The carbonate system and air-sea CO2 fluxes in coastal and open-ocean waters of the Macaronesia

The CO2 system, anthropogenic carbon (Cant) inventory and air-sea CO2 fluxes (FCO2) were analysed in the archipelagic waters of the Macaronesian region. The (sub)surface data were collected during POS533 (February and March, 2019) in coastal areas leeward of Cape Verde (CV), Canary Islands (CA) and Madeira (MA) and through the vessel track. The CO2 variability was controlled by changes in temperature, biological activity and advection processes forced by spatial heterogeneities in the Canary Upwelling System, the mixed layer depth, the mesoscale activity and the circulation patterns. The surface fCO2,sw variability was driven by biological production and CO2-rich water injection in tropical waters and by temperature fluctuations in subtropical waters. The factors controlling the upper ocean changes in the total inorganic carbon normalized to a constant salinity (NCT) were assessed. The uptake and storage of anthropogenic carbon, calculated by using the TrOCA 2007 approach described, as an upper limit, > 60% (>90% above the MLD) of the NCT increase from preformed values. The organic carbon pump accounted 36.6-40.9% for tropical waters and lose importance for subtropical waters (7.5-11.6%), while the carbonate pump has a minimal contribution (<4.2%). The upper-ocean Cant inventory in coastal areas of CV (8,570 Km2), CA (7.960 Km2) and MA (1,250 Km2) was 7.57 x 103, 9.26 x 103 and 8.86 x 103 µmol kg-1, respectively (0.51, 0.58 and 0.09 Tg C, respectively). In terms of FCO2, the CV, CA and MA behaved as a winter CO2 sink (-4.74, -3.90 and -8.34 mmol m-2d-1, respectively) while a strong outgassing was detected over the Cape Blanc filament (20-25 mmol m-2d-1). The total average FCO2 for the ocean area of the three archipelagos (371,250 Km2) was -28.27 Gg CO2 d-1. The POS533 data were compared and compilated with SOCAT and GLODAP data and a new set of equations was provided to calculate the fCO2,sw, Cant and FCO2 in the Macaronesian region based on physical and biogeochemical properties. Graphical Abstract


Introduction
The Eastern Boundaries Upwelling Systems, transitional areas to subtropical gyres waters and coastal regions play an important role in the global CO 2 cycle because they have high nutrient inputs that lead to high primary production and high recycling/export rates of carbon (e. g. Huntsman and Barber, 1977;Schulz, 1982;Jewell, 1994). These regions are characterized by highly variable seawater properties but have received limited study and are poorly represented in global models. Thus, the field monitoring and development of new local and regional scale studies are required in these zones.
The Macaronesian region ( Figure 1) in the eastern boundary of the North Atlantic (sub)tropical circulation is a key zone in terms of upper ocean CO 2 distribution, natural and anthropogenic carbon inventory and air-sea exchange. This area is influenced by the Canary Upwelling System following the equatorward Canary Current (e. g. Wooster et al., 1976;Mittelstaedt, 1991;Van Camp et al., 1991;Cropper et al., 2014) and the intense mesoscale activity mainly driven by upwelling filaments and both coastal and island generated eddies (e. g. Barton et al., 2004;Garcıá-Muñoz et al., 2004;Sangrà et al., 2009). The Cape Verde Frontal Zone (CVFZ) extended from Cape Blanc (21.5-22.5°N) and crossing the Cape Verde archipelago also represents an important source of latitudinal heterogeneity in the CO 2 distribution by separating the low-nutrient content and oxygen-rich subtropical waters from the nutrient-rich and oxygen-depleted tropical waters (Pelegrı́and Peña-Izquierdo, 2015a), which is referred to as a subsurface oxygen minimum zone (OMZ) (Stramma et al., 2008a;Stramma et al., 2016).
The temporal variability of the CO 2 system in the Eastern boundary of the North Atlantic subtropical gyre has been studied north of the Canary archipelago at the European Station of Oceanic Time Series (ESTOC) (Gonzaĺez-Davila et al. 2003, Gonzaĺez-Davila et al., 2010Bates et al., 2014;Santana-Casiano et al., 2007), located in a windward and open-ocean area not affected by islands generated eddies and which normally does not receive upwelling filaments (Davenport et al., 1999;Davenport et al., 2002). The results obtained show the seasonality of fCO 2 in the upper water-column, its interannual rate of increase linked with a decrease in pH and the net annual ocean CO 2 sink behaviour explained by high ingassing rates during the cold months. In addition, the monitoring of physical and biogeochemical properties of the eastern tropical North Atlantic at the windward Cape Verde Ocean Observatory (CVOO) north of the Cape Verde archipelago has allowed the characterization of the oxygen-depleted eddies and its contribution to the formation of the shallow OMZ in the coastal transitional area (Fiedler et al., 2013;Karstensen et al., 2015;Fiedler et al., 2016;Schütte et al., 2016b;Pietri and Karstensen, 2018). The eddies also introduce spatio-temporal variability in the biological processes and carbon cycles (e. g. Benitez-Nelson, 2000;Fiedler et al., 2013;Romero et al., 2016).
Several studies based on data collected aboard volunteer observing ships and research vessels have evaluated the effect of coastal upwelling in the CO 2 distribution and air-sea fluxes along the coastal transition region north of the Canary Islands (Pelegrı́et al., 2005a;Padin et al., 2010;Curbelo-Hernańdez et al., 2021). This zone has been reported as an annual CO 2 sink driven by lower sea surface temperatures (Pelegrı́et al., 2005a;Curbelo-Hernańdez et al., 2021) and an enhanced biological CO 2 uptake compared with the warmer and less biological productivity Mauritanian-Senegalese upwelling area (Lachkar and Gruber, 2013;Cropper et al., 2014) which acts as an annual CO 2 source (Gonzaĺez-Davila et al., 2017). The evaluation of the island-generated eddies effect on the CO 2 system south of the Canary Islands shows that these mesoscale features drive an enhancement of nutrient supply and increase the inorganic carbon in the upper layers (Gonzaĺez-Davila et al., 2006;Ucha et al., 2010). However, the CO 2 distribution, anthropogenic carbon content and air-sea CO 2 fluxes through the inter-islands, coastal and leeward regions of the Macaronesian archipelagos are poorly known.
The present study analyses the upper water-column CO 2 system and air-sea fluxes in the Macaronesian region and quantifies the anthropogenic carbon inventory and its spatio-temporal changes based on data collected during the POS533 cruise. The vertical profiles have been performed mainly in the leeward coastal regions of the islands close to the 100 m isoline, and at the ESTOC and CVOO sites. The surface underway sampling through the vessel track allows the monitoring of CO 2 properties over upwelling filaments and equatorward mesoscale eddies and through the inter-island routes.
This study contributes to a better understanding of the CO 2 cycle in coastal transitional regions of the (sub)tropical Northeast Atlantic.

