Volume 124, Issue 11 p. 8414-8428
Research Article
Free Access

Regional Modeling of Antarctic Bottom Water Flows in the Key Passages of the Atlantic

D. I. Frey

Corresponding Author

D. I. Frey

Shirshov Institute of Oceanology, Russian Academy of Sciences, Moscow, Russia

Correspondence to: D. I. Frey,

[email protected]

Search for more papers by this author
E. G. Morozov

E. G. Morozov

Shirshov Institute of Oceanology, Russian Academy of Sciences, Moscow, Russia

Search for more papers by this author
V. V. Fomin

V. V. Fomin

Physics Department, Zubov State Oceanographic Institute, Moscow, Russia

Institute of Numerical Mathematics, Russian Academy of Sciences, Moscow, Russia

Search for more papers by this author
N. A. Diansky

N. A. Diansky

Physics Department, Zubov State Oceanographic Institute, Moscow, Russia

Institute of Numerical Mathematics, Russian Academy of Sciences, Moscow, Russia

Faculty of Physics, Lomonosov Moscow State University, Moscow, Russia

Search for more papers by this author
R. Y. Tarakanov

R. Y. Tarakanov

Shirshov Institute of Oceanology, Russian Academy of Sciences, Moscow, Russia

Search for more papers by this author
First published: 11 November 2019
Citations: 32

Abstract

The goal of this research is to modify and apply a version of high-resolution three-dimensional numerical model for simulations of bottom circulation and to study the flows of Antarctic Bottom Water in abyssal channels of the Atlantic Ocean using this model. We adjusted the Institute of Numerical Mathematics Ocean Model σ-level ocean circulation model for several regions with intense bottom currents in abyssal channels. High vertical resolution near the seafloor allowed us to study the abyssal part of the ocean circulation, while high horizontal resolution is necessary for modeling currents in narrow underwater channels and fracture zones. We used our direct velocity measurements carried out at key points of the currents in the channels for verification of the model. This approach was applied in the regions with different seafloor topography: in the long and narrow Vema Channel with a strong bottom current and in several fracture zones of the Mid-Atlantic Ridge with rough bathymetry. On the basis of simulated three-dimensional velocity fields, we analyzed the spatial structure of the bottom currents along the entire length of the channels, determined maximum velocities at different sections, investigated the influence of the Ekman flux on the structure of the flows, and compared our model results with in situ observations. We also calculated the total transports of Antarctic Bottom Water through the fractures in several underwater ridges of the Atlantic Ocean.

Key Points

  • Regional modeling with high near-bottom resolution allows us to study the properties of bottom circulation
  • We simulate velocities and water transports of the bottom flows in abyssal channels of the Atlantic
  • Ekman flux caused by bottom friction transforms the thermohaline structure of the deep-water flows in narrow channels

Plain Language Summary

Deep and bottom waters of the ocean are formed in the polar regions and occupy a significant part of the ocean volume. Propagation of these coldest waters plays an important role in the heat transport of the ocean and influences the Earth's climate. At the same time, the lower part of oceanic circulation is less investigated than its upper part. The most intense bottom currents are formed in the narrow abyssal channels connecting ocean basins. Direct deep-water measurements at these points are technically complicated and time consuming; modern computer models are usually focused mostly on the simulations of circulation in the upper layer. The goal of this work is to study intense bottom currents in the key passages of the Atlantic using a numerical model. We adjust our ocean circulation model based on available velocity measurements in the most interesting parts of the currents and compute the spatial structure of the currents in the entire channel. These calculations allow us to get the entire pattern of bottom water motions through underwater ridges and to study some hydrodynamic features of deep-water flows.

1 Introduction

Antarctic Bottom Water (AABW) occupies deep basins of the major part of the Atlantic Ocean up to the midlatitudes of the Northern Hemisphere (Wüst, 1936). This water is formed at a few locations around the Antarctic continent, in particular, in the Weddell Sea (Baines & Condie, 1998; Orsi et al., 1999). Owing to the high density and meridional inclination of the upper boundary of AABW, part of this water propagates to the north in the bottom layer flowing from one ocean basin to another and gradually mixes with upper layer of North Atlantic Deep Water. The main pathways of AABW propagation in the Atlantic are determined by the bottom topography (Morozov et al., 2010; Sandoval & Weatherly, 2001) (Figure 1a). In the Atlantic Ocean, AABW is usually defined as a layer with potential temperature θ < 2 °C (Wüst, 1936). The northward spreading of AABW contributes to the meridional overturning circulation (Haarsma et al., 2008). It plays an important role in the heat transport of the ocean and influences the Earth's climate (Herrford et al., 2017).

