Water Mass Structure and Mixing Fractions in the Deepwater Region of the Gulf of Mexico

Abstract

Here, we estimate water mass fractions in the deepwater region of the southern Gulf of Mexico from samples collected during June 2016. Past water mass characterizations have relied mainly on temperature, salinity, and dissolved oxygen concentrations, although the degree of mixing and the effects on the distribution of nutrients and other dissolved constituents remain uncertain. Based on nine water types defined in the eastern Caribbean Sea and North Atlantic and excluding surface waters, five dominant water masses with contributions between 50% and 93% in their cores were identified. North Atlantic Subtropical Underwater contributed 42% ± 9% between 49 and 105 m. Modal water with origins in the northern subtropical gyre (Eighteen Degree Water) were ubiquitous in the gulf, with average contributions of 62% ± 25% between 110 and 153 m and 51% ± 12% between 248 and 306 m, while the core of Tropical Atlantic Central Water (TACW; 70% ± 5%) dominated between 355 and 510 m. Water at the nitrate maximum consisted of TACW (48% ± 13%) and Antarctic Intermediate Water (AAIW; 32 ± 14%), with the local remineralization of organic matter adding ∼10% to the nitrate pool. Below ∼800 m, water transitioned from AAIW to a domain of two components of Labrador Seawater. Mesoscale eddies controlled the observed depth variability of the water masses from the surface down to ∼1,000 m. This study provides a quantitative approximation to evaluate the addition of nutrients to the different water masses by biogeochemical processes driven by an epipelagic ecosystem generally considered to be oligotrophic.

Key Points

• Water mass mixing fractions influence the distribution and concentration of water-column properties in the Gulf of Mexico

• The relative dominance of nine source water types and their mixing by mesoscale processes in the first 1,000 m of the water column are explored

• The origin and mixing of source water types constitute a source of variability in the relationship between dissolved oxygen and nitrate

Plain Language Summary

Waters present at different depths in any region of the ocean may have different origins. Temperature, salinity, nutrient, and oxygen content are commonly used to determine water mass origins. The water masses present in the Gulf of Mexico originate (in order of increasing depth below 70 m) in the central North Atlantic, Sargasso Sea, the West African coast, the sub-Antarctic region of the South Atlantic, and Labrador Sea in the North Atlantic. In our study, we evaluated the degree of mixing among water masses in the deepwater region of the southern Gulf of Mexico for the first time. Our results suggest that water from the Sargasso Sea is present throughout the gulf (between 110 and 306 m depth), constituting an important source of nutrients that sustain the food web in the gulf. We show the importance of rotating features called eddies in deepening or shoaling water masses in the upper 1,000 m of the water column and the importance of evaluating water mass mixing to understand the biology and chemistry of the gulf.

1 Introduction

The Gulf of Mexico (GoM) is a semienclosed system that includes a semipermanent cyclonic eddy in the Bay of Campeche and anticyclonic circulation in the central and western regions (Nowlin et al., 2000; Pérez-Brunius et al., 2013; Tenreiro et al., 2018). These circulation patterns are associated with the interactions among rotational wind stress, the bathymetry, and anticyclonic Loop Current eddies (LCEs) that detach from the Loop Current (LC) within the gulf. These LCEs have diameters of ∼300 km and separate from the LC in the eastern region of the gulf every 5–19 months, advecting waters from the Caribbean toward the central and western regions of the GoM (Hall & Leben, 2016; Tenreiro et al., 2018; Vidal et al., 1994; Vukovich, 2007). As LCEs propagate toward the interior of the gulf, they may join with other eddies or interact with the continental slope, dissipating as smaller eddies (Alvera-Azcárate et al., 2009; Hamilton et al., 2014; Vukovich, 2007). Moreover, LCEs coherently transport water in the upper ∼1,000 m for periods of 6–11 months (Hamilton et al., 2014; Leben, 2005) at speeds of 4.4 ± 2.9 km day⁻¹ (Vukovich, 2007), redistributing heat and promoting the mixing of dissolved constituents in the interior of the gulf (Hamilton et al., 2018; Kolodziejczyk et al., 2012; Tenreiro et al., 2018). As a result, detectable differences are present among the hydrographic profiles of the region under the influence of the LC and those from the western region of the GoM (Morrison et al., 1983; Vidal et al., 1994).

The identification of water masses in the GoM has been based on the extreme characteristics of temperature–salinity (T–S) diagrams and on dissolved oxygen (DO) and nutrient concentrations (e.g., Morrison & Nowlin, 1977; Morrison et al., 1983; Nowlin et al., 2000; Vidal et al., 1994). In the surface layer (<50 m), thermohaline gradients have been identified as a result of river discharge over the continental shelf and the intrusion of salty, warm, and tropical Caribbean Surface Water (CSW; Portela et al., 2018). North Atlantic Subtropical Underwater (NASUW), Eighteen Degree Water (EDW), and Tropical Atlantic Central Water (TACW) make up the waters of the thermocline and are characterized by a linear trend in T–S diagrams. In the upper part of the thermocline, NASUW, which originates in the central region of the subtropical gyre, is characterized by a salinity maximum in the potential density isopycnal of ∼25.5 kg/m³ (Qu et al., 2016; Yu et al., 2018). The westward advection of LCEs is responsible for distributing these thermocline water masses within the GoM (Meunier et al., 2018). The NASUW salinity maximum is diluted within the gulf due to vertical mixing processes that are the result of the collision of LCEs with the continental slope and/or shelf and vertical diffusion or winter mixing, which result in the formation of Gulf Common Water (GCW; Nowlin et al., 2000; Portela et al., 2018; Sosa-Gutiérrez et al., 2020; Vidal et al., 1994).

EDW is formed under conditions of relatively low salinity and uniform temperature (36.5°C ± 0.1°C and 18°C ± 0.5°C) in two regions of the Sargasso Sea (Joyce, 2012; Kwon & Riser, 2004). Although the presence of EDW in the Caribbean is distinguished by a subsurface DO maximum (Jochens & DiMarco, 2008; Nowlin et al., 2000), this maximum is only intermittently observed within the GoM (Portela et al., 2018) and thus may not always be useful as an EDW tracer. This water mass has also been recognized in the subtropical Atlantic Ocean by a deviation from the canonical Redfield N:P ratio, denoted by the parameter N* (Gruber & Sarmiento, 1997; Hansell et al., 2004). Positive N* values may result from a variety of characteristics and processes at work within the region of EDW formation, the most important of which is the composition of exported and mineralized organic matter associated with diazotrophic prokaryotes (Landrum et al., 2011) and/or nondiazotrophic prokaryotes (Singh et al., 2013). Approximately 50% of the volume of the permanent thermocline of the northwestern Atlantic is EDW (Joyce, 2012).

Due to its proximity to the study region, it was expected that the contribution of EDW to the volume and ventilation of waters that constitute the GoM thermocline would be notable. In the deep part of the thermocline, TACW is characterized by a DO minimum (<3 mL/L, ∼130 μmol/kg, and a σ_θ value of 27.15 kg/m³; Morrison et al., 1983). It is possible that this water mass comes from the oxygen minimum zones (OMZs) off the West African margin (Metcalf, 1968, 1976; Stramma & Schott, 1999). Although both EDW and TACW are identified in the LC by their contrasting DO concentrations, the DO minimum in TACW is partially eroded as this water mass is transported westward in the GoM (Jochens & DiMarco, 2008; Morrison & Nowlin, 1977; Morrison et al., 1983).

Antarctic Intermediate Water (AAIW) can be identified by its salinity minimum and maximum nitrate and phosphate concentrations (Morrison et al., 1983). On its journey through the Caribbean, the AAIW core is eroded due to mixing with waters of different origin, showing increased salinity values at its core in the Yucatan Channel of 34.85‰ (Wüst, 1963). AAIW is also transported from the LC to the western region of the GoM by LCEs. Along its trajectory, AAIW experiences mixing, which is reflected in the horizontal distribution of salinity within the core of this water mass (Hamilton et al., 2018). Metcalf (1976) suggested the presence of another water mass, Caribbean Intermediate Water (characterized by a silicic acid maximum), below AAIW. However, recent studies on water masses have not mentioned Caribbean Intermediate Water in their descriptions of the GoM (Hamilton et al., 2018; Portela et al., 2018).

The deep water (below 1,000 m) in the GoM is dominated by North Atlantic Deep Water (NADW), which preferentially enters the Cayman basin in the Caribbean through the Windward Channel (Smith, 2010). This water mass is mainly formed in the Labrador Sea and Irminger Sea (Herrford et al., 2017; Rhein et al., 2015) and shows lower nutrient concentrations than those of AAIW (Carder et al., 1977; Metcalf, 1976; Nowlin et al., 2000; Stalcup et al., 1975). In the western Atlantic Ocean, two components of Labrador water have been recognized: an upper, lighter, and saltier component (Upper Labrador Sea Water; uLSW) that is formed by mixing between Labrador water and intermediate Mediterranean Outflow Water (MOW; 20%–40%) and a denser component, Classical Labrador Sea Water (cLSW; van Sebille et al., 2011). In theory, both uLSW and cLSW could be transported to the GoM, whereas the deeper components of NADW, namely Iceland Scotland Overflow Water and Denmark Strait Overflow Water, would not enter the GoM, as the depths of the Anegada–Jungfern (1,815 m) and Windward (1,680 m) passages constitute physical barriers of entry to the Caribbean Sea (Morrison & Nowlin, 1982; Smith, 2010; Sturges, 2005; Wüst, 1963).