Study area
The Macaronesian region (Figure 1) in the Northeast Atlantic is constituted by the Azores, Madeira, Canary and Cape Verde volcanic islands archipelagos, which are located at 100-750 km off the African coast in a coupling area between the Canary Upwelling System along the Northwest African coast and the oligotrophic open-ocean waters of the Northeast Atlantic subtropical gyre. The temperature-salinity relationship provided by Mehlmann et al. (2020) during POS533 identifies the water masses in the Cape Verde, Canary and Madeira archipelagos up to 4000 m depth as well as the mixed layer depths (MLDs), which were considered in this investigation. In the upper layers, a Surface Water mass (SW) up to~250 m deep above the low saline South Atlantic Central Water (SACW) was found in Cape Verde, while surface waters around the Canary and Madeira archipelagos included the saltier Madeira Mode Water (MMW) above the Eastern North Atlantic Central Water (ENACW) and were influenced by Mediterranean Water (MW).
The circulation pattern in this region has been widely studied (e. g. Mittelstaedt, 1983;Peŕez-Rodrıǵuez et al., 2001;Stramma et al., 2005;Meunier et al., 2012). The Canary Upwelling System follows equatorward the Canary Current, transporting surface waters with low temperatures above the ENACW. It leaves the coast at the latitude of Cape Blanc (21.5-22.5°N) and flows southwestward to become the North Equatorial Current (NEC), while a poleward current transporting warmer surface waters above the SACW recirculates cyclonically into the open ocean between the Cape Verde Archipelago and Cape Blanc. Hence, cooler surface waters and the warmer, saltier and low-nutrient ENACW are separated from the warmer surface waters and the cooler, fresher and nutrient-rich SACW forming the CVFZ (e. g. Zenk et al., 1991;Peŕez-Rodrıǵuez et al., 2001;Meunier et al., 2012). The CVFZ crosses southwestward the Cape Verde archipelago and is characterized by a strong thermohaline and gradients of both nutrients and oxygen (Pelegrı́et al., 2015). It is a permanent front strongly influenced by the seasonal meridional migration of the Intertropical Convergence Zone (ITCZ; Mayer and Weisberg, 1993;Stramma and Schott, 1999), which reaches its northeastmost position in summer and fall (Laźaro et al., 2005). The weak ocean ventilation and enhanced respiration in the Northeast tropical Atlantic deplete the oxygen content in both ENACW and SACW and result in an extended horizontal OMZ south of Cape Blanc, which is the strongest south and east of Cape Verde Islands in the depth range of 200-800 m (Karstensen et al., 2008;Stramma et al., 2008a;Stramma et al., 2016).
The heterogeneous interaction of the Canary Current with the complex morphology of the Northwest African coastline and continental slope results in substantial latitudinal variability in terms of intensity and seasonality of the upwelling, circulation, nutrients supplies, biological production and carbon distribution (e. g. Pelegrı́et al., 2005a;Pelegrı́et al., 2005b;Valdeś and Deńiz-Gonzaĺez, 2015;Pelegrı́and Peña-Izquierdo, 2015a;Pelegrı́and Benazzouz, 2015b;Bonino et al., 2021). These spatial differences separate the permanent coastal upwelling between Cape Blanc and the Strait of Gibraltar from the seasonal upwelling south of Cape Blanc occurred only in winter and identifies the more intense upwelling throughout the year between Cape Ghir (30.7°N) and Beddouza (32.5°N) and south of Cape Yubi (27.9°N) (e. g. Wooster et al., 1976;Mittelstaedt, 1991;Van Camp et al., 1991;Cropper et al., 2014). The physical and biogeochemical seawater properties of the Macaronesian region are strongly linked with those of coastal upwelling due to the offshore transport hundreds of kilometres off the African coast, especially though the main upwelling filaments of Cape Ghir (Hagen et al., 1996;Garcıá-Muñoz et al., 2005;Pelegrıé t al., 2005a;Pelegrı́et al., 2005b;Laiz et al., 2012) and Cape Blanc (Bonino et al., 2021;Gabric et al., 1993;Pastor et al., 2008;Meunier et al., 2012;Ohde et al., 2015;Lovecchio et al., 2017). The surface horizontal advection by filaments is combined with the upwelling/ downwelling of water in the core of cyclonic/anticyclonic eddies throughout The Canary Eddy Corridor and the small zonal corridors south of Madeira and near Cape Blanc Sangrà et al., 2009;Cardoso et al., 2020). The coupling of these processes represents a source of spatio-temporal heterogeneity in the physical and biogeochemical properties in the tropical and subtropical Northeast Atlantic coastal transitional area.

Data collection
The CO 2 system was studied in the Macaronesian region based on data collected on board the RV Poseidon during the POS533 cruise (February 28 to March 19, 2019). Both surface underway measurements and depth seawater samples were taken during the cruise mainly downwind in the archipelagos of Cape Verde (CV), Canary (CA) and Madeira (MA) and through the latitudinal transects that connect them (Figure 1). A detailed cruise overview is given by Mehlmann et al. (2020) and Arnone et al. (2022).
The pH in total scale was underway monitored in surface waters along the vessel track by using a spectrophotometric pH sensor (SP101-SM) developed by the QUIMA group (IOCAG-ULPGC) and SensorLab Gonzaĺez-Davila et al., 2016) and based on the method described by Clayton and Byrne (1993). The SP101-SM pH sensor uses 4 wavelengths analysis for the m-cresol purple, includes auto-cleaning steps, performs a blank for pH calculation immediately after the dye injection and removes any dye effect in each pH reading (the accuracy concerning a TRIS seawater buffer was ±0.002 units). The spectrophotometric system was placed in the onboard lab and measured the pH of the seawater pumped from the main seawater intake of the vessel (at around 6 m depth) with a frequency of 5 minutes. The change in temperature between the seawater intake and the multiple seawater outlets in the onboard lab (0.3 ± 0.1°C) was considered to correct the total scale pH values to the in situ conditions (~0.015 ± 0.001 units°C -1 ). A SeaCat SBE21 thermosalinograph placed at the same location of the seawater immersed pump was used to monitor at the main intake the sea surface temperature (SST) and salinity (SSS) with an accuracy of 0.01°C and 0.001, respectively. The surface underway fluorescence was monitored by a not-calibrated Wetlabs ECO fluorometer and is given as chlorophyll a (Chl-a) data in arbitrary units.
The characterization of the water column was carried out by selecting 58 sampling stations ( Figure 1) ranging from the surface to around 250-300 meters deep. In the CV archipelago, 15 stations were selected leeward of the islands of Sao Antao, Sao Nicolau, Fogo, Santiago, Boa Vista and Sal. In the CA archipelago, 25 stations were selected around the islands of El Hierro, La Gomera, Tenerife and Gran Canaria, mainly on their lee side and followed eastward the coastlines. In the MA archipelago, 15 stations were selected leeward of Madeira at two sections and at the Ilhas Desertas. In addition, two windward stations were selected North of CV and CA at the CVOO and ESTOC sites, respectively.
Depth water sampling and in situ measurements were performed using a bottle rosette sampler containing twelve 10 L Niskin bottles with a CTD and additional sensors of pressure, temperature, salinity, dissolved oxygen and not-calibrated fluorescence data given as Chl-a (Mehlmann et al., 2020).