Details are in the caption following the image
Scheme of AABW pathways in the Atlantic Ocean (a) and our model domains for simulating the bottom currents in the main underwater channels (b). The chart of AABW propagation is based on Morozov et al. (2010); the bottom topography is based on the GEBCO2014 data.

The deep-water part of oceanic circulation is less investigated than its upper part. The global aspects of abyssal circulation are usually studied based on the distribution of temperature, salinity, and other properties of near-bottom waters. Direct velocity measurements performed over abyssal plains do not represent the global spreading of AABW due to low mean velocities of its propagation. For example, the age of bottom waters in the Argentine Basin is estimated at approximately 30 years (Smythe-Wright & Boswell, 1998) which gives the mean velocity of AABW propagation in this basin less than 0.01 m/s. However, the bottom flows accelerate significantly in the narrow abyssal channels, which connect the deep ocean basins (Morozov et al., 2010; Whitehead, 1998). Studying bottom currents in such channels allows us to estimate the integral characteristics of AABW propagation.

Relatively high velocities the in narrow channels allow for the study the abyssal circulation based on direct velocity measurements and ocean models. Field observations are widely used for describing abyssal circulation (Naveira Garabato et al., 2002; Zenk, 2008; Zenk & Visbeck, 2013); global circulation models are not so often used for the specific studies of the bottom flows. High vertical resolution in the modern global models is usually used only in the upper layer. The horizontal resolution of these models is not sufficient for describing processes in the narrow abyssal channels. Even special models for studying AABW propagation (Santoso & England, 2008) do not resolve the motions in the lower part of the Antarctic waters through the narrow channels and fracture zones. Some particular models for the most intense flow in the Vema Channel (Wadley & Bigg, 1995) cover only a part of the current. Thus, the model studies with high resolution in the major part of the fracture zones in the Mid-Atlantic Ridge have never been realized.

It should be noted that flows through the studied abyssal channels exist in the regime of hydraulically controlled overflows (Pratt & Whitehead, 2007; Whitehead, 1998). The theory of the hydraulics of currents in a rotating channel allows calculations of volume fluxes through these channels based on known geometry of the topographic obstacles and thermohaline structure of the bottom waters (Whitehead et al., 1974). We use AABW transports predicted by this theory together with the existing experimental estimates for the verification of our model.

The goal of this paper is to analyze the bottom flows of AABW through the main passages in the ocean topography north of the Argentine Basin using a high-resolution circulation model adjusted for several regions of the Atlantic and compare these simulations with field observations and theoretical calculations to understand the spatial structure of bottom water flows in the Atlantic, to clarify pathways of AABW propagation, and to estimate velocities and volume transport of these waters. We simulated the bottom currents in the Vema Channel, Vema Fracture Zone, and equatorial Romanche and Chain fracture zones of the Mid-Atlantic Ridge (Figure 1b). Our direct velocity measurements with the lowered acoustic current Doppler profiler (LADCP) were used for verification of the model.

2 Model Realizations

We adjusted the Institute of Numerical Mathematics Ocean Model (Zalesny et al., 2010) for simulating the bottom water circulation in several regions of the Atlantic. This model is based on the full system of primitive thermo-hydrodynamic equations in spherical coordinates with the hydrostatic and Boussinesq approximations (Zalesny et al., 2010; Zalesny et al., 2012). The model has been well tested in different parts of the Atlantic Ocean including the coordinated ocean-ice reference experiments (COREs) international program (Danabasoglu et al., 2014). The Institute of Numerical Mathematics Ocean Model uses a vertical coordinate σ = σ(x,y,t) scaled by the depth of the ocean at the given point. It is specified by relation:
urn:x-wiley:21699275:media:jgrc23733:jgrc23733-math-0001
where z is the ordinary vertical coordinate in meters measured down from the undisturbed sea surface, H(x,y) is the ocean depth, ζ(x,y,t) is the sea surface height, (x,y) are the longitude and latitude, respectively. The σ-coordinates, which are also called isobathic, follow the topography more exactly and allow simulation of bottom circulation in a wide range of depths (Griffies et al., 2005; Jungclaus & Vanicek, 1999). The best resolution was specified in the layer of AABW (Figure 2), simulated by 17 levels with a step of 50 m over the ocean depth equal to 4,850 m. We also added a few additional levels near the sea surface for better modeling of the air-sea interaction.
Details are in the caption following the image
Schematic view of vertical σ-level distribution along the Vema Channel in the Southwest Atlantic including the Vema Sill at 31°12' S in the middle of the figure. The bottom is shown with gray color. High resolution in the lower layer of the ocean is specified in both deep basins and in all relatively shallow channels of abyssal ridges regardless their depths.