Optimum multiparameter (OMP) analysis is used to investigate the contributions and distributions of water masses (Karstensen & Tomczak, 1998; Tomczak, 1981) and is based on an inverse model defined by a linear system of mixing equations, in which the contributions or fractions of several predefined source water types (SWTs) are estimated from a set of water-column observations. The SWTs describe the thermohaline and chemical characteristics of water masses at their places of formation. An extended optimum multiparameter (eOMP) analysis incorporates stoichiometric relationships to quantify the effect of oxygen utilization on nutrient mineralization. Both methods have been implemented in the Atlantic Ocean to evaluate water mass circulation (e.g., Álvarez et al., 2014; Fontela et al., 2016; Jenkins et al., 2015; Poole & Tomczak, 1999), the temporal variability of water masses (Leffanue & Tomczak, 2004), the effects of biogeochemical processes on metal and nutrient distributions (Middag et al., 2018, 2019, 2020), and the composition of particulate organic matter (Álvarez-Salgado et al., 2014). To the best of our knowledge, there are no published OMP analyses for the main thermocline and deep waters of the GoM, although eOMP has been used to evaluate the contributions of SWTs associated with the surface layer in the coastal zone (Kim & Min, 2013; Xue et al., 2015). Kim and Min (2013) and Xue et al. (2015) evaluated the contributions of NASUW and of the Mississippi and Atchafalaya Rivers to the formation of a hypoxic zone over the Texas–Louisiana shelf and the temporal variation associated with the origin and utilization of organic matter.

In this study, the contributions of various SWTs in the main thermocline and deepwater domains were estimated for summer 2016 in the southern GoM (south of 26°N) to evaluate the role of lateral advection and mixing in the spatial variation of the concentrations of nutrients and DO. We considered that six water masses, represented by eight SWTs, enter the GoM through the Yucatan Channel below CSW. The SWTs used in this study were defined with data from World Ocean Circulation Experiment (WOCE)/CLImate VARiability and Predictability (CLIVAR) cruises in the Caribbean Sea, the Rapid Climate Change program (RAPID) of the Atlantic (24–26°N), and the Gulf of Mexico Ecosystems and Carbon Cycle (GOMECC)-2007 cruise (Figure 1a). The distributions of water masses associated with mesoscale eddies were also examined. Moreover, the role of LCEs in the transport and mixing of CSW, NASUW, and EDW was inferred. We also estimated nitrate addition and DO removal due to organic matter mineralization in the interior of the gulf after evaluating the contribution of each water mass due to mixing.

Figure 1

Map of the stations sampled in the XIXIMI-5 (2016) campaign within the Gulf of Mexico (GoM) and the location of sections used to define source water types (SWTs) in the mixing analysis (a). These locations were the RAPID and southern A22 (Caribbean) sections and a set of northern GoM stations corresponding to the GOMECC-2007 summer cruise (gray points). The stations of the XIXIMI-5 cruise are denoted by their alpha-numerical numbering. The brown line shows the transect in Figure 4. In addition, the location of station 56 in the Anegada–Jungfern passage complex (magenta square point) of the A22 Caribbean section is marked. (b) Geostrophic velocities and nonsteric absolute dynamic topography (ADT) averages during the XIXIMI-5 campaign in the GoM. The geostrophic velocity field represents the vector average during the cruise period. The color field represents the ADT without the steric portion (cm). The thermal expansion effect was removed from each of the individual daily sea level anomaly (SLA) maps and was calculated as the daily area-average SLA within the GoM at depths greater than the 200-m isobaths. Subsequently, the resulting SLA value was added to the long-term mean dynamic topography to obtain the daily ADT without the steric effect. The contours of negative sea level (red line) and 32 cm (green line) are presented. The 200-, 1,000-, and 3,500-m depth contours indicate the continental slope and deep basin (Sigsbee Deep; black lines). The velocity vectors and ADT values were omitted from depths less than 200 m.

2 Materials and Methods

2.1 A Hydrographic and Nutrient Data in the GoM

The XIXIMI-5 oceanographic cruise took place from 10 to 25 June 2016. A total of 35 stations were sampled in the southern region of the GoM on board the oceanographic vessel BO/Justo Sierra administered by the National Autonomous University of Mexico (UNAM). The sampling region (20°–26°N, 86°–97°W) comprised the deep waters of the GoM, including the continental slope and abyssal plain (Figure 1a). At 17 stations, CTD (SeaBird 911, Seabird Electronics Inc., Bellevue, WA) casts (0–1,000 m) were conducted with a rosette equipped with 12 20-L Niskin bottles, while casts at depths >1,000 m were conducted at 18 stations. The deepest samples were collected at 3,719 m (B13). The CTD sensors were previously calibrated by the manufacturer, and the accuracy of the measurements was ±0.001°C for temperature, ±0.002 for salinity, and 2% for DO.

Data from the DO sensor were calibrated with data from the Niskin bottle samples, which were analyzed with the microWinkler method. Calibration of CTD data was performed via a linear regression model (DO_CTD_calibrated [mL/L] = 0.995 × DO_CTD−0.063), after which the data were transformed to μmol/kg. The accuracy and precision of the Winkler method were 0.1% and ∼1.3 μmol/kg, respectively.

Samples collected for nutrient analysis from 0 to 200 m were filtered through combusted Whatman GF/F filters, whereas the deeper samples were not filtered. Nitrate + nitrite (hereinafter referred to as nitrate) and phosphate analyses were performed with a SEAL-AA3-HR auto-analyzer (SEAL Analytical Ltd, Norderstedt, Germany), following the guidelines described in the GO-SHIP Repeat Hydrography Manual (Hydes et al., 2010). The method of Armstrong et al. (1967) with modifications was used for nitrate analysis, while the procedure of Murphy and Riley (1962) was followed for phosphate analysis. The limits of detection (3 SD, n = 11) for nitrate and phosphate were 0.037 and 0.028 μmol/kg, respectively. Accuracy and precision were estimated with measurements of certified reference material (CRM) for nutrients (lots CD and CC; Kanso Co. Ltd, Osaka, Japan). The two CRMs were analyzed repeatedly in each run for each nutrient, and the results of a Student t test indicated that the average CRM measurements were equal to those of the certified values (nitrate t-cal = 0.078 and 0.265 for the CD and CC lots, respectively; phosphate t-cal = 0.186 and 1.008 for the CD and CC lots, respectively; degrees of freedom = 50; α = 0.05). The mean values obtained for nitrate from the CD and CC lots were 5.514 μmol/kg (CV = 0.490%) and 30.958 μmol/kg (CV = 0.336%), respectively. The mean values obtained for phosphate from the CD and CC lots were 0.447 μmol/kg (CV = 5.157%) and 2.092 μmol/kg (CV = 1.311%), respectively. Absolute salinity (S_A) and conservative temperature (Θ) were determined with the GSW toolbox v. 3.05 (IOC et al., 2010).

2.2 Definition of SWTs for the OMP Analysis

The SWTs for the mixing model were defined with data from the RAPID zonal section (24.5°–26.5°N) in the Atlantic, from the WOCE/CLIVAR A22 section in the Caribbean, and from the GOMECC cruise in the northern portion of the GoM (Figure 1a and Table S1 in Supporting Information S1). Data from stations located between 77° and 69°W (spanning ∼ 815 km) of the RAPID section (McCarthy et al., 2015) collected during winter (1998–2010) and spring (2004) cruises were used to define the SWTs of EDW_Rapid, uLSW, and cLSW (Table 1). Data collected from the southern A22 section (11°–18°N) during summer 1997, autumn 2003, and spring 2012 were used to obtain the corresponding SWTs for Tropical Surface Water (TSW), NASUW, EDW_A22, TACW, and AAIW. Additionally, 28 northern GoM stations corresponding to the GOMECC-2007 summer cruise were analyzed to define the characteristics of remnant Caribbean Surface Water (CSWr). The data were downloaded from the Global Ocean Data Analysis Project v. 2 (GLODAPv2; Olsen et al., 2016).

