Figure 1. The measurement lines (A) and distinct sampling locations (B) throughout the seasonal campaigns, encompassing winter (black), autumn (yellow), and summer (red). The underway measurements are only available for winter and summer campaigns.
2.2 Discrete Water Column Measurements
Water samples were collected using 10 L Niskin or FreeFlow bottles (HydroBIOS) attached to a CTD/Rosette system (SBE 911plus, USA) during all expeditions. The carbonate system is typically assessed using pH, total alkalinity (AT), seawater partial pressure of CO2 (p CO2), and total dissolved inorganic carbon (CT), of which at least two have to be measured. We sampled for pH and CT, filling two 250 mL glass ground-stoppered flasks from each depth. During winter and summer campaigns, pH was measured immediately after collection. The HydroFIA pH system (4H Jena Engineering) was used, three measurements were executed from each sample bottle, pH values were calculated on the total scale following Müller and Rehder (2018), and the average of the three valid measurements was reported. The accuracy of results was validated through measurements of in-house buffer solutions after Müller et al. (2018) to identify any drift. During the autumn cruise, separate samples for pH and CT were collected and poisoned with 200 µL of saturated HgCl2 solution immediately after sampling for later analysis ashore. The precision of pH measurements during all surveys was 0.002 between the three replicates.
For CT analysis, around 5 ml of each discrete sample was acidified with 10% phosphoric acid to release carbon content. The acidified sample flowed through an automated inorganic carbon analyzer (AIRICA, Marianda) equipped with an infrared gas detector (LICOR 7000) with a flow rate of 180 ml min-1, using a carrier gas (99.999% Nitrogen) and finally passing through a NAFION dryer and Peltier cooler. The LICOR system tracked the released CO2 content over time, corresponding to CT in the water sample. The final calculated CT concentration is determined as the average of three consecutive measurements from the system. Certified reference materials (Scripps Institution of Oceanography) were used to eliminate blank impurities and allow for drift correction, ensuring precision in the range of 1.50 - 2.00 µmol kg-1 with a maximum deviation of +/- 3.00 µmol kg-1 in triplicate measurements. Measurements for the autumn cruise were performed ashore upon return.
AT and p CO2 were computed following the methodology outlined by Dickson et al. (2007) during both winter and summer campaigns. The carbon system calculations were executed using the CO2SYS EXCEL program. Dissociation constants K1 and K2 were determined after Millero (2010), and the sulfate contribution was derived from Dickson et al. (2007). We applied the total boron parameterization proposed by Uppström (1974). In the autumn campaign, AT from poisoned samples were measured by potentiometric titration (using a glass electrode type LL, Electrode Plus, 6.0262.100; Metrohm AG, Filderstadt, Germany) in the open-cell configuration, following Dickson et al. (2007). The system underwent calibration using the same CRMs employed for CT, resulting in consistent precision. Prior to commencing the measurement sequence, a reference water was analyzed to determine its precise target value (in µmol kg-1), allowing for a deviation of up to 5 µmol kg-1. The measurement result was adjusted to align precisely with the target value during calculation. Towards the end of the series, a final standard water was examined to assess changes in the measuring system’s sensitivity. The calculation of results between initial and final standard measurements incorporated these adjustments and accounted for the system’s drift behavior.
For CH4 and N2O, bubble-free sampling was conducted by using a silicon tube to transfer water from the CTD Niskin bottles into 200 ml glass vials. These vials were then sealed with rubber butyl stoppers and poisoned with 200 µL of saturated HgCl2 solution immediately after. Determination of CH4 and N2O was carried out by means of a dynamic headspace method as outlined in Sabbaghzadeh et al. (2021). This method ensures an average precision better than 0.5% for CH4 and 0.6% for N2O by purging 20 ml of the water sample, along with detection limits below 0.3 nmol for both gases. Measurements were performed by means of a custom-designed purge and trap system integrated with a gas chromatograph (Agilent 7890B) as described previously (Sabbaghzadeh et al. 2021). To ensure accuracy and detect drift, different volumes of standard gases (9.26 ppm ± 0.2% CH4 and 1992 ppb ± 0.2% N2O in synthetic air) were measured before and after each set of analysis on a daily basis. The measurements were consistently maintained within the calibration limits, thereby assuring the reliability of the results.