The horizontal resolution of the model is determined by the width of the narrowest part of the studied abyssal channels. We choose the resolution by specifying at least 10 model points across any section of the channel and specify equal spatial steps by latitude and longitude. The resolution varied from 0.02° or approximately 2 km in the relatively wide Vema Channel up to 0.002° of coordinate degrees or 200 m in the narrow fracture zones of the Mid-Atlantic Ridge. After determining the horizontal resolution, we specified the model domain, which included the entire bottom current in the channels and parts of the deep basins for correct modeling of the inflow and outflow from the channel.

The bottom topography data were taken from digital atlases GEBCO2014 and Global Multi-Resolution Topography (GMRT) (Ryan et al., 2009). The data were interpolated to the model domain and smoothed using the Tukey filter. Zero velocity and sea surface fields and temperature and salinity fields from the climatological World Ocean Atlas (Locarnini et al., 2013; Zweng et al., 2013) were specified as initial conditions. These data were also used in a buffer zone with a width of 10 points at the liquid boundaries (Klinck, 1995). In this zone, the model fits the ocean level and currents to the specified density of the ocean water practically according to the geostrophic relations. The atmospheric forcing was taken into the account by specifying the fluxes of heat, freshwater, and momentum at the boundary using the data from the CORE database (Large & Yeager, 2009). The horizontal resolution of these data is 1.875° × 1.875° by latitude and longitude. The atmospheric surface characteristics included air temperature, humidity, and wind velocity at a height of 10 m, precipitation, downward short- and long-wave radiation, and atmospheric pressure.

Large-scale horizontal turbulent diffusion of temperature and salinity was parameterized using a second-order operator with a coefficient of 1 m2/s for the model of the Vema Fracture Zone and 10 m2/s for the models of the Vema Channel and Romanche Fracture Zone. For the horizontal viscosity, we used a fourth-order operator with a coefficient of 5 × 105 m4/s in the model of the Vema Fracture Zone, 2.6 × 108 m4/s in the Romanche Fracture Zone, and 9 × 107 m4/s in the Vema Channel. As for the vertical viscosity and diffusion, we used the Philander-Pacanovsky parameterization (Pacanowski & Philander, 1981); in accordance with this work, the viscosity coefficients ν and diffusion k were calculated by the formulas:
urn:x-wiley:21699275:media:jgrc23733:jgrc23733-math-0002
urn:x-wiley:21699275:media:jgrc23733:jgrc23733-math-0003
where, ν0 = 10−2 m2/s, νb = 10−4 m2/s, kb = 5 × 10−6 m2/s are the background values of the viscosity and diffusion, Ri is the Richardson number, α = 5.0, n = 2. In the case of unstable stratification, the coefficient of vertical diffusion was set to 5 × 10−2 m2/s for the parametrization of convective mixing.

We simulated three-dimensional fields of the horizontal velocities, potential temperature, and salinity at σ-levels, and sea surface height. The numerical experiment was carried out in the regime of diagnosis and adjustment. At the first stage, the simulation is carried out with the “frozen” temperature and salinity fields. At this stage, only the sea level and current velocity components are prognostic variables of the model. This simulation is necessary to restore the structure of the flow field, since the initial velocity fields are set to zero. Further, when the current velocity field reaches the quasi-geostrophic balance with the density field and the kinetic energy stops to change significantly, we continue simulation in the adjustment regime. At this stage, temperature and salinity change together with the sea level and velocities; they are predicted variables of the model. Thus, the inevitable errors introduced into the initial temperature and salinity fields of low spatial resolution (and also into the density gradients) are significantly reduced in the model solution. The method of hydrophysical adjustment of density, velocity, and bottom topography is described in detail in (Sarkisyan & Sündermann, 2009). The total duration of simulation varied from 25 to 45 days depending on the study area; the model time step varied from 30 to 180 s. In our analysis, we used the estimates on the final day of calculation. The described method allows us to restore the ocean circulation based on the known temperature and salinity fields.