CO 2 system variables
The Total Alkalinity (A T ) and Total Inorganic Carbon (C T ) in the water column were determined onboard from the depth water samples using a VINDTA 3C and following Mintrop et al. (2000). Samples were potentiometrically titrated with HCl to the carbonic acid endpoint for A T determination, while C T was coulometrically determined. Both A T and C T values were corrected using CRM bottles (batch #177, provided by A. Dickson at Scripps Institution of Oceanography), giving values with an accuracy of ±1.5 and ±1.0 μmol kg -1 , respectively. They were normalized (NA T and NC T , where NX= X/S·36.4) to a constant salinity of 36.4, which is the average salinity of the upper layer (0-250 m depth) in the entire Macaronesian region.
The surface A T values was also calculated for the longitude and latitude of the surface underway data using the A T -SSS relationship obtained from the surface samples (Eq. 1, r 2 = 0.94) (temperature was not found to improve the fitting). The average A T values calculated along the cruise tracks (2398.5 ± 20.0 μmol kg -1 ) agreed with those calculated according to the global relationship of A T with SSS and SST in surface waters of the Atlantic (sub)tropics (2401.6 ± 19.5 μmol kg -1 ) given by Lee et al. (2006).
The ocean values of CO 2 fugacity (fCO 2,sw ) and pH were computed from the determined A T and C T . The surface values of fCO 2,sw and C T throughout the cruise track were computed from the underway measured pH and determined A T . The Excel program CO 2sys was used for computation, with the carbonic acid dissociation constants of Lueker et al. (2000), the HSO 4 dissociation constant of Dickson (1990) and the value of [B] T determined by Lee et al. (2010). The relevance of thermal and non-thermal processes on the variability of the fCO 2,sw and pH was evaluated in this study. The fCO 2,sw and pH explained by non-thermal processes (mainly biological activity and horizontal/vertical advection) were studied by removing the temperature effect through a normalization of the fCO 2,sw and pH to 21°C (fCO 2,T21 and pH T21 ). The changes in fCO 2,sw and pH due to thermal processes (DfCO 2,thermal and DpH thermal ) were assumed as the differences between the observed and normalized values (DfCO 2, thermal = fCO 2,sw -fCO 2,T21 and DpH thermal = pH -pH T21 ).

Oxygen
Seawater samples for dissolved oxygen determination were collected in pre-calibrated glass wide-neck bottles (previously washed 3 times with the seawater sample) and avoiding bubbles formation. The temperature of the water was recorded during the sampling. The reagents 1 (MnCl +4H 2 O) and 2 (NaOH + NaI) were then added and thoroughly mixed with the seawater sample. The already mixed samples were kept in a dark box for 6 h to allow the precipitate to settle at the bottom of the bottles.
The WINKLER method introduced by Winkler (1888) and optimized by Carpenter (1965) and Carrit and Carpenter (1966) was used to determine the dissolved oxygen of the seawater samples. A Metrohm 888 Titrando operated with the software Tiamo and an amperometric electrode to determine the endpoint was used for the titration (Culberson and Huang, 1987). The reagent 3 (H 2 SO 4 ) was added immediately before starting the titration to acidify the sample. Thiosulfate 0.01N was used as a titrant and a solution of KIO 3 0.01N as a standard solution. All the reagents and solutions used during the cruise for DO determination were prepared following the procedures described by Dickson and Goyet (1994). Standardization of the thiosulphate was performed every two days. The possible impurities of the reagents were controlled by determining a blank every 2 days.

Anthropogenic carbon
There are several indirect-based methods for the calculation of the excess of carbon due to anthropogenic input (C ant ) explained in detail by Sabine and Tanhua (2010). The methods assumed that ocean circulation and the biological pump have operated in a steady state in terms of seasonal and interannual variability of the natural carbon cycle since preindustrial times (before 1750) and considered that C ant could be estimated by removing the contribution of the biological and physical pumps and the preindustrial C T from the determined C T .
In this study, the C ant was calculated by using the TrOCA backcalculation technique and the improved TrOCA 2007 approach (Touratier et al., 2007). The potential temperature (q,°C) and the A T , C T and O 2 concentrations (in μmol kg -1 ) were considered (Eq. 2).
The changes in C ant due to differences in air-sea CO 2 exchange rates and O 2 content were reflected in the variation of this tracer. The TrOCA method provided overestimates C ant values by about a factor of 2, which were considered as upper limit values (Yool et al., 2010).

A T and C T components
The general processes that control the vertical distribution of A T and C T were studied by applying the initial model proposed by Chen and Millero (1979) and previously used, in the Northeast Atlantic (Gonzaĺez-Davila et al., 2010) and the Pacific Subarctic (Andreev et al., 2009) (Eq. 3 and 4). This set of calculations provided the drivers of the observed A T and C T in the coastal waters of the three archipelagos. The model considers that the measured concentrations of A T and C T ( A meas The A pre T values were computed using the function given by Wanninkhof et al. (1999) and considered correct for the water masses in the North Atlantic by Peŕez et al. (2002). This A pre T model Curbelo-Herná ndez et al. 10.3389/fmars.2023.1094250 Frontiers in Marine Science frontiersin.org was reported with an uncertainty of ±5.2 μmol kg -1 and taken into consideration the salinity and the concentration of NO 3 and O 2 (Eq. 5).
The NO 3 was spectrophotometrically determined (Grasshoff et al., 1999) with a QuAAtro auto-analyser (SEAL Analytical, UK) by Kastriot Qelaj (GEOMAR). The DA org T was estimated by using the apparent oxygen utilization (AOU) computed from the measured temperature, salinity and O 2 and the stoichiometric coefficients for the change in nitrate (DN/DO 2 = 16/170) given by Anderson and Sarmiento (1994) (Eq. 6). The DA carb T was directly calculated from Eq. 3.
A pre T = 278:4 + 57:01 · Salinity + 0:0074 · (10:6 · NO 3 + O 2 ) (5) The change in C T due to the biological contribution through the organic carbon and carbonate pumps was taken into consideration. The DC org T was computed from the AOU and the stoichiometric coefficients for the change in carbon (DC/DO 2 = 117/170) given by Anderson and Sarmiento (1994) (Eq. 7), while DC carb T is half of the contribution estimated for DA carb T according to Chen and Millero (1979)

Air-sea CO 2 fluxes
The air-sea CO 2 fluxes (FCO 2 ) were calculated from the surface underway data collected through the vessel tracks and using Eq. 9. The solubility (S), the difference between the surface seawater and low atmosphere fCO 2 (DfCO 2= fCO 2,sw -fCO 2,atm ) and a conversion factor of 0.24 mmol m -2 d -1 were considered in the calculation to express the FCO 2 results in mmol m -2 d -1 . Negative fluxes indicate that the ocean behaves as an atmospheric CO 2 sink, while positive ones indicate that it behaves as a source. FCO 2 = 0:24 · S · k · Df CO 2 (9) The parameterization of Wanninkhof (2014) was used in this study, with k being the gas transfer rate expressed in Eq. 10: where w is the wind speed (m s -1 ) and Sc is the Schmidt number (kinematic viscosity of seawater, divided by the gas diffusion coefficient). The underway low atmospheric CO 2 concentration and wind speed used for the calculation of FCO 2 were collected on board the vessel and provided by Weiß et al. (2020) in the World Data Center PANGAEA ® . The low atmospheric CO 2 concentration was measured with a cavity ring-down spectrometer (CRDS, Picarro G2301-f) and a GEOMAR's 'Atmospheric Intake System' (AIS). The wind speed was continuously monitored by an onboard meteorological station. Both underway measurement systems were installed at 7.5 m above the sea surface.

