The Ventilation of the Deep Gulf of Mexico

1. Introduction

Beneath about 1000 m, the Gulf of Mexico is filled with water whose characteristics correspond to North Atlantic Deep Water (NADW; Nowlin et al. 2001), which is cold, saltier, and quite rich in oxygen (see Figs. 1 and 2). Remarkably, the maximum values of oxygen are not found near the surface or at intermediate levels but at the greatest depths (Fig. 1a). In addition, this deep water mass is capped by a layer of Antarctic Intermediate Water (AAIW), which is much poorer in oxygen, so that any mixing or diffusion between the two would have the effect of extracting oxygen from the deeper water mass, as would the effects of the oxidation of organic matter at depth. This oxygen minimum at about 500-m depth thus isolates the deep Gulf from any diffusive inflow of oxygen from the surface layers. Moreover, no local formation in the classical sense, starting with surface waters, of the very deep water can exist in the Gulf of Mexico since it would imply extreme cooling (>10°C) and salinity increases (>5 psu). Very clearly, the deepest water mass must be renewed at a rather efficient rate in order to sustain the near-bottom oxygen maximum. Since the sill at the Florida Straits lies at a depth of 740 m, well above the level of the oxygen minimum of the AAIW, the only renewal of deep water, or ventilation of the deep Gulf of Mexico, can be through the Yucatan Sill, whose sill at 2040 m is deep enough to allow the NADW to flow from the depths of the Caribbean into the deep Gulf, with the renewal water sliding down the northern slope of the Yucatan Channel into the abyss of the Gulf of Mexico. The purpose of this paper is to show that the above schematics are part of a consistent hypothesis on the processes that bring about the renewal of the deep water in the Gulf of Mexico and sustain the higher levels of dissolved oxygen at depth. This ventilation is caused by sporadic inflow through the deepest layers of the Yucatan Channel and appears to be associated with bottom-trapped motions.

The paper is organized as follows: Section 2 summarizes the background on the water masses in the Gulf of Mexico and Yucatan Channel. Section 3 describes the data and methods used for their analysis. Section 4 shows the results on the budgets of heat, oxygen, and salts in the deep Gulf of Mexico, and the description of the events that ventilate the deep basin. In section 5, the implications of the ventilation process are discussed. Section 6 summarizes the main results.

2. Water masses in the Gulf of Mexico and Yucatan Channel

In the upper layers of the Yucatan Channel, a rapid current enters the Gulf of Mexico from the Caribbean and feeds the Loop Current, which sheds large warm rings at intervals from 6 to 12 months (Sturges et al. 1993; Sturges and Leben 2000). The Loop Current exits the Gulf of Mexico through the Florida Straits as the main contribution to the Gulf Stream. Sporadic “counterflows” (from the Gulf of Mexico into the Caribbean) take place near Cuba as the Cuban Countercurrent, and at depth on the Yucatan and Cuban sides of the channel, as the Yucatan and Cuban Undercurrents. No such counterflows have been documented in the Florida Straits. But in the deeper layers, below the level of the Florida Sill (∼800 m), all exchanges from the Gulf of Mexico must take place with the Caribbean over the 2040-m-deep sill, and their long-term average vanishes exactly (Sheinbaum et al. 2002; Ochoa et al. 2001). The average flow distribution in the Yucatan Channel below the level of the Florida Sill consists of a generalized inflow into the Gulf of Mexico that extends from the base of the incoming Yucatan Current to the bottom, but flanked by outflows, one sliding off the western slope as the Yucatan Undercurrent (Fig. 3a; see also Sheinbaum et al. 2002; Abascal et al. 2003) and the other off the Cuban slope. Transports of the Yucatan and deep Cuban countercurrents are 0.63 and 0.75 Sv (1 Sv ≡ 10⁶ m³ s⁻¹) toward the Caribbean (Table 1). The mean deep exchange in the Yucatan Channel thus appears as two lobes of upper and lower inflow into the Gulf of Mexico and two lobes of western and eastern outflow, with a velocity saddle point near the center, or also as three approximate top-to-bottom bands from side to side of the deep channel (Fig. 3a). This deep circulation plays a crucial role in renewing the deep water in the Gulf of Mexico.

The water masses present in the Gulf of Mexico are mainly those that enter through the Yucatan Channel from the northwestern Caribbean Sea (see, e.g., Morrison and Nowlin 1977, 1982; Mooers and Maul 1998; Nowlin et al. 2001), which are depicted in Fig. 2 and described below. As usual, each of these water masses is distinguished by one or more extrema or inflection points in the distribution of property values with depth. The somewhat fresher near-surface water mass reflects the influence of the Amazon and local dilution in the Caribbean Sea. The Subtropical Underwater (SUW; also known as Tropical Water), derived from the subtropical surface waters north and south of the equator, where evaporation exceeds precipitation, is characterized by a pronounced salinity maximum (S ≈ 36.7) at T ≈ 23°C and depths of 150–200 m, also present in the Loop Current and in the eddies shed from it. The water in the Gulf’s interior shows a decrease in salinity at the core of the SUW, probably because mixing decreases that layer’s salinity maximum at the western edge of the Yucatan Current, in the Loop Current, and within the core of the eddies as they age and spin down. Beneath the SUW is the 18°C Sargasso Sea Water, best characterized by an upper-level oxygen maximum of about 3.4 ml L⁻¹ at depths of 200–400 m (Kinard et al. 1974), S ≈ 36.3, and low stratification. It is worth noticing that the low stratification near 18°C and 36.3 psu water, with a relative oxygen maximum, is only found sporadically in patches across the Yucatan Channel. The next layer is formed by the Tropical Atlantic Central Water (TACW), characterized by an oxygen minimum at depths of 400–600 m, and thermohaline characteristics (8 < T < 16°C; 35 < S < 36.1) between the more saline North Atlantic Central Water and less saline South Atlantic Central Water (Metcalf 1976). At depths of 600–900 m, the AAIW can be identified by its characteristic salinity minimum close to S ≈ 34.8 and T ≈ 7°C. Below 1000 m and thus filling most of the deep Gulf of Mexico is a large amount of NADW, characterized by a salinity maximum of S ≈ 35.0 and T ≈ 4°C; this water mass enters the Venezuelan and Colombian Basins from the Atlantic through the Anegada–Jungfern Passage and between Jamaica and Hispaniola (Morrison and Nowlin 1982), but it also enters the Cayman and Yucatan Basins directly through the Windward Passage (Sturges 1965). These water characteristics are reflected across the section over the Yucatan Sill as a stronger temperature gradient in the upper 500 m, an oxygen minimum near 600-m depth, a salinity minimum near 700-m depth, and oxygen maximum in the lower 1000 m (Fig. 4).

