Fluctuations of the Atlantic North Equatorial Undercurrent and Associated Changes in Oxygen Transports

Although the core velocity of the Atlantic North Equatorial Undercurrent (NEUC) is low (0.1−0.3 m s−1), it has been suggested to act as an important oxygen supply route towards the oxygen minimum zone in the eastern tropical North Atlantic. For the first time, the intraseasonal to interannual NEUC variability and its impact on oxygen are investigated based on shipboard and moored velocity observations around 5°N, 23°W. In contrast to previous studies that were mainly based on models or hydrographic data, we find hardly any seasonal cycle of NEUC transports in the central Atlantic. The NEUC transport variability is instead dominated by sporadic intraseasonal events. Only some of these events are associated with high oxygen levels suggesting an occasional eastward oxygen supply by NEUC transport events. Nevertheless, they likely contribute to the local oxygen maximum in the mean shipboard section along 23°W at the NEUC core position.


Introduction
The circulation of the upper tropical Atlantic Ocean is characterized by a complex current system which takes part in the wind-driven equatorial gyre circulation, the shallow subtropical and tropical overturning cells and the basin-wide Atlantic meridional overturning circulation (e.g., Hazeleger & Drijfhout, 2006;Schott et al., 2004). Zonal currents play a key role in the basin wide distribution of water mass properties and affect the transport of heat, salt, and biogeochemical components such as oxygen (e.g., Brandt et al., 2015;Hazeleger & Drijfhout, 2006;Schott et al., 2004). The North Equatorial Undercurrent (NEUC) is an eastward flowing subsurface (here defined in the depth range 65-270 m) current centered at 5°N. Its upper limit is commonly defined as the 24.5 kg m −3 neutral density layer, which separates the tropical surface water from the subtropical underwater (e.g., Bourlès et al., 1999;Goes et al., 2013;Schott et al., 1995). Although the NEUC core velocity (0.1 to 0.3 m s −1 ) is one of the lowest among the wind-driven off-equatorial currents in the tropical Atlantic, the NEUC has been suggested to act as an important oxygen supply route towards the Oxygen Minimum Zone (OMZ) in the Eastern Tropical North Atlantic (Brandt et al., 2010;Stramma et al., 2008). Hahn et al. (2017) observed a shoaling of the Eastern Tropical North Atlantic OMZ since 2006 which is associated with a thinning of the available habitat of tropical pelagic fishes (Stramma et al., 2012). The intraseasonal to interannual variability of the NEUC causes oxygen variability along the OMZ boundary. Under global warming with increasing mean oxygen gradients, oxygen variance might enhance the impact on tropical pelagic ecosystems and biogeochemical cycles.
Model studies generally agree that the NEUC is mainly in geostrophic balance, but its driving mechanism is under discussion. Potential mechanisms that were described for the NEUC or similar subsurface currents in other tropical basins are the conservation of angular momentum of the tropical overturning cells (Marin et al., 2000), the Eliassen-Palm flux associated with the propagation of Tropical Instability Waves (TIWs Jochum & Malanotte-Rizzoli, 2004) or the pull of upwelling within domes in the eastern basin or at the eastern boundary (Furue et al., 2007(Furue et al., , 2009McCreary et al., 2002).
Until now, the seasonal to long-term variability of the NEUC has been investigated based on model output (Burmeister et al., 2019;Hüttl-Kabus & Böning, 2008) or geostrophic velocities derived from a combination of hydrography and satellite data (Goes et al., 2013). In these studies, a seasonal cycle of the NEUC with amplitudes of 1 to 3.5 Sv was identified. In general, NEUC transport estimates derived from meridional ship sections are obscured by mesoscale activity (Weisberg & Weingartner, 1988) and interannual variability (Hüttl-Kabus & Böning, 2008). So far, studies based on shipboard velocity observations have not been able to detect a seasonal cycle of the NEUC (Bourlès et al., 1999(Bourlès et al., , 2002Burmeister et al., 2019;Schott et al., 1995Schott et al., , 2003Urbano et al., 2008).
For the first time, we will investigate the NEUC variability using direct velocity observations. In this study we reconstruct the eastward transport associated with the NEUC at 5°N, 23°W using moored velocity observations from June 2006 to February 2008 and November 2009 to January 2018 in combination with 24 meridional ship sections taken between 21°W and 26°W. This study aims to investigate the intraseasonal to interannual variability of the NEUC and its impact on oxygen levels.