Results
The distributions of temperature, salinity, fluorescence and CO 2 system variables were analysed up to 250 m depth in the CV ( Longitudinal distribution of (A) temperature, (B) CO 2 fugacity in seawater (fCO 2,sw ), (C) normalized total inorganic carbon (NC T ) and (D) pH in the first 250 m depth leeward of Cape Verde archipelago. Isolines were added in plots every 1°C for temperature, 50 µatm for fCO 2,sw , 25 µmol kg -1 for NC T and 0.05 units for pH.
(Fogo, Santiago, Maio and Boa Vista; CV-SE hereinafter) as well as in the CA between the Western most area (El Hierro, La Gomera and the West coast of Tenerife; CA-W hereinafter) and the Eastern most area (the East coast of Tenerife and Gran Canaria; CA-E hereinafter). The longitudinal surface changes in the physical properties and surface CO 2 were analysed in each archipelago along the cruise tracks ( Figure 5) and the latitudinal surface gradients from CV to MA were also considered (Supplementary Figure 4). The regional averages of the study variables from the surface to 180 m depth are shown in Table 1.

Temperature, salinity and fluorescence
Latitudinal temperature differences were found in the upper water column up to 100 m depth, with maximum average values in the CV Longitudinal distribution of (A) temperature, (B) CO 2 fugacity in seawater (fCO 2,sw ), (C) normalized total inorganic carbon (NC T ) and (D) pH in the first 250 m depth leeward of Madeira. Isolines were added in plots every 1°C for temperature, 10 µatm for fCO 2,sw , 10 µmol kg -1 for NC T and 0.01 units for pH.

FIGURE 3
Longitudinal distribution of (A) temperature, (B) CO 2 fugacity in seawater (fCO 2,sw ), (C) normalized total inorganic carbon (NC T ) and (D) pH in the first 250 m depth leeward of Canary Islands. Isolines were added in plots every 1°C for temperature, 10 µatm for fCO 2,sw , 10 µmol kg -1 for NC T and 0.01 units for pH.  Longitudinal distribution of (1) SST (red) and SSS (blue), (2) fCO 2,sw (blue) and fCO 2,T21 (red), (3) pH (blue) and pH T21 (red) and (4) 1 Regional average of temperature, salinity, Chlorophyll a, fCO 2,sw pH, C T and NC T in the nothern and southeastern section of Cape Verde (CV-N and CV-SE, respectively), in the western and eastern section of Canary Islands (CA-W and CA-E, respectively) and in the Madeira archipelago (MA) in surface waters and at 10, 30, 50, 100 and 180 m depth.

