Due to its tropical character, temperatures along the northeastern Brazilian coast do not vary significantly, with precipitation being the main factor responsible for determining the two main seasons: rainy and dry. Our study was carried out in the late rainy season, in July of 2013. In FN, one station was done off its south side (station 14) and two stations off its north side (stations 15 and 16), comprising an area of approximately 100 Km² (Figure 1(b)). At least five depths were sampled at each station. On the way to FN, an additional single oceanographic station was done in the open ocean (depth > 4000 m) southeast off FN, nearly halfway from ITA, where the deepest sampling was done, with ten sampled depths, the deepest at 500 m (Figure 1(a)). In ITA, seawater sampling was carried out along its shelf, one satiation down south (Station 10) and other two up north (stations 11 and 12), with at least three sampled depths at each one, comprising an area of approximately 36 Km² (Figure 1(c)). In REC there were nine oceanographic stations (stations 1 - 9), most of them with at least three sampled depths, comprising an area of approximately 28 Km² in front of the Port of Recife, which is also where the Capibaribe River meets the ocean (Figure 1(d)). See Table 1 for detailed information on all stations and sampled depths.

Seawater was collected aboard R/V Alpha Delphini (Oceanographic Institute of the University of São Paulo) during its inaugural oceanographic campaign using

a Seabird^® rosette with 12-bottles and a CTD module attached (model SBE 19 plus). Seawater temperature was retrieved from the CTD (±0.001˚C). Salinity determination was conducted using a Beckman^® RS-10 salinometer (±0.001) and the Dissolved Oxygen concentration (DO) was determined by potentiometric determination using al Titrando, Metrohm^®, with precisions of ±0.02 mL∙L⁻¹ for levels up to 2 mL∙L⁻¹ and ±0.04 mL∙L⁻¹ for higher levels, following the recommendations of Grasshoff et al. (1999). The pH was measured in the total scale using a portable combined glass electrode (model PHM 203) in a Thermo Orion pH meter model P-210A Radiometer^®, calibrated with standard buffers and stored in a KCl solution, following Aminot and Chaussepied (1983) and the recommendations of Dickson et al. (2007) with a precision of ±0.001.

Seawater for dissolved nutrients analysis was filtered through 0.45 μm GF/F

Whatman^® glass fiber membranes and kept frozen at −20˚C before chemical analyses. Dissolved silicate (Si-Si(OH)₄) and dissolved phosphate ( P-PO 4 3 − ) were determined by the colorimetric method described by Grasshoff et al. (1999). The absorbance was measured using a Genesis II Bauch & Lomb^® digital spectrophotometer and precisions were ±0.01 µmol∙L⁻¹ and ± 0.02 µmol∙L⁻¹, respectively. Dissolved inorganic nitrogen (DIN) was obtained from the sum of N-ammonium, N-nitrite and N-nitrate. N-ammonium was determined by the colorimetric method as described by Solorzano (1969) with the proposed adaptations by Tréguer and Le Corre (1975). The absorbance was read at 630 nm using a Genesis II Bauch & Lomb^Ò digital spectrophotometer with a precision of ±0.05 μM. Dissolved nitrite and nitrate were determined as described by Tréguer and Le Corre (1975). The analysis was determined in a continuous flux system using a Bran-Luebbe^® AutoAnalyser II, following recommendations proposed by Grasshoff et al. (1999), Tréguer and Le Corre (1975) and Braga (1997a, 1997b) with precisions for nitrite and nitrate of ±0.01 μM and ±0.02 μM, respectively.

Total Alkalinity (TA) analysis was carried out by an open-cell potentiometric titration with a solution of 0.01 M HCl in NaCl media to account for seawater ionic strength and to ensure stable activity coefficients (Dickson et al., 2007). Titrations were performed using a Methrom^® Dosimat 700 doser, a Methrom^® 827 pH lab and a Methrom^® Unitrode glass electrode. The titrant solution was calibrated using an ultrapure TRIS primary standard solution following the recommendations of Grasshoff et al. (1999) and a SRM 423 reference. This method is cost-effective while also maintaining uncertainties at acceptable levels of 1.5% (Anes et al., 2018). Readings of pH, electric potential and temperature of seawater samples were obtained throughout the titration in a temperature-controlled dark chamber and then used to determine the equivalence volume (Veq) on a Gran plot based on inflexion points. The seawater density was calculated from T and S using the seawater state equation (UNESCO, 1981) and TA was calculated in gravimetric units of μmol∙kg⁻¹.