Data
Moored Acoustic Doppler Current Profiler (ADCP) velocity, hydrography, and oxygen data are available at 5°N, 23°W (June 2006to February 2008, November 2009 to January 2018) as well as at 4. 6°N, 22.4°W and4.5°N, 23.4°W (November 2012 to April 2014). All instruments were set to a sampling rate of 2 h or higher. The upper limit of the ADCP observations varies between 65 and 85 m, the lower limit is at least 755 m. The moored velocity data were linearly interpolated onto a regular time-depth grid (12 h × 10 m), and a 40-day low-pass Butterworth filter was applied to remove tides from the time series. As the NEUC is located approximately between 3.5°N and 6°N and in a depth of about 65 to 270 m at 23°W (Burmeister et al., 2019), the moored velocity observations do not always cover the entire extent of the NEUC which can result in an underestimation of its mean transport. Observations of dissolved oxygen, temperature, conductivity, and pressure were used from respective sensors that were installed at the mooring at 5°N, 23°W in 100, 200, and 300 m depth. The moored hydrography and oxygen data were interpolated onto a 12-h time grid. This data set is an extension of the one used in Hahn et al. (2014, see supporting information Text S1 for more details).
In addition to the mooring time series, we use data from 24 meridional ship sections taken between 21°W and 26°W in the time period 2002 to 2018 (supporting information, Table S1). Only shipboard ADCP , hydrography and oxygen sections that cover at least the upper 350 m between 0°a nd 10°N are used. The ship sections are an extension of the data set used in Burmeister (2019, see Text S1 for more details).

Observed Velocity Variability at 5°N, 23°W
Moored ADCP measurements at 5°N, 23°W show a weak mean eastward velocity with maximum values of 9 cm s −1 at the upper limit of the ADCP range (85 m), while the meridional velocity varies around zero (Figures 1a and 1b). In the upper limit (300 m), zonal and meridional velocities exhibit anomalies of comparable magnitude. The periodogram of the horizontal velocity components indicates variability over a range of frequencies, in particular in the intraseasonal band ( Figure 1c). While the zonal velocity exhibits variability mainly for periods greater than 70 days, the meridional velocity is dominated by variability with periods between 35 and 75 days which is associated with TIWs (Jochum & Malanotte-Rizzoli, 2003). The zonal velocity, while eastward in the mean, occasionally changes to westward. Its variability is characterized by strong eastward anomalies with a duration of about 1 to 5 months occurring without a clear seasonal preference. Unexpectedly, the seasonal cycle of the zonal velocity is much weaker than found in previous studies (Burmeister et al., 2019;Goes et al., 2013;Hüttl-Kabus & Böning, 2008). The annual harmonic fit with a maximum amplitude of 3 cm s −1 only explains between 2% (85 m) and 9% (300 m) of the zonal velocity variability. For the semi-annual harmonic fit, we derived a maximum amplitude of 5 cm s −1 and an explained variance between 11% (75 m) and 1% (300 m).

