1. Introduction

Oxygen minimum zones (OMZs) are present in the eastern tropical oceans between 100 and 700-m depth (e.g., Karstensen et al. 2008). These zones are caused by a combination of weak ventilation responsible for the shadow zones of the ventilated thermocline (Luyten et al. 1983) and elevated oxygen consumption due to enhanced respiration, a consequence of their proximity to coastal upwelling regions that are hotspots of biological productivity. The recently reported expansion of OMZs in the World Ocean (Stramma et al. 2008b) has, in large ocean areas, major implications for biogeochemical cycles, particularly those of carbon and nitrogen, and might also have adverse effects on ecosystem function and services (Chan et al. 2008; Keeling et al. 2010). As proposed by Joos et al. (2003) and confirmed in a recent review by Keeling et al. (2010), deoxygenation can now be regarded as a near-ubiquitous feature of all ocean basins that is indicative of the effects of global warming. Although the science of ocean deoxygenation is still in its infancy, a clear convergence of observational and model evidence suggests that more significant changes are looming. These effects are further aggravated by concurrent oceanic invasion of anthropogenic CO₂, which further tightens the limits to marine life (Brewer and Peltzer 2009).

Despite the potentially serious consequences of ocean deoxygenation, the physical and/or biogeochemical processes that are responsible for the observed changes in the oxygen content of the OMZs are largely unknown. There are several possible explanations how anthropogenic forcing can cause a decrease in the oxygen content. Among them are: (i) an increase in heterotrophic respiration along the pathways of ventilating water masses due to excess organic carbon formed at higher CO₂ levels (Oschlies et al. 2008), (ii) an increase of organic matter oxidization in shallow waters due to reduction of mineral-ballast fluxes (Hofmann and Schellnhuber 2009), (iii) an increased stratification and a corresponding reduction of ventilation, or (iv) solubility changes associated with a warming of subducted water masses (Bopp et al. 2002; Matear and Hirst 2003). However, a recent study by Gnanadesikan et al. (2007) showed using a simple advection–diffusion model that with increased stratification and resulting reduced ventilation the average water mass age within the OMZ can still decrease. This counterintuitive result, which emerged from a changed relative contribution of old deep waters and young near-surface waters, points toward the importance of different processes, such as vertical and lateral mixing, for local oxygen balances. Recently observed oxygen changes in the subtropical and subpolar water masses of the North Atlantic Ocean have mainly been attributed to changes in circulation and water mass properties associated with decadal changes of the North Atlantic Oscillation (Johnson and Gruber 2007), the primary climate mode of the Northern Hemisphere. Stramma et al. (2008b) showed that there are substantial differences between the oxygen changes within the OMZs of the World Ocean. The oxygen trend of −0.34 ± 0.13 μmol kg⁻¹ yr⁻¹ in the 300-to-700-m layer of the North Atlantic OMZ is twice as large as trends obtained for the OMZs of the South Atlantic, Pacific, and Indian oceans (Stramma et al. 2008b). Thus, there might be different processes contributing to the deoxygenation of the OMZs in the different ocean basins.

The velocity field in the tropical Atlantic Ocean is characterized by vigorous surface and thermocline current branches (Fig. 1). Among them are the North Equatorial Current (NEC), northern and central branches of the South Equatorial Current (nSEC; cSEC), the North Equatorial Countercurrent (NECC), its northern branch (nNECC), the North Brazil Current (NBC), the North and South Equatorial Undercurrents (NEUC; SEUC), and the Equatorial Undercurrent (EUC). Below the thermocline, the flow is generally weaker. Near-equatorial observations revealed the existence of a mean westward flow at the equator, the Equatorial Intermediate Current (EIC), and eastward current branches near 2°N and 2°S, the Northern and Southern Intermediate Countercurrents (NICC; SICC), respectively (Tsuchiya et al. 1992; Schott et al. 1995; Boebel et al. 1999; Bourlès et al. 2002; Brandt et al. 2006, 2008). More recently, it became evident that these zonal current branches are part of a system of multiple latitudinally alternating zonal jets with meridional wavelengths of 2°–4° latitude. The direct observation of these jets at intermediate depths within the OMZ is hindered because of dominant eddy motions in the region. However, their existence has been verified using global altimetry (Maximenko et al. 2005) as well as with subsurface floats drifting near 800- and 1000-m depths in the tropical Atlantic (Ollitrault et al. 2006). These jets, which are believed to be generated by the eddy field in the presence of a planetary potential vorticity gradient, are also reproduced in high-resolution models of the Pacific (Nakano and Hasumi 2005; Richards et al. 2006) and Atlantic oceans (Eden 2006).