Table 1. Characteristics (Mean ± SD) of the Nine Source Water Types (SWTs) That Were Used to Solve the Mixing Model in the Gulf of Mexico (GoM)

SWT	SA (g/kg)	Θ (°C)	PO4* (μmol/kg)	NO (μmol/kg)	DO (μmol/kg)	NO3 (μmol/kg)	PO4 (μmol/kg)	Neutral density (kg/m^3)	Data source
CSWr	36.856 ± 0.177	31.222 ± 2.286	1.114 ± 0.084	195.000 ± 10.914	206.661 ± 10.324	0.03 ± 0.078	0.04 ± 0.037	22.98 ± 0.83	GOMECC-2007
TSW	34.318 ± 0.394	29.697 ± 0.195	1.193 ± 0.014	195.276 ± 2.015	194.658 ± 1.949	0.06 ± 0.039	0.01 ± 0.005	21.19 ± 0.30	A22-Caribbean
NASUW	37.271 ± 0.108	23.545 ± 0.731	1.179 ± 0.050	195.636 ± 6.490	177.492 ± 13.064	1.87 ± 1.176	0.10 ± 0.068	25.35 ± 0.19	A22-Caribbean
EDWA22	36.662 ± 0.121	18.007 ± 0.570	1.437 ± 0.077	242.007 ± 10.816	157.157 ± 15.859	8.75 ± 2.496	0.48 ± 0.160	26.44 ± 0.07	A22-Caribbean
EDWRapid	36.497 ± 0.049	16.717 ± 0.287	1.523 ± 0.060	256.585 ± 9.708	183.579 ± 6.552	7.52 ± 0.951	0.41 ± 0.059	26.63 ± 0.03	RAPID
TACW	35.162 ± 0.077	9.067 ± 0.470	2.552 ± 0.072	395.073 ± 9.848	120.685 ± 1.932	28.29 ± 1.085	1.82 ± 0.077	27.19 ± 0.03	A22-Caribbean
AAIW	34.947 ± 0.027	6.416 ± 0.510	2.897 ± 0.043	438.585 ± 5.422	133.709 ± 8.522	31.43 ± 0.736	2.09 ± 0.043	27.44 ± 0.08	A22-Caribbean
uLSW	35.233 ± 0.023	5.427 ± 0.614	2.668 ± 0.062	416.485 ± 11.190	220.366 ± 19.425	20.26 ± 1.543	1.33 ± 0.094	27.80 ± 0.07	RAPID
cLSW	35.157 ± 0.019	3.958 ± 0.231	2.764 ± 0.043	433.713 ± 6.819	259.833 ± 5.420	17.92 ± 0.640	1.19 ± 0.034	27.93 ± 0.02	RAPID
R^2	1.000	1.000	0.982	0.983	0.590	0.959	0.960
RMSE	0.007	0.061	0.132	16.070	33.086	3.029	0.154
N	397	397	397	397	397	397	397

• Note. The model uses conservative temperature (Θ), absolute salinity (S_A), the parameter NO (NO = [DO] + R_O/N × [NO₃]; R_O/N = 9.7), and the parameter PO₄* (PO₄^∗ = [PO₄] + ([DO]/R_O/P); R_O/P = 165). The average concentrations of dissolved oxygen (DO) and nutrients are also presented for the calculation of the parameters. The fit of the model was evaluated using the coefficients of determination (R²) and the mean square error (RMSE). The F test, with a Type I error (α) of 0.05, was used to establish the significance of all the regression models (F_cri = 3.865).

The SWT characteristics were defined by the extreme features of property–property graphs (e.g., S_A–Θ diagrams) and their vertical distributions. A total of nine SWTs were proposed to quantify the water mass fractions in the GoM, characterized by S_A, Θ, and the semiconservative parameters of NO and PO₄* (Broecker, 1974). NO is defined in Equation 1 (Broecker, 1974) as

(1)

where DO is the dissolved oxygen concentration (μmol/kg), NO₃ is the nitrate concentration (μmol/kg), and R_O/N is the stoichiometric ratio between DO consumption and nitrate release during respiration. PO₄* is defined in Equation 2 (Broecker et al., 1985; Rae & Broecker, 2018) as

(2)

where DO is the dissolved oxygen concentration (μmol/kg), PO₄ is the phosphate concentration (μmol/kg), and R_O/P is the molar ratio between DO consumption and phosphate release during organic matter remineralization. The characteristics for the nine SWTs are presented in Table 1.

The distribution of the physical (Θ and S_A) and biogeochemical (DO, nitrate, phosphate, and silicic acid) variables allowed for the characteristics and positions of eight SWTs in the Caribbean and Atlantic sections to be determined (Figures 2 and 3 and Figure S1 in Supporting Information S1). In the southern A22 section during October 2003, TSW (Kirchner et al., 2009) was defined by σ₀ < 22 kg/m³ and a freshwater signal (S_A < 35 g/kg; Figure S2 in Supporting Information S1). The TSW layer with a thickness and width of 26 m and ∼450 km, respectively, was warm (29°C) and had an excess silicate concentration compared to that of nitrate, indicated by Si* values (Si* = [silicic acid] − [nitrate]; Sarmiento et al., 2004) between 1.84 and 4.8 μmol/kg. The water masses of the thermocline are located below this layer, namely NASUW, EDW, and TACW.

Figure 2

Conservative temperature (Θ)–absolute salinity (S_A) diagram (a), and the relationships between salinity and dissolved oxygen (DO), nitrate and phosphate concentrations (b, c, and d, respectively) are presented for the stations sampled in the XIXIMI-5 (2016) campaign within the Gulf of Mexico (GoM) and for the RAPID-2010 (yellow dots) and southern A22-2012 (Caribbean; green dots) sections. In addition, data for station 56 of the A22 Caribbean section are marked to show the properties of the SWTs that enter/exit the Anegada–Jungfern passage complex (magenta squares-line). CTD sensor data for the XIXIMI-5 cruise (panels a and b, gray points) were superimposed on the discrete GoM data (blue points).

Figure 3

Oceanographic properties that were used to resolve the mixing of the different source water types (SWTs) in (a) section A22 of the Caribbean in 2012, (b) stations of lines 24° and 25°N of the XIXIMI-5 cruise in the Gulf of Mexico (GoM), and (c) in the RAPID section in the Atlantic Ocean in 2010. The variables include absolute salinity (S_A; g/kg, upper panel), dissolved oxygen (DO; μmol/kg, middle panel), and nitrate (μmol/kg, bottom panel). To increase the resolution of the SWT properties associated with waters above 1,000 m, the depth axis was stretched at the top. Neutral densities (kg/m³) were overlapped on nitrate panels to show the core layers for water masses in each region (NASUW = 25.35 kg/m³, EDW = 26.63 kg/m³, TACW = 27.19 kg/m³, AAIW = 27.44 kg/m³, uLSW = 27.80 kg/m³, and cLSW = 27.93 kg/m³).

In the Caribbean, NASUW was characterized by a salinity maximum between 37.12 and 37.45 g/kg (Figures 2a and 3a) in the σ₀ interval of 25–25.7 kg/m³, a change in the sign of Si* from positive to negative, and a PO₄* minimum between 1.1 and 1.32 μmol/kg (Figure S3 in Supporting Information S1). We defined NASUW with average data from the three campaigns in the A22 section because the salinity maximum observed in the RAPID section was 37 g/kg, which contrasts with the salinity maximum of the LC within the GoM of 37.1 g/kg (Portela et al., 2018) and suggests a likely propagation route from the eastern Caribbean (Figures 2a and 3c). Advection and mixing resulted in a reduction of the thickness of NASUW (denoted by the isohaline of 37 g/kg), which occurs as the water mass is advected from the Anegada–Jungfern passage (between 100 and 200 m) toward the coast of Venezuela, where this layer also becomes shallower (∼80 m; Figure 3a and Figure S3 in Supporting Information S1).

We used two SWTs to represent EDW. Conceptually, we assumed that this water mass is transported from the Atlantic through both the Windward passage (EDW_Rapid) and the Anegada–Jungfern (EDW_A22) passage, with each of these SWTs having different properties (Table 1). Both EDW_A22 and EDW_Rapid were defined by excess dissolved nitrate concentrations [DINxs] (0–2 μmol/kg; Hansell et al., 2004) between the 26.3 and 26.6 kg/m³ isopycnals. EDW_A22 showed the same pattern of southward thinning and uplift in the Caribbean as that of NASUW, with vertical displacements of ∼100 m (Figure S4 in Supporting Information S1). The SWT of this water mass presented a DO concentration of 157 μmol/kg (lower than that of SUW, 178 μmol/kg), salinity of 36.66 g/kg, and temperature of 18°C. The SWT of EDW_Rapid presented a DO concentration of 184 μmol/kg, salinity of 36.50 g/kg, and temperature of 16.7°C (Figure S5 in Supporting Information S1). EDW_Rapid was oxygenated, cold, and less salty than Caribbean SWT_A22 and closer to the formation site and may have spread to the Caribbean through the Windward passage.

The characteristic feature of TACW in the Caribbean was the DO minimum between 117 and 125 μmol/kg, although it also presented minimum values (−12 to −13 μmol/kg) in Si*. Minimum Si* and DO values occurred between the 27.05 and 27.15 kg/m³ isopycnals. In contrast to the thinning of NASUW and EDW_A22 due to the mixing that occurred toward the American coast, the TACW core thinned from the continental slope of Venezuela toward the north of the section (Figure S6 in Supporting Information S1). The three remaining water masses were distributed in intermediate and deep layers of the Caribbean and Atlantic, although their characteristic features may have been absent or modified in one of the two sections.