Total inorganic carbon
The surface distribution of C T was linked to the SSS, with a higher concentration towards the north of Cape Blanc (Supplementary Figure 4). Maximum C T surface values were obtained in coastal CA-W areas (~2120-2140 μmol kg -1 ) followed by CA-E (~2115-2130 μmol kg -1 ) and MA (~2105-2120 μmol kg -1 ), while the minimum values were found in CV (~2035-2095 μmol kg -1 ) and especially leeward of Sao Antao and Sao Vicente (~2060-2075 μmol kg -1 ) and throughout CV-SE (~2035-2075 μmol kg -1 ) ( Figures 5A.4,  B.4, C.4). The surface and upper water column C T were normalized (NC T ) to constant salinities of 36.6 and 36.4, the average salinities of the Macaronesian region for surface and first 250 m depth, respectively, to remove the effect of evaporation/precipitation, the horizontal/vertical advection and the influence of different water masses.
The highest NC T surface values in the entire region were found around Cape Blanc (~2140-2165 μmol kg -1 ) (Supplementary Figure 4E), while a strong NC T depletion was observed south of Cape Blanc at CV ( Figure 5A.4). Minima NC T surface values in CV were found southwest of Fogo and Maio (~2070-2080 μmol kg -1 ), while maxima were observed through the Boa Vista-Sal-Sao Vicente northeast most transects (~2085-2105 μmol kg -1 ). In CA ( Figure 5B.4), the surface NC T -rich waters were found on the East coast of El Hierro (~2133-2136 μmol kg -1 ) and La Gomera (~2134-2139 μmol kg -1 ), in more open-ocean waters between both islands (~2130-2136 μmol kg -1 ) and at certain locations along the southwest and southeast coast of Tenerife (~2127-2142 μmol kg -1 ). The lowest NC T surface values were obtained across the most oceanic transect south of La Gomera and El Hierro (~2106-2113 μmol kg -1 ), in CA-E (~2105-2120 μmol kg -1 ) and north of Gran Canaria around the ESTOC site (~2105-2112 μmol kg -1 ). In MA, higher NC T surface concentrations were measured through the coastal transect compared with those of the open-ocean transect ( Figure 5C.4). The maximum values were observed East of Madeira and along the West coast of Ilhas Desertas (~2113-2115 μmol kg -1 ).
The C T and NC T remained depleted in the most productive euphotic layer and increased toward the remineralized deep waters below the MLD ( Table 1). The highest NC T values in the upper water column of the Macaronesian region were collected in CV-N (~2187-2246 μmol kg -1 ) and CV-SE (~2263-2281 μmol kg -1 ) around 125-150 m depth ( Figure 2C). These NC T values were~100-150 μmol kg -1 greater than the maximums of CA (~2127-2169 μmol kg -1 ; Figure 3C) and MA (2110-2160 μmol kg -1 ; Figure 4C) in deeper areas between 150 and 250 m depth.  (Figures 2-4) collected during POS533. Their spatial variability was found to be strongly linked with differences in the intensity of the coastal upwelling along the African coast, the intense mesoscale activity acting in this region, the influence of different circulation patterns and water masses and the effect of trade winds (e. g. Pelegrıé t al., 2005a;Pelegrı́et al., 2005b;Valdeś and Deńiz-Gonzaĺez, 2015;Pelegrı́and Peña-Izquierdo, 2015a;Pelegrı́and Benazzouz, 2015b;Bonino et al., 2021). These processes drove horizontal and vertical transports which modified the observed temperature, salinity and Chl-a patterns and biogeochemical cycles. Minimal changes of surface fCO 2,sw and pH with latitude given by linear relationships were observed between CV and CA (-0.4 μatm (r 2 = 0.02) and 0.001 units (r 2 = 0.15) per degree of latitude, respectively) despite the strong measured decrease of SST with latitude between both archipelagos (-0.33°C per degree of latitude; r 2 = 0.87) (Supplementary Figures 4C, D). The fCO 2,sw increased with SST by 7.93 μatm°C -1 (r 2 = 0.31) between Cape Blanc and the south of CA along a permanent upwelling area (22.5-28.0°N), while an inverse relationship of -6.04 μatm°C -1 (r 2 = 0.23) was encountered between the north of CV and Cape Blanc along a winter upwelling area (17.0-21.0°N). The highest decrease of fCO 2,sw with SST was observed at 21.5-22.5°N around Cape Blanc (-95.9 μatm°C -1 ; r 2 = 0.58). Latitudinal gradients were stronger from CA to MA (-3.2 μatm (r 2 = 0.49) and 0.003 units (r 2 = 0.49) per degree of latitude, respectively), where the theoretical changes of fCO 2,sw with SST were observed (14.26 μatm°C -1 ; r 2 = 0.35) (Supplementary Figure 4). These findings agreed with previous studies along the African coastal transitional region through the latitudinal ranges of 10-28°N (Gonzaĺez-Davila et al., 2017) and 28-36°N (Curbelo-Hernańdez et al., 2021) and were explained by the seasonal and spatial variability of the Canary Upwelling System (Mittelstaedt, 1991;Cropper et al., 2014).
The maximum fCO 2,sw and minimum pH values encountered around Cape Blanc at~470 km away from the African coast (Supplementary Figures 4C, D) were explained by the highestintense offshore transport of CO 2 -rich upwelled water not compensated by biological uptake (Gonzaĺez-Davila et al., 2017) through the cold and fresh Cape Blanc giant filament (21.5-22.5°N ) (e. g. Bonino et al., 2020;Gabric et al., 1993;Lovecchio et al., 2017). The low intensity of the permanent annual upwelling in winter north of Cape Blanc weakened the injection in the euphotic zone of CO 2 and nutrient-rich water, which explained the increased relevance of SST fluctuations in the fCO 2,sw changes from Cape Blanc to Madeira (Supplementary Figures 4A, C). The highest average changes of fCO 2, sw and pH were explained by thermal processes in this area (-24.64 μatm and 0.023 units, respectively) compared with the area most influenced by non-thermal processes south of Cape Blanc (6.69 μatm and -0.007 units, respectively) (Supplementary Figure 4C). The less intense offshore transport through the Cape Ghir filament (Hagen et al., 1996;Garcıá-Muñoz et al., 2005;Pelegrı́et al., 2005a;Pelegrıé t al., 2005b;Laiz et al., 2012), together with the dominance of respiration over primary production which compensates the decrease in fCO 2,sw drove by the cooling of the water at this time of the year (Curbelo-Hernańdez et al., 2021), explained a lower increase in surface fCO 2,sw encountered at Cape Ghir latitude (30.7°N) at~650 km away from the African coast (Supplementary Figures 4A, C). The strong influence of the Canary Upwelling System through the easternmost parts of CV and CA introduced longitudinal differences in the surface distribution of fCO 2,sw ( Figures 5A.2, B.2) and pH ( Figures 5A.3, B.3). The low fCO 2,T21 values in CV-SE showed that the biological production actively reduced the excess of CO 2 in the upwelled waters. The fCO 2,sw and fCO 2,T21 decreasing eastward with the SST and SSS indicated that both the enhancement of the biological activity and the cooling of the water due to the influence of the African coastal upwelling accounted for the formation of the longitudinal physical and biogeochemical gradient in CV-SE. The minimal longitudinal changes in fCO 2,T21 and pH T21 in CA indicated that the West-to-East decrease of fCO 2,sw and increase in pH was mainly controlled by the cooling of the upwelled waters around the closer islands to the African coast ( Figures 5B.2, B.3). The weaker influence of the coastal upwelling in MA due to its greater distance from the African coast and the low intensity of the offshore transport by the Cape Ghir filament at this time of the year explains the minimal observed longitudinal changes of fCO 2,sw and pH ( Figures 5C.2, C.3).
Several differences in the surface distribution of fCO 2,sw and pH were encountered between CV-N and CV-SE ( Figures 5A.2, A.3) and explained by the influence of the CVFZ, which reaches its southernmost position in winter (Zenk et al., 1991;Peŕez-Rodrıǵuez et al., 2001). The high surface fCO 2,sw values encountered in CV-N can be explained by the influence of cold but low-productive and CO 2 -rich surface waters transported by the NEC. Although the upwelling-driven eastward decrease of SST was observed in CV-N, the fCO 2,sw and fCO 2,T21 values were higher in the eastern part ( Figures 5A.1, A.2). It indicated that the CO 2 enrichment of these waters through the injection of deep and remineralized waters in the closest area to the coastal upwelling was the main factor controlling the fCO 2,sw and pH distributions. In contrast, minimum fCO 2,sw values in the easternmost part of CV-SE around Santiago, Maio and Boa Vista ( Figure 5A.2) indicated that the CO 2 enrichment was weaker and that the fCO 2,sw distribution was controlled by the enhancement of the biological uptake in warmer but higherproductive surface water transported through the NECC and recirculated in the cyclonic Guinea Dome (Siedler et al., 1992;Faye et al., 2015).
The horizontal advection processes can be coupled with divergent/convergent movements of water in mesoscale eddies Sangrà et al., 2009;Cardoso et al., 2020), which introduces local increases/decreases in surface fCO 2,sw , as it was observed leeward of the islands ( Figure 5) and through the CV-CA-MA track (Supplementary Figure 4) Figures 4A, C). The opposite occurred at 18.97°N (374.5 μatm and 21.55°C, respectively). These local changes were explained by the upwelling of cold and CO 2rich deep-water in the core of cyclonic eddies and downwelling of surface waters in anticyclonic eddies, which has been previously detected south of Cape Blanc (e. g. Schütte et al., 2016a;Karstensen et al., 2017;Cardoso et al., 2020). Mesoscale islands-generated eddies were most frequently observed from Cape Blanc to the southern part of CA through The Canary Eddy Corridor (Sangrà et al., 2009) and introduced stronger local differences in the physical and biogeochemical seawater properties which modified the fCO 2,sw and pH through the CV-CA track (Supplementary Figure 4). The highest local increase in surface fCO 2,sw occurred at 23.79°N (Supplementary Figure 4C) and coincided with a major low signal of SST and a high signal of Chl-a associated with a cyclonic eddy (Supplementary Figures 4A, B). The mesoscale variability and its effect on the local changes in surface fCO 2,sw weakened through the CA-MA track, where cyclonic eddies detected by Couvelard et al. (2012) south of Madeira were only observed in minimal decreases in SST and SSS between 31.3 and 32.8°N (Supplementary Figures 4A, C).
The shallowest position of the MLD in the tropical CV area enhanced the vertical mixing processes and deep-water injection in surface layers, which contributed to change strongly the fCO 2,sw and pH in the first 250 m depth (Figures 2B, D). These changes were lower in CA (Figures 3B, D) and MA (Figures 4B, D) due to the deepening of the MLD. Several heterogeneities in the upper-ocean distribution were found between the leeward coastal areas, the inter-island waters and the windward open-ocean waters. The trade winds drove intense vertical mixing processes and raised the MLD in the windward areas of each archipelago, while the suppression of the wind effect in the leeward areas weakened the vertical mixing and decreased the MLD. These differences were observed at inter-islands windward stations compared to CV and CA downwind stations (Figures 2, 3). The effect of trade winds cools the upper water column in the windward areas, while the leeward coast show higher temperature values. However, the increase in surface fCO 2,sw with temperature in leeward coastal upperocean waters was compensated by the enhancement of biological production, as was observed with higher Chl-a values (data not shown).
The high spatial variability of fCO 2,sw in the study area has a great influence on the distribution of ocean CO 2 along the African coastal transitional zone and should be studied on a regional scale. Thus, the fCO 2,sw data during POS533 were compared and compilated with 52328 surface fCO 2,sw winter data in the Macaronesian region available in the Surface Ocean CO 2 Atlas database (SOCAT v2022; Bakker et al., 2016) and 2120 fCO 2,sw upper water-column data (first 250 m depth) in the Northeast Atlantic available in the GLODAP database (GLODAP v2022; Key et al., 2015;Olsen et al., 2016). The location of the sampled stations for SOCAT and GLODAP data is shown in Supplementary Figure 5.
The spatio-temporal changes in surface fCO 2,sw were analysed in CV, CA, MA and for the entire Macaronesian region using POS533 and SOCAT data. These changes were also assessed from 10 to 100 m depth and compared with the changes observed between 100 and 250 m depth using POS533 and GLODAP data. Multiparametric regressions were applied to the compilated datasets (POS533 +GLODAP datasets and POS533+SOCAT datasets) to analyse the distribution of fCO 2,sw based on changes in the seawater physical properties (SST and SSS). The combination of the POS533 with the SOCAT and GLODAP datasets provided a set of equations statistically significant at the 95% level (r value< 0.05) that can be used to estimate the fCO 2,sw in the upper layers of the entire region by considering temperature and salinity fluctuations and both temporal (seasonal and interannual) and longitudinal changes ( Table 2). The interannual increase of surface fCO 2,sw in CA (2.01 ± 0.01 μatm yr -1 ) and MA (2.11 ± 0.01 μatm yr -1 ) was similar to the observed rate at the ESTOC site (1.92 μatm yr -1 ; Bates et al., 2014). The highest interannual increase around CV (2.36 ± 0.02 μatm yr -1 ) coincided with those obtained at the subpolar Irminguer Sea sites (2.37 μatm yr -1 ; Bates et al., 2014), where vertical mixing processes were enhanced and contributed to the increased observed trend. The lower change of surface fCO 2,sw with SST in CV compared to CA and MA ( Table 2) was indicative of the relevance of thermal processes in the change of fCO 2,sw north of Cape Blanc while non-thermal processes (mainly the horizontal and vertical advection of CO 2 -rich water) played a key role in the distribution of surface fCO 2,sw south of Cape Blanc. This set of equations improved the understanding of the carbon cycle along the African coastal transition area in the Northeast Atlantic.