Both latitudinally alternating zonal jets and eddy fluxes contribute to the isopycnal ventilation of the OMZ in the eastern tropical North Atlantic. While eastward jets transport high-oxygen waters from the western boundary toward the OMZ, the relative importance of zonal and meridional eddy fluxes depends on the anisotropy of both horizontal eddy diffusivities and horizontal oxygen gradients (Fig. 1). Here we analyze long-term changes in the hydrographic fields and the associated geostrophic flow between the periods 1972–85, a period characterized by particularly good data coverage, and 1999–2008 and relate them to possible changes in the ventilation of the OMZ.

After describing the available data used for this study in section 2, the mean and variability of currents and hydrographic properties for the period 1999–2008 are discussed in section 3. Changes in the zonal flow field and hydrographic properties between 1972–85 and 1999–2008 are investigated in section 4. In section 5, we present a conceptual model for the ventilation and oxygen consumption in the OMZ. This model is aimed at corroborating a new hypothesis, in which the mean oxygen level in the OMZ is affected by changes in the strength of alternating zonal jets. Summary and discussion are presented in section 6.

2. Data

For the period 1999–2008, we make use of current and hydrographic shipboard measurements obtained from various research cruises into the central tropical Atlantic (Table 1). An updated mean zonal velocity section between 28.5° and 23°W is derived after Brandt et al. (2006). For direct current measurements, vessel mounted and lowered acoustic Doppler current profilers (ADCPs) have usually been in use. Accuracy of 1-h vessel-mounted ADCP averages is estimated to be better than 2–4 cm s⁻¹ (Fischer et al. 2003) and the accuracy of lowered ADCP data is assumed to be better than 5 cm s⁻¹ (Visbeck 2002). During postprocessing, all available current measurements are merged taking into account the higher accuracy and better horizontal resolution of the vessel mounted ADCP. For each section, the current and hydrographic data are mapped on a regular meridional (0.05° latitude) and vertical (10 m) grid using a Gaussian interpolation scheme. The respective mean sections¹ are derived by averaging all existing data at each grid point, finally smoothed by a Gaussian filter [horizontal and vertical influence (cutoff) radii: 0.05° (0.1°) latitude and 10 m (20 m), respectively]. When calculating the mean velocity field from a large number of velocity sections, the measurement error is believed to be small compared to the error of the mean velocity because of noise associated with oceanic fluctuations, such as, for example, the mesoscale eddy field and internal waves. Near the core of the OMZ; that is, between 7° and 11°N and between 300- and 500-m depth, the local standard deviation of the zonal velocity averages to 4.3 cm s⁻¹, which yields a standard error of 1.4 cm s⁻¹ for 8–12 independent measurements from the individual ship sections.

For the earlier period from 1972 to 1985, we have assembled hydrographic profiles (salinity, potential temperature, and dissolved oxygen) measured within the region 0°–14°N, 22°–24°W (Table 2) from the database that was used to produce the World Ocean Circulation Experiment (WOCE) Global Hydrographic Climatology (WGHC; Gouretski and Koltermann 2004). The data accuracy, especially of the older datasets, can only be estimated. A substantial number of profiles was collected following the standards set by WOCE. Older profiles are likely to have achieved a lower accuracy. Overall we conservatively assume that the accuracy of any data point falls within twice the WOCE standard for water samples; that is, 0.01°C, 0.004, and 5 μmol kg⁻¹ for temperature, salinity, and dissolved oxygen, respectively.

To ensure a high data quality for the earlier period, obvious outliers are removed by visual inspection and a standard deviation criterion is applied for latitude (neutral density) [1°–2° (0.1)] or latitude (depth) [0.5° (50 m)] boxes; that is, values deviating more than two standard deviations from a preliminary box median are excluded. In addition, the distribution of hydrographic properties along different sections is taken from the relatively smooth climatology obtained by Gouretski and Koltermann (2004), hereinafter referred to as WGHC.

While we use data from ship sections between 28.5° and 23°W for the general description of velocity and hydrographic fields, we use only data acquired between 22° and 24°W for the analysis of trends and variability of hydrographic properties (see Table 1). This narrowing of the region is necessary to avoid biases due to the presence of zonal gradients in the distribution of hydrographic properties, particularly oxygen and salinity.