Further hydrographic parameters, including seawater temperature, salinity, and oxygen concentration, were also assessed during the campaigns with a Seabird SBE911 CTD. The instrumentation featured a double sensor system with Digiquartz pressure sensor, SBE 3 temperature sensor, SBE 4 conductivity sensor, and SBE 43 dissolved oxygen sensor. To mitigate potential artifacts from ship heave, these sensors were arranged within a tube system, referred to as the TC-duct, ensuring a continuous flow of seawater. Data were recorded at a rate of 24 scans s-1 and subsequently processed using Seasoft V2. Temperature readings were reported on the ITS-90 temperature scale, while salinity values were obtained through the Practical Salinity Scale. Prior to the expeditions, sensor calibration was conducted at the Leibniz Institute for Baltic Sea Research calibration laboratory. During the expeditions continuous monitoring of sensor stability was carried out using an SBE 38 thermometer for temperature, and an AUTOSAL 8400B instrument for the salinity comparison measurements respectively. Oxygen concentration was calculated utilizing the Sea-Bird Scientific implementation of the oxygen saturation concentration. Regular samples were taken in unstratified water layers, and oxygen concentration was determined using the Winkler titration method. High resolution hydrographic observations along cross shelf transects were also performed by a towed CTD system, the ScanFish. This system undulates vertically between the surface and about 120 m depth, while it is towed behind the ship. The ScanFish was equipped with a Seabird SBE911 CTD, consisting of SBE 3 temperature sensor, SBE 4 conductivity sensor, and SBE 43 dissolved oxygen sensor. Data quality was checked with the validated data of the standard CTD described above. The fraction of SACW below the mixed surface layer was estimated from the CTD and ScanFish data using the water mass definitions and procedure described in details elsewhere (Mohrholz et al., 2008).
2.3 Continuous Surface Water and Atmospheric Measurements
The p CO2 and dissolved CH4 and N2O in surface water were also continuously measured using the Mobile Equilibrator Sensor System (MESS; Sabbaghzadeh et al. (2021)) during winter and summer expeditions only. In brief, the system consists of a custom-made bubble-type/showerhead equilibrator and a control unit, both connected to two Los Gatos Research off-axis laser absorption spectroscopy (OA-ICOS) analyzers (Model # 908-0011-0001 for CO2/CH4/H2O and Model # 908-0014-0000 for N2O/CO/H2O). Seawater was supplied by a deep-well pump (CAPRARI Desert E4XP30-4 with CAPRARI XPBM1 control unit, ~ 100 L min-1, Italy) located in the moon-pool at a depth of about 5.6 m on R/V METEOR during winter and about 6.2 m on R/V Maria S. Merian during summer campaign. The water flow rate was approximately 5 – 6 L min-1. Concurrently, the air flow rate was adjusted to roughly 4 – 5 L min-1. Regular control measurements of all analyzers was conducted using three standard gases (Table S1) measured regularly when the ship was on station. This process ensured data accuracy through recalibration and drift correction. Additionally, an occasional measurement of a ”zero” gas (Nitrogen 5.00, Linde GmbH, Germany) was performed during the surveys to identify any system deficiencies, such as potential leakage. The average sensor drift observed was around 0.03 ppb d-1 for CH4 and approximately 0.26 ppb d-1 for N2O during winter, while it was 0.10 ppb d-1 for CH4 and 0.24 ppb d-1 for N2O during summer.