NEUC Transport Estimates at 23°W
In this study, the NEUC transport was estimated by four different methods using only eastward velocities. (i) We calculated the NEUC transport between the 24.5 and 26.8 kg m −3 neutral density surfaces from 24 ship sections using a path following algorithm developed by Hsin and Qiu (2012; Figure 2, green diamonds). This method follows the current core, thereby avoiding artifacts in the transport calculation if the current is meridionally migrating (see supporting information Text S2 for more details). Using this method, we estimate a mean NEUC transport of 2.7 ± 0.4 Sv. Uncertainties are given in terms of the standard error. In the following, we consider this estimate as a reference NEUC transport keeping in mind that the NEUC transport
(ii) The second approach is also based on ship sections only, but with a fixed smaller integration box to be consistent with NEUC transports calculated from a combination of ship mounted and moored observations. We integrate the meridional sections of zonal velocity between 4.25°N and 5.25°N, 65 and 270 m (Figure 2, black circles). The mean fixed box integrated transport derived from ship sections is 1.4 ± 0.2 Sv. The fixed box integrated method underestimates the NEUC strength by 1.3 Sv. However, it represents the variability of the reference NEUC transport well (R=0.91, supporting information, Figure S2a).
(iii) We reconstruct the NEUC transport combining moored velocity observations at 5°N, 23°W as well as 4.6°N, 22.4°W and 4.5°N, 23.4°W and ship sections following the optimal width method described in Brandt et al. (2014). This method aims to find an optimal latitude range W i for each mooring position i to reconstruct the latitudinally integrated zonal velocity U(z,t) by For this purpose, we latitudinally integrate the eastward velocities of each ship section from 4.25°N to 5.25°N. The latitudinally integrated velocity is then regressed onto the eastward velocities of the ship sections at the three mooring position between 65 and 270 m depth to obtain W i . We estimated that the moorings at 5°N, 4.6°N, and 4.5°N correspond to latitude ranges, W i , of 0.46°, 0.18°, and 0.37°, respectively. The NEUC transport is then reconstructed by integrating Equation 1 over the same depth range as for method (ii; Figure 2, blue line). Note that data from all three mooring positions are only available for the time period November 2012 to April 2014. The root mean square error of the reconstructed transport using method (iii) and the box integrated transport (method ii) using the shipboard data is 0.16 Sv. The mean reconstructed transport is 0.9 ± 0.3 Sv. For a validation of the method see supporting information, Text S2.
(iv) This method is similar to method (iii) but uses moored velocity observations only at 5°N, 23°W to obtain a longer transport time series (Figure 2 grey line). Here the 5°N mooring corresponds to a latitude range of 0.88°. Although the root mean square error between the reconstructed transport and the box integrated transport increases (0.51 Sv) when using only one mooring, the reconstructed transport based on one mooring agrees well with the one reconstructed using three moorings (R=0.89). The mean NEUC transport is the same as estimated for method (iii), that is, 0.9 ± 0.2 Sv.
In the following, we will analyze the NEUC variability based on the reconstructed time series using only the mooring at 5°N, 23°W. Although the reconstructed transport time series from moored velocity observations tends to underestimate the mean current strength of the NEUC, we still consider the variability to be captured to a large extent. This is supported by transport estimates based on all three moorings combined with zonal velocity sections between 3.5°N and 6.0°N which agree reasonably well with the transport time series reconstructed by method (iii) accounting for velocities between 4.25°N and 5.25°N only (supporting information, Figures S2-S5, Text S2).

NEUC Variability
Similar to the zonal velocity, the NEUC transport is dominated by strong eastward anomalies that persist between 1 and 5 months. The transport time series do not exhibit a seasonal cycle. The explained variance of the annual and semi-annual harmonic is only 2% and 3%, respectively. We therefore focus the analysis on the intraseasonal eastward flow events. We define a strong eastward flow event rather arbitrary as the time period in which the NEUC transport exceeds 2.5 times the standard deviation of the complete monthly mean time series. We find eight such events which take about 1 to 5 months to develop, peak, and fade away (duration from local minimum to minimum in the transport time series). They have a maximum monthly mean transport between 2.3 and 3.8 Sv and peak without a clear seasonal preference (3 in January and 1 each in April, May, June, August, and December). During these events the 12-hourly transports reach maximum values from 3.8 to 7.2 Sv.
The NEUC at 5°N, 23°W appears to be rather weak, and it is likely that there are different generation mechanisms for the strong eastward flow events. The short period of intraseasonal events implies that the upwelling within the Guinea Figure S6 and Text S3). Easterly wind stress anomalies in the eastern equatorial basin are leading the NEUC transports by 1 to 2 months. We hypothesize that these zonal wind stress anomalies may force equatorial Kelvin waves reflecting at the eastern boundary into westward propagating Rossby waves and poleward propagating coastal trapped waves. The northward propagating waves may shed Rossby waves when arriving at the exit of the Gulf of Guinea where the coastline turns northward. Such remotely forced Rossby waves may reach 5°N, 23°W, causing the observed eastward flow events. Rossby waves at 5°N, 23°W can also be forced locally by wind stress curl anomalies of small meridional scale (Burmeister et al., 2016;Foltz & McPhaden, 2010) or remotely generated by the radiation of Rossby waves from coastal trapped waves generated by local wind anomalies in the Gulf of Guinea (Chu et al., 2007). However, in a composite analysis of sea level anomalies, we could not clearly identify a reflection of an equatorial Kelvin wave at the eastern boundary or a Rossby wave propagating from the eastern basin to 5°N, 23°W prior to any eastward flow event.