Seawater samples for Chlorophyll-a (Chl-a) determination were filtered in 0.45 μm GF/F Whatman^® glass fiber membranes. Chl-a was extracted from the filters using a 90% acetone solution and concentrations were determined by spectrophotometry as recommended by Strickland and Parsons (1972), using a Genesis II Bauch & Lomb^® digital spectrophotometer, with a precision of ±0.01 mg∙m³. Seawater samples for dissolved Calcium (Ca²⁺) were also filtered using 0.45 μm GF/F Whatman^® glass fiber membranes and the Ca²⁺ concentration was determined by chelatometric titration using EDTA following Grasshoff et al. (1999) with a precision of ±2 mg∙L⁻¹.

The parameters of the carbonate system were estimated using CO₂ calc (Robbins et al., 2010), in which the inputs were pH and TA, along with pressure, salinity and dissolved nutrients. The set of equilibrium constants used was the one proposed by Millero (2010) for the CO₂ species, Dickson (1990) for the silicic acid and Lee et al. (2000) for total boron. Values of DIC, pCO₂sw, R, Ω_ar and CO₂ flux were generated by the software. The atmospheric partial pressure of CO₂ was required (pCO_2atm) as an input along with wind speed to obtain the air-sea CO₂ flux (CO₂ flux). The pCO_2atm was estimated as follows:

p CO 2 atm = CO 2 mixingratio × ( P a t m − P v a p o r ) (3)

where: the CO_{2 mixing ratio} is the relative concentration of CO₂ in the atmosphere, P_atm is the atmospheric pressure at sea level and P_vapor is the water vapor pressure.

The atmospheric CO₂ mixing ratio (CO_{2mixing ratio}) was set at 393 ppm, which represents the 20 m-high averaged atmospheric CO₂ concentration obtained from the time series record for the city of Natal, the closest to REC at the time of sampling, approximately 250 Km away (NOOA/ESRL in: Dlugokencky et al. (2015)) (Figure 2). Values of atmospheric pressure (P_atm) and wind speed were obtained from the meteorological station of R/V Alpha Delphini. The negative contribution of water vapor pressure (P_vapor) was calculated according to Weiss and Price (1980).

The data were processed using the Grubb’s test (significance level α = 0.05) and

the outliers were removed. Statistical treatment was applied to the dataset by conducting a one-way ANOVA (α = 0.05) to parameters that followed normal distribution, which was checked previously via Anderson-Darling probability plots (α = 0.05). If differences in the means were statistically significant, Tukey tests were applied to identify significant pair-wise differences. R version 3.5.3 was used to perform statistical analyses and to generate Box-Whisker plots.

Recife (REC)—Coastal Waters

The largest range of T and S was found in REC (25.46˚C - 27.08˚C and 33.93˚C - 36.49˚C, respectively). REC was also the region where the lowest values of S were registered (<35) (Figure 3(a) and Figure 3(b)). This is common in an environment where river inputs are constant, forming a salinity gradient of distribution. Santiago et al. (2010) detected values much lower in samples collected 4 km upstream the channel of the Recife port, where the riverine influence is even greater than in the river mouth (S maxima ~9.50). This riverine salinity clearly has an influence on the components of the carbonate system, however it is still unclear to which extent in REC when compared to S-driven fluctuations registered by studies on other impacted tropical estuaries, such as in the estuaries of the Yangtze by Zhai et al. (2007) and the Pearl River estuary, as showed by Guo et al. (2008).

