also thanks to COLCIENCIAS and the University of Magdalena, for the financial, technical and logistical support received trough the the project "Marine recreational fishing as alternative for the development of ecological and socioeconomic tourism in Santa Marta” code 074–2017. The authors also thank the financial support of the National Investment Projects Bank of Colombia – BPIN, project code 20170110000113 “Investigaci� on científica hacia la generaci� on de informaci� on y con­ �n”, ocimiento de las zonas marinas y costeras de interes de la nacio carried out in the Marine and Coastal Research Institute - INVEMAR. INVEMAR contribution No. 1238.

15

M. Correa-Ramirez et al.

Deep-Sea Research Part I 155 (2020) 103145

References

Messi� e, M., Ledesma, J., Kolber, D.D., Michisaki, R.P., Foley, D.G., Chavez, F.P., 2009. Potential new production estimates in four eastern boundary upwelling ecosystems. Prog. Oceanogr. East. Bound. Upwelling Ecosyst.: Integr. Comp. Approaches 83, 151–158. https://doi.org/10.1016/j.pocean.2009.07.018. Montoya-S� anchez, R.A., Devis-Morales, A., Bernal, G., Poveda, G., 2018. Seasonal and intraseasonal variability of active and quiescent upwelling events in the Guajira system, southern Caribbean Sea. Cont. Shelf Res. 171, 97–112. https://doi.org/ 10.1016/j.csr.2018.10.006. Muller-Karger, F., Varela, R., Thunell, R., Scranton, M., Bohrer, R., Taylor, G., Capelo, J., Astor, Y., Tappa, E., Ho, T.-Y., Walsh, J.J., 2001. Annual cycle of primary production in the Cariaco Basin: response to upwelling and implications for vertical export. J. Geophys. Res. 106, 4527–4542. https://doi.org/10.1029/1999JC000291. Muller-Karger, F.E., McClain, C.R., Fisher, T.R., Esaias, W.E., Varela, R., 1989. Pigment distribution in the Caribbean sea: observations from space. Prog. Oceanogr. 23, 23–64. https://doi.org/10.1016/0079-6611(89)90024-4. Mu~ noz, E., Busalacchi, A.J., Nigam, S., Ruiz-Barradas, A., 2008. Winter and summer structure of the caribbean low-level jet. J. Clim. 21, 1260–1276. https://doi.org/ 10.1175/2007JCLI1855.1. Oschlies, A., Dietze, H., K€ ahler, P., 2003. Salt-finger driven enhancement of upper ocean nutrient supply. Geophys. Res. Lett. 30, 2204. https://doi.org/10.1029/ 2003GL018552. Paramo, J., Correa, M., Nú~ nez, S., 2011. Evidencias de desacople físico-biol� ogico en el sistema de surgencia en La Guajira, Caribe colombiano. Rev. Biol. Mar. Oceanogr. 46, 421–430. https://doi.org/10.4067/S0718-19572011000300011. Park, Y.-H., Lee, J.-H., Durand, I., Hong, C.