3. Circulation and hydrographic fields

b. Oxygen–salinity variability

Local hydrographic distributions exhibit a large variability on small space and time scales. This variability is thought to be generated by the mesoscale eddy field and by zonal jets acting on background gradients of the different hydrographic properties. The result is the generation of fine-scale variability; that is, the formation of filaments characterized by strong lateral gradients of hydrographic properties, which ultimately must be dissipated by turbulent diffusion (Ferrari and Polzin 2005). In the following, the relationship between anomalies of oxygen ΔO₂ and salinity ΔS, is analyzed (Figs. 4 and 5). These anomalies are calculated for 2° latitude by 2° longitude boxes by subtracting the respective box averages. To quantify the range of possible ΔO₂–ΔS relations, we calculated variance ellipses in the ΔO₂–ΔS space (encompassing 95% of the data points) that are obtained by rotating their axes into the direction of maximum variance. These ellipses are then compared with zonal and meridional ΔO₂–ΔS relations taken from the WGHC along the 23°W section from the equator to 15°N (dashed curves in Fig. 4) and along zonal sections from the western to the eastern boundary cutting through the box centers (solid curves in Fig. 4). The WGHC relations are derived by subtracting the WGHC oxygen and salinity values at the box centers from the WGHC oxygen and salinity values along the meridional and zonal sections, respectively. In this way, we account for different mean values of the WGHC and recent shipboard data. The intersection between the WGHC curves and the variance ellipses defines the source region of the fine-scale variability; that is, its latitudinal and longitudinal ranges. The range of the WGHC ΔO₂–ΔS relations covers the recent hydrographic variability except for the very low oxygen values probably originating near the eastern boundary and the very low salinity values probably originating in the Southern Hemisphere. If the WGHC curves do not intersect the variance ellipses twice, the corresponding end points of the WGHC meridional (0° or 15°N) and zonal (eastern or western boundary) sections are taken as boundary of latitudinal and longitudinal ranges, respectively. Strictly speaking, these ranges can only be obtained unambiguously if the WGHC zonal and meridional ΔO₂–ΔS relations are linearly independent, which is only roughly the case. The obtained ranges are between 15° and 25° in longitude and between 5° and 10° in latitude indicating significantly more zonal than meridional stirring (Fig. 5).

The ratio between latitudinal and longitudinal range can be used to estimate the relative strength of zonal versus meridional stirring. According to the mixing length theory after Armi and Stommel (1983), the effective diffusivity can be estimated by k_e ≈ c_eU_eL_e, where c_e is an efficiency factor, U_e is the rms eddy velocity, and L_e is a measure of the lateral transfer scale (see Ferrari and Polzin 2005). The rms eddy velocity is proportional to the rms surface-height anomaly associated with the geostrophic eddy field ζ_e divided by the corresponding length scale; that is, U_ex ∼ ζ_e/L_ey and U_ey ∼ ζ_e/L_ex. When assuming an isotropic efficiency factor (i.e., c_ex = c_ey), the resulting ratio of the effective diffusivities is

View Expanded

Here we obtained a ratio of longitudinal to latitudinal range of about 3 (slightly increasing with neutral density and toward the equator) suggesting strong zonal–meridional anisotropy of the effective eddy diffusivity with the ratio of k_ex/k_ey to be about 9. It is important to note that the ratio of effective diffusivities derived here also includes the stirring effect of east- and westward jets. These jets contribute to the local variability in the ΔO₂–ΔS relation. The anisotropy of zonal and meridional diffusivity induced by the eddy field (without the effect of zonal jets) should thus be smaller than 9.

4. Long-term changes

Stramma et al. (2008b) reported an expansion of the OMZs of the tropical oceans. They particularly showed that since 1960 minimum oxygen values declined continuously and that the OMZ expanded vertically in the tropical North Atlantic (10°–14°N and 20°–30°W). The downward oxygen trend of the tropical North Atlantic OMZ was estimated to be twice as large as trends obtained for the OMZs of the South Atlantic, Pacific, and Indian oceans. In comparison to the recent data presented in the previous section, we use here hydrographic data taken between 0° and 14°N as well as 22° and 24°W during 1972–85. This period was chosen as it is characterized by high oxygen levels and by exceptionally good data coverage.