The Caribbean Sea deep-water renewal through the Jungfern Passage (whose depth is close to 1900 m, the deepest entrance to the Caribbean Sea) has been documented by Sturges (1970, 1975). They showed that as the deep water from the Atlantic flows through the passage during a tidal cycle, the portion that reaches over the sill sinks into the basin, and the return flow is formed by less-dense water above; this renewal water sinks only an intermediate depth, but months or years later reaches the bottom as it mixes along its path, so that a “bottom water” is eventually formed. A similar process may be responsible for the ventilation of the Gulf of Mexico. The observation series reported by Sturges (1970, 1975) lasted only a few days so that subtidal frequencies could not be resolved adequately. In our case, the series lasted 21 months, and we show that events lasting from a few days to about one month dominate over the tidally modulated exchange, which, although quite energetic, contributes less than 3% of the exchange. The statistical analysis of historical data reported by Sturges (2005) shows a connection of the return flow that compensates the deep-water renewal of the Caribbean by sporadic flow over its two deepest sills (Anegada–Jungfern and Windward Passages) from the Atlantic, and the compensation flow from the Gulf of Mexico through the northwest Caribbean.

A few decades ago, McLellan and Nowlin (1963) reported data from 52 deep hydrographic stations (below 1500 m) distributed throughout the Gulf of Mexico. The resulting potential temperature and salinity distributions proved to be horizontally very uniform and with nearly neutral stratification (vertical gradients were observed but were small). This is consistent with the concept of a deep basin isolated by a sill whose depth is close to 2000 m (Yucatan Channel) and also that either the deep waters of the entire deep basin have a common source and a rapid redistribution mechanism or that their residence time is sufficiently long for exchange processes to obliterate horizontal gradients. In contrast with the horizontal uniformity of temperature and salinity, these authors’ deep observations of dissolved oxygen displayed considerable lateral variation, with a cell of relatively low values in the central Gulf and higher values near the Yucatan Channel. Nowlin et al. (1969) questioned the horizontal variation of dissolved oxygen in the deep water of the Gulf of Mexico and found no discernible evidence of that variation in deep-water observations, so they argued that the horizontal distribution of oxygen reported by McLellan and Nowlin (1963) might be attributed to poor sampling or inaccurate analysis of the data. Nonetheless, more recent observations (Fig. 1a) show that the deep distribution of dissolved oxygen of McLellan and Nowlin (1963) is in fact correct.

In the Gulf of Mexico, the oxygen values observed at depths below 3000 m exceed 5.0 ml L⁻¹, with a maximum near the bottom (see Fig. 1a). Beneath 1000 m, the amount of oxygen increases from the western Gulf toward the Caribbean, which suggests a recent inflow into the Gulf through the Yucatan Channel. Near 1800 m, an inflection in the oxygen profile caps a bottom layer of higher oxygen (Fig. 1a), and these higher values are also found in the Yucatan Basin below sill depth, indicating that water from deeper levels in the Caribbean is present over the Yucatan Sill.

3. Data

Seven moorings with Aanderaa RCM7 current meters provided 21 months, from September 1999 to June 2001, of hourly velocity and temperature measurements in the lower levels of the Yucatan Channel (Fig. 5). These data series correspond to two consecutive deployments, analyzed by Sheinbaum et al. (2002) for the first period (September 1999 to June 2000) and Ochoa et al. (2003) for the second period (July 2000 to June 2001). The distribution of current meters was based on the availability of sensors, a desire for a uniform coverage, and the alignment with hydrographic features such as the extrema of salinity, oxygen, or isotherms (Ochoa et al. 2003; see Figs. 1 and 2) in order to best document the motions of the various water masses present in the Channel. The mean-transport maps of the first and second deployments are very much in agreement (see Sheinbaum et al. 2002; Ochoa et al. 2003). The series at 1250 m was lost during the second deployment at mid-Channel and excluded from the present 21-month analysis (see Fig. 5) but is included in the discussion of the velocity fluctuations at the end of section 4c.

Short gaps in the time series were filled by harmonic fitting, including tidal frequencies. Hourly velocity and temperature data were low passed with a Lanczos filter with a cutoff period of 48 h and decimated to 6-hourly values. Transports were computed by interpolating the (filtered) velocity normal to the section onto a 0.05° by 15 m grid using a six-parameter objective mapping as described by Ochoa et al. (2001). Large length scales of 1.5° (≈150 km) in the horizontal and 1500 m in the vertical and a signal-to-noise ratio of 0.1 were used to estimate the mean background of each map; short length scales of 0.7° (≈70 km) and 400 m and a signal-to-noise ratio of 0.05 were used to map the anomalies. The scales of the mapping are consistent with the straightforward estimates of the correlation scales reported by Ochoa et al. (2003).