-S., 2014. Validation of Thorpe-scale-derived vertical diffusivities against microstructure measurements in the Kerguelen region. Biogeosciences 11, 6927–6937. https://doi.org/10.5194/bg-11-6927-2014. Richardson, P.L., 2005. Caribbean current and eddies as observed by surface drifters. Deep Sea Research Part II: topical studies in oceanography, direct observations of oceanic flow. A tribute to Walter Zenk 52, 429–463. https://doi.org/10.1016/j. dsr2.2004.11.001. Ruddick, B., 1983. A practical indicator of the stability of the water column to doublediffusive activity. Deep Sea Res. Part A. Oceanographic Research Papers 30, 1105–1107. https://doi.org/10.1016/0198-0149(83)90063-8. Rueda-Roa, D.T., Muller-Karger, F.E., 2013. The southern Caribbean upwelling system: sea surface temperature, wind forcing and chlorophyll concentration patterns. Deep Sea Res. Oceanogr. Res. Pap. 78, 102–114. https://doi.org/10.1016/j. dsr.2013.04.008. Santos, F., G� omez-Gesteira, M., Varela, R., Ruiz-Ochoa, M., Días, J.M., 2016. Influence of upwelling on SST trends in La Guajira system. J. Geophys. Res. Oceans 121, 2469–2480. https://doi.org/10.1002/2015JC011420. Schmitt, R.W., 1981. Form of the Temperature-Salinity Relationship in the Central Water: Evidence for Double-Diffusive Mixing. Journal of Physical Oceanography 11 (7), 1015–1026. https://doi.org/10.1175/1520-0485(1981)011<1015: FOTTSR>2.0.CO;2. Schmitt, R.W., 2005. Enhanced diapycnal mixing by salt fingers in the thermocline of the tropical Atlantic. Science 308, 685–688. https://doi.org/10.1126/science.1108678. Schmitt, R.W., Perkins, H., Boyd, J.D., Stalcup, M.C., 1987. C-SALT: an investigation of the thermohaline staircase in the western tropical North Atlantic. Deep-Sea Res. Part A Oceanogr. Res. Pap. 34, 1655–1665. https://doi.org/10.1016/0198-0149(87) 90014-8. Thorpe, S.A., 2005. The Turbulent Ocean [WWW Document]. Cambridge Core. https:// doi.org/10.1017/CBO9780511819933. Troupin, C., Barth, A., Sirjacobs, D., Ouberdous, M., Brankart, J.-M., Brasseur, P., Rixen, M., Alvera-Azc� arate, A., Belounis, M., Capet, A., Lenartz, F., Toussaint, M.-E., Beckers, J.-M., 2012. Generation of analysis and consistent error fields using the Data Interpolating Variational Analysis (DIVA). Ocean Model. 52 (53), 90–101. https://doi.org/10.1016/j.ocemod.2012.05.002. Turner, J.S., 1973. Buoyancy Effects in Fluids, Cambridge Monographs on Mechanical and Applied Mathematics. University Press, Cambridge. Wang, C., 2007. Variability of the caribbean low-level jet and its relations to climate. Clim. Dyn. 29, 411–422. https://doi.org/10.1007/s00382-007-0243-z. Wüst, G., 1964. Stratification and Circulation in the Antillean-Caribbean Basins. Columbia University Press, New York. Zhang, J., Schmitt, R.W., Huang, R.X., 1998. Sensitivity of the GFDL modular ocean model to parameterization of double-diffusive processes. J. Phys. Oceanogr. 28, 589–605. https://doi.org/10.1175/1520-0485(1998)028<0589:SOTGMO>2.0.CO; 2. Zhu, Y., Zhang, R., 2018. An argo-derived background diffusivity parameterization for improved ocean simulations in the tropical pacific. Geophys. Res. Lett. 45, 1509–1517. https://doi.org/10.1002/2017GL076269.