Because no direct velocity observations are available for the period 1972–85, we calculated geostrophic zonal velocities from hydrographic data (Fig. 6a). Error estimates of geostrophic velocities referenced to 1000 m, calculated using standard errors of the geopotential anomalies, yield an uncertainty of about 1 cm s⁻¹ between 7°–11°N and 300–500-m depth. The geostrophic calculation results generally in an underestimation of the jet strength due to horizontal averaging and due to the chosen reference level at 1000 m, at which zonal velocities associated with east-/westward jets could have nonzero contributions (e.g., Ollitrault et al. 2006). Data coverage does not allow deeper reference levels. Details of the geostrophic calculation and error estimates are given in the appendix.

The general structure of the zonal velocity and oxygen distributions for the period 1972–85 is similar to those for the period 1999–2008. The eastward jets at 6° and 9°N are more pronounced and deeper reaching during the earlier than during the later period. The mean oxygen concentration for the period 1972–85 (Fig. 6b) is generally higher than the oxygen concentration found for the period 1999–2008 (Fig. 2b), and, associated with the stronger jets, the high-oxygen tongues near 6° and 9°N are similarly more pronounced and deeper reaching (Fig. 6b). The comparison between the oxygen distribution in the core of the OMZ on the neutral density surface γ_n = 27.1 for both periods (Fig. 7c) reveals an average reduction of the oxygen concentration by 16.6 μmol kg⁻¹ between 2° and 14°N from the earlier to the later period, with a maximum difference of 26 μmol kg⁻¹ at 3°N. Near the equator, however, the changes are weak. At the same time, the mean depth of the neutral density surface γ_n = 27.1 stays almost constant, with stronger undulations during the earlier compared to the later period (Fig. 7d). Accordingly, zonal jets on the neutral density surface γ_n = 27.1 show substantially stronger amplitudes in the region between 4° and 10°N (Fig. 7a) and are associated with stronger oxygen undulations, particularly visible when subtracting the background oxygen curvature (Fig. 8).

Between the equator and the Cape Verde frontal zone at 15° to 20°N, water masses of Southern and Northern Hemisphere origin form an extended transition zone that is reflected in an almost linear salinity increase from south to north on neutral density surfaces near the core of the OMZ (Fig. 7b). To avoid biases by local variations in the partitioning of waters of Southern and Northern Hemisphere origin, apparent oxygen utilization (AOU, defined as the air saturation value of dissolved oxygen at a given temperature and salinity minus the measured dissolved oxygen) is now used instead of dissolved oxygen itself. Here AOU is analyzed as a function of salinity for the two periods 1972–85 and 1999–2008 on the neutral density surface γ_n = 27.1 (Fig. 9). In the core of the OMZ, a significant mean oxygen decrease (or AOU increase) between the two periods of 14.7 μmol kg⁻¹ is found for salinities larger than 35. Over the same time interval, the salinity at the oxygen minimum (or AOU maximum) increased by about 0.1 indicating a relatively stronger contribution of Northern Hemisphere waters.

For the time periods 1972–85 and 1999–2008, the salinity and AOU data are mapped onto regular 1° latitude × 0.1 neutral density grids. To reduce further the noise inherent in the data, we actually averaged over 1.5° wide intervals in latitude (i.e., ± 0.75° of the nominal latitude), before we calculated the differences between the two periods (Fig. 10). Along the 23°W section, salinity changes from 1972–85 to 1999–2008 vary for different neutral density surfaces. While we observe an increase in salinity in the core layer of the OMZ north of 6°N, the salinity changes only weakly or even decreases above the OMZ core (Fig. 10a). A simultaneous increase of AOU is evident in the core layer of the OMZ across the whole tropical North Atlantic between the equator and 14°N, as well as above the OMZ core south of 10°N (Fig. 10b). In particular within the latitudinal range of the eastward flowing current band at around 9°N (see Fig. 2a), the salinity decreases in the neutral density layers γ_n = 26.6–26.8, while it increases in deeper layers (Fig. 10a). As the boundary between North and South Atlantic water masses is directed from southwest to northeast in the central water layer but from northwest to southeast in the AAIW layer (Kirchner et al. 2009), zonally inflowing oxygen-rich water masses may transport positive (negative) salinity anomalies above (below) γ_n = 26.9 into the OMZ, respectively.