In REC, pH variability was the highest (8.08 - 8.20) in relation to ITA and FN (Figure 3(c)), reef and open ocean areas respectively. The lowest limit highlights the interplay between the intrusion of seawater into the estuarine system and ebb flow carrying continental runoff. In addition to the riverine input in REC, the relatively low pH values might also result from ammonium-rich sewage discharge and excessive remineralization of organic matter from the metropolitan area (Braga et al., 2018a). N-ammonium comprised the major fraction

of the total DIN pool at the Capibaribe River mouth, where most of the stations registered values higher than 50%.

The intensified remineralization in REC was confirmed by the decreasing DO and Chl-a minima found at the bottom, close to the river mouth (3.87 mL∙L⁻¹ and 0.10 mg∙m⁻³, respectively). On the other hand, at the surface and subsurface levels, specially at station 2, values for DO and Chl-a maxima were respectively 5.38 mL∙L⁻¹ and 4.9 mg∙m⁻³ (Figure 3(d) and Figure 3(h)) showing the positive effect of the nutrient fertilization at the mentioned point, northward.

In terms of total dissolved nutrients, REC showed the highest values associated with the lowest salinities (DIN/ PO 4 3 − /Si(OH)₄: maxima of 16.10/0.58/35.91 μmol∙kg⁻¹), which is expected from estuarine systems (Figures 3(e)-(g)). Santiago et al. (2010) reported even higher nutrient concentrations inside the mooring area of the port of REC DIN/ PO 4 3 − /Si(OH)₄: 59.75/16.40/133.39 µM, where riverine influence is pronounced with anthropogenic contribution. This indicates that nutrients are diluted by seawater and consumed on their way out of the port channel, possibly fuelling algal blooms.

In general, Chl-a values in REC were higher than in ITA and FN, with concentrations > 2.00 mg∙m⁻³ (Figure 3(h)) and increasing closer to the river mouth and towards the surface. Such vertical variability in terms Chl-a as well as DO content in REC can be explained by relatively high photosynthetic activity within the upper layer followed by contribution of pronounced remineralization rates at depth. REC also showed the largest range of pCO₂sw (~200 μatm), with the highest mean (282 μatm) and a maximum of 340 μatm (Figure 4(c)). Sutton et al. (2019)

reported a range from 60 to 337 matm in coastal areas measured in different mooring locations. The range of results in REC is influenced by freshwater input associated with tidal action on the seafloor associated with the intrusion of remineralized nutrients even under a strong salinity gradient, which contribute to the retention of newly introduced CO_2(aq) into the water column and fuels new production locally. Similar correlations were also found by Guo et al. (2008) in the outer estuary of the Pearl River.

Additionally, a relatively high Revelle factor (R) in REC (Figure 4(d)), suggests the water column in this region is less efficient in terms of buffering pCO_2atm changes into seawater through an increase in DIC when compared to other two regions studied (ITA and FN). However, it is important to note that the waters studied in REC were twice as efficient when it comes to converting CO₂atm into DIC in relation to other estuarine systems, especially those located in temperate regions, such as the Juan de Fuca Strait studied by Feely et al. (2010), where typically high R values reached around 15.6.

Itamaracá Island (ITA)—Coastal waters

Seawater temperatures in ITA, the zone with calcareous benthic community, were the most homogenous, with values ranging from 26.45˚C - 26.68˚C, while salinity values varied from 35.4 - 36.3, comprising the highest values amongst the studied regions (Figure 3(a) and Figure 3(b)).

DO concentrations in ITA were similar to the values registered in REC and slightly lower when compared to open sea area (FN) (Figure 3(d)). Results for dissolved nutrients in ITA showed slightly lower mean values of DIN/ PO 4 3 − /Si(OH)₄: 0.72/0.08/3.81 μmol∙kg⁻¹, when compared to the values registered in REC (Figures 3(e)-(g)). Relatively low values of Chl-a (range 0.64 - 1.90 mg∙m⁻³) found in ITA indicate productivity partially based on regenerated processes limited by inorganic N availability.