We verified the simulated velocity fields using our direct velocity observations. The velocity profiles were measured by the LADCP RDI WorkHorse Sentinel 300 kHz. Velocity measurements were accompanied by conductivity, temperature, and depth (CTD) measurements using the SBE SeaCat 19plus profiler. The measurements were performed from the sea surface almost to the ocean floor (usually 5 m above the bottom) from the R/V “Akademik Sergey Vavilov” of the Russian Academy of Sciences. The raw LADCP data were processed using the LDEO Software version IX.10 (Visbeck, 2002); the TPXO7.2 model (Egbert & Erofeeva, 2002) was used for subtracting the barotropic tidal velocities at the moment of measurements. We analyzed the lower part of the measured velocity profiles for verification of the model.

3 The Vema Channel

The Vema Channel is the main channel that allows transport of the coldest bottom waters to the tropical Atlantic. The major part of bottom waters formed in the Weddell Sea propagates to the north in the western part of the Atlantic, flowing through the Argentine and Brazil basins (Figure 1a). The Rio Grande Rise and Santos Plateau separate these deep basins from each other and prevent the free propagation of bottom water to the north. The deepest pathway in this underwater ridge is the Vema Channel, which was formed due to the long-term erosion of the bottom by the abyssal currents (Gamboa et al., 1983). The channel is approximately 4,650 m deep along its entire length, while the depths of the surrounding Rio Grande Rise are about 4,200 m. The bottom of the channel is relatively flat in comparison with the fractures of the Mid-Atlantic Ridge. The shallowest point of the Vema Channel is the Vema Sill at 31°12'S; the maximum depth of this sill measured by multibeam echo sounder is 4,614 m, and the width of the channel here is 18 km (Zenk et al., 1993). The transport of Antarctic waters through the channel was estimated at 2.5–3.5 Sv (Morozov et al., 2010).

We selected a rectangular domain between 37.0° and 22.8° S and between 45.2° and 29.0° W for computation of the bottom current in the channel (Frey et al., 2017). The horizontal resolution was 0.02°, which gives 814 × 764 points in the horizontal plane. Simulated velocities from the lowest σ-level are presented in Figure 3a. The model estimates of currents are up to 0.05–0.1 m/s in the Argentine and Brazil basins. On the contrary, the velocities in the Vema Channel are relatively high and reach 0.3 m/s. They are observed along the entire length of the channel from 33° to 26° S. A slight decrease in velocities is revealed in the wide middle part of the channel between 28° and 29° S, the maximum velocities here are about 0.2 m/s.

Details are in the caption following the image
Velocities in the bottom layer in the Vema Channel based on numerical modeling (a) and corresponding direct measurements at a few locations in the channel (b,c). The velocities are presented from the data of the deepest σ-level (approximately 50 m above the seafloor) for the depths greater than 4,550 m.

We made LADCP velocity measurements at three sections across the channel at 32°17', 31°12' (Figure 3c), and 26°23' S (Figure 3b). The bottom velocities based on both experimental observations and numerical modeling are 0.1–0.15 m/s at 32°17' S and 0.25–0.3 m/s at 31°12'; the currents south of 32°30'S are much slower. As for the outflow of AABW from the Vema Channel at 26°23'S, the speed of the bottom current is up to 0.3 m/s, which is the same as in the southern part. However, unlike the inflow of bottom waters from the Argentine Basin, the velocities exceeding 0.1 m/s are observed in the Brazil Basin even in 200 km to the north from the end of the channel (Figure 3a). Characteristic decrease of the velocities is about 0.05 m/s per 100 km of the flow. The estimates of the measured and simulated velocities as well as the volume transports of AABW are presented in Table 1. One can see that our model gives 2.0–2.7 Sv for transport of bottom waters with potential temperature less than 2 °C depending on the location of sections. The previous estimates are: 4.0 Sv based on long-term measurements on moorings at the Vema Sill (Hogg et al., 1999); 1.3–5.3 Sv based on measurements on different ocean sections (Holfort & Siedler, 2001; Mcdonagh et al., 2002; Zenk & Hogg, 1996); 1.1–3.1 Sv based on our LADCP measurements (Morozov et al., 2008; Morozov et al., 2019); 2.9–8.6 Sv based on theoretical predictions (Pratt & Whitehead, 2007; Killworth, 1992). It should be noted that measurements on moorings at the Vema Sill (Hogg et al., 1999) seems to be the best measurements for the transport calculations as they were taken over a long time period using several current meters at different depths. Strong differences of these estimations can be explained by several factors, including different locations of the sections and different time intervals for averaging of the data. For example, we take mean climatic data for simulations of velocities in our model; the use of instantaneous fields of temperature and salinity could improve the correlation with the measured estimates of transport.