Upper water column distribution of C T components
The shallow ocean C T content was studied leeward of each archipelago by considering the non-conservative factors controlling its vertical distribution. The variations of NC T from preformed NC T values (NC Tpre ) were explained by fluctuations in the organic carbon pump (biological production/respiration processes), the carbonate pump (carbonate dissolution/formation) and the uptake of anthropogenic carbon (Chen and Millero, 1979;Feely et al., 2002;Chou et al., 2007;Andreev et al., 2009) and were described at any given depth by Eq. 4. The results allow to quantify the relative contribution of each term involved in the temporal evolution of NC T from NC T pre values above and below the MLD (up to 250 m depth) ( Figure 6; Table 3). The increase in NC T from NC T pre above the MLD was around 75-90 μmol kg -1 throughout the region. The increment below the MLD was approximately twice in CV than in CA and MA (Table 3). The highest increase with depth of NC T below the MLD occurred in CV-N and CV-SE (1.4 ± 0.8 and 1.9 ± 1.3 μmol kg -1 m -1 , respectively) and was explained by the maximum NC T pre values (2051.1 ± 25.3 and 2050.4 ± 29.2 μmol kg -1 , respectively) and organic carbon pump contribution (52.3% and 57.43%, respectively) encountered throughout the study region. Minor changes of NC T with depth below the MLD were observed in CA-W (0.5 ± 0.3 μmol kg -1 m -1 ), CA-E (0.4 ± 0.2 μmol kg -1 m -1 ) and MA (0.4 ± 0.1 μmol kg -1 m -1 ) due to the lower NC T pre values (2025.1 ± 3.8, 2028.2 ± 3.1 and 2036.3 ± 7.4 μmol kg -1 , respectively) and the weakened of the biological contribution (16.2%, 18.93% and 16.9%, respectively). These differences were mainly explained by spatial heterogeneities in the ventilation and remineralization rates through the Northeast Atlantic ITCZ. The longer residence time of tropical waters south of Cape Blanc originated in the South Atlantic compared with the well-ventilated subtropical waters that originated in the high-latitude of the North Atlantic enhanced remineralization (Pelegrı́and Peña-Izquierdo, 2015a) and led a higher NC T increase in (sub)surface waters around CV compared with CA and MA. The high primary production through the coastal Mauritania-Senegalese upwelling at this time of the year and the upwelling of water in the Guinea Dome also accounted for the increase of remineralization rates south of Cape Blanc (Karstensen et al., 2008;Stramma et al., 2008b). This fact was observed in a slightly higher contribution of the organic carbon pump above the MLD in CV-N and CV-SE (6.11% and 12.75%, respectively) compared with CA-W, CA-E and MA (4.4%, 2.6% and 1.9%, respectively). The change in NC T from NC T pre in the entire region was mainly controlled by the anthropogenic carbon input and storage, which increased the NC T content throughout the region around 72-80 and 68-74 μmol kg -1 above and below the MLD, respectively. The uptake of C ant in surface waters contributed more than 90% of the NC T change above the MLD. The C ant storage below the MLD account for 79.1% in CA-W, 78.2% in CA-E and 80.8% in MA of the NC T increase, while its contribution fell to 47.4% in CV-N and 42.35% in CV-SE. It is related to the weakening of mixed-down processes due to higher vertical stratification and the enhancement of remineralization processes in the ITCZ (Mayer and Weisberg, 1993;Stramma and Schott, 1999). The relevance of the carbonate pump in the change of NC T through the upper layers was considerably lower in the entire region, coinciding with previous observations in the first 500 depths at the ESTOC site (Gonzaĺez-Davila et al., 2010). The dissolution of CaCO 3 in (sub)surface waters increased the NC T by 3.7-4.8% and 1.4-3.1% in CA and MA, respectively. The decrease in CaCO 3 solubility toward warm tropical waters  explained the minimal contribution of NC T carb on the increase of NC T below the MLD in CV-N (0.3%) and CV-SE (0.2%). Negative DNC T carb values above the MLD in CV (Table 3) indicated that the formation processes of CaCO 3 exceed those of dissolution and accounted for the depletion of NC T by 2.9-3.0%.  All the values are given in μmol kg -1 .