Another possibility to explain the observed changes on neutral density surfaces would be a change of source water properties. In fact, calculations of salinity trends from hydrographic data for the period from the 1960s to 1990s revealed a general increase in salinity in the central water masses of subtropical origin and a decrease in salinity in intermediate water masses of subpolar origin, suggesting a change in the freshwater balance of the global oceans (Curry et al. 2003; Boyer et al. 2005). These changes, estimated for a period shifted backward in time by about 10 years, are of reversed sign than observed here. Moreover, changes in salinity of about 0.1 on the neutral density surface γ_n = 27.1 (Fig. 9) are an order of magnitude larger than those reported by Curry et al. (2003) and are not likely produced by changes of the properties of the source waters.

The ΔO₂–ΔS relations for the period 1972–85 reveal an elevated scatter compared to the period 1999–2008 that is well above the estimated accuracy of the historical data (cf. Figs. 4 and 11). The resulting latitudinal and longitudinal ranges exceed on average the ranges for the later period by a factor of 1.3, even though very similar values are obtained at 10°N (Fig. 12). However, the ratio between latitudinal and longitudinal range remains almost constant. The elevated scatter in the ΔO₂–ΔS relations during 1972–85 additionally suggests enhanced ventilation during the earlier period.

5. Conceptual model of the OMZ

A conceptual model of the ventilation of the OMZ is used here to corroborate the hypothesis that alternating zonal jets contribute to the ventilation of the OMZ and that reasonable variations of their strength may result in changes of the mean oxygen levels in the OMZ, comparable in magnitude to observed long-term changes. The model is representative only for a single eastward jet and its westward return flow, superimposed on the background meridional curvature of the oxygen distribution. It simulates the oxygen supply by mean zonal and meridional advection associated with the jet, by zonal and meridional eddy diffusivity, and by vertical (diapycnal) eddy diffusivity. The model is aimed at calculating the oxygen balance in an OMZ composed of multiple alternating zonal jets on a given density surface. We will show that reasonable jet strengths yield an advective oxygen supply that is comparable in magnitude to the diffusive supply. We chose parameters that correspond to the core of the North Atlantic OMZ; that is, the neutral density surface γ_n = 27.1. With this conceptual model, we want to address the question: What are the necessary changes in the strength of zonal jets or lateral eddy diffusivity that would yield similar changes in the oxygen concentration as observed between the two periods 1972–85 and 1999–2008.

a. Model

The conceptual model is an advection–diffusion model that will be integrated until equilibrium is reached:

View Expanded

Here C is the dissolved oxygen concentration; J the constant dissolved oxygen consumption coefficient; u and υ the zonal and meridional velocity components, respectively; k_x and k_y the coefficients of the zonal and meridional eddy diffusivities, respectively; k_z the coefficient of the vertical eddy diffusivity; C_bg the constant large-scale background oxygen distribution; and F_corr a correction factor to the background oxygen curvature depending on the simulated oxygen concentration. In the model [Eq. (2)], the oxygen tendency on the left-hand side is attributed to the following seven terms on the right-hand side: 1) oxygen consumption, 2) zonal advection, 3) meridional advection, 4) zonal eddy diffusivity, 5) meridional eddy diffusivity associated with east- and westward jets, 6) meridional eddy diffusivity associated with the large-scale oxygen distribution between the equatorial region and the northern subtropics, and 7) vertical eddy diffusivity associated with the vertical oxygen distribution. The dissolved oxygen consumption constant is taken from van Geen et al. (2006), who obtained an estimate by constraining a one-dimensional advection–diffusion model for the North Pacific OMZ with chlorofluorocarbon data. Their best-fit value was J = 0.041 yr⁻¹. The model will be solved numerically for a region covering simplified east- and westward zonal flows between 7°N (y = 0) and 11°N (y = l_y) and between the western boundary regime at 50°W (x = 0) and the eastern boundary at 15°W (x = l_x). We assume the following prescribed background flow field:

View Expanded

where u₀ is the amplitude of the zonal jets at the western boundary. The zonal flow decreases linearly toward the east. The meridional velocity field is calculated using the continuity equation. The meridional eddy diffusivity associated with the large-scale oxygen distribution is approximated using a second-order fit to the observed oxygen distribution along 23°W between 4° and 14°N (see Fig. 7c). We obtained very similar values for the distributions of the two time periods as well as for the WGHC distribution. In the model we use ∂²C_bg/∂y² = D_y = 1.2 × 10⁻¹⁰ μmol kg⁻¹ m⁻². The vertical eddy diffusivity associated with the vertical curvature of the oxygen distribution is similarly obtained using second-order fits. These fits are applied to the observed oxygen distribution between 300 and 500 m at each latitudinal grid point between 7° and 11°N. The obtained mean second order fits are ∂²C_bg/∂z² = D_z = 1.5 × 10⁻³ μmol kg⁻¹ m⁻² for the period 1972–85 and D_z = 1.7 × 10⁻³ μmol kg⁻¹ m⁻² for the period 1999–2008. In the following, we will use the latter value for the more recent period. We assume that both terms associated with the background oxygen distribution are constant for the whole model domain. Additionally, we apply the following boundary conditions:

View Expanded

The dissolved oxygen concentration at the western boundary C₀ is chosen to be 120 μmol kg⁻¹. The correction factor for the background meridional diffusivity is defined as follows:

View Expanded

Here C₁ is the observed mean oxygen concentration along 23°W between 7° and 11°N; that is, 50 μmol kg⁻¹, and C_23W is the corresponding simulated value. If C_23W becomes close to C₀ the factor tends to 0; that is, no meridional background eddy diffusivity; if C_23W is equal to C₁ the factor is 1. This correction factor acts to increase/decrease the oxygen supply because of the background eddy diffusivity for reduced/increased oxygen concentration in the OMZ. For the range of observed oxygen concentrations, this factor only slightly changes the obtained solution (resulting in a slight damping of oxygen changes). This correction factor is also applied to the vertical diffusion as the vertical oxygen curvature should similarly decrease/increase for periods with enhanced/reduced oxygen levels in the core of the OMZ.

The diffusive solution of Eq. (2) with u₀ = 0 cm s⁻¹ can be obtained for F_corr = 1:

View Expanded

This solution yields with the above assumptions for J, D_y, and D_z, with k_y = 200 m² s⁻¹, vanishing zonal eddy diffusivity at the eastern boundary, and k_z = 10⁻⁵ m² s⁻¹ (Ledwell et al. 1993), a minimum oxygen concentration at the eastern boundary of 32 μmol kg⁻¹.

6. Summary and discussion

A large number of recent research cruise data (Table 1) allows a detailed description of the mean circulation and hydrographic properties as well as their variability along the 23°W meridian from 6°S to the Cape Verde archipelago. The mean flow field in the upper 1000 m is characterized by east- and westward current bands (or alternating zonal jets). In general, eastward jets are associated with enhanced oxygen levels and westward jets with reduced oxygen levels. From the scatter in the ΔO₂–ΔS relations, we estimated an effective zonal diffusivity that is about 10 times larger than the effective meridional diffusivity. In that calculation, we did not account for advection and the resulting effective zonal diffusivity includes the effect of the alternating zonal jets. Without the effect of alternating zonal jets, the ratio between zonal and meridional diffusivity would be smaller.

The comparison of hydrographic properties for the periods 1972–85 and 1999–2008 revealed the following changes: (i) a decrease of the oxygen concentration in the core of the tropical North Atlantic OMZ of about 15 μmol kg⁻¹ and an increase of salinity at the oxygen minimum of about 0.1; (ii) a shoaling of oxygen tongues observed at about 5°–6° and 9°N; (iii) a decrease of the oxygen concentration for neutral density levels γ_n = 26.6–27.3 within the eastward jet at about 9°N, while the salinity decreased (increased) above (below) γ_n = 26.9, respectively; and (iv) a decrease of the scatter in the local ΔO₂–ΔS relations.

Stramma et al. (2008b) revealed a negative oxygen trend since 1960 in the 300- to 700-m layer for selected tropical ocean areas. However, the oxygen depletion in the OMZ of the tropical North Atlantic is about twice as large as in all other tropical regions. Here our new hypothesis is that local circulation changes contribute to long-term changes in the oxygen concentration of the OMZs. One possible explanation for the documented changes in oxygen and salinity on density surfaces between the two analyzed time intervals is that, besides the oxygen depletion common to the OMZs of the World Ocean, in the North Atlantic OMZ a transition occurred from a situation characterized by strong zonal jets to one characterized by weak zonal jets: During 1972–85 a strong eastward jet at about 9°N was responsible for the transport of low-salinity oxygen-rich AAIW from the western boundary toward the OMZ below its core. Above the OMZ core, high-saline oxygen-rich North Atlantic Central Water was transported eastward. The existence of a stronger eastward jet during 1972–85 compared to 1999–2008 would also explain the observed increased scatter in the ΔO₂–ΔS relations (cf. Figs. 4 and 11) as well as the increased amplitude of the oxygen undulations along the 23°W section at γ_n = 27.1 (Fig. 8), both are larger than the estimated uncertainty of the observations. Note that the scatter in the ΔO₂–ΔS relations represents a filamentation of hydrographic properties generated by mesoscale eddies and zonal jets acting on the background hydrographic gradients.