Table 1. Maximum Velocities and AABW Transports in Three parts of the Vema Channel
Section of the channel Measurements Modeling
Maximum velocity, meters per second 32°17' S 0.12 0.14
31°12' S 0–0.65 0.26
26°23' S 0.3 0.31
AABW transport, sievert 32°17' S - 2.5
31°12' S 1.6–2.7 2.0
26°23' S - 2.7

A detailed map of the simulated currents in the bottom layer is shown in Figure 4. The best studied part of the channel is the Vema Sill at 31°12'S (Figure 4a). More than 20 CTD sections across the channel were made in this area since the 1970s. Long-term current measurements performed near the bottom at the sill show that the speed of the current varies from 0 to 0.65 m/s with a mean value of 0.3 m/s (Hogg et al., 1999). The LADCP measurements performed at 31°12' show the northward current over the entire section. Two branches of the current are observed in the middle part of the channel (Figure 4b). The main jet is located in the deepest part of the channel with depths of approximately 4,700 m; it is the main pathway for the coldest waters. The additional southern jet with velocities up to 0.2 m/s is observed at depths of 4,300–4,400 m on the southern (right) slope of the channel. The jets are specified in Figure 4b. The lengths of the jets (from south to north) are 200 km, and the maximum distance between them is 50 km.

Details are in the caption following the image
Simulated velocities in the bottom layer near the Vema Sill (31°12' S) (a) and in the wide middle part of the channel (b). Thick line indicates the experimentally well studied repeated section of the Vema Channel at 31°12' S (Morozov et al., 2010). Two branches of the current are observed in the middle part.

The distribution of potential temperature in the channel based on different data sets is shown in Figure 5. Direct CTD measurements (Figure 5c) reveal the physical phenomenon that the jet of the coldest waters displaces to the eastern slope. The difference in the bottom temperatures near the western and eastern walls of the channel is about 0.02°–0.03 °C. The eastern displacement of the coldest core is explained by the Ekman flux due to the bottom friction (Ekman, 1905). In the Vema Channel, this displacement was studied numerically in (Jungclaus & Vanicek, 1999). This effect is not observed in the World Ocean Atlas (WOA) data (Figure 5a) due to the low resolution. However, the simulation (Figure 5b) performed on the basis of the WOA data revealed the displacement of the cold waters to the eastern part of the channel similarly to our field measurements (Figure 5c). Both the model and measurements show inclination of the isotherms in the upper part of the current and a thermocline at a depth of 4,300–4,400 m near the eastern part of the channel. We examined this effect using two different versions of the model, with zero bottom friction (Figure 5d) and zero Coriolis force (Figure 5e). In the model without friction, the core of the coldest waters displaces to the west due to the Coriolis force only, which confirms that the Ekman flux is the key cause of the observed temperature shift. As can be seen from Figure 5e, the inclination of the isotherms is explained by the influence of the Coriolis force. Our model shows that this inclination and the cross-current Ekman flux appears in the bottom layer at any section of the Vema Channel; the simulated difference of bottom temperatures is 0.02 °C.

Details are in the caption following the image
Potential temperature sections across the Vema Channel at 31°12' S (location of the section is shown in Figure 4a with a thick line) based on different data: Interpolated WOA data were specified as the initial field for simulations (a); numerical modeling (b); direct CTD measurements (c); simulations without friction (d); and without Coriolis force (e).

4 The Vema Fracture Zone

Part of AABW from the Brazil Basin penetrates to the Guyana Basin through the Equatorial Channel (Hall et al., 1997). These waters propagate further to the East Atlantic through numerous fracture zones of the Mid-Atlantic Ridge north of 3°N latitude (Morozov et al., 2018). The other part flows to the north up to the deep ocean east of the Newfoundland Bank. The coldest part of these waters spreads to the east through the Vema Fracture Zone in the Mid-Atlantic Ridge located approximately at 11°N. In 2014–2016, several CTD/LADCP sections were occupied across the fracture at the main sills (Morozov et al., 2018). On the basis of these results, it was confirmed (after (Mantyla & Reid, 1983)) that the Vema Fracture Zone plays the most important role in transporting bottom waters to the Northeast Atlantic among all fracture zones of the Mid-Atlantic Ridge in the Northern Hemisphere. The total transport of AABW through the fracture was estimated at 0.7–1.2 Sv.