In the Caribbean section, AAIW (600–1,000 m) propagates to the northwest with the Caribbean Current. The AAIW SWT was delimited by salinity minima (S_A < 35 g/kg) between the 27.2 and 27.5 kg/m³ isopycnals from the continental slope to 16°N (∼300 km). The maximum phosphate and nitrate concentrations of this water mass were 2.15 and 32.43 μmol/kg, respectively (Figures 2c,2d, and 3a and Figure S1 in Supporting Information S1). The thinning of AAIW toward the Greater Antilles is the product of gradual mixing with uLSW of North Atlantic origin, which is saltier and more oxygenated and has lower nutrient concentrations than those of AAIW (Figures S7 and S8 in Supporting Information S1). The uLSW SWT, which represents mixing between MOW and LSW (van Sebille et al., 2011), was defined in the waters of the RAPID section by S_A values of 35.1–35.27 g/kg and DO values of 220 μmol/kg between the 27.55 and 27.75 kg/m³ isopycnals. The uLSW SWT was defined in the RAPID section and not in the Caribbean because its salinity and DO characteristics were diluted in the A22 section. In addition, the bathymetry (the Jamaican ridge) restricts the passage of deep waters between the eastern basins of the Caribbean and those to the west (see thresholds of the Jamaican and Haitian ridges, Figure S10 in Supporting Information S1). In addition, the deep component, cLSW, was defined in the RAPID section (Figure S9 in Supporting Information S1). A silicic acid minimum of ∼13 μmol/kg occurs at ∼1,600 m, with DO values of ∼260 μmol/kg associated with the flow of the Western Deep Boundary Current, carrying Labrador and Irminger water south. The SWT characteristics of cLSW were defined by salinity, temperature, and NO concentration values of 35.16 g/kg, 3.96°C, and 433 μmol/kg, respectively. Waters below 1,700 m (>27.8 μmol/kg), which included the other two northern components of NADW, did not enter the Caribbean because of the sill depths of the two deepest passages, namely the Anegada–Jungfern (1,649 m) and Windward (1,608 m; Figure S10 in Supporting Information S1) passages.

2.3 The OMP Analysis

In the OMP analysis used to quantify the SWT fractions in the samples collected during the XIXIMI-5 cruise, S_A, Θ, NO, and PO₄* parameters were used to construct a system of linear mixing equations. In our analysis, four mixing diagrams were established with a maximum of five SWTs each (Table 2 and Figure S11 in Supporting Information S1). The SWT contributions were resolved by a nonnegative least squares approach. This optimization process always guarantees that the SWTs positively contribute to the mixing equations, which are conditioned to have strictly conserved mass. The samples were assigned to the mixing diagram in which their total residuals were minimal (see details in Supporting Information S1). The reliability of the model was examined for each parameter using the coefficient of determination (R²), which assesses the fit of the observed and reconstructed data, and mean square error (RMSE), which provides a measure of the average differences between the observed and reconstructed values.

Table 2. Mixing Diagrams Analyzed by the Optimum Multiparameter (OMP) Analysis Inside the Gulf of Mexico (GoM)

	Diagram	SWT	Parameters	n
Mixed layer and seasonal thermocline	1	CSWr, TSW, NASUW, EDWA22, TACW	SA, Θ, NO, PO4*	166
Permanent thermocline	2	EDWA22, TACW	SA, Θ, NO, PO4*	37
Intermediate waters	3	EDWRapid, TACW, AAIW, uLSW	SA, Θ, NO, PO4*	72
Deep and bottom waters	4	EDWRapid, AAIW, uLSW, cLSW	SA, Θ, NO, PO4*	122
Total	4	9		397

• Note. The average number of observations (n) of each diagram was established as the result of the “best solution” that was obtained from 100 random perturbations of the characteristics of the source water types (SWTs), with an R_O/N = 9.7 and an R_O/P = 165.

To evaluate the stability and sensitivity of the proposed mixing model, perturbation tests were performed on the five sources of error that make up the OMP analysis, according to the methods of Álvarez et al. (2014). The initial matrix of SWT characteristics was modified by adding a normally distributed random number to each element that was multiplied 4 times by the standard uncertainty (or propagated error) assigned to the variable before solving the system of equations. The combination of the standard uncertainty of the measurements and of the phosphate () and nitrate reference materials is shown in Equations 3 and 4:

(3)

and

(4)

These uncertainties were used to calculate the propagated error of the NO (u_ΔNO) and () parameters with Equations 5 and 6:

(5)

and

(6)

The procedure was repeated 100 times, and 100 variations of the initial SWT matrix were obtained to calculate the contributions due to mixing. The data were also perturbed 100 times to account for variations in the accuracy of the measurements. In each test, random numbers that were normally distributed within their nominal values and the standard uncertainty of each variable were generated. Additionally, weights were assigned to each variable (CT and S_A: 1–10; NO and : 1–6). The greatest weighting of the equations occurred for the conservation of mass, with a value of 100 that remained constant in the 3,600 weight perturbation tests. The R_O/N value that was used to calculate the NO parameter was modified 41 times (iteratively between 8 and 12). In addition, the molar ratio used to calculate was modified between 140 and 180. The contributions of the SWTs presented in this study, which are referred to as the best solutions, are the averages of 100 perturbations of the SWT matrix (Table 1) with weights of six for Θ and S_A and 100 for the conservation of mass, with R_O/P and R_O/N values of 165 and 9.7, respectively. Although the use of two SWTs (EDW_A22 and ED_Rapid) to represent EDW resulted in lower residuals in the OMP analysis, the results of the two SWTs were merged into a single EDW to simplify the discussion.

The sensitivity of the results to the perturbations that were carried out was explored based on the reproducibility of the parameters, the difference between the average contribution of the disturbed system and that of the best solution, and the number of samples in each SWT whose values did not differ from those of the best solution. The degree of stability of the system was evaluated by the standard deviation (STD) of the mixing fractions obtained for each perturbation analysis.

Finally, absolute dynamic topography (ADT) and geostrophic velocities were used to analyze the role of the LC and the detachment of anticyclonic LCEs in the distribution of water masses. The orientation of the geostrophic velocity vectors was used to classify the stations into either cyclonic (counterclockwise rotation) eddies or anticyclonic (clockwise) LCEs. The ADT was evaluated by combining mean dynamic topography (MDT) and the sea level anomaly, the latter without the steric signal (Liu & Weisberg, 2012). To remove the steric or thermal expansion signal, the methodology of Weisberg and Liu (2017) was employed based on the procedure described by Dukhovskoy et al. (2015) and Hall and Leben (2016).

3 Results

3.1 Circulation Based on Altimetry

During the XIXIMI-5 campaign, the LC was observed in a retracted position within the GoM, with a northern limit at ∼24°N. A large LCE named Poseidon, which was released from the LC in April 2016 (see http://www.horizonmarine.com/loop-current-eddies.html), was one of the main circulation features during the sampling campaign. Only one station, PO1 (24.5°N–96°W; Figure 1b), was located within the core of this LCE. Poseidon was separated from the LC by a relatively large cyclonic eddy; station B18 was located at the edge of this cyclonic feature and presented the most negative ADT value (<−10 cm, Figure 1b). In the western gulf, the LCE Olympus was centered at 24°N–96°W (station B11 and adjacent) and was released from the LC in October 2015, influencing various stations near its core, including B11 and C21. In the southern region, a front separated Olympus from the Campeche cyclone, which was centered at 20°N–95°W and included stations F37 and H46 among others.

3.2 Water Masses in the GoM

The surface waters of the GoM with a density anomaly <23.5 kg/m³ showed relatively high salinity (35.81 and 36.89 g/kg; Figure 2a) and temperatures (26.91°C and 29.95°C). In these waters, DO values were between 177 and 207 μmol/kg (Figure 2b) and nutrients were almost depleted, with nitrate <0.21 μmol/kg and phosphate <0.13 μmol/kg (Figures 2c and 2d). Three water masses were present between densities of 23.75–26 kg/m³ (from 20 to ∼190 m, excluding LCE Poseidon stations PO1, A8, and A10), namely CSWr, GCW, and NASUW. In the salinity interval of 36.03–36.79 g/kg (with temperatures of 19.5°C–27.3°C), in the domain of CSWr and GCW (Figure 2a), wide ranges in DO (220–124 μmol/kg, Figure 2b), nitrate (0.04–8.84 μmol/kg, Figure 2c), and phosphate (0.03–0.37 μmol/kg, Figure 2d) values were observed. This variability in biogeochemical properties in the GoM contrasts with that of the A22 and RAPID sections, where DO concentrations decreased and nutrient concentrations increased with decreasing salinity (Figures 2b–2d). It is worth noting that DO values >202 μmol/kg (Figure 2b) correspond to a subsurface maximum occurring at depths of 31–68 m, where nutrients were depleted (nitrate <0.24 μmol/kg and phosphate <0.08 μmol/kg, Figures 2c and 2d), whereas the layer where DO decreased below 202 μmol/kg and nutrients increased corresponds to the domain of GCW.