Geostrophic velocities show in general larger amplitudes of latitudinally alternating zonal jets for the period 1972–85 than for the period 1999–2008 (cf. Figs. 2a and 6a). The error of the geostrophic velocity that was estimated to be about 1 cm s⁻¹ is mainly due to the larger velocity error at the reference level (1000 m). Jet amplitudes are likely underestimated in the geostrophic calculation for the later period because of unavoidable horizontal averaging and nonzero velocities at the reference level. Further evidence for the existence of strong jets during 1972–85 and weak jets during 1999–2008 can be found in the different meridional inclinations of the neutral density surface γ_n = 27.1 (Fig. 7d) that is not affected by sparse data coverage at larger depths. During the later period the neutral density surface γ_n = 27.1 was flat, while during the earlier period strong undulations can be observed on this density surface.

To determine the necessary changes in the strength of zonal jets or lateral eddy diffusivity that would yield similar changes in the oxygen concentration as observed between the two periods 1972–85 and 1999–2008, we developed a conceptual model for the ventilation and oxygen consumption in the OMZ. The model suggests that the main balance in the OMZ is between oxygen consumption, meridional eddy diffusivity associated with the large-scale oxygen distribution, vertical eddy diffusivity, and advection associated with zonal jets. Zonal jets with amplitudes of 5 and 2.5 cm s⁻¹ at the western boundary lead to similar oxygen concentrations in the OMZ and similar amplitudes of oxygen undulations on isopycnal surfaces along 23°W between 7° and 11°N as observed for the two periods 1972–85 and 1999–2008, respectively. Assuming that parts of the oxygen depletion since 1960 are due to other processes not studied here would reduce the necessary change in the strength of alternating zonal jets.

Unfortunately, there exist no direct observations of long-term changes in the strength of latitudinally alternating zonal jets or eddy kinetic energy. Therefore, the outlined picture of long-term ventilation changes due to changes in the strength of the zonal jets is based only on the analysis of hydrographic properties and geostrophic velocity calculations. High-resolution model simulations, able to reproduce narrow equatorial zonal current branches and alternating zonal jets in the tropics, could be used to identify possible mechanisms for the suggested changes in the strength of the jets.

A possible explanation for such changes could be the atmospheric and oceanic variability in the Atlantic region on multidecadal time scales that is associated with the Atlantic multidecadal oscillation (AMO). Minimum values of the AMO have been observed during the 1970s with a steady increase afterward. The AMO, typically measured as the sea surface temperature (SST) anomaly of the whole North Atlantic (after subtracting the long-term trend), is thought to be associated with variations in the Atlantic meridional overturning circulation (Knight et al. 2005; Sutton and Hodson 2007). However, the SST variability associated with the AMO represents a differential warming of the Northern Hemisphere compared to the Southern Hemisphere. During high phases of the AMO, the SST gradient from south to north is enhanced (Knight et al. 2005; Sutton and Hodson 2007; Delworth and Mann 2000) and correspondingly the northeasterly (southeasterly) trade winds are found to be reduced (increased; e.g., Foltz and McPhaden 2008). Such wind changes act on similar time scales as the observed changes in the tropical North Atlantic OMZ. As the forcing of latitudinally alternating zonal jets is given by the zonal background flow and associated vertical shear (see, e.g., Berloff et al. 2009), the variability of the Sverdrup circulation in the tropical North Atlantic driven by the cross-equatorial meridional SST gradient (Joyce et al. 2004) represents a likely candidate responsible for changes in the jets’ strength.

The proposed reduction of the OMZ ventilation because of a weakening of latitudinally alternating zonal jets adds a further mechanism to the list of biogeochemical and physical mechanisms that could explain the ongoing oxygen depletion in the tropical North Atlantic. The detection of anthropogenic climate change in the available observational record of oceanic oxygen distribution, however, strongly depends on a quantitative understanding of intrinsic and anthropogenic oxygen variability on decadal to multidecadal time scales.