Alonso, J., Bl� azquez, E., Isaza-Toro, E., Vidal, J., 2015. Internal structure of the upwelling events at Punta Gallinas (Colombian Caribbean) from modis-sst imagery. Cont. Shelf Res. 109, 127–134. https://doi.org/10.1016/j.csr.2015.09.008. Amador, J.A., 2008. The Intra-Americas Sea low-level jet: overview and future research. Ann. N. Y. Acad. Sci. 1146, 153–188. https://doi.org/10.1196/annals.1446.012. Andrade, C., Barton, E., Mooers, C., 2003. Evidence for an eastward flow along the central and south American caribbean coast. J. Geophys. Res. Ocean. 108 https:// doi.org/10.1029/2002JC001549. Andrade, C.A., Barton, E.D., 2005. The Guajira upwelling system. Cont. Shelf Res. 25, 1003–1022. Arcos-Pulido, M., Rodríguez-Santana, A., Emelianov, M., Paka, V., Arístegui, J., Benavides, M., Sangr� a, P., Machín, F., García-Weil, L., Estrada-Allis, S., 2014. Diapycnal nutrient fluxes on the northern boundary of Cape Ghir upwelling region. Deep Sea Res. Oceanogr. Res. Pap. 84, 100–109. https://doi.org/10.1016/j. dsr.2013.10.010. Beier, E., Bernal, G., Ruiz-Ochoa, M., Barton, E.D., 2017. Freshwater exchanges and surface salinity in the Colombian basin, Caribbean Sea. PLoS One 12. https://doi. org/10.1371/journal.pone.0182116. �reas de surgencia Castellanos, P., Varela, R., Muller-Karger, F., 2002. Descripci� on de las a al sur del Mar Caribe examinadas con el sensor infrarrojo AVHRR. Mem. - Soc. Ciencias Nat. La Salle 154, 55–76. Centurioni, L.R., Niiler, P.P., 2003. On the surface currents of the Caribbean Sea. Geophys. Res. Lett. 30, 1279. https://doi.org/10.1029/2002GL016231. Chavez, F.P., Messi�e, M., 2009. A comparison of eastern boundary upwelling ecosystems. Prog. Oceanogr. 83, 80–96. https://doi.org/10.1016/j.pocean.2009.07.032. Cherubin, L.M., Richardson, P.L., 2007. Caribbean current variability and the influence of the Amazon and Orinoco freshwater plumes. Deep Sea Res. Oceanogr. Res. Pap. 54, 1473. Corredor, J.E., 1979. Phytoplankton response to low level nutrient enrichment through upwelling in the Columbian Caribbean Basin. Deep-Sea Res. 26A, 731–741. Diaz-Pulido, G., Garzon-Ferreira, J., 2002. Seasonality in algal assemblages on upwelling-influenced coral reefs in the Colombian Caribbean. Bot. Mar. 45, 284–292. Dietze, H., Oschlies, A., K€ ahler, P., 2004. Internal-wave-induced and double-diffusive nutrient fluxes to the nutrient-consuming surface layer in the oligotrophic subtropical North Atlantic. Ocean Dyn. 54, 1–7. https://doi.org/10.1007/s10236003-0060-9. Eidens, C., Bayraktarov, E., Hauffe, T., Pizarro, V., Wilke, T., Wild, C., 2014. Benthic primary production in an upwelling-influenced coral reef, Colombian Caribbean. PeerJ 2, e554. https://doi.org/10.7717/peerj.554. �n, E., Chouci~ Fern� andez-Castro, B., Mouri~ no-Carballido, B., Mara~ no no, P., Gago, J., Ramírez, T., Vidal, M., Bode, A., Blasco, D., Royer, S.-J., Estrada, M., Sim� o, R., 2015. Importance of salt fingering for new nitrogen supply in the oligotrophic ocean. Nat. Commun. 6 https://doi.org/10.1038/ncomms9002. Gargett, A., Garner, T., 2008. Determining Thorpe scales from ship-lowered CTD density profiles. J. Atmos. Ocean. Technol. 25, 1657–1670. https://doi.org/10.1175/ 2008JTECHO541.1. Glessmer, M.S., Eden, C., Oschlies, A., 2009. Contribution of oxygen minimum zone waters to the coastal upwelling off Mauritania. Prog. Oceanogr. East. Bound. Upwelling Ecosyst.: Integr. Comp. Approaches 83, 143–150. https://doi.org/ 10.1016/j.pocean.2009.07.015. Gordon, A.L., 1967. Circulation of the Caribbean sea. J. Geophys. Res. 72, 6207–6223. https://doi.org/10.1029/JZ072i024p06207. Hales, B., 2005. Irreversible nitrate fluxes due to turbulent mixing in a coastal upwelling system. J. Geophys. Res. 110 https://doi.org/10.1029/2004JC002685. Hamilton, J.M., Lewis, M.R., Ruddick, B.R., 1989. Vertical fluxes of nitrate associated with salt fingers in the world’s oceans. J. Geophys. Res. 94, 2137. https://doi.org/ 10.1029/JC094iC02p02137. Hern� andez-Guerra, A., Joyce, T.M., 2000. Water masses and circulation in the surface layers of the Caribbean at 66� W. Geophys. Res. Lett. 27, 3497–3500. https://doi. org/10.1029/1999GL011230. Hidalgo, H.G., Dur� an-quesada, A.M., Amador, J.A., Alfaro, E.J., 2015. The caribbean low-level jet, the inter-tropical convergence zone and precipitation patterns in the intra-americas sea: a proposed dynamical mechanism. Geogr. Ann. Ser. A Phys. Geogr. 97, 41–59. https://doi.org/10.1111/geoa.12085. Jouanno, J., Sheinbaum, J., Barnier, B., Molines, J.-M., Debreu, L., Lemari� e, F., 2008. The mesoscale variability in the Caribbean Sea. Part I: simulations and characteristics with an embedded model. Ocean Model. 23, 82–101. https://doi.org/ 10.1016/j.ocemod.2008.04.002. McDougall, T.J., Ruddick, B.R., 1992. The use of ocean microstructure to quantify both turbulent mixing and salt-fingering. Deep Sea Res. Part A. Oceanographic Research Papers 39, 1931–1952. https://doi.org/10.1016/0198-0149(92)90006-F.

16