NEUC and Oxygen
The NEUC is thought to transport oxygen-rich water from the western boundary towards the generally poorly ventilated eastern tropical North Atlantic (Brandt et al., 2010;Stramma et al., 2008). The mean ship section indicates a local oxygen maximum which is associated with the NEUC and single ship sections often show maxima in the area of the NEUC (Figure 3). Here we will investigate if the strong eastward NEUC events are associated with an increased eastward oxygen transport using moored observations.
We calculated oxygen anomalies on isopycnals. First, a temporal mean oxygen profile as a function of density was calculated from the mooring time series. Next, for each time step, the oxygen anomaly was calculated with respect to the mean oxygen value of the respective density. Finally, we applied a second-order 30-day low-pass Butterworth filter to both, the transport and oxygen time series (Figure 3).
In 100 and 200 m depth, we found a significant positive correlation between the NEUC transport and oxygen anomalies (R max,100 m =0.35, NEUC leads by 16 days; R max,200 m =0.35, NEUC leads by 2.5 days), while in 300 m depth, which is just below the NEUC, the correlation is not significant (R 300 m =0.05, zero lag). If data are available, positive oxygen anomalies can be found during most of the strong eastward flow events except for (g-i) Thirty-day low-pass filtered NEUC tranport (red lines) and oxygen anomalies (blue lines) at 5°N, 23°W at a depth of (g) 100, (h) 200, and (i) 300 m. Grey bars mark strong NEUC events. R is the correlation coefficient of zonal velocity and oxygen anomalies at zero lag (black/grey color indicates that R is significant/not significant on a 95% confidence level).
in 100 m depth during the 2012 and 2014 event.
Year 2017 seems to be an exceptional year in the moored time series with the two strongest eastward flow events occurring consecutively within 6 months. Additionally, these two events are associated with some of the highest oxygen anomalies observed. The correlation between oxygen and NEUC transport anomalies during these two events is much higher compared to the rest of the time series. The correlation between the time series for the period from September 2016 to February 2018 is 0.63 (NEUC leads by 13.5 days) and 0.56 (NEUC leads by 1.5 days) in 100 and 200 m depth, respectively.
Positive oxygen anomalies also occur independently from strong eastward flow events and may be associated with changes in the meridional velocity. On intraseasonal timescales, zonal and meridional velocity anomalies are of similar strength and the dominant frequency band of the meridional velocity (30 to 70 days) overlaps with the frequency of the strong eastward flow events (1 to 5 months). It is thus not possible to clearly differentiate between the effect of meridional and zonal velocity on oxygen values at the mooring position. Yet we find no significant correlation between the 30-day low-pass filtered meridional velocity and oxygen anomalies (−0.12 ≤ R ≤ 0.06 at zero lag; supporting information, Figure S7).