C ant inventories
The coastal regions and marginal seas have been found to store more C ant than the open oceans in proportion to their areas (Park et al., 2006;Olsen et al., 2010;Sabine and Tanhua, 2010;Schneider et al., 2010), but have been largely ignored in global ocean studies of the C ant storage due to the limited data available (Khatiwala et al., 2013). This study provides a new analysis of the vertical distribution and inventory of C ant in shallow coastal and inter-islands waters of the Macaronesian region ( Figure 7A). The upper-ocean profiles of C ant were evaluated in CV, CA and MA by applying a smoothing spline with a smoothing parameter p=0.999 to the total of C ant values calculated from POS533 data in each archipelago ( Figure 7A). The smoothing spline provided vertical profiles well-fitted to the calculated C ant values (r 2 >0.98) and with standard error of estimate of 0.12 μmol kg -1 for CV, 0.03 μmol kg -1 for CA and 0.06 μmol kg -1 for MA. The entry of C ant into the ocean through air-sea exchange and the large time scales of mixing processes explained the highest C ant values in near-surface waters and its decrement with depth . The range of variation of C ant in the first 250 depths was lower in CA and MA (between 65 and 83 μmol kg -1 ) compared with CV (40-87 μmol kg -1 ). These differences in the upper water column distribution of C ant were related to differences in the exposition time of waters to the atmosphere and to the Revelle factor (Revelle and Suess, 1957;Takahashi et al., 1993), which value is inversely proportional to the capacity of oceans to uptake atmospheric CO 2 . The lower ventilation rates of the ocean interior in the tropical Northeast Atlantic (Karstensen et al., 2008;Stramma et al., 2008b) and slightly higher (in average) Revelle factors in CV-N and CV-SE (10.77 ± 1.76 and 11.06 ± 2.02, respectively) compared to CA-W, CA-E and MA (10.14 ± 0.23, 10.28 ± 0.42 and 10.34 ± 0.42, respectively) explained the strongest decrease of C ant with depth in CV ( Figure 7A).
The upper water-column inventory of C ant in the sampled stations was calculated by integrating its vertical distribution through the first 250 m depth. The C ant storage in the subsurface waters of CV, CA and MA was 7.57 x 10 3 , 9.26 x 10 3 and 8.86 x 10 3 μmol kg -1 , respectively. The C ant storage downwind of each archipelago was calculated by considering the leeward area that encloses all the stations of CV (8,570 Km 2 ), CA (7,960 Km 2 ) and MA (1,250 Km 2 ) and the neutral density of 26.0 kg m -3 for (sub)surface waters . The total amount of C ant stored in the upper water column leeward of CV, CA and MA was 0.51, 0.58 and 0.09 Tg C (1 Tg = 10 12 g), respectively.
These results were compared with C ant values calculated from 9,870 GLODAP data available for the Macaronesian region during the period 1989-2020. The vertical distribution in subsurface waters of both POS533 and GLODAP C ant values was analysed by averaging the data every 25 m depth ( Figure 7B). The GLODAP C ant values in the first 250 m depth of the Macaronesian region increased on average by 10.47 ± 3.38 μmol kg -1 from 1989-1999 to 2000-2009 and by 8.08 ± 1.68 μmol kg -1 from 2000-2009 to 2010-2020, which were in agreement with those of C T and C ant in the subtropical North Atlantic during the periods 1992-1998 (8.8 ± 0.5 and 8.0 ± 0.2 μmol kg -1 , respectively) and 1998-2004 (8.6 ± 0.5 and 6.8 ± 0.3 μmol kg -1 , respectively) (Brown et al., 2010) and in the ESTOC site between 1995 and 2004 (9.4 and 8.1 μmol kg -1 over a decade in surface and 200 m depth waters, respectively) (Gonzaĺez-Davila et al., 2010). It indicated that the physical and biological conditions could be assumed as constants and that the increase in C T in the Macaronesian region is mainly driven by the rate of C ant uptake by the ocean (Gonzaĺez-Davila et al., 2010).
C ANT = −1993 + 1:03 · year − 0:08 · depth + 1:09 x 10 −4 · depth 2 (11) An area in the subtropical Northeast Atlantic of 5.4 x 10 3 km 2 was considered to calculate the total amount of C ant in the first 250 m depth of the entire Macaronesian region. The C ant inventories in this parcel of water were 0.24 Pg C (1 Pg = 10 15 g) for the period 1989-1999, 0.30 Pg C for the period 2000-2009 and 0.34 Pg C for the period 2010-2020. The C ant inventory estimated for the year 2050 was 0.53 Pg C.

Air-sea CO 2 fluxes in the Macaronesian region
The spatial distribution of FCO 2 was analysed based on changes in the DfCO 2 , the wind speed and the solubility (linked with temperature) to evaluate the air-sea CO 2 exchange in the Macaronesian region during winter (Figures 8, 9). The atmospheric CO 2 uptake was found to dominate over outgassing processes at this time of the year, coinciding with the CO 2 sink behaviour encountered in the Northeast Atlantic during the cold months (Gonzaĺez-Davila et al., 2003;Curbelo-Hernańdez et al., 2021).
The wind forcing has an important role in the variability of the surface physical and biogeochemical properties by controlling the spatio-temporal changes in the intensity of the African coastal upwelling and contributing to the formation of mesoscale eddies. The intense trade winds run the injection of cold deep-water in the surface layers favouring the solubility of the atmospheric CO 2 . It explained the high ingassing rates encountered in areas most exposed to wind stress between Sao Antao and Sao Vicente in CV-N, in interisland waters of CA and southwest of Madeira (Figure 8), where the wind speed reached 15-20 m s -1 . It shows the relevance of the climatology on the air-sea exchange. However, the wind speed is only directly involved in the CO 2 gas transfer and influences the intensity of the ingassing/outgassing processes, while the formation of CO 2 source/sink is controlled by changes in DfCO 2 throughout the Macaronesian region (Figures 8, 9).
The average winter ingassing rate in MA (-8.34 mmol m -2 d -1 ) was twice that of CV (-4.74 mmol m -2 d -1 ) and CA (-3.90 mmol m -2 d -1 ), which could be explained by its colder surface waters favouring CO 2 solubility and by the lower influence of both the coastal upwelling and the arrival of CO 2 -rich surface filaments due to its furthest location from the African coast. However, the leeward side of the MA region showed the strongest spatial variability in the air-sea CO 2 exchange ( Figure 8C) due to the highly heterogeneous interaction between the atmosphere and the surface ocean in this area (Pullen et al., 2017), with FCO 2 ranged between 0 and -28 mmol m -2 d -1 . These spatial differences were weaker south of CA (between 3 and -15 mmol m -2 d -1 ; Figure 8B) and CV (between 0 and -15 mmol m -2 d -1 ; Figure 8A), where the vertical movement of water in the core of the islandinduced eddies influenced the changes of the ingassing/outgassing rates. The upwelling of deep CO 2 -rich water in cyclonic eddies decreases the absolute value of DfCO 2 and weakened the ingassing, while the opposite occurs in anticyclonic eddies through the Canary Eddy Corridor (Sangrà et al., 2009) and downwind of Cape Verde (Cardoso et al., 2020). These deep-water rise movements caused surface fCO 2,sw to exceed fCO 2,atm (the DfCO 2 became positive), which indicate in several locations between CV and Cape Blanc and south of CA that behaved as a CO 2 source. The maximum outgassing was detected at the Cape Blanc latitude (20-25 mmol m -2 d -1 ) ( Figure 9) and was explained by the offshore transport of cold but high CO 2 -saturated waters through the giant Cape Blanc filament (Bonino et al., 2020;Gabric et al., 1993;Lovecchio et al., 2017).
The FCO 2 coastal values during POS533 were compared with FCO 2 calculated for the entire Macaronesian region (13-36°N, 6-30°W) by using winter SOCAT fCO 2,sw data referred to 2019 (an interannual increase of 1.8 μatm per elapsed year was considered; Bates et al., 2014). The fCO 2,atm data used in this calculation were collected at the Izaña Atmospheric Research Center (Tenerife, Canary Islands) and provided by the Agencia Estatal de Meteorologıá (AEMET). The satellite surface ocean 6-hourly and 0.25°grid wind speed data are derived from the Advanced Scatterometer (ASCAT) and provided by the IFREMER-CERSAT. The average FCO 2 and DfCO 2 through the monitored vessel track during POS533 were -4.40 mmol m -2 d -1 and -19.5 μatm, respectively. This ingassing rate was considerably higher than that obtained throughout the entire Macaronesian region based on SOCAT data (-0.53 mmol m -2 d -1 and -5.6 μatm, respectively). It is indicative of a stronger winter CO 2 sink in island shelves compared to open-ocean areas in the Northeast Atlantic (Figures 8, 9). Similar behaviour was determined in previous studies of the air-sea exchange in other coastal regions and continental shelves (e. g. Borges et al., 2005;Borges et al., 2006;Cai et al., 2006;Chen and Borges, 2009;Laruelle et al., 2010;Cai, 2011;Chen et al., 2013;Laruelle et al., 2013;Laruelle et al., 2018). The POS533 and SOCAT databases were compilated to analyse the air-sea CO 2 exchange in each archipelago and in the entire Macaronesian region. The total FCO 2 data available was well-fitted (r 2 = 0.76) to a multiparametric regression statistically significance at 95% level (r value< 0.05) given by Eq. 12 to assess the variability of the air-sea CO 2 exchange and the different factors involved in it. The latitudinal and longitudinal distribution of FCO 2 and the changes in the physical properties of the surface ocean (SST and SSS), in the climatology (wind speed) and in the DfCO 2 were considered. This equation fits the winter air-sea exchange in the Macaronesian region and can be used to estimate the FCO 2 with a standard error of estimate of 1.80 mmol m -2 d -1 .
The winter average FCO 2 for the ocean area of CV (186,000 Km 2 ), CA (145,000 Km 2 ) and MA (40,250 Km 2 ) were -11.42, -12.40 and -4.45 Gg CO 2 d -1 (1 Gg = 10 9 g), respectively. The entire ocean area of the Macaronesian region (4,190,000 Km 2 ) acted as a CO 2 sink during the cold months, with an average FCO 2 of -107.30 Gg CO 2 d -1 . Thus, the winter ingassing rate of the archipelagic ocean waters of CV, CA and MA represented 26.4% of the atmospheric CO 2 absorbed by the ocean in the Macaronesian region despite occupying only 8.9% of its total ocean area. These results highlighted the relevance of coastal regions and inter-island waters on atmospheric CO 2 uptake as well as the importance of their incorporation in regional and global-scale biogeochemical studies. The assessment of the FCO 2 in this study improved the knowledge of the air-sea exchange in the Northeast Atlantic.  Latitudinal changes in FCO 2 , DfCO 2 and wind speed along the CV-CA-MA transect.