The maximum salinities inside the GoM, ranging from 36.86 to 37.07 g/kg and that characterize NASUW, were observed in CTD casts in the eastern gulf stations associated with LCE Poseidon (i.e., A7, A8, PO1, and A10). NASUW showed DO concentrations between 170 and 144 μmol/kg but was only detected in bottle samples of nitrate and phosphate in two stations (A7 and A10), with nitrate values of 2.94 and 6.30 μmol/kg and phosphate values of 0.10 and 0.29 μmol/kg. The maximum salinity values associated with NASUW were observed in CTD data at ∼250 m depth in the core of Poseidon (PO1; Figure 2a). This maximum was eroded to the west, leaving a trail of salinity values around ∼36.6 g/kg (25.75 kg/m³ isopycnal; Figures 2a and 3b) at depths of 90–160 m. A similar trend in the salinity distribution was observed in the Caribbean, with maximum salinity values ∼37.30 g/kg observed in stations near the Anegada–Jungfern passage complex that decreased due to mixing toward the south in the A22 section (<14°N) where isohalines uplifted. The Caribbean NASUW was also characterized by higher DO values (∼174 μmol/kg) and lower nutrient concentrations (nitrate: ∼1.11 μmol/kg; phosphate: ∼0.04 μmol/kg) than those in the GoM. In the RAPID section (Figure 3c), NASUW showed maximum salinity values of 37.0 g/kg that were lower than those in the GoM (see Figure 2b).

At densities greater than 26.26 kg/m³, the Θ–S_A characteristics observed in the GoM were similar among stations. EDW and TACW, with densities between 26.26 and 27.25 kg/m³ (from ∼100 to 640 m, excluding PO1), were associated with the linearity of the Θ–S_A diagram in salinity and temperature intervals between 35.12 and 36.78 g/kg and 8.0°C–18.9°C, respectively. These water masses showed a gradual decrease with depth in DO (148–105 μmol/kg) and an increase in the concentrations of nitrate (9.90–30.00 μmol/kg) and phosphate (0.51–1.93 μmol/kg). The low oxygen layer (DO < 130 μmol/kg) expanded from east to west (from ∼350 to 600 m), and the minimum values (<117 μmol/kg) occurred at depths of ∼400 m in the interior of the gulf, although at station PO1, minimum values were found at depths down to 630 m. Both EDW and TACW in the GoM showed lower DO concentrations and higher nitrate concentrations than those in the Caribbean and RAPID (Figures 2b and 2c). The low oxygen layer that characterizes TACW thinned from south to north in the Caribbean section (between 400 and 600 m; Figure S6 in Supporting Information S1), which is similar to what was observed for AAIW with regard to minimum salinity and maximum nitrate values (between 600 and 900 m; Figure S7 in Supporting Information S1). In the RAPID section, there was an apparently low contribution of TACW as indicated by minimum DO values ∼150 μmol/kg at 750 m depth (Figure 3c).

In the Caribbean south of 17°N, the core of AAIW was identified with salinity minima (<35 g/kg) and maximum values of nitrate (∼32.43 g/kg) and phosphate (∼2.15 g/kg; Figure 3 and Figures S1a and S7a in Supporting Information S1). Due to mixing with adjacent water masses, AAIW became more saline and nutrient concentrations decreased toward the GoM and RAPID section (Figure 3). The salinity minimum of AAIW was observed in the GoM between the 27.35 and 27.51 kg/m³ isopycnals (from 512 to ∼930 m, excluding PO1), with salinity values of 35.06–35.08 g/kg and temperatures of 5.81°C–6.93°C. Minimum AAIW salinity values (35.06 g/kg) were observed between 585 and 940 m depth in the eastern region of the GoM, showing slight salinity enrichment (35.07–35.08 g/kg) toward the interior of the gulf between 626 and 896 m. At the salinity minimum, DO concentrations were between 128 and 146 μmol/kg and nitrate values were between 27.50 and 29.66 μmol/kg; however, the maximum nitrate concentrations in the water column varied from 29.16 to 30.25 μmol/kg and frequently (i.e., in 26 of 33 stations) did not coincide with the salinity minima. This means that maximum nitrate concentrations were ∼200 m shallower than the salinity minimum (Figure 3b). In contrast, maximum phosphate concentrations ranged from 1.76 to 2.01 μmol/kg and coincided with the salinity minima. In the RAPID section, the remnant of this water mass can be identified in the change in slope of the Θ–S_A diagram, and its nitrate and phosphate concentrations were ∼25 and 1.68 μmol/kg, respectively, lower than those in the GoM.

Between the 27.58 and 27.765 kg/m³ isopycnals (787–3,720 m depth) corresponding to uLSW and cLSW, the waters showed increasing salinity (35.09–35.15 g/kg) and DO concentrations (158–206 μmol/kg) with depth and decreasing temperature (5.33°C–4.06°C), nitrate (28.48–21.96 μmol/kg) and phosphate (2.02–1.11 μmol/kg; see Figure 3b and Figure S1b in Supporting Information S1) concentrations. However, at depths >2,000 m (>27.754 kg/m³), hydrological and chemical variables presented low variability, with values ranging from 35.14 to 35.15 g/kg for salinity, 4.06°C–4.12°C for temperature, 199–206 μmol/kg for DO, 21.96–23.63 μmol/kg for nitrate, and 1.12–1.52 μmol/kg for phosphate (Figures 2b–2d). In the GoM and Caribbean, uLSW and cLSW presented lower DO values and higher nutrient concentrations than those in the RAPID section. The highest density observed within the GoM during the XIXIMI-5 cruise was 27.765 kg/m³ (PO1 station, 3,277 m depth), whereas the maximum values in the A22 and RAPID sections were 27.82 and 27.895 kg/m³, respectively. The maximum density value found in the GoM was observed in the RAPID section at depths of 1,300–1,500 m with typical DO, nitrate, and phosphate values of ∼261, ∼18, and ∼1.2 μmol/kg, respectively.

3.3 Mixing Fractions Evaluated With the OMP Analysis and Vertical Modifications of Water Mass Cores Due To Circulation

In the discrete samples, the core of each water mass was indicated by fractions greater than 50% with regard to the mixture of the SWTs defined for the OMP analysis (see Table 1 and Figure 4). CSWr was dominant in layers above 50 m depth, with an average of 54%. In two stations located near the continental shelf (TS1 and A3), the dominant component was TSW, with fractions of ∼76% and 83%, respectively, in the surface layer (<15 m). The dominance of NASUW occurred from 49 to 105 m in most stations, although in stations influenced by the LCE, this dominance extended to 150 m with a maximum mixing fraction of 71% in A10 at 150 m (see Figures 4 and 5).

Figure 4

Depth distribution of source water type (SWT) fractions occupying the seasonal and permanent thermocline, the intermediate layer, and the deep layer of the Gulf of Mexico. The color scale represents the nitrate concentrations measured (μmol/kg) in all of the XIXIMI-5 stations (upper axis; see Figure 1a) during June 2016. Colored dots represent samples with SWT fractions >50% (except uLSW showing fractions >25%), and the colors of the station labels indicate the effects of mesoscale circulation due to Loop Current eddies (LCEs, red letters) or cyclones (blue letters). Small black dots indicate samples where no SWT showed a fraction >50%. The CSWr contribution was generally observed up to 50 m depth (magenta dots). Abbreviations: CSWr, Caribbean Surface Water remnant; NASUW, North Atlantic Subtropical Water; EDW, Eighteen Degree Water; TACW, Tropical Atlantic Central Water; AAIW, Antarctic Intermediate Water; uLSW, Upper North Atlantic Deep water; and cLSW, Classical Labrador Sea Water.

Figure 5

Variability in the mixing fractions of the different water masses in layers defined by the sample collection depth. The data shown with box plots represent the fraction in percentiles (25%, 75%, and the median) of each source water type (SWT) in the depth interval; the circles represent the average value. SWTs with zero contributions are not represented in the figure. The figure was made with the IoSR Matlab Toolbox (Hummersone, 2020).

The dominance of EDW with maximum contributions of up to 85% occurred ubiquitously within the GoM at depths of 70–306 m, with shallower/deeper depths observed in stations associated with cyclonic/anticyclonic eddies. The core of TACW was present at ∼440 m depth, showing a variability of ±200 m that was also controlled by mesoscale eddies. The contribution of this water mass at the sampling depth where the oxygen minimum was detected at each station ranged between 57% and 83% in between 355 and 510 m. In these samples, EDW contributed between 7% and 32% and AAIW contributed <26% (Figure 5). As a result of mixing, this TACW core became warmer (10.04°C ± 0.57°C) and more saline (35.39 ± 0.08 g/kg) when compared to SWT in the Caribbean (Table 1). At ∼600 m where the maximum nitrate concentration between 29.16 and 30.25 μmol/kg was frequently observed, a dominant contribution of TACW (up to 72%) was found in 16 stations (Figures 4 and 5).

The highest contribution of the AAIW SWT was observed at the nominal depth of 800 m, with a contribution between 39% and 68%. The greatest contribution of the uLSW SWT was also observed at the nominal depth of 800 m, with a recurring contribution of ∼22% (maximum of 30%). At the nominal depth of 1,000 m, a cLSW SWT contribution greater than 50% was observed in 18 stations while in the remaining stations, no SWT was dominant, with the exception of AAIW (57%) in station P01 (LCE Poseidon). At 1,000 m, uLSW contributed an average of 14%, AAIW between 20% and 44%, and cLSW between 19% and 72%. cLSW was dominant at the deepest layers with 59%–81% at 1,200 m and between 90% and 93% below 2,000 m. From 2,000 m to the maximum sampling depth of this campaign (3,720 m), both AAIW and uLSW each contributed ∼4%.