We also conducted measurements of atmospheric CO2, CH4 and N2O in ambient air (Table S2) at various locations, employing an inlet situated at a height of 35 meters. To ensure data quality, a comparison was made with data from NOAA’s nearest atmospheric sampling station (Station NMB [Gobabeb, Namibia], located at 23.58°S, 15.03°E). During winter, the calculated mean dry atmospheric values for CO2, CH4and N2O were ± σ; 410.50 ± 3.27 ppm, ± σ; 1844.81 ± 31.89 ppb and ± σ; 331.51 ± 0.99 ppb respectively. These values were compared to NOAA’s mean values of ± σ; 411.02 ± 0.01 ppm for CO2, ± σ; 1842.62 ± 9.75 ppb for CH4 and ± σ; 331.84 ± 0.37 ppb for N2O. During the summer survey, the ship-borne daily mean atmospheric gas levels (CO2; ± σ; 414.16 ± 1.51 ppm, CH4; ± σ; 1830.85 ± 13.81 ppb and N2O; ± σ; 337.52 ± 1.31 ppb respectively) closely aligned with the atmospheric mean values provided by NOAA (CO2; ± σ; 414.10 ± 0.00 ppm, CH4; ± σ; 1850.74 ± 7.05 ppb and N2O; ± σ; 335.40 ± 0.43 ppb).
3 Data Analysis
The concentrations of dissolved CH4 and N2O were computed after Sabbaghzadeh et al. (2021). To further explore the N2O cycle in the BUS, the excess N2O (ΔN2O) was calculated, representing the difference between the observed N2O concentration (N2Oin situ ) and the equilibrium concentration of N2O with the atmosphere (N2Oeq .) (e.g. Nevison et al., 1995 and Weiss and Price, 1980).
We determined gas flux densities (F ) for CO2 in mmol m−2 d−1 and for CH4 and N2O in μmol m−2 d−1 as follows:
F = kw * ΔC (obs . –eq .) (1)
where ΔC represents the difference between gas concentrations in seawater and the equilibrated atmosphere (C(obs. )- C(eq. )), and k w (in cm h-1) is the gas transfer velocity computed according to Wanninkhof (2014). For sea-air CO2 fluxes, the bulk equation is expressed in terms of the partial pressures of CO2 in equilibrium with surface water and in the overlying atmosphere, respectively, denoted as Δp CO2 or p CO2wp CO2a Wanninkhof (2014).
To standardize the comparison of GHG emissions in a common metric, we also computed CH4 and N2O emissions in terms of CO2 equivalent (CO2-e) emissions, following the methodology outlined by Eyre et al. (2023);
Total CO2-e source = CH4CO2-e + N2OCO2-e (2)
CH4CO2-e (mg CO2m-2 y-1) = CH4flux (µmol m-2 d-1) * 106 * 365 * 16 * 79.7 (GWP20)
or . 27.0 (GWP100) (3)
N2OCO2-e (mg CO2m-2 y-1) = N2Oflux (µmol m-2 d-1) * 106 * 365 * 44 * 273 (GWP20,100) (4).
We firstly determined the average fluxes for each sub-region during winter and summer and subsequently, CO2-e values were determined using the 20-year and 100-year global warming potentials (GWP) for each gas (Eyre et al. 2023). The GWP20 values for CH4 and N2O stand at 79.7 and 273, while the GWP100 values for CH4 and N2O are 27 and 273 respectively, according to IPCC (2021).
The significance of temperature and biological influences onp CO2 variations is gauged through a ratio that considers both factors (Takahashi et al., 2002). The biological impact is assessed by the seasonal amplitude of temperature-normalizedp CO2, while the temperature effect is measured by the seasonal amplitude of the annual mean p CO2, adjusted for seasonal temperature changes. The fluctuations inp CO2 resulting from biological and thermal influences are calculated using the following equations, respectively:
p CO2 at Tmean =p CO2obs. * exp [0.0423 (Tmean − Tobs. )] (5)
p CO2 at Tobs. =p CO2mean * exp [0.0423 (Tobs. − Tmean )] (6)
where Tmean and Tobs.represent the annual mean and observed temperatures in °C, respectively. These calculations are applied separately to coastal and offshore sampling sites. The biological impact on surfacep CO2 variations, denoted as Δp CO2Bio. , is also quantified by the seasonal amplitude of p CO2, adjusted using the mean annual temperature (p CO2 at Tmean ),
Δp CO2Bio. = (p CO2at Tmean ) max - (p CO2 at Tmean )min (7)
where Tmean (max ) and Tmean (min ) indicate the average seasonal maximum and minimum temperature. The influence of temperature fluctuations on the mean annual p CO2 value, Δp CO2temp , is
Δp CO2Temp. = (p CO2at Tobs. ) max - (p CO2 at Tobs. )min (8).