The mean temperature, salinity, and oxygen distributions in the Channel’s section (Fig. 4) are an average of 11 complete hydrographic crossings (108 casts) across the Yucatan Channel during three oceanographic cruises on August–September 1999 (4 crossings and 36 casts), June–July 2000 (4 crossings and 44 casts), and May–June 2001 (3 crossings and 28 casts), corresponding to the deployment, turnaround, and final recovery of the current-meters array. The casts also provided velocity profiles with depth, by lowering a Sea Bird SBE911 plus CTD, together with a 300-kHz RDI Acoustic Doppler Current Profiler (a probe usually referred to as LADCP).

The moored current meters recorded only velocity and temperature, but both salinity and oxygen are highly correlated with the temperature in the deep water of the Yucatan Channel (Fig. 2). Thus temporal variations of oxygen and salinity can be inferred from temperature variations with an accuracy of ±0.15 ml L⁻¹ for oxygen and ±0.008 for salinity. Near the bottom, a decrease in temperature is related to an increase in both oxygen and salinity. Temperature–oxygen and temperature–salinity relations [O₂(T) and S(T)] were obtained from third-degree and fourth-degree polynomials fitted to the hydrographic data shown in Fig. 2, from the 10°C isotherm to the bottom (Fig. 6).

4. Results

a. Mean heat budget

Given the close relation between the temperature, oxygen, and salinity in the depths of the Yucatan Channel, an evaluation of the contributions of heat transports through the Channel to the heat budget of the Gulf of Mexico, very interesting in itself, is also a useful way to study the ventilation of the deep Gulf. The heat budget can be conveniently assessed for the deeper layers in the Channel, below the level corresponding to the sill in the Florida Straits, and an equivalent calculation can be made for oxygen and salts, using temperature as a proxy variable. Elsewhere, the deep Gulf of Mexico is bounded by either the bottom or slopes through which exchanges are negligible, and waters cannot sink from the surface, as the water transformations required would be excessive. The heat budget (actually the temperature budget) for the deep water in the Gulf of Mexico, assuming no long-term warming or cooling, is then

View Expanded

a balance between the mean and eddy fluxes of heat through the Yucatan Channel, and an eddy vertical flux of heat across the horizontal surface A_top bounding the deep basin. In Eq. (1), C_p ≈ 4000 J kg⁻¹ °C⁻¹ is the heat capacity of seawater, ρ₀ ≈ 1025 kg m⁻³ is its density, T is the temperature, υ is the velocity normal to the Yucatan section A, and w is the vertical velocity. The overbars represent an average over the 21-month data series, the angle brackets represent spatial averages over the Yucatan cross-section area, and the primes represent fluctuations about the time average. The purely molecular diffusion across A_top is neglected because it is estimated to contribute less than 0.33 GW, which, as will be shown below, is negligible relative to the other terms in balance Eq. (1).

The first term in Eq. (1) is the heat flux associated with the mean circulation through the Yucatan Channel,

View Expanded

An unambiguous evaluation of this quantity can be made because the products of terms with brackets, which represent the unit-dependent “temperature flux” (Bryden et al. 1980; Richman and Badan-Dangon 1983), vanish conveniently when A is chosen such that the mean transport vanishes across the section (i.e., 〈υ〉 = 0). Only then does a calculation of heat (temperature) transport make sense (Montgomery 1974; Warren 1999). In fact, the vertically integrated transport across the deep channel vanishes not at one, but two, levels, near the 6.85° and 4.39°C isotherm depths: about 750 and 1350 m (see Figs. 7a and 3). Within error bounds, the choice of temperature or potential temperature makes no difference. Close to the bottom, the mean current flows into the Gulf of Mexico; but somewhat above that, mean countercurrents flowing out from the Gulf (into the Caribbean) dominate, exceeding the balance of the deepest inflow, above the 4.39°C isotherm. Still farther up the water column, the mean currents that form part of the base of the Yucatan Current again flow into the Gulf. The result is that the integrated mean transports vanish in two layers extending above (up to 6.85°C) and below the Yucatan Undercurrent core (at 4.39°C), and Eq. (2) can be evaluated in a meaningful way either for both layers separately or for both layers lumped into a single one. For each layer, the heat budget is as expressed in Eq. (1), a balance between the heat flux carried by the mean circulation, the eddy heat flux driven by oscillations, and the vertical exchanges within the Gulf, through the interfaces that define the layers.

Table 2 shows the mean heat flux through the Yucatan Channel area within each layer, integrated up from the bottom in the lower layer and up from the 4.39°C isotherm depth for the intermediate layer. The mean heat flux in the intermediate layer is then simply

View Expanded

gained by the Gulf because warmer water flows into the Gulf of Mexico at the base of the Yucatan Current, but cooler water leaves within the Yucatan and Cuban Undercurrents. In the lower layer (T ≤ 4.39°C), both the mean current and the temperature difference between the inflowing and outflowing layers are smaller, so only 153.0 GW are lost from the Gulf of Mexico as a result of the cold water that flows near the bottom into the Gulf and the comparatively warmer water that leaves in the undercurrent. With the two layers taken together, the deep Gulf below about 750-m depth gains on the order of 2.9 × 10³ GW by virtue of the mean exchange circulation.

Heat can also be carried by covariant fluctuations of the temperature and velocity fields at the Yucatan section, an eddy heat (temperature) flux of the form

View Expanded

The 21-month averaged eddy heat flux amounts to −48.6 GW for the intermediate layer and −59.2 GW for the lower layer, a loss for the Gulf in both layers (Table 2). The eddy flux of heat pumps about 110 GW out of the Gulf; it is less than one order of magnitude smaller than the heat exchanged by the mean flow in the intermediate layer but is nonnegligible for the heat balance of the lower layer. In addition, the eddy exchange of properties in the lower layer is crucial to the renewal of oxygen in the deep Gulf, whose distribution in the lower layer will be examined in the following section. The heat transport due to the mean exchange is clearly dominant in both layers; the eddy heat gain is a negligible addition to the gain of the intermediate layer but contributes about one-third more to the mean heat loss in the lower layer.