ITA waters were the most alkaline of the three studied regions, with a TA maximum of 2470 μmol∙kg⁻¹ (Figure 4(a)). The relatively high values in ITA can be explained by the higher contribution of dissolved carbonate ( CO 3 2 − ) as a member of carbonate system. This is supported by the relatively high values of pH as well as by Ω_ar, with slightly higher Ω_a registered in ITA especially closer to the bottom (~4.8) followed by FN (4.6, see the next section) and REC (4.0) (Figure 4(e)).

The 15% higher TA in ITA is likely to be a result of a series of synergic factors, all of which result in an increase of Ω_ar. An insignificant freshwater input combined with a shallow and well-ventilated water column, as well as slightly higher T and S contributed to the reduction of Kspar, while the constant supply of CO_2(aq) from the atmosphere is coupled to an efficient biogenic precipitation, leading to the excess of Ca ligands (i.e. carbonate ions) with respect to dissolved ionic Ca²⁺ that is removed from the water column due to the local calcareous algae banks. This is reflected in the mean ratio of dissolved CO 3 2 − /Ca²⁺, which was relatively higher in ITA (2.35 × 10⁻²), in relation to 2.06 × 10⁻² and 2.27 × 10⁻² in REC and FN respectively (Table 2).

Results for pCO₂sw in ITA showed a narrower range when compared to REC (Figure 4(c)), which reflects its oceanic and more oligotrophic conditions supported by little remineralization of organic matter inducing a new injection CO2aq. Furthermore, the relatively high TA and DIC signals in ITA indicate a water column prone to CO_2(aq) saturation, leading to shifts in the carbonate system towards the formation of HCO 3 – and CO 3 2 − .

Overall, these results suggest that ITA is the system that sits at the optimal part of the spectrum for CO₂ take-up when compared to the other regions investigated. This is mainly attributed to new production and fixation of C by calcareous algae closely coupled with wind-driven and turbulent mixing in a shallow water column, which seem to be the main processes influencing the cascading of atmospheric CO₂ (see wind speed record in Figure 5(c)).

Fernando de Noronha (FN)—Oceanic Waters Around the Archipelago

Seawater temperatures were significantly higher in FN than in ITA and REC (one-way ANOVA, F(2.52) = 143.13, P ≤ 0.01; post-hoc Tukey: P_FN-ITA ≤ 0.01 and P_FN-REC ≤ 0.01), where the warmest waters were found (FN mean temperature of 27.60˚C) (Figure 3(a)). This is a result of its equatorial oceanic location, where solar radiation is constant throughout the year. Salinity values in FN were the most homogeneous, with little variation detected across depths and stations (35.90 - 36.15) (Figure 3(b)).

Chl-a concentrations were the most homogenous, with values ranging from 0.15 - 0.50 mg∙m⁻³ (Figure 3(h)), the lowest between the three studied regions. A study at the Rocas Atoll, located 145 km east of FN showed a similar range in terms of Chl-a (0.30 - 1.07 mg∙m⁻³) (Jales et al., 2015). However, de Souza et al. (2013) reported patches of high Chl-a (>0.70 mg∙m⁻³) nearby the archipelago of FN, which might indicate some occasional nuclei of more evident production. In general, the lowest values of dissolved nutrients were found in FN (DIN/ PO 4 3 − / Si ( OH ) 4 : maxima of 0.28/0.08/4.99 μmol∙kg⁻¹) (Figures 3(e)-(g)). The low Chl-a results combined with low dissolved nutrients suggest that primary production is maintained by regenerative processes (i.e., regenerated primary production) and likely supported by nitrogen fixers. A study by Queiroz et al. (2015)on the composition of phytoplankton in the Equatorial Atlantic around the archipelago of Saint Peter and Saint Paul, located 600 km NE of FN (0˚55'10"N 29˚20'33"W—Figure 1(a)) showed that nitrogen fixers (Trichodesmium thiebautii) comprised an important fraction of the phytoplankton biomass in the Equatorial Atlantic.