3.4 Stability and Reliability of the Mixing Model

All variables included in our OMP analysis (conservative and semiconservative parameters) showed a good linear fit between the observations and the reconstructed values (RMSE, Table 1). The vertical distribution of the total residuals showed a better fit between the reconstructed values and the values measured at depths >2,000 m compared to those of the upper layer (Figure S13a in Supporting Information S1). The reconstructed values tended to be slightly overestimated (negative values, lower axis; Figures S13b–S13e in Supporting Information S1), with uncertainty attributed mainly to , which showed similar variation to those of the total residuals. The stability of the solution decreased in NASUW, in which maximum variations of 11% were observed with regard to the mean of the standard deviation matrix (SD, Table S2 in Supporting Information S1).

The mixing fractions of the SWTs were stable (Table S2 in Supporting Information S1) and robust (Table S3 in Supporting Information S1) with regard to the five sources of error in the OMP. Changes in the fractions of SWTs to our best solution were <6%, with the exception of the weight perturbation (Table 3). When obtaining the best solution, if the SWT characteristics were perturbed, the average solution would consistently be almost identical to the best solution, with maximum average differences of ∼2% in NASUW (Table 3). The perturbation to the observed data compared to the best solution and the perturbation of the SWTs increased the stability of all constituents of the water column. These modifications decreased the total residual values (negative values in Figure 6), improved the ability of the model to reproduce the measured NO values (due to the reduction in RMSE and increased R² values; Table S3 in Supporting Information S1), and modified the values between 1% and 6% of the average contributions of the best solution.

Table 3. Proportion of Change in the Mixing Fractions When Modifying the Five Sources of Error in the Optimum Multiparameter (OMP) Analysis, Namely the Source Water Type (SWT), the Measured Data, the Weights, and the Stoichiometric Coefficients of the Consumption of Oxygen to Nitrate (R_O/N) and Oxygen to Phosphate (R_O/P)

	SWT	Data	Weights	RO/N	RO/P
CSWr	0.762 ± 0.717	1.768 ± 1.853	2.349 ± 2.723	2.715 ± 2.566	1.517 ± 1.354
TSW	0.263 ± 0.216	0.716 ± 0.576	1.735 ± 1.121	0.963 ± 0.785	0.667 ± 0.504
NASUW	1.680 ± 1.642	4.002 ± 3.385	7.940 ± 6.345	5.928 ± 5.369	2.836 ± 2.778
EDW	0.618 ± 1.005	2.253 ± 2.644	4.745 ± 5.267	2.705 ± 3.824	2.291 ± 2.589
TACW	0.368 ± 0.417	2.351 ± 3.354	4.853 ± 5.627	2.586 ± 2.994	2.468 ± 2.615
AAIW	0.521 ± 0.967	4.448 ± 5.260	9.104 ± 8.937	2.141 ± 2.990	3.795 ± 4.176
uLSW	0.466 ± 0.510	5.603 ± 6.544	15.345 ± 9.074	2.040 ± 1.673	5.228 ± 5.857
cLSW	0.484 ± 0.543	5.593 ± 6.683	20.274 ± 14.313	1.771 ± 1.717	5.314 ± 5.959

• Note. The mean and standard deviation (both in percentage) were obtained by subtracting the absolute mean fractions from each perturbation test and the best solution.

Figure 6

Sensitivity analysis of the proposed mixing model (the best solution) against the different sources of variability throughout the full-depth distribution of sampling. The sources of variability were the characteristics of the source water type (SWT), the measured data, the weights, and the stoichiometric coefficients of the consumption of oxygen to nitrate (R_O/N) and oxygen to phosphate (R_O/P). The lower axis shows the difference between the average total residuals obtained with the different perturbation tests in relation to the total residuals of the best solution.

The modification to the data was the perturbation test that least altered the SWT fractions of TACW, AAIW, and cLSW. In contrast, the perturbation to the weights altered the SWT fractions of TACW and the deepwater components to the greatest extent. The variation of weights assigned to each variable resulted in a maximum difference with regard to the best solution between 9% (AAIW) and 20% (cLSW) and also increased the values of the total residuals due to a decrease in the reproducibility of the conservative variables. However, with variable weighting, the ability of the model to reproduce NO and notably improved. In contrast to the perturbation of the weights, the R_O/N ratio was the perturbation that resulted in the greatest variability of the NO parameter. The modification of the quotient increased the total residual values between 600 and 1,200 m depth, where AAIW occurs and where two deepwater components were modified by ∼2% in our best solution. With the perturbation to R_O/N, the fractions of the best solution from NASUW and EDW were very stable and changed only by 6% and 3%, respectively. The effect of the perturbation to R_O/P was similar to the perturbation of the data. Even with the perturbation to R_O/P, a maximum reproducibility of the conservative variables was observed. In summary, perturbation tests were found to be able to modify the best solution by a maximum of 20% for cLSW, which was the SWT that showed the greatest change and sensitivity to weight. In contrast, the perturbations to the SWTs, data, and stoichiometric ratios in the mixing model resulted in differences of up to 6% in the best solution, particularly in NASUW and in the two deepwater components of the GoM.

4 Discussion

In this study, the mixing fractions of nine SWTs in the deepwater region of the GoM were estimated. This quantitative evaluation was based on the hydrographic and biogeochemical characteristics of SWTs that were defined in the eastern Caribbean (A22) and Atlantic Ocean (RAPID) and constitutes a preliminary step toward quantifying the contributions of the respiration of organic matter to dissolved inorganic nutrient concentrations and DO consumption on local and regional scales. This numerical approximation using an OMP analysis allowed for a robust identification of the vertical cores of the water masses and of the variation in their realms due to mesoscale processes. The predominant water masses with mixing fractions >50% between the thermocline and the bottom were NASUW, EDW, TACW, AAIW, and cLSW. However, other water masses with mixing fractions <50%, which were transformed by mixing as they were advected toward the interior of the GoM, are important for the reproducibility of the conservative and semiconservative variables used in the OMP analysis. The numerical evaluation allowed us to conclude that the surface and the thermocline waters of the GoM were the ones to preserve a higher fraction of SWTs as defined in the Colombia and Venezuela basins (A22-Caribbean). In contrast, waters in intermediate and deep layers showed a large dilution of uLSW and a noticeable reduction of the AAIW fraction (<60%). The composition of these water masses depended on mass exchange among the northwest Atlantic, the Caribbean, and the GoM through the Windward passage and Yucatan Channel. Our analysis also shows a decoupling in the GoM between the depths of the maximum nitrate concentration and the salinity minimum.

The sensitivity and stability of the OMP analysis indicate that the water mass fraction estimates of our best solution were robust (Table 3). The sensitivity to the variability of SWT properties resulted in a maximum fraction variability of 11%, which was associated with NASUW, followed by a value of 7%, associated with uLSW (Table S2 in Supporting Information S1). The SWT fractions obtained with our best solution were stable, varying by a maximum of ∼2% in NASUW (Table 3). In the Caribbean, TSW with σ₀ < 22 kg/m³ acquires its identity due to precipitation and continental runoff in the tropical region (Kirchner et al., 2009; Schott et al., 1998). However, in the deepwater region of the GoM, water with σ₀ < 22 kg/m³ as observed in stations TS1 and A3 is CSW, with low salinity due to continental runoff that may be transported from the continental shelf by mesoscale circulation (Hamilton et al., 2018). For example, during summer, low salinity water from the Mississippi–Atchafalaya system can be carried into the deep GoM along the edges of anticyclonic eddies (Brokaw et al., 2019). In our analysis, CSWr was the dominant component of the surface layer (<50 m) of the GoM. This SWT was defined with data from the interior of the GoM, and its characteristics are the result of winter mixing between the surface waters (e.g., TSW and CSW) and subsurface water (NASUW) that arrive with LCEs (Portela et al., 2018; Vidal et al., 1994) and an increase in temperature and salinity that reach maximums in summer. In this study, a SWT corresponding to GCW was not included in the OMP analysis, as it is a product of the winter mixing of CSWr, NASUW (Sosa-Gutiérrez et al., 2020; Vidal et al., 1994), and EDW, which is indicated by our results and can be inferred from other studies (Nowlin & Parker, 1974). In a recent study, Cervantes-Díaz et al. (2022) described the formation of GCW and concluded that it results from the mixing of CSWr and TACW. However, these authors did not include EDW in their water mass analysis. GCW is characterized by relatively low temperature and a homogeneous salinity distribution that reaches depths between 100 and 150 m. In the summer, this water is isolated from the surface layer by seasonal stratification. Based on the temperature, salinity, and DO intervals reported by Portela et al. (2018) that characterize the water masses inside the GoM, the GCW was located at 115 ± 27 m (n = 43) during the XIXIMI-5 campaign and was absent in the LCE Poseidon. According to our results, GCW water could have a greater contribution of EDW than previously thought (Figure 5).