Figure 3b shows the 21-month averaged eddy heat flux per unit area in the Channel’s section. In the deepest part of the Channel, there is a layer of 200–300-m thickness through which heat is lost by the Gulf but mass enters it. That is, cold and dense near-bottom water reaches the sill and is expected to slide down the northern slope of the Yucatan Sill and thus to sink into the Gulf, as suggested by Sturges (1970, 1975) for the Jungfern Passage, although the precise dynamics of the downslope flow remain to be explored.

The net heat deficit of the lower layer must then be compensated for by a downward contribution from the intermediate layer by means of an equivalent eddy diffusion of heat. However, the considerable excess heat brought into the intermediate layer cannot be compensated for by the downward flux into the lower layer. It must then be removed into the upper layer of the Gulf by an upward, countergradient heat flux at the base of the Yucatan Current. There is no way to transport heat through the Florida Straits in either layer.

An illustrative formulation of this vertical exchange can be made in terms of an equivalent heat diffusivity, such that

View Expanded

where κ is an eddy diffusivity that parameterizes vertical exchanges. From the transport through the Yucatan Channel below a given temperature (T_r), we can compute the mean vertical velocity w occurring at the surface of the corresponding isotherm. Within two isotherms, by volume conservation,

View Expanded

where A_G is the area of the Gulf. Since w = 0 below the minimum temperature observed in Yucatan Channel,

View Expanded

As usual, an eddy diffusivity can be estimated as

View Expanded

where w_m = 2.3 × 10⁻⁷ m s⁻¹ is the mean lower-layer vertical velocity, from Fig. 7c; θ_z = 5.9 × 10⁻⁴ °C m⁻¹ and θ_zz = 1.0 × 10⁻⁶ °C m⁻² are estimates of the first and second vertical derivatives of the potential temperature evaluated at T = 4.39°C (the interface between the lower and the intermediate layers) from hydrographic profiles, and we suppose the eddy diffusivity to be horizontally and vertically uniform everywhere in the basin. This is a reasonable value for vertical diffusion in a deep oceanic basin (Munk 1966; Gregg 1987; Hopfinger 1987; Bryden and Nursen 2003), and the diffusion of heat computed in this manner across the top of the lower layer A_top ∼ 10¹² m² shows that a coefficient κ = 7.9 × 10⁻⁵ m² s⁻¹ suffices to set this balance, which is of the same order of magnitude.

A similar exercise for the top of the intermediate layer would yield a downward heat flux about one order of magnitude larger than that driven into the lower layer, since the temperature gradient is that much larger above the intermediate layer, further enhancing the heat surplus of the intermediate layer. Instead, heat must be extracted upward, against the vertical gradient, by an upward eddy flux of heat. By setting the balance of Eq. (5) to extract about 2.9 × 10³ GW of excess heat upward, given the temperature gradient near the upper portion of the intermediate layer as (∂T/∂z)_T=6.85°C ≈ 8.2 × 10⁻³ °C m⁻¹, and a similar size of the interface A_top ∼ 10¹² m², yields a negative diffusivity of κ ∼ −8.6 × 10⁻⁵ m² s⁻¹, a considerable countergradient flux. Such upgradient fluxes have been shown to exist elsewhere in geophysical systems (Starr 1968). Whatever mechanism might be responsible for this countergradient flux of heat [such as that present in the Gulf Stream (Oort 1964) and in the atmospheric boundary layer (Holtslag and Moeng 1991)], it requires an investment of energy, as does the motor of a refrigerator.

b. Mean oxygen and salts budget

As mentioned in section 3, dissolved oxygen and salts can be estimated from temperature in order to obtain proxy time series of oxygen and salinity fluctuations. An oxygen budget can be evaluated in a manner similar to that of heat, with the addition of a consumption term; thus

View Expanded

where O₂ is the dissolved oxygen and F_O2 is an oxygen consumption rate expressed in a way that supposes that organic matter is oxidized once it reaches the ocean floor, describes the balance between mean and eddy oxygen transports across the Yucatan Channel, and both the vertical eddy and molecular diffusion rates, plus the oxygen consumption associated with biochemical activity in the deep basin. For oxygen or salt budgets the requirement of 〈υ〉 = 0 is unnecessary because absolute transports are unambiguously defined, but we examine the calculations in the previously defined lower (between the bottom and the 4.39°C isotherm) and the intermediate (between the 4.39° and 6.85°C isotherms) layers.

Inasmuch as temperature and oxygen are inversely related within the NADW that occupies the lower levels of the Yucatan Channel, the oxygen budget includes similar contributions but operates in reverse fashion. In this way, the deep Gulf gains 138.4 m³ s⁻¹ by the mean exchange through the lower layer, and an additional 12.3 m³ s⁻¹ through covariant fluctuations of oxygen concentration and velocity (Table 2). This total gain is balanced by the oxidation of organic matter at the ocean floor. Rowe et al. (2003) report an organic carbon remineralization (estimated from a benthic chamber) of approximately 4.0 mg C m⁻² day⁻¹ (as CO₂) at 3650-m depth on the Sigsbee Plain (northwestern Gulf of Mexico). This value is equivalent to an oxygen consumption of 0.37 ml O₂ m⁻² h⁻¹, given a respiratory quotient of 0.85. Applying this consumption to the entire Gulf yields a rate of removal of oxygen of approximately 103 m³ s⁻¹. This leaves a small amount, maybe one-half that much, that needs to be diffused upward into the intermediate layer, which is probably a correct assessment of the lower-layer balance. The spatial distribution of the eddy oxygen transport per unit area in the channel’s section is shown in Fig. 3c, with flux directions roughly opposite to those of heat (Fig. 3b).