Studies on the so-called “island mass effect” (IME) have highlighted its role in driving the spatial distribution of Chl-a and nutrients at some parts of the Fernando de Noronha Ridge (FNR) in the Tropical Western Atlantic, such as Jales et al. (2015) and Cordeiro et al. (2013), both studies carried out at the Rocas Atoll, located 145 km east off FN (Figure 1(a)). Regarding the presence of this effect in FN, de Souza et al. (2013) reported there was evidence of upwelling occurring at seamounts with summits located below the thermocline close to FN. This suggests a topography-flow interaction can occur and disrupt thermohaline structures, consequently causing upward entrainment of deep nutrient-rich water to the top of the thermocline. Even though the intrusion of nutrient-rich water around FN leading to enhanced nutrient assimilation was not investigated in this study. Silva et al. (2019) showed evidence for an uplift of the thermohaline structure during the late rainy season of July 2010, which was associated with an increase of Chl-a concentrations of up to 1 mg∙m⁻³ downstream the Archipelago. However, the same authors highlighted that the observed IME was not strong enough to significantly change the carbonate system to further induce biological carbon uptake or cause a supply of CO₂-rich deep waters.

The obtained results for TA show similar means between FN and ITA (Figure 3(a); one-way ANOVA, F(2,52) = 4.44, post-hoc Tukey: P_FN-REC = 0.11). However, a larger spread was detected in FN with water tending to be less alkaline (mean TA 2294 μmol∙kg⁻¹, Figure 4(a)) despite the oceanic character of FN and the lowest variability both in terms of S and pH (Figure 2(b) & Figure 2(c)). The lower end values of the TA range (~2150 μmol∙kg⁻¹) in FN indicate that the carbonate system is partially influenced by other processes associated with the release of hydrogen protons (H⁺) within the subsurface layer and the top of the thermocline in FN, consequently reacting with proton acceptors such as CO 3 2 − and decreasing overall TA values.

Three main processes involving nitrogen fixation at surface levels and a rapid assimilation of nutrients at subsurface levels could help explain the observed mean difference in TA observed between ITA and FN (ΔTA of ~70 μmol∙kg⁻¹). The nitrification of recycled nitrogen from NH 4 + to NO 3 − followed by rapid NO 3 − assimilation is responsible for a decrease in TA of 2 moles per mole of NO 3 − formed. Secondly, the remineralization of newly formed NH 4 + produced via N fixation followed by nitrification is responsible for a further decrease in TA of 1 mole per mole of N fixed. Another process that also reduces TA is the assimilation of NH 4 + itself, which results in a decrease in TA of 1 mole per mole of NH 4 + used, following the so-called nutrient-H⁺-compensation principle (Wolf-Gladrow et al., 2007).

Results for the DIN pool composition in FN show relatively high fractional values of N-ammonium particularly at shallower depths. NH 4 + comprised more than 60% of the DIN pool at Station 14 and more than 40% at Stations 15 and 16 for surface and subsurface depths. Recent studies also show there is enhanced recycling of NH 4 + associated with productive regimes and a TA drawdown around the main island in FN and close to the shore (Braga et al., 2018a; Braga et al., 2018b).

Furthermore, a quantitative assessment as to whether these nutrient-assimilation mechanisms could have caused such a drawdown is presented based on results shown here and values found in the literature.

Considering nitrification of upwelled NH 4 + at station 13 and given that the observed mean concentration of NO 3 − below the thermocline was ~20 μmol∙kg⁻¹, it is hypothesized that at least 20 μmol of recycled NH 4 + would have been nitrified to NO 3 − in the upper thermocline and readily assimilated in the oligotrophic water column of FN, producing a ΔAT of ~40 μmol∙kg⁻¹. N fixation estimates in the equatorial South Atlantic are very scarce. However, considering that most studies in the equatorial North Atlantic have estimated the integrated N fixation at approximately ~200 μmol N∙m⁻²∙d⁻¹ (Gruber & Sarmiento, 1997; Capone et al., 2005; Moore et al., 2009), values for the upper thermocline in the southern portion of the equatorial Atlantic would likely agree. Therefore, as a first approximation, a 100 m homogenous upper thermocline in the equatorial South Atlantic would represent a ΔAT of 2 μmol∙kg⁻¹ due to remineralization of fixed N into NH 4 + followed by nitrification. Lastly, to account for the total ΔAT of ~70 μmol∙kg⁻¹, the direct assimilation of a further 28 μmol∙kg⁻¹ of remineralised nitrogen in the form of NH 4 + would have to have occurred. This is plausible given that NH 4 + constituted a significant part of the DIN pool, which reflects that productivity in FN could also rely on ammonium-based carbon fixation. However, there might be additional contributions driving the observed decrease in TA, which can be, for example, via the release of sulphates during remineralization (Wolf-Gladrow et al., 2007), nevertheless this needs further investigation as the local processes driving the sulphur as well as the nitrogen and carbon cycles in this part of the Atlantic are still poorly understood.