The ratio of Δp CO2Temp. p CO2Bio. > 1 signifies the prevalence of temperature’s impact on the annual meanp CO2, while a ratio < 1 signifies the prominence of biological effects (Takahashi et al., 2002).
To further investigate the predominant factors influencing seasonal-spatial outgassing in the nBUS, Principal Component Analyses (PCA) were conducted, accompanied by the corresponding biplots using GraphPad Prism 10. The biplots facilitates a comprehensive understanding of the interplay among different variables and their collective impact on the observed phenomena. They effectively presenting both sample scores (depicted by different colors) and variable loadings (represented by arrows). The methodology includes interpreting Factor 1 in environmental terms, where it is perceived as a synergy of biological factors (associated with oxygen and gas concentrations) and physical factors (highlighted by wind speed). Factor 2 also elucidates the total variance, primarily associated with salinity and temperature. This approach aids in grasping the underlying mechanisms influencing gas emissions. These analyses were carried out individually for each gas, incorporating ancillary data, and performed separately for the winter and summer seasons to account for seasonal variations. To address the diverse measurement scales of the variables, we standardized them using a correlation matrix, ensuring a uniform scale for comparison. For determining the number of principal components, we opted for the parallel analysis method, chosen for its robustness in identifying significant components.
Additional hydrographic data, including wind speed, air temperature, and ambient air pressure, were obtained from the DSHIP data system for all campaigns. Instantaneous wind speeds from the ship’s meteorological data were normalized to a height of 10 meters above sea level (U10n) following Large and Pond (1982). Daily average wind speeds were further computed and reported to enable a seasonal comparison within our specified regions. Underway hydrographic data, encompassing sea surface temperature (SST) and sea surface salinity, were compiled from information acquired through the ship’s thermosalinograph (Seabird Micro SBE35).
4 Results
4.1 Seasonal Hydrographic Features in Upwelling Events
In general, the intensity of remote equatorial forcing and local upwelling controls the distinct distributions of dominant central water masses along the Namibian shelf (Mohrholz et al., 2008, Siegfried et al., 2019). Strong upwelling leads to robust cross-shelf movement that restricts the southward flow of SACW along the coast. This process can result in the ESACW being confined to higher latitudes, exerting a notable influence and governing the composition of subsurface water masses in regions like Walvis Bay and Lüderitz, especially during the winter. Previous interannual mooring observations in the region have highlighted distinct patterns: ESACW pulses dominate during winter (June to October), indicating intensified upwelling due to favorable winds. In contrast, summer (January-February) experiences a prevalence of nutrient-enriched SACW intrusions, indicating reduced upwelling (Mohrholz et al., 2008; Junker et al., 2017).
The most significant variation in upwelling intensity was observed along the inner shelf (within about 20 km from the coastline) of Walvis Bay during winter. This region experienced temperature fluctuations between 11°C and 14°C, along with salinity changes ranging from 34.90 to 35.06. In contrast, the uppermost recorded temperature in summer reached 17.50°C, and a salinity of 35.14. ScanFish observations during our research campaigns also revealed a moderate level of upwelling near Walvis Bay during winter. This coincided with lower temperatures and increased oxygen levels. Interestingly, even in the presence of notable ESACW intrusion, the waters directly over the shelf off Walvis Bay experienced near-oxygen depletion (< 25.00 µmol kg-1) during winter. This finding could be attributed to a strong local demand for oxygen caused by the organic-rich mud belt area, as indicated by prior research (Monteiro et al., 2006; van der Plas et al., 2007). Conversely, summer profiles displayed no signs of active upwelling, with over 50% of SACW and corresponding oxygen levels of 50.00 µmol kg-1 or lower (Figure S1).