The calculation for the intermediate layer shows that the mean exchange circulation contributes a mean loss toward the Caribbean of 605.4 m³ s⁻¹, compensated for in small part by an eddy flux gain to the Gulf of 41.9 m³ s⁻¹ (Table 2) and a similar amount that might be diffused upward from the lower layer. As in the case of heat, even neglecting an upward diffusion of oxygen into the oxygen minimum layer above still requires a considerable downward countergradient flux of oxygen to balance the budget of the intermediate layer. One would suppose that if a plausible mechanism were to be found, it would be the same that is responsible for the vertical eddy flux of heat and, as we will show next, of salts. The existence of such a mechanism is consistent with the sustained presence of a midwater oxygen minimum layer throughout the western Gulf, immediately overlying the intermediate layer defined here.

Last, salts, computed like oxygen from their relationship with temperature, provide similar results. The lower layer gains 5.2 t s⁻¹ of salts by virtue of the mean exchange circulation, compensated for by an eddy loss of 0.4 t s⁻¹ of salts (Table 2; Fig. 3d), leaving the near totality of the gain to be diffused upward into the intermediate layer. The salts budget for the intermediate layer consists of a loss of 35.8 t s⁻¹ through the mean circulation, compensated for in part by an eddy gain of 8.0 t s⁻¹ (Table 2; Fig. 3d) and by the 4.8 t s⁻¹ diffused from the lower layer. This leaves an imbalance of 23.0 t s⁻¹ to be drawn from the levels immediately above the intermediate layer, in the same countergradient fashion as that of heat and oxygen.

c. Ventilating events and the bottom boundary layer

The near-bottom flow in the Yucatan Channel fluctuates greatly, as observed in the series plotted in Fig. 8: 0.32 ± 0.84 Sv in the mass transport (Fig. 8a), −29.7 ± 122.3 GW in the eddy heat flux (Fig. 8b), 17.5 ± 31.0 m³ s⁻¹ in the eddy oxygen transport (Fig. 8c), and 1.1 ± 1.7 t s⁻¹ in the eddy salts transport (Fig. 8d). The series show events during which the mass transport and the eddy oxygen transport are toward the Gulf, but the eddy heat flux is opposite; that is, colder and oxygen-rich water is entering the Gulf, such as during April–May 2000, one of the most important Gulf cooling–ventilating events in the series. By contrast, during November 1999, warmer and oxygen-poor water was exiting the Gulf, the sort of event that cools the deep Gulf but does not ventilate it, since it corresponds to deep water exiting the Gulf. These cooling events have variable intensities and last from a few days to about 1.5 months. Then, there are also numerous smaller warming events (either warm and oxygen-poor water entering the Gulf or cold and oxygen-rich water exiting from it), but the cooling–ventilating ones predominate in the series since the averaged eddy heat flux is 29.7 GW toward the Caribbean (Fig. 8b); the averaged mass, oxygen, and salts transports are 0.32 Sv, 17.5 m³ s⁻¹, and 1.1 t s⁻¹, all toward the Gulf (Figs. 8a,c,d). Figure 9 shows the variance-preserving spectrum of the eddy temperature flux υ′T ′ at the first and second deepest current meters from the bottom, at 2043- and 1630-m depths, of the central mooring (see Fig. 5). A considerable intensification at the deepest series is evident, with a highest, narrow peak in the spectrum near the diurnal frequencies and several broader peaks at lower frequencies. The tidally modulated exchange of heat is energetic but contributes less than 3% of the exchanged heat, as can be verified by comparing the time averages of the filtered (subtidal) and the unfiltered series of the deep temperature flux . Most of the subtidal variance is concentrated at periods longer than 15 days, with peaks near 23 days and 60–100 days.

At the beginning and at the end of the series, deep hydrographic casts were made near the central mooring during cooling and ventilating (mass transport toward the Gulf) events, whose along-channel velocity, temperature, dissolved oxygen, and salinity near-bottom profiles document the shape of the bottom boundary layer (Fig. 10). Unfortunately, casts are unavailable during other important events in the series.

A bottom boundary layer about 200 m thick, with cooler, saltier, and oxygen-rich water flowing into the Gulf is observed; this denser water must slide down the northern slope of the Yucatan Sill and sink to the bottom of the Gulf to sustain the near-bottom oxygen maximum. Figure 11 (left) shows the difference of density Δρ between the deep channel and the central Gulf at the same depth (below 1600 m). Fig. 11 (right) shows the difference of potential density Δσ_θ in the deepest 400 m of the channel (below 1600-m depth) and in the deepest 400 m of the central Gulf (below 3100-m depth); the density difference is small, but the densities are larger in the channel. Figure 12 shows the deepest portion of casts made during the 11 crossings of the Yucatan Channel during the cruises described in section 3; most profiles show a near-bottom inflection, attributable to the existence of a bottom boundary layer. The bottom boundary layer occupies the coldest 0.15°C of the water column; the ventilation of the lower layer is then modulated by the variability of the inflow within this boundary layer.

It is interesting to compare these values with theoretical or empirical formulations of oceanic boundary layers present in the literature. A bottom frictional stress is usually parameterized as

View Expanded

where ρ₀ = 1025 kg m⁻³, C_D = 1.6 × 10⁻³ is a common drag coefficient (see, e.g., Badan-Dangon et al. 1986), and υ_b is the total current velocity near the bottom, say at the deepest instrument. A “friction velocity,”

View Expanded

expresses an equivalent velocity relevant for frictional processes. From it, the thickness of the logarithmic layer, that is, the distance in which velocity decreases near the bottom following the well-known logarithmic law, is

View Expanded

where f = 5.3 × 10⁻⁵ s⁻¹ is the Coriolis parameter and G is the midwater “geostrophic” velocity measured well above the boundary layer.