We chose a model domain between 10.5° and 11.25° N and between 42.7° and 39.5° W for studying the structure variability of the bottom flows along the fracture. The horizontal resolution was 0.002° for describing the current in the narrowest sections of the channels. The bottom topography here was studied by a multibeam echo sounder; hence, we used the bottom topography from the GMRT database (Ryan et al., 2009). The main topographic obstacle for the AABW flow is located at 41°10'–40°50' W (Figure 6a); the flow splits into three parts here. The main and coldest part of AABW propagates through the central channel. Its maximum depth at the sill is 4,690 m; the width of the channel at a depth of 3,850 m (the 2 °C isotherm) is 6.5 km. The southern channel is shallower and narrower than the central one, its depth is 4,500 m, and the width is 4 km. The northern channels are relatively wide and shallow; the depths of their main sills do not exceed 4,350 m.

Details are in the caption following the image
Simulated (a) and measured (b) velocities in the bottom layer of the Vema Fracture Zone at a height of 50 m above the seafloor. The data presented here correspond to the deepest σ-level in the regions of the fracture deeper than 4,250 m. The bottom topography is based on the GMRT database (Ryan et al., 2009), which assimilates multibeam echo sounder measurements. The velocity scales are the same for the measured and simulated values.

Simulated velocities and the bottom topography of the fracture are shown in Figure 6a. The velocities reach 0.4 m/s in the central and southern channels and 0.2 m/s in the northern channels (the estimates are given in Table 2). Acceleration of the flow occurs in a narrow region of the central channel between 41°08' W and 41°00' W. The maximum velocities are gained between 41°00' W and 40°55' W; then the flow gradually slows down. The velocities decrease by a factor of two over the 30-mile part of the channel. High velocities in the southern channel are observed only near the sill at 41°05' W. The channel leads to a depression with low bottom currents. A warmer part of AABW can propagate further to the east over the walls of the depression at higher levels. The currents in the northern channels are directed to the east and north depending on the local features of topography. The velocities reach 0.2 m/s only in the western part of the area with relatively narrow channels; further propagation to the east occurs trough the wide channel, and velocities here do not exceed 0.1 m/s. The estimates of AABW transport based on our model and measurements are shown in Table 2. The total transport through all channels of the fracture zone varies from 0.7 to 1.1 Sv; our estimates significantly differ from the value of 2.1 Sv calculated in (McCartney et al., 1991). These observations were performed near the outflow of bottom waters to the abyssal plains of the northeastern Atlantic. This part of the fracture is wider than the part of our measurements which partly explains the difference in transport estimates.

Table 2. Maximum Velocities and AABW Transports in Three Parts of the Vema Fracture Zone
Channels of the Vema FZ Measurements in 2014 Measurements in 2015 Measurements in 2016 Modeling
Maximum velocity, meters per second Central 0.34 0.46 0.37 0.38
Southern 0.25 0.40 0.29 0.42
Northern 0.21 - 0.21 0.19
AABW transport, sievert Central 0.86 0.50 0.52 0.43
Southern 0.26 0.30 0.14 0.16
Northern 0.08 - 0.05 0.15
Total transport, sievert 1.20 >0.80 0.71 0.61

The transversal structure of the bottom flow was studied at the sill of the main (central) channel (Figure 7). The temperature and velocities were measured over the meridional section at 41°01' W. The simulated temperature and velocity fields were interpolated to the same section. The effect of the displacement of the potential temperature minimum is observed near the bottom of the channel. Unlike the Vema Channel located in the Southern Hemisphere, the Ekman friction in the bottom layer in the Vema Fracture Zone displaces the cold dense core to the left.

Details are in the caption following the image
Measured (a) and modeled (b) potential temperature and measured (c) and modeled (d) zonal velocity at the section across the main channel of the Vema Fracture Zone at 41°01' W.