The OMP results suggest a ubiquitous distribution of EDW within the GoM, with an average fraction of 62% ± 25% between 110 and 150 m; however, this water mass is often overlooked in descriptions regarding water masses in the gulf. The presence of EDW has been associated with subsurface DO maximums observed in the eastern GoM and the Caribbean (Morrison & Nowlin, 1977, 1982; Morrison et al., 1983; Portela et al., 2018; Rivas et al., 2005); however, in the deepwater region of the GoM, this maximum is erased by the consumption of DO associated with local organic matter respiration. Thus, DO cannot be used to trace EDW. In our study, the excess of nitrate [DINxs] in the upper thermocline was a useful tracer for this water mass, which was magnified in the GoM when compared to that present in the RAPID and Caribbean A22 sections. Nitrate enrichment along the way from the RAPID section to the Caribbean and GOM may explain the concomitant trend of decreasing DO concentrations (Figures 2b and 2c). The excess of nitrate in the North Atlantic is a product of the remineralization of nitrogen-enriched organic matter produced by diazotrophs in the North Atlantic subtropical gyre (Fernández-Castro et al., 2016) and/or by the local mineralization of nitrogen-enriched nondiazotrophic (prokaryotic) particles in the region of formation north of Bermuda (Singh et al., 2013). Within the Caribbean and in the GoM, the sources of nitrate enrichment are probably similar to those of the North Atlantic given that both diazotrophs (Trichodesmium sp. and unicellular) and nondiazotrophs (Prochlorococcus and Synechococcus) may be abundant in these basins (Linacre et al., 2019; McManus & Dawson, 1994; Mulholland et al., 2006).

In the GoM, EDW also mixes with TACW, which makes water with minimum DO concentrations more saline and warmer compared to that of the Caribbean. Samples A10 and PO1 showed TACW contributions of 72% and 76%, reflecting mixing during transit from the eastern Caribbean to the eastern gulf. TACW acquires its biogeochemical characteristics in the hypoxic OMZ (DO > 42 μmol/kg, Brandt et al., 2015) of Mauritania and Senegal in the eastern North Atlantic (Metcalf, 1976) and/or in the Namibia OMZ of the South Atlantic (Bower et al., 2019; Stramma & Schott, 1999). During its transit through the Atlantic, TACW oxygenates by diapycnal mixing and enters the Caribbean through the passages to the south of the Lesser Antilles where the DO concentration is approximately threefold higher than that at its origin. Within the GoM, the samples targeted for collection at the core of TACW (355–510 m) showed average contributions of 70% ± 5%, indicating less mixing compared to other SWTs defined in the Caribbean (Figure 5). Within the GoM, a zonal (east-west) gradient in DO concentrations was also observed, with higher concentrations (DO < 117 μmol/kg) at station PO1 and A10, reflecting the influence of more oxygenated LC waters of Caribbean origin within the LCE Poseidon that culminated in a minimum value of 105 μmol/kg, which is a biogeochemical pattern that reflects DO consumption within the basin (Figure 3b).

Below the TACW core, samples collected during the XIXIMI-5 campaign at the nominal depth of 600 m showed TACW as the dominant SWT in half of the stations (Figures 4 and 5). The maximum nitrate concentration, which is a characteristic generally associated with AAIW, occurred regularly at this depth (26 out of 33 stations). However, the highest mixing fraction of the AAIW SWT (39%–68%) was observed at 800 m, the nominal depth where the salinity minimum was observed. This observation indicates a decoupling in the GoM between the depth of the nitrate maximum and the depth of the salinity minimum, which was also noticed by Morrison et al. (1983), although these authors did not provide an explanation for such decoupling. Five of the seven stations where the maximum nitrate concentration and the minimum salinity value coincided at 800 m were located in LCEs (B11, B12, C20, C21, and A10). Wüst (1963) estimated that during its passage through the Caribbean Sea before entering the GoM, the proportion of AAIW is reduced by at least 25% compared to what is observed outside the arc of the Antilles.

Outside the Caribbean (9.5°N), maximum nitrate concentrations can reach ∼34 μmol/kg and are coupled to the salinity minimum (S_A 34.82 g/kg; sta. 35, GEOTRACES GA02 transect; http://geotraces.org/dp/idp2017). In the eastern Caribbean basins in the A22 section, maximum nitrate and minimum salinity values continued to occur at the same depth, although nitrate decreased (<33 μmol/kg) and salinity increased (>34.88 g/kg). A further increase in salinity between the Venezuela and Cayman basins was documented by Osborne et al. (2014), who described a gradient of salt addition from 34.75 to 34.85 (practical salinity) to the salinity minimum of AAIW (see their Figures 2b and 2d). The increase in salinity of >0.1 units between the eastern Caribbean basins and the Cayman basin suggests the presence of a greater volume of saltier uLSW (Smith, 2010) that enters from the Northern Hemisphere through the Windward passage to the Cayman basin, where it mixes with AAIW. The salinization and decrease in the nitrate concentration of AAIW in the Cayman basin and in the deep region of the GoM could result from mixing between this water mass and deeper uLSW (Morrison & Nowlin, 1982; Smith, 2010; Sturges, 2005). However, the contribution of nitrate due to the remineralization of nitrogen-rich organic matter above the salinity minimum (Figure 2c) is probably the mechanism for the decoupling in depth between the nitrate maximum and salinity minimum.

A transition in the realm of the intermediate to deepwater masses was observed at 1,000 m depth. This boundary separates the upper layer dominated by mesoscale eddies from the deep layer dominated by barotropic circulation (Chang & Oey, 2011; Rivas et al., 2005). During their movement to the west, LCEs transport physical and chemical properties, affecting their vertical distributions below 1,000 m of the water column (Hamilton et al., 2018, Figure 4). In stations in which the incidence of LCEs dominated, the cLSW core was observed at 1,200 m depth, in contrast to the other stations in which it was observed at 1,000 m depth. Compared to that of the cLSW, the uLSW contribution was less than 30% at depths of 800–1,200 m. This small uLSW contribution is probably a result of mixing during its transit from the RAPID section to the Caribbean through the Windward passage (Smith, 2010). Also, there is flow out the GoM via the Yucatan Channel that probably increases uLSW mixing at approximately 800–1,300 m (e.g., Candela et al., 2019; Chang & Oey, 2011; Rivas et al., 2005; Sheinbaum et al., 2002). The lower temperatures and higher DO concentrations observed between 1,200 and 2,000 m in the eastern Caribbean basins compared to those of the GoM indicate that these basins are not connected through an east-west circulation at this depth range and reflect the journey of cLSW from the Atlantic through the Windward passage and its transport through the Cayman and Yucatan basins (Figure 3 and Figure S1 in Supporting Information S1). After passing through the Yucatan Channel, the cLSW contribution below 2,000 m is homogeneous (with an average of 92%). The cLSW is transformed by the dynamics that originate from bidirectional flows through the Yucatan Channel and deep circulation within the GoM (>1,000 m). The exchange of this deepwater body with the Caribbean Sea and the bidirectional flow paths represent future research targets needed to understand nutrient and DO distributions within the GoM.

In summary, the water masses that dominated the GoM from the thermocline to the deep layer were NASUW, EDW, TACW, AAIW, and cLSW. The relative dominance of these water masses is also reflected in the relationship between DO and nitrate concentrations that showed three relatively linear segments with changing slopes throughout the water column below the upper layer (Figure 7a). The linearity in the relationships between these nonconservative properties probably reflects the strong effect of water mass mixing on their vertical distributions. In the upper layer where CSWr and NASUW dominated, nitrate was depleted due to phytoplankton consumption and DO showed concentrations greater than 190 μmol/kg. Below the mixed layer, at depths of 31–68 m, concentrations were >202 μmol/kg, which were above DO saturation and corresponded to a subsurface maximum that probably resulted from photosynthetic activity (Lee-Sánchez et al., 2022). Below the nitrate-depleted layer, an oxycline delimited by concentration values decreasing from 190 to 130 μmol/kg and a nitracline delimited by concentration values increasing from ∼0.5 to 7 μmol/kg were observed. While this layer represents a depth range of intense organic matter remineralization, the linear trend reflects intense mixing between NASUW and EDW. The layer where mixing between EDW and TACW dominates corresponds to the linear segment where DO shows a relatively small decrease from 125 to ∼110 μmol/kg compared to the relatively large increase in nitrate from ∼8 to ∼23 μmol/kg. The layer dominated by mixing between TACW, AAIW, and uLSW was nonlinear and corresponds to the segment of nitrate concentrations above 27 μmol/kg where the nitrate maximum (∼30 μmol/kg) was observed. Finally, the linear segment below the nitrate maximum reflects mixing between nitrate-rich AAIW and oxygen-rich cLSW, with maximum DO concentrations of 206 μmol/kg registered in near bottom waters. It is worth noting that the lump at the edge of this segment, with nitrate concentrations around 23 μmol/kg, was composed of samples collected between 2,000 and 3,720 m depth, indicating that the deep waters of the GoM are homogeneous and composed of cLSW.