The bottom boundary layer proper is the classical depth of frictional influence, which, without stratification (Wimbush and Munk 1970), reads

View Expanded

A correction for constraints of the stratification was proposed by Weatherly and Martin (1978) as

View Expanded

where N ²₀ = 4.4 × 10⁻⁵ s⁻² is the ambient Brunt–Väisälä frequency squared, computed from hydrographic casts. Figure 13 shows the temporal evolution of the boundary layer estimations described by Eqs. (12), (13), and (14); in general the boundary layer thickness is underestimated. The h_mb layer best approaches the boundary layer observed in Fig. 10, whose estimated thicknesses are of about 90 and 60 m, differing by a factor of 2 or 3. The drag coefficient for this region should then be corrected to a somewhat greater value as C*_D ∼ 4C_D = 6.4 × 10⁻³ in order to increase the estimated h_mb thicknesses by a factor of 2. This is in agreement with Bryden and Nursen (2003), who suggest that strait mixing associated with overflows across deep sills is much larger than the interior mixing in abyssal basins and is probably a function of bottom roughness.

Last, we examine the velocity fluctuations, whose time-dependent structure differs along the water column in the Yucatan Channel. The deepest currents show little coherence with signals beyond the vicinity of the bottom, except during the two most important events shown in Fig. 8. The strong cooling event (warm water exiting) of October–November 1999 extends up from the bottom to 800-m depth (Fig. 14). The largest ventilating event that occurred on April–May 2000 (see Fig. 8) is equally intensified near the bottom, beneath 823-m depth (Fig. 14). Otherwise, the fluctuations at the deepest current meter appear unrelated to the currents above the lower layer, which supports the notion that motions that are bottom trapped or restricted to a bottom boundary layer are responsible for inducing a renewal inflow.

5. Discussion

We have addressed, with the help of a heat balance, the origin of the cold, oxygen-rich water of North Atlantic origin (NADW) that fills the deep basin of the Gulf of Mexico. Exchanges beneath the level of the Florida Sill can only take place through Yucatan Channel and by mixing and diffusive exchanges with the upper layers. It appears that most of the heat exchange is by virtue of the mean exchange over the sill. This balance requires heat to be pumped upward from the deep basin, against the mean vertical gradient. The cooling of the deep Gulf by intermittent ventilating events and of the intermediate levels through an upward countergradient pumping of heat should set up a baroclinic pressure gradient toward the Caribbean at about 750- to 1000-m depth, a process that may drive the Yucatan and Cuban Undercurrents. This has indeed been confirmed by Sturges (2005), from his analysis of historical hydrographic data.

The vertically integrated transport across the deep channel vanishes at two levels, at the 6.85° and 4.39°C isotherm depths (about 750 and 1350 m), which define two layers that extend above and below the Yucatan Undercurrent, within which the heat balance can be evaluated separately. Quite interestingly, the levels of zero net transport across the deep Yucatan Channel correspond to a change of slopes in the profiles of oxygen, temperature, and salinity in the deep Gulf of Mexico (see Fig. 1), which implies that they must be sustained by different diffusive regimes and, hence, express different balances. The processes that lead to the distributions of velocity across the deep section as described in Fig. 3a and the diffusive regimes evidenced by the vertical profiles in Fig. 1 appear to be a deep inflow near the bottom of the Yucatan Channel into the Gulf of Mexico. This fills the deep basin of the Gulf and interacts with a midlevel inflow near the top of the deep layers, through vertical exchanges that conform the deep countercurrents outflowing into the Caribbean. This mean circulation pattern is shown schematically in Fig. 15, in good agreement with Fig. 9 of Sturges (2005), which shows an upper layer flowing into the Gulf, above 800 m, with strong vertical shear; immediately below, at depths of ∼800–1100 m, a second layer is flowing into the Gulf; and a third layer contains a mean return flow. The mechanisms by which the two inflows shown in Fig. 15 interact to collectively feed the outflowing undercurrents evidently take place within the Gulf of Mexico and must involve mixing, interactions with the boundaries of the basin, and the collection of cyclonic flows along the periphery of the Gulf, in manners described by DeHaan and Sturges (2005). This topic deserves much further research.

Most important, the deep inflow near the bottom of the Yucatan Channel must be sliding down the slope to fill the deepest portions of the Gulf of Mexico, since the coldest, oxygen-rich waters are found in the deepest portions of the basin (Fig. 1a). The detailed processes by which the undercurrent forms and the characteristics of the gravity current that fills the deep basin are beyond the scope of this paper but constitute very important subjects that need to be investigated; both are parameterized here simply as a vertical diffusion.

Then, the eddy diffusion of heat at the top of the intermediate layer must be upward, implying a negative diffusivity. This countergradient heat exchange may be an upward eddy heat pumping due to baroclinic instability of the Yucatan Current and in the Loop Current eddies, as observed in the Gulf Stream. Measurements in the surface Gulf Stream show a flow of kinetic energy from disturbances (meanders) to the mean flow (Webster 1961) and an eddy transport of heat from colder to warmer water (Oort 1964), against the ordinarily expected diffusion. Similar countergradient fluxes have been shown to provide adequate closure to turbulent balances of heat and other scalars in the convective atmospheric boundary layer (Holtslag and Moeng 1991). This energy transformation from eddy to mean energy is the “negative viscosity effect” discussed by Starr (1968), who explains that the deeper layers must supply eddy available potential plus internal energy to be converted into kinetic energy in sufficient amounts to drive the flows near the surface. This result was obtained later by Orlanski (1969) and Orlanski and Cox (1973) in their theoretical analyses of baroclinic stability of the Gulf Stream. This possibility constitutes a very important subject to be investigated for the Loop Current and Loop Current eddies, and will be examined in a forthcoming paper.