Our model also allowed us to simulate the transformation of thermohaline properties of the waters along the fracture (Figure 8). White dots indicate locations of all stations with measurements up to the seafloor, the values of measured bottom potential temperature are given in Figure 8. The temperature and salinity increase from west to east; there are two different mechanisms of this increase. First, the coldest and densest bottom waters cannot overflow the topographic obstacles at the sills of the channel, and only relatively warm and more saline waters spread further. Second, temperature and salinity increase due to the mixing of the jet with overlying waters. Our simulation shows that the most significant increase of thermohaline properties occurs east of the main sill in the jet. Hence, the mixing of the flow is manifested in the temperature and salinity rise. The simulated values of temperature increase were 0.4 °C and those of salinity were 0.05 PSU.

Details are in the caption following the image
Simulated potential temperature (a) and salinity (b) distributions in the bottom layer of the Vema Fracture Zone at the ocean depths greater than 4,250 m. Depths less than 4,250 m are shown with gray color. The measured values are indicated at the locations of stations shown with white dots.

5 The Romanche Fracture Zone

Propagation of bottom waters to the East Atlantic occurs mainly through three deep-water channels: the Romanche, Chain, and Vema fracture zones (McCartney et al., 1991; Messias et al., 1999). The first two of them are located near the equator. We simulated bottom currents in the domain, which completely covers the Romanche and Chain fracture zones: from 28° to 8° W and from 4° S to 5° N. The horizontal resolution in this region was equal to 0.02° or approximately 2 km. The Romanche Fracture Zone is 800 km long and 10–40 km wide. There are a few sills along the fracture. We analyzed the AABW flow in a few narrow places with the local intensification of the bottom currents.

The chain of sills and flow contractions in the Romanche Fracture Zone form a giant deep cataract (Whitehead, 1989). The inflow of AABW occurs in the western part of the Romanche Fracture Zone (Tarakanov et al., 2018) (Figure 9). The main flow in this region crosses the first zonally oriented sill approximately at 22°26' W, 1°06' S (southern entrance), then turns to the east and propagates through the second sill located at 22°20' W, 1°00' S. The maximum velocities are 0.25 m/s at the first and 0.28 m/s at the second sill. An additional inflow is located at 22°35' W, 1°03' S west of the main southern entrance; the maximum velocities here do not exceed 0.2 m/s. Our direct LADCP velocity measurements were performed at a few locations in different parts of the region (Tarakanov et al., 2013; Tarakanov et al., 2018) (Figure 9b); characteristic velocities measured by LADCP and moored current meters vary from 0.2 to 0.4 m/s (Tarakanov et al., 2018). It should be noted that the jet displaces to the northern wall of the fracture after crossing all the sills. This effect is pronounced both in the simulated and measured (meridional section at 22°10' W) data: The velocity at Station 2470 is the highest among five stations at this section (Figure 9b).

Details are in the caption following the image
The inflow of AABW to the Romanche Fracture Zone in its western part based on numerical modeling (a) and direct measurements (b). The velocities are presented at a height of 100 m above the seafloor for the ocean depths greater than 4,300 m.

High zonal velocities are observed at the sills in the eastern part of the Romanche Fracture Zone. We analyzed sections of zonal velocity over two sills at 16°03' W and 15°08' W; the simulated velocities reach 0.45 and 0.35 m/s, respectively (Figure 10). In 2005 and 2009, we made CTD/LADCP measurements across the Romanche Fracture Zone at 16°03' W. The along-channel velocity distribution was quite similar to the one shown in Figure 10с with the maximum velocities of 0.27 m/s. The velocity estimates for all studied sections are presented in Table 3. We also simulated the transports of AABW through the fracture at different sections; their positions are shown and indicated in Figures 9 and 10. The total transport of waters with potential temperature below 2 °C varies from 1 to 1.5 Sv based on our simulations. Calculation of the transport based on long-term measurements gives 0.66 Sv (Mercier & Speer, 1998); the calculation was made for the waters with potential temperature θ < 1.9 °C, which partly explains high values of our estimates. Note that the location of this section is different from the locations studied in this work.

Details are in the caption following the image
Simulated velocity distributions across the Romanche Fracture Zone at 16°05' W (a,b) and 15°08' W (c,d). The bottom is shown with gray color. The sections of measurements are indicated with vertical black lines (a,b).
Table 3. Maximum Velocities and AABW Transports at a Few Sections Across the Romanche Fracture Zone
Location of section Maximum velocities, meter per second Modeled AABW transports, Sv (θ < 2 °C)
Measurements Modeling
1 01°05' S 0.25 0.19 1.55
2 22°20' W 0.15 0.42 1.11
3 22°06' W 0.16 0.10 0.96
4 16°03' W 0.27 0.41 -
5 15°08' W - 0.35 -
  • Note. Locations of sections are shown with thick lines in Figures 9 and 10.