Surface and subsurface pCO₂sw data from REC showed relatively high values (>280 μatm) in front of the Capibaribe river mouth (stations 1 and 2), causing the lower differences between estimated pCO_2atm and pCO_2sw (~100 μatm). These high pCO₂sw values reflect the role of the river plume in injecting DIC-rich water into the system. Additionally, this coincides with the observed Chl-a maxima at surface levels, showing that this region favours patches of high productivity to develop. In contrast, at stations more offshore (6, 7 and 8) the lowest pCO₂sw values and the highest differences between estimated pCO_2atm and pCO_2sw were registered (~180 μatm). This region of the estuarine system is relatively less dynamic in terms of water movement, which allows for the settlement of organic matter that had been produced at the river mouth. This means that, even with moderate to low primary production levels happening at the surface layers, there is significant remineralization in this portion of the system close to the bottom that drives the observed increase in pCO₂sw. In ITA, pCO₂sw values ranged from 250 - 300 μatm. Station 10 showed the maximum value at the surface, followed by station 12 (300 μatm and 280 μatm, respectively). These points differed from the estimated pCO_2atm by approximately 80 and 100 μatm, respectively. Even though, the range of pCO₂sw observed in REC was larger than that of ITA, no significant difference in mean pCO₂sw between REC and ITA was found. On the other hand, pCO₂sw values differed significantly between REC and FN, this is mainly due to surface values in FN being significantly lower, highlighting the relatively less saturated and oligotrophic water column of FN (Figure 5(a)).

Overall, CO₂flux estimates suggest that the three regions acted as sinks (negative fluxes) during the sampling campaign in the late rainy season of July 2013 (Figure 5(b)). However, values of CO₂flux from REC were significantly lower than in ITA and FN, with data from stations 1,5, 8 and 9 indicating practically atmospheric equilibrium (fluxes > −2 mmol C∙m⁻²∙d⁻¹). This suggests that the CO₂flux into the ocean in REC coastal waters is attenuated by high pCO₂sw, which might be a result of the combined effects of eutrophication, remineralization, low pH and high freshwater and DIC inputs, all of which counteract a further dissolution of CO_2atm, shifting the carbonate system towards CO_2(aq) formation and outgassing. Most studies concerning CO₂ fluxes in estuarine systems along the Brazilian northeastern coast show that these systems typically act sources of CO₂

when their entire catchment area is considered (Araujo et al., 2013; Noriega & Araujo, 2014; Noriega et al., 2015). Nonetheless, local studies investigating the interface between estuary and shelf waters are scarce and the complexity of biological processes taking place in these areas makes the carbonate system variability in the downstream portion of estuarine environments poorly understood. There is also seasonality, which can be an important agent contributing to the variability in the carbonate system (Sutton et al., 2019), adding to the complexity of this area.

Additional uncertainty occurs when pCO₂ estimates and other parameters of the carbonate system are derived from linear relationships with, for example, sea surface salinity (SSS). Bonou et al. (2016) emphasized that this is particularly the case when SSS-based pCO₂ values are lower than 250 µatm in regions off the northeastern coast of Brazil as SSS expressions fail to consider the high biological activity observed along the shelf. These deviations might be understood as noise when first analysed altogether in the context of the linear relationships based on large collated datasets. However, by looking closely at the processes that might be driving these deviations, the role of high biological activity becomes evident in differentiating each of these coastal systems, ultimately setting them aside from the general trends. The results presented here suggest that both ITA and FN are systems where biological processes underpinning the carbonate biogeochemistry are very distinct from other areas along the shelf and more offshore. This means that studies involving estimates for the carbonate system parameters relying on SSS in these locations should be considered with caution and factor in biogeochemical processes and seasonal conditions.