It is interesting to compare the amount of heat introduced into the deep Gulf (exported in its majority to the upper layer) mentioned above with the air–sea heat exchange at the Gulf’s surface. Some authors report an annual heat loss for the Gulf, of uncertain magnitude: 15.1 (Hastenrath 1968), 24.1 (Etter 1983), 2.2 (Adem et al. 1993), and 12.0 W m⁻² (Sánchez Montante 1999). Estimations reported by Zavala-Hidalgo et al. (2002) show instead an annual gain of 9.0 W m⁻². Our budget estimates would require the upper layer of the Gulf to gain heat from the intermediate layer at a rate ∼10³ GW, about one order of magnitude smaller than the uncertainty in the available estimates of heat exchange at the sea surface with the atmosphere and thus negligible in terms of the upper-layer budget.

The oxygen and salts budgets are roughly opposite to the heat budget. The intermediate layer loses oxygen and salts to the Caribbean through the mean flow (and gains a negligible amount by the eddy flow), which are replaced in part from the lower layer, which gains oxygen and salts from the Caribbean by the mean flow, mainly by means of vertical eddy diffusion. As in the heat budget, the eddy diffusion of oxygen in the intermediate layer is related to a negative diffusivity. A fact that seems to support the hypothesis of a countergradient eddy diffusivity at the base of the Yucatan Current is the existence of the oxygen minimum above the intermediate layer (Fig. 1a), which may be sustained by a downward oxygen pumping, similar to the upward heat pumping. The single value of oxygen consumption in deep water (Rowe et al. 2003), integrated to the entire abyssal Gulf, is comparable to the oxygen transported from the Caribbean through the Yucatan Channel, which emphasizes the efficiency of the mechanisms that sustain the high levels of oxygen in the lower layer.

Indeed, the net mass transport across the deepest portion of the lower layer (0.32 Sv) would replace the volume of the deep Gulf (∼2.5 × 10¹⁵ m³) in about 250 yr, which implies that the deep water is relatively recent, and the renewal quite strong. To estimate an error of the residence time, confidence limits on the mean near-bottom transport (Fig. 8a), the equivalent degrees of freedom were estimated following Priestley (1981). The 90% confidence limit is ±0.11 Sv (34% the mean value of 0.32 Sv). On the other hand, samples of 100 and 1000 bootstrapped means of random samples (see, e.g., Gentle 2002) taken from the series of near-bottom transport (Fig. 8a) have a standard deviation of ±0.02 Sv (5% the mean), a much more optimistic estimation of the error of the mean. Then, the error in the residence time of 248 yr is estimated to be from 12 to 84 yr, assuming the mean near-bottom transport of 0.32 Sv as representative for longer periods. The question of whether the 21-month average reported here is close to values for scales of order of centuries or not cannot be answered with the series analyzed here. The ratio of ventilating events depends on the eddy activity through the Yucatan Channel, which can vary over time. Ochoa et al. (2003) suggest that the transport series in the Yucatan for the period 1999–2001 are representative of longer periods of time, but those series are not necessarily related to the near-bottom eddy flow, so that the uncertainty in our values remains.

The ventilation of the Gulf of Mexico is then modulated by the variability of the ventilating events that last from a few days to about 1.5 months. During these events, colder, denser, and oxygen-rich water from deeper levels (beneath 2000 m) in the Caribbean is pushed over the sill within a bottom boundary layer close to 200 m thick. Near-bottom profiles show that during a ventilating event, the water in the Channel is slightly denser than that of the central Gulf (a density difference of ∼0.02 kg m⁻³; see Fig. 11), which is also observed in the difference of potential density between the deepest levels (below the 1600-m depth) of the Channel and the deepest levels (below 3100 m) of the central Gulf (Fig. 11). This slight difference should allow the dense water to slide down the northern slope of the Yucatan Sill as a gravity current. This current may initially fall rapidly before turning rightward (looking upslope) under the action of Coriolis force, but after approximately a deformation radius (which is about 40 km), the along-slope velocity may become almost steady and the rate of depth increase would be constant (Killworth 2001). Along its path, the current can develop wavelike disturbances, or even cyclonic eddies, and mixing can increase (Cenedese et al. 2004). A few years later, the renewal water reaches the bottom of the basin.

The origin of the fluctuations of the deep transports through the Yucatan Channel remains to be clarified. Abascal et al. (2003) suggest that flow variability in the Yucatan Channel is dominated by the passage of anticyclonic and cyclonic eddies (or anomalies) advected by the mean flow propagating northwestward through the channel. Altimetry observations over the Caribbean Sea support the notion that O (200 km) eddies can be found south of the Yucatan Channel, and a few have been detected as they are being advected into the Gulf of Mexico (Ochoa et al. 2005). However, a question still remains regarding how much of the variability seen in the channel could be locally generated by instability of the mean currents (Abascal et al. 2003). On the other hand, in their observational results, Bunge et al. (2002) found a strong relation between the variation of the surface area extension of the Loop Current and the transport through the Yucatan Channel beneath the 6°C isotherm such that the Loop Current volume increases are compensated for by deep outflows in the Yucatan Channel, in order to maintain constant the total Gulf of Mexico water volume. This result is also supported by the numerical results obtained by Ezer et al. (2003), who report that pulses of strong return flows below 800 m usually occur between eddy-shedding events, as the Loop is growing. The accumulation of Caribbean waters in the Loop Current need the outflows described by Bunge et al. (2002), regardless of the shape (Loop, or Loop plus the eddy shed) that they take. The shedding event itself does not require any extra compensation at depth; it only requires a different distribution of the waters that already occupy the Gulf’s interior.