Summary and Discussion
The NEUC is an eastward flowing current centered around 5°N in the tropical Atlantic. Although its core velocity is weak (below 0.3 m s −1 ) it is thought to act as an important oxygen supply route towards the eastern tropical North Atlantic oxygen minimum zone (Brandt et al., 2010;Stramma et al., 2008). For the first time, we reconstructed a time series of the NEUC transport based purely on direct velocity observations. By combining moored zonal velocities at 5°N, 23°W and meridional ship sections along ∼23°W we obtained a NEUC transport time series from June 2006 to February 2008 and November 2009 to January 2018. In contrast to previous studies (Burmeister et al., 2019;Goes et al., 2013;Hüttl-Kabus & Böning, 2008), neither the moored zonal velocity at 5°N (Figure 1) nor the reconstructed eastward NEUC transport ( Figure 2) exhibits a pronounced seasonal cycle. We find that both time series are dominated by strong intraseasonal eastward flow events which can peak throughout the year (Figures 1 and 2). Although in the mean meridional ship sections the zonal velocity maximum associated with the NEUC coincides with a local oxygen maximum, we find that the eastward flow events are only occasionally associated with high oxygen levels ( Figure 3).
We reconstructed the NEUC transport at 23°W based on velocity observations from three moorings and 24 meridional ship sections (Figure 2). To obtain a longer time series, we then reconstructed the NEUC variability based on data from only one mooring position. The comparison with the three-mooring solution indicates that the variability is dominated by a meridionally homogeneous structure covering the complete integration box ( Figure S3). However, the method has some limitations. The used integration box does not cover the entire NEUC region ( Figure S1). Consequently, our method is underestimating the true NEUC transport. A comparison of the box integrated transport calculated from ship sections with a reference transport calculated from ship sections using a path following algorithm indicates that the box integrated transport represents the variability of the NEUC reasonably well (R=0.9, Figures S2 and S5). An additional uncertainty of the reconstructed transport is due to the upper range of the moored velocity observations that varies between 65 and 85 m, that is, the moored observations do not always cover the entire NEUC depth range. In summary, the used method is underestimating the true NEUC strength, but it is still capable of representing its variability, as this is dominated by a homogeneous structure in latitude and depth.
Within the scope of this paper, we could not identify a clear forcing mechanism for the strong eastward flow events. We found that easterly wind stress anomalies in the eastern equatorial basin are leading the NEUC transport by 1 to 2 months. Zonal wind stress anomalies and associated wind stress curl anomalies could directly (Burmeister et al., 2016;Foltz & McPhaden, 2010) or remotely (Chu et al., 2007) force Rossby waves which reach 5°N, 23°W to cause the observed eastward flow events. However, we could not clearly identify a Rossby wave propagating from the eastern basin to 5°N, 23°W prior to any event. Other potential mechanisms might include nonlinear processes resulting, for example, in the development of multiple cores that were discussed by Furue et al. (2007). These cores that were superimposed on the otherwise linear arrested front dynamics (Dewar, 1991) were identified in simulations with enhanced horizontal resolution.
Finally, we investigated the relationship between the NEUC transport events and oxygen. In 2017, eastward transport and oxygen anomalies agree very well. For other events, the correlation is weaker or non-existent 10.1029/2020GL088350

Geophysical Research Letters
( Figure 3). One possible explanation is that the local oxygen maximum associated with the NEUC is not located directly at 5°N as visible in the meridional ship sections of March 2017 (Figures 3b and 3e). In that case, the higher oxygen concentrations are simply not captured by the mooring. Another explanation may be that some strong eastward flow events are not connected to the well-ventilated western boundary but are instead supplied out of the westward flow of low-oxgen waters north and south of the NEUC. For example, in the meridional section of February 2018, we find strong eastward velocities between 4°N and 5°N at 50 and 100 m depth that are associated with low oxygen concentrations, which might be due to a recirculation of low-oxygen waters typically present south of the NEUC (Figures 3c and 3f Burmeister et al., 2019). As the zonal velocity events with a duration between 1 and 5 months overlap with the dominant frequency band of meridional velocity (30 to 70 days), it is not possible to clearly differentiate between the effect of meridional and zonal velocity on oxygen. However, we did not find any correlation between the meridional velocity and oxygen anomalies on intraseasonal time scales at 5°N, 23°W. Furthermore, Hahn et al. (2014) found that the mean meridional eddy flux at the mooring position is not significantly different from zero. Consistent with previous studies (Burmeister et al., 2019;Hahn et al., 2014;Weisberg & Weingartner, 1988), we suggest that a continuous flow of oxygen-rich water from the western boundary toward the eastern Atlantic basin by the NEUC is regularly altered by TIWs or other mesoscale recirculations. This can explain why not all NEUC transport events are associated with higher oxygen levels. Nevertheless, the significant positive correlation between the 30-day low-pass filtered NEUC transport and oxygen anomalies at 5°N, 23°N (Figure 3) indicates that the NEUC transport events likely result in an elevated mean eastward oxygen transport and contribute to the oxygen maximum observed in ship sections at the NEUC core position.