The present values for CO₂flux were generally lower than −12 mmol C∙m⁻²∙day⁻¹ in ITA and FN. Stations 14 in FN and 10 in ITA, showed the highest influxes (~−20 mmol C∙m⁻²∙day⁻¹). Rain, wind speed and sea roughness are thought to be the main factors responsible for the injection of atmospheric CO₂ at these stations, as these contribute to the dissolution of CO₂ into seawater by lowering the salinity of the air-sea layer and forming undulations and turbulence. In fact, at the time waters of ITA and FN were sampled at stations 10 and 14, wind speed and sea roughness were the highest registered (wind speeds > 10 m∙s⁻¹). Relatively high wind gusts were also observed at stations 15, 12 and 11 (>8 m∙s⁻¹, Figure 5(c)). The observed fluxes result from a water column saturated with atmospheric CO₂ being efficiently coupled with biological processes that remove DIN species. Particularly, results suggest that ITA is marked by efficient biological precipitation of calcium carbonate by the calcareous algae matts while the uptake in FN is mainly characterized by ammonium-based productivity and intermittent bursts of moderate production following the entrainment of nutrient-rich water within the thermocline.

An estimated uptake of approximately 9.46 Gt C∙yr⁻¹ was found when all surface in situ values of CO₂flux were averaged within each region and then integrated in space (Table 3). This estimate would represent only a 1.7% offset on the so-called Brazilian Current marginal catchment area outgassing (+0.567

TgC∙yr⁻¹) reported by Laruelle et al. (2014). However minimal this offset might represent when compared to the overall outgassing estimation, it is important to note that this study covered an area over 3000 times smaller than that considered by the referred authors. This highlights how important regional-scale observations are to improving our understanding of the biogeochemical processes which alter the carbonate system. They also contribute to fine tuning large-scale approaches in tropical shelf seas.

It is important to highlight that measurements were based on single observations during the late rainy season, thus they do not represent net fluxes over seasonal cycles and that results were compared to studies based on monthly or seasonally-averaged measurements. This also reinforces the need of further in situ seasonal research to be conducted in the tropical part of the Western South Atlantic to assess whether the patterns observed remain year-round. The need for monitoring efforts addressing the state of Brazilian waters in face of increasing atmospheric CO₂ and ocean acidification has been recognized and initiatives are being implemented towards the establishment of a more cohesive and integrative network, able to share information at regional, national, and international stages (Kerr et al., 2016). This study further emphasizes the role of regional-scale studies in the identification of important sites to gain a better understanding of the carbon cycle in shelf seas and long-term shifts in the carbonate system in the equatorial Atlantic off northeastern Brazil.

In FN, even though low Chl-a and DO values indicated the occurrence of remineralization associated with high temperatures, which are factors that induce CO₂ outgassing, relatively high CO₂ fluxes into the ocean were also detected where typically high seawater temperatures occur throughout the year. de Souza et al. (2013) suggested that, as greater evaporation takes place in FN, surface water salinity tends to increase thus becoming denser. As a result, surface water parcels tend to sink to the base of the thermocline, where they meet lower temperatures and become even denser. This might indicate there are other factors participating in the CO₂ sink in the waters surrounding FN and further investigations as to how this sink is driven must be carried out. Additionally, while this study highlights possible effects caused by the North Brazil Undercurrent bifurcation along and off the shelf up to FN’s location, observations from Olivier et al. (2021) also highlight the importance of biogeochemistry, biological activity, low-salinity plumes and the role of the North Brazil Current rings in driving variability in productivity, affecting the carbonate system and modulating CO₂ fluxes in this portion of the Atlantic.