In the winter months, Kunene exhibits hydrographic features characterized by warm and saline tropical water, leading to subsurface oxygen depletion. The minimum observed oxygen concentration surpasses 40 µmol kg-1 within the depth range of 200 m to 400 m. Contrastingly, the autumn season in Kunene reveals a convergence of the poleward undercurrent with the nBUS, accompanied by a substantial pool of oxygen-depleted water below 100 m. Lower salinity and colder coastal waters, associated with an equatorward-directed coastal jet, mark this season. Off Kunene, oceanic water with elevated salinity and a deep chlorophyll maximum dominates. As summer arrives, Kunene maintains the prevalence of warm and saline tropical water, akin to winter conditions, with continued subsurface oxygen depletion.
The winter hydrography off Walvis Bay is characterized by well-ventilated surface layers with medium to strong oxygen depletion below the mixed surface layer, heightened turbidity due to phytoplankton abundance, and sea surface temperatures (SST) ranging from 12°C to 16°C, indicative of upwelling (Figure S1). The dominance of ESACW on the inner shelf contributes to high oxygen content, preventing anoxic conditions. However, increased oxygen demand in the deep inner shelf during winter is linked to elevated organic carbon concentrations in the sediment. Offshore, oxygen depletion extends below 200 m, with the OMZ core at approximately 300 m, and a substantial SACW fraction (> 40%) at the shelf edge. In autumn, Walvis Bay also exhibits distinct evidence of coastal upwelling, highlighted by spreading isotherms and water masses originating partly from tropical regions penetrating poleward. In summer, offshore SST exceeds 20°C, reflecting the South Atlantic trade wind system, and significant temperature gradients with a north-to-south surface salinity gradient are notable south of Walvis Bay. Despite a general surface salinity gradient from north to south, a detached salinity minimum occurs at Walvis Bay, potentially due to weak winds hindering a coastal jet. The central region of the nBUS at Walvis Bay indicates coastal upwelling, resulting in a thin layer of warm water near the coast. Oxygen depletion below 200 m persists, with near-complete exhaustion on the broad shelf, featuring sulfidic waters in some coastal stations. The colder yet less saline upwelling water undergoes rapid heating and stabilization, overlaying more saline water as it drifts offshore with the Ekman transport (Mohrholz et al. 2014). Oxygen depletion is observed below 200 m, with near-complete exhaustion on the broad shelf with observed sulfidic waters in a few coastal stations.
In winter, transitioning to Lüderitz reveals significant changes in salinity and oxygen levels, indicating a poleward flow near the coast. Decreased salinity and oxygen depletion suggest a distinct hydrographic regime, particularly at the southernmost point (27°S), where stratification signals the presence of an upwelling regime. Oxygen depletion in the bottom layer of the shelf and a discernible signal of poleward transport of saltier and oxygen-depleted water near 14°E at depths of 200 m to 300 m further characterize the winter hydrography off Lüderitz. In autumn, the hydrographic section off Lüderitz was situated where the northward-flowing water carried by the poleward undercurrent has already undergone upwelling. Despite this, salinity maxima at specific stations, coupled with corresponding oxygen minima, suggest the persistence of poleward flow reaching Lüderitz. Chlorophyll concentrations confined to the sea surface, with maximum values near the coast, indicate active upwelling during autumn. In summer, off Lüderitz, the variability of salinity on the shelf is considerably lower compared to Walvis Bay. Less saline water forming the top ocean layer is displaced far offshore, and prominent oxygen depletion is observed on the shelf. A hypoxic bottom layer with oxygen concentrations below 40 µmol kg‑1 extending almost to the sea surface near the coast. These seasonal variations in Lüderitz underscore the diverse hydrographic conditions, ranging from poleward flow and upwelling in winter to persistent poleward flow with active upwelling in autumn and distinct oxygen depletion during summer (e.g. Hutchings et al., 2009).
During our study, the distinct seasonal variations in upwelling patterns were also evident in the spatial distribution of SST. The SST signals show that cool upwelling waters covered the entire coast during the winter season, characterized by the SST values ranging between 11.33 °C to 15.50 °C. In contrast, during the summer campaign cool water related to coastal upwelling is observed only south of Walvis Bay (Figure 2).