We compare potential temperature distributions across the studied channels at locations with intense bottom currents in the Vema Channel, Vema,and Romanche fracture zones (Figure 11). We can see that temperature minimums at these cross-sections are located near the right (Figure 11a) and left (Figure 11c) slopes of the channels. It is clear that these displacements of the coldest water cores depends on the position of the channel relative to the equator. Isotherms of potential temperature are practically horizontal in the Romanche Fracture Zone (0.5°N), while in the Southern and Northern hemispheres, we observe the displacement of the cold jet across the channel to the right and left walls, respectively. This occurs due to change of the Coriolis force direction and corresponding change of the Ekman circulation in the bottom layer.

Details are in the caption following the image
Comparison of modeled potential temperature distributions across the Vema Channel in the Southern Hemisphere at 31° S (a), the Romanche Fracture Zone near the equator at 0.5° N (b), and the Vema Fracture Zone in the Northern Hemisphere at 11° N (c).

6 Summary and Conclusions

Abyssal currents in the narrow deep-water channels and fracture zones were studied using combination of modeled velocity fields and direct point measurements. We focused on the key channels for AABW propagation in the Atlantic: the Vema Channel in the Southwestern Atlantic; the Romanche Fracture Zone near the equator and the Vema Fracture Zone of the Mid-Atlantic Ridge in the tropical North Atlantic. The results of this work are listed below:
  1. A three-dimensional model of the intense gravitational currents was adjusted and tested for abyssal channels with different regimes of bottom water motion. We gained good correlation between the model and our direct measurements. It was also shown that this approach allows us to study unknown features of the bottom circulation such as additional jets of the abyssal currents, evolution of the flows along the channels, and features of the outflows from these channels.
  2. Strong bottom current in the Vema Channel is observed along its entire length; the maximum velocities vary from 0.2 to 0.3 m/s at different cross-sections of the channel. Splitting of the current into two branches was found in the wide middle part of the channel at 29 S. It was confirmed that the bottom friction causes the cross-channel Ekman flux, which modifies the transversal structure of the flow. It was also shown that inclinations of the isotherms in the upper layers of the bottom current are caused by the Coriolis force. Combination of these two effects leads to the formation of a sharp thermocline in the eastern part of the channel approximately at a height of 200 m above the seafloor.
  3. An eastward flow with velocities up to 0.4 m/s is observed in the main channel of the Vema Fracture Zone. Based on the modeling, this flow exists between 41°30' and 40°30' W; on the contrary, the velocities of the currents in the northern channels become less than 0.1 m/s at 41°45' W after passing the narrow sills. Velocities reach 0.4 m/s in the narrowest part of the fracture; we estimate the AABW transport here at approximately 1 Sv. Analysis of simulated thermohaline fields shows that the increase in temperature and salinity along the current occurs mainly due to mixing; the main sill does not prevent the coldest bottom waters to propagate through this topographic obstacle.
  4. The bottom current in the Romanche Fracture Zone changes significantly along the channel. Velocities up to 0.3–0.5 m/s are observed at a few narrow sills; the current slows down up to zero velocities in the wide parts of the channel. We estimated transports of bottom waters with different potential temperatures (which were selected as the upper boundary of the AABW current) through the fracture; the total transport of AABW with potential temperature θ < 2 °C exceeds 1 Sv.
  5. Bottom friction and the Coriolis force cause the Ekman flux in the deepest layer of all studied underwater channels and significantly transforms the lateral structure of the bottom flows. This phenomenon is not observed near the equator; the direction of the displacement depends on the hemisphere; the difference in the bottom temperatures at a fixed level reaches 0.02–0.03 °C.

Acknowledgments

The work supported by the State Task of Russia 0128-2019-0008, the Russian Science Foundation 16-17-10149 (field studies), and Russian Foundation for Basic Research 18-05-01107 (adjustment of the numerical model) and 19-57-60001 (verification of the model and analysis). We thank the editor and two reviewers for their time and efforts and for the very important comments to the manuscript that helped us to make the presentation better. The results of numerical simulations and all experimental data used in the publication are available in open access through Pangaea (https://doi.pangaea.de/10.1594/PANGAEA.907919). Calculations have been performed by means of MVS-100K supercomputer resources of the Joint Supercomputer Center of the Russian Academy of Sciences.