Figure 7

Relationship between dissolved oxygen (DO) and nitrate in the XIXIMI-5 campaign. The relationship indicates the presence of three relatively linear segments differentiated by the mixture between the realms of the water masses (mixing fractions >50%) quantified by the mixing model (a). Boxplots of the residuals (Δ) of nitrate and DO concentrations at different depth layers (b and c). The residuals are obtained from the observations minus the data reconstructed with the OMP analysis, where positive residuals indicate net addition and negative residuals indicate net removal from organic matter production/degradation. For each depth layer, the boxplot of the potential density anomaly with respect to the surface (σ₀; d) is presented.

The residuals obtained for DO and nitrate with the OMP analysis reflect the balance between the addition and removal of these variables due to biogeochemical processes that occurred both inside the GoM and during the journeys of the water masses from the SWT source regions as defined in our study. In surface waters with σ₀ values <25.5 kg/m³, negative nitrate residuals with a median of 3.7 μmol/kg were observed between 49 and 105 m, which coincided with positive DO residuals with a median of 21 μmol/kg (Figures 7b and 7c). In this depth range, the deep chlorophyll maximum (DCM) was present, thus the residuals reflect a nitrate deficit and a DO excess that is probably associated with photosynthetic activity. A nitrate deficit observed in the layer where nitrate is depleted indicates that the nitrate concentration predicted from SWT mixing was 3.7 μmol/kg. Based on the average water mass fractions obtained at this depth range (31%, 42%, and 3%, respectively, Figure 5) and nitrate concentrations in the SWT at its source (1.9, 8.7 and 28.3 μmol/kg, respectively, Table 1), NASUW, EDW, and TACW would have contributed 0.8, 2.7, and 0.8 μmol/kg, respectively. In other words, EDW would contribute the largest amount of nitrate in the DCM through water mass advection and mixing; however, the nitrate contribution from TACW cannot be ignored.

A noticeable transition from negative to positive nitrate residuals and from positive to negative DO residuals occurred from 110 to 153 m (Figures 7b and 7c), where EDW presented the largest fraction (mean 62%). Samples collected within this depth range showed a large dispersion of Δnitrate and ΔDO values, partially due to the effect of mesoscale eddies lifting or deepening isopycnals (note the wide range in potential density anomalies observed in these samples; Figure 7d). In this depth range, negative or near zero residuals were observed in stations within LCEs, and the remaining stations showed a predicted nitrate concentration of 9.2 μmol/kg (0.3, 6.7, and 2.3 μmol/kg, respectively) based on the average contributions of NASUW, EDW, and TACW (14%, 77%, and 8% respectively), while the observed value was 11.4 μmol/kg, reflecting a nitrate addition of 2.2 μmol/kg. The dispersion in the residuals also reflects that samples at these depths were located within the upper part of the nitracline and the oxycline, a layer of intense remineralization of organic matter where substantial changes in nitrate and oxygen concentrations occur with small changes in depth.

The net consumption of DO and addition of nitrate between 248 and 805 m was relatively constant, as indicated by the narrow range in the residual medians (Δnitrate 2.1–2.9 μmol/kg and ΔDO −29 to −34 μmol/kg). The values of the residuals indicate that ∼10% of the total observed nitrate concentration was added by remineralization, with a corresponding DO consumption of ∼20% of the value predicted from STW mixing. Maximum median values of ∆nitrate of 3.9 μmol/kg and of ∆OD of −50 μmol/kg were observed between 1,000 m and the bottom, indicating that ∼16% of the observed nitrate concentration was added and ∼20% of the predicted DO removed by remineralization in deep waters of GoM. The profiles of the DO and nitrate residuals reflect a balance between the residence time of the water at each depth and the addition of nitrate (and DO removal) due to the remineralization of organic matter. The higher residuals obtained in waters below 1,000 m depth likely reflect the higher residence time of deep waters in the GoM (∼231 years; Chapman et al., 2018) rather than an intensification of organic matter remineralization. In general, particulate organic matter fluxes decrease exponentially in the deep open ocean, becoming more refractory with depth (Marsay et al., 2015; Weber et al., 2016). The effect of the residence time on the DO deficit was emphasized below 2,000 m, a layer with the highest density in the gulf (Figure 7d) and with relatively high and vertically homogeneous DO concentrations (202 ± 1 μmol/kg). In this layer, DO showed the largest negative residual while nitrate residuals remained similar compared to those in the overlying layer between 1,000 and 2,000 m. In these well-ventilated waters (Rivas et al., 2005), an increase in the DO deficit was not followed by an increase in excess nitrate, suggesting that the organic particulate matter remineralized at this depth has a higher carbon:nitrogen ratio than those of the overlying layers.

The sources of particulate organic matter that remineralize in the GoM probably originated to a large extent in surface oligotrophic waters, where Prochlorococcus and Synechococcus dominate the biomass of autotrophic picoplankton (Linacre et al., 2019) and where diazotrophic phytoplankton occasionally bloom (Holl et al., 2007; Mulholland et al., 2006). The results of Howe et al. (2020) suggest that the transport of particulate organic matter that is produced in coastal areas of the northern GoM associated with discharge from the Mississippi–Atchafalaya River system to the deep region of the GoM is limited. However, the supply of particulate organic matter produced in coastal areas associated with river discharge by lateral advection cannot be ruled out, as suggested by studies with sediment traps in the northern gulf (Chanton et al., 2018; Giering et al., 2018; Liu et al., 2018). Additional organic matter sources that contribute to the residuals observed in the GoM are those in the western Caribbean Sea. In order to evaluate the contributions from remineralization to nutrient and DO concentrations in the water masses of the GoM, studies of both basins that evaluate vertical fluxes and the elemental compositions of particulate organic matter as well as the residence times of the different water masses are required.

5 Conclusions

Our study quantified the mixing fractions of the different SWTs that comprise the deepwater region of the GoM. The physical and biogeochemical characteristics of the water column in the GoM, the Caribbean Sea, and the RAPID section are described based on S_A–property diagrams and vertical sections of nutrient and DO profiles. We identified several water masses and the spatial variability controlled by mixing and biogeochemical processes. Evidence is provided for the ubiquitous presence of EDW within the GoM and its role in the formation of GCW. A decoupling of the nitrate maximum from the salinity minimum in the gulf, in contrast to what is observed in the AAIW core in the Caribbean Sea (A22), is also shown. From the surface to ∼1,000 m depth, mesoscale eddies control the spatial variability of the water mass fractions. This study further describes how NADW within the GoM has two components: uLSW and cLSW. uLSW contributes salt during its propagation from the North Atlantic toward the GoM, whereas cLSW is dominant within the GoM at depths >1,000 m. Due to their residence times and the remineralization of particles that arrive from the surface layers, the deep waters increase their nitrate concentrations by ∼4 μmol/kg and show DO deficits of ∼50 μmol/kg compared to those of their defined SWTs in the Atlantic (RAPID section). This study does not address the potential sources of nitrate enrichment from the Atlantic and Caribbean to the GoM, but it does provide a quantitative approximation that may be used to evaluate the addition of nutrients due to organic matter remineralization in this oligotrophic marginal sea.

Acknowledgments

This research was funded by the Mexican National Council for Science and Technology - Mexican Ministry of Energy - Hydrocarbon Fund (project 201441). This is a contribution of the Gulf of Mexico Research Consortium (CIGoM). We acknowledge PEMEX’s specific request to the Hydrocarbon Fund to address the environmental effects of oil spills in the Gulf of Mexico. A postdoctoral grant to Valencia-Gasti was provided by CIGoM. We thank Mauricio Muñoz-Anderson and María del Carmen Ávila-López for dissolved inorganic nutrient analyses. This study uses data from RAPID, provided by the British Oceanographic Data Centre and funded by the Natural Environment Research Council of the UK, the National Science Foundation of the USA, and the National Oceanic and Atmospheric Administration (NOAA). We thank the crew and scientists of the research vessels involved in sampling and processing. MDT_CNES-CLS18 data were produced by Collecte Localisation Satellites (CLS) Space Oceanography Division and distributed by the Archiving, Validation, and Interpretation of Satellite Oceanographic Data (Aviso+), with support from the Centre National d’Etudes Spatiales (CNES; https://www.aviso.altimetry.fr/). Satellite altimeter data were provided by Ssalto/Duacs with the support of CNES and distributed by the Copernicus Marine and Environment Monitoring Service (CMEMS). We thank the two anonymous reviewers for the positive comments and careful reviews, which helped to improve the manuscript.

Open Research

Data Availability Statement

The hydrographic and nutrient data for XIXIMI-5 campaign in the deep waters of the Gulf of México are available from PANGAEA: https://doi.org/10.1594/PANGAEA.934036. Data from World Ocean Circulation Experiment (WOCE)/CLImate VARiability and Predictability (CLIVAR) cruises in the Caribbean Sea, from the Rapid Climate Change program (RAPID) of the Atlantic (24°–26°N), and from the Gulf of Mexico Ecosystems and Carbon Cycle (GOMECC)-2007 cruise were obtained from the Global Ocean Data Analysis Project v. 2 (GLODAPv2; Olsen et al., 2016). The figures were created with Ocean Data View software (Schlitzer, 2018). Figure 5 was created with a Matlab code Hummersone (2020).