The deep flows through the Yucatan Channel induced by the Loop Current fluctuations, reported by Bunge et al. (2002), do not seem to be the main forcing of the ventilation of the Gulf of Mexico. Figure 8 shows the periods of extension, growth, and retraction of the Loop Current (denoted by the Bs and the Ss, respectively), as reported by Candela et al. (2002). During periods S₁ and S₄, the retraction was abrupt and the Loop Current shed an anticyclonic eddy; these events are the only two eddy-shedding events that occurred during our observation period (Candela et al. 2002). Apparently, there is little relation between the variation of the surface area extension of the Loop Current and the evolution of the near-bottom eddy flow in Yucatan Channel (see Fig. 8), except probably during October–November 1999. This period corresponds to a strong cooling event (warm water exiting), which coincides with the second half of the Loop retraction period S₁ (when an eddy shed) and the entire Loop growth period B₁. This cooling event might be then attributable to a compensating deep outflow of the Caribbean waters upper inflow (independent of the shedding). This return flow is clearly visible below 800-m depth (Fig. 14), with an intensification toward the bottom. By contrast to this strong cooling event, the large ventilating event that occurred on April–May 2000 (see Fig. 8) corresponds to an intense deep inflow beneath 823-m depth (Fig. 14) that is stronger toward the bottom and appears to be unrelated to the Loop Current fluctuations.

Spectral analysis of the near-bottom eddy temperature flux υ′T ′ shows the highest peak at frequencies corresponding to tides, but, as mentioned above, their contribution to the eddy heat exchange is negligible. At subtidal frequencies, the spectrum shows a major peak at about 23 days, slightly weaker peaks at 60–100 days, and minor peaks at about 10 days (Fig. 9). The first two peaks are consistent with motions attributed to topographic Rossby waves (TRWs), as reported by Hamilton (1990). Also, the very strong intensification at the deepest series observed in the spectrum is a characteristic of TRWs. Similarly, Dengler et al. (2004) have shown that the deep western boundary current in the South Atlantic breaks in deep, possibly bottom-trapped eddies south of 8°S. The variance concentrated at the 10-day period might be somehow associated to vertically coherent meanders of the Yucatan Current, as it occurs in the Gulf Stream off Cape Hatteras (Johns and Watts 1986; Savidge 2004). Figure 9 then suggests that the inflow of the Gulf renewal water is mainly modulated by bottom-trapped motions or eddies formed in the bottom layer. The TRWs have wavelengths of 150–250 km and a group velocity of about 9 km day⁻¹ (Hamilton 1990) and are characterized by bottom intensification with vertical trapping scales from 600 to 300 m (Oey and Lee 2002). The eddies observed in the deep western boundary current (Dengler et al. 2004) have diameters of ∼120 km, height of ∼1000 m, depth of the center of ∼2100 m, maximum velocities at the core of ∼50 cm s⁻¹, and translation velocity of ∼4 cm s⁻¹. The TRWs are probably better candidates to be responsible for the modulation of the near-bottom renewal inflow into the Gulf, possibly by means of a topographic rectification. The denser water that enters close to the bottom flows into a weakly stratified region where the bottom drops abruptly to ∼3500 m, resulting in vortex stretching and inducing a cyclonic spin of the entering fluid. When water leaves the Gulf, the ambient stratification in the Gulf, though it is weak, would tend to restrict the source of outgoing fluid to depths above the sill, so that much less vortex compression is expected (Sturges et al. 2004).

6. Conclusions

Observational results show that the maximum values of oxygen in the Gulf of Mexico are found at the greatest depths, rather than toward the surface, below a layer of water of Antarctic origin that is poor in oxygen. The water in the deep Gulf must be renewed at a rather efficient rate in order to sustain the near-bottom oxygen maximum. Since the Florida Sill is well above the level of the oxygen minimum, the only path for this renewal or ventilation of the deep Gulf of Mexico can only take place through the Yucatan Sill. The near-bottom net transport measured over the sill (∼0.32 Sv) suggests a residence time for the Gulf’s deep waters of about 250 yr, an extremely short time, consistent with a vigorous renewal rate.

Beneath the Florida Sill, the Gulf of Mexico shows two layers of zero net transport. The intermediate layer (4.39°C < T ≤ 6.85°C) gains heat by the mean flow through the Yucatan Channel (∼3000 GW). Most of this heat gain (∼2800 GW) is exported to the upper levels by a countergradient, negative eddy diffusion, and a smaller portion (∼200 GW) is transferred to the lower layer (T ≤ 4.39°C) by a positive eddy diffusion and is thus exported toward the Caribbean by the mean and the eddy fluxes. The two layers taken together, the deep Gulf beneath 750-m depth, gain on the order of 2.9 × 10³ GW by virtue of the mean exchange circulation.

The eddy transfer of heat from the deeper layer of the Gulf toward the Caribbean (∼60 GW) occurs mostly in the form of ventilating events that last from a few days to about 1.5 months, during which colder, denser, and oxygen-rich water from deeper levels (beneath 2000 m) of the Caribbean flows over the Yucatan Sill within a bottom boundary layer of the order of 200- m depth and is thus expected to slide down the northern slope of the sill as a gravity current. The origin of the ventilating events remains unclear. A possible mechanism causing these events may be a rectified near-bottom flow associated to TRWs, which can be induced by meandering of the mean currents in the Yucatan Channel and the passage of eddies through it.

Topics deserving of further research are the assessment of the vertical eddy heat flux in the intermediate levels of the Gulf of Mexico and its relation to the eddy activity of the upper layers, the propagation of TRWs in the Gulf’s interior, the nature of velocity fluctuations in the deep Yucatan Channel, and the dynamics of the renewal flow from the sill into the Gulf’s abyss. Direct observations are now available in the central Gulf but will be reported elsewhere.