Conclusion
The CO 2 cycle was evaluated in (sub)surface coastal transitional, archipelagic and open-ocean waters of the Macaronesian region. The spatial heterogeneity in the intensity of the Canary Upwelling System and the southernmost position of the CVFZ in winter introduced latitudinal differences in the physical and biogeochemical properties. The non-thermal processes were found to have high relevance in tropical waters south of Cape Blanc, where the MLD reaches its shallowest position and the African coastal upwelling shows its maximum intensity at this time of the year. The biological uptake in this region decreased the high surface CO 2 content expected in this warm area and controlled the fCO 2,sw distribution. The opposite occurred north of Cape Blanc where the non-thermal processes do not compensate the northward decrease of fCO 2,sw drove by the cooling of the surface water. The advective processes were found to have high relevance in the spatial distribution of CO 2 through the area of maximum intensity of the coastal upwelling and represent an important source of local variability in the core of mesoscale eddies generated leeward of each archipelago. The strongest injection of deep CO 2 -rich and low-productive waters in surface layers near Cape Blanc and its offshore recirculation through the NEC explained the maximum surface fCO 2,sw and minimum pH values in the entire Macaronesian region. The contribution of these horizontal and vertical transports were lower north of Cape Blanc due to the weakening in the intensity of the coastal upwelling and filaments (Supplementary Figure 4). The vertical mixing processes were also an important source of local variability in the core of mesoscale eddies generated leeward of each archipelago. The high spatial variability of fCO 2,sw in the Macaronesian region was evaluated by compilating the POS533 collected data with the SOCAT and GLODAP databases. A new set of equations was provided in this study to describe the temporal and longitudinal variability of fCO 2,sw based on temperature and salinity fluctuations ( Table 2).
The input and storage of C ant in the (sub)surface archipelagic waters was the main driver of the average increase in NC T from NC T pre accounting for 60.0% in CV-N, 64.2% in CV-SE, 88.2% in CA-W, 84.7% in CA-E and 86.4% in MA. The contribution of the C ant uptake was higher than 90% above the MLD throughout leeward coastal areas of the three archipelagos and higher than 78% below the MLD in CA and MA. The lowest changes in NC T explained by C ant storing below the MLD were found in CV (42.3-47.4%), where the capacity of the ocean interior to sequestrate C ant was lower and the contribution of the organic carbon pump was higher (52.3-57.4%). The maximum average contribution of the organic carbon pump was encountered in CV-N and CV-SE (36.6 and 40.9%, respectively) and explained the high stratification and low ventilation of the tropical waters weakening the mixed down processes and enhancing remineralization. The average contribution of the organic carbon pump fell to 7.5-11.6% toward the low-stratified and well-ventilated subtropical waters around CA and MA. The role of the carbonate pump in the change of NCT was found to be minimal in the entire Macaronesian region (0.8-4.2% on average). The C a nt inventory in coastal waters leeward of the Macaronesian archipelagos was analysed for the first-time in this study and compared with Northeast Atlantic open-ocean data available in the GLODAP database. The important role of coastal areas in the C ant storing was shown in higher average C ant values obtained leeward of the islands based on POS533 data (8.74 x 10 3 μmol kg -1 ) than in open-ocean waters based on GLODAP data between 2010 and 2020 (7.99 x 10 3 μmol kg -1 ). The POS533 and GLODAP databases were compilated to consider the coastal areas in the calculation of the average upper-ocean C ant inventory in an area of 5.4 x 10 3 km 2 in the Macaronesian region (0.34 Pg C). The amount of C ant was described by a new polynomial equation which described its distribution as a function of time and depth changes and that can be used to estimate the C ant values between 10 and 250 m depth in the Macaronesia region with a standard error of estimate of 6.60 μmol kg -1 .
The analysis of the air-sea CO 2 exchange showed that the CO 2 sink behaviour dominated the outgassing processes in the entire study region at this time of the year. The average FCO 2 calculated for the monitored transect based on POS533 data (-4.40 mmol m -2 d -1 ) and for the entire Macaronesian region based on SOCAT data referred to 2019 (-0.53 mmol m -2 d -1 ) showed the higher-intense ingassing rate in island shelves compared to open-ocean areas and proved the need of include coastal regions in air-sea CO 2 exchange evaluations. The winter average FCO 2 calculated for the total ocean area of the three archipelagos (371,250 Km 2 ) was -28.27 Gg CO 2 d -1 , which represented the 26.4% of the winter ingassing rate calculated for the ocean area of the Macaronesian region (4,190,000 Km 2 , -107.30 Gg CO 2 d -1 ).
The analysis of the CO 2 distribution, the C ant inventory and the air-sea CO 2 exchange developed in this investigation allows a better understanding of the role in the climate change of archipelagic waters, coastal regions and island/continental shelves. The high relevance of these areas in the ingassing/outgassing processes, C ant uptake and storing and (sub)surface distribution of the CO 2 system variables of the entire Macaronesian region were evaluated and highlighted the importance of developing new regional scale studies to include these areas in global models. The monitoring and assessment of the CO 2 variability in these biogeochemical active and exposed to humanpressure areas, especially in ultraperipherical populated territories with large areas covered by the ocean such as the archipelagos of the Macaronesia, powerfully contributes to the achievement of the goals indicated in the most recent IPCC reports 2007 and 2021 (IPCC, 2007;.

Data availability statement
The datasets presented in this study can be found in online repositories. The names of the repository/repositories and accession number(s) can be found below: PANGAEA data repository (https:// www.pangaea.de/).

Author contributions
Sampling and CO 2 variables measurements were performed by MG-D and JMS-C during the POS533 cruise (AIMAC project). All authors contributed to the article and approved the submitted version.

Funding
This study received funding from the European Union's Horizon 2020 research and innovation program under grant agreement N°8 20989 (project COMFORT, our common future ocean in the Earth system-quantifying coupled cycles of carbon, oxygen, and nutrients for determining and achieving safe operating spaces with respect to tipping points). The European Commission and their executive agency are not responsible for any use that may be made of the information the work contains. The POS533 cruise was funded by the GEOMAR through the AIMAC project. This work also received partial funding from the European Union INTERREG V-A MAC 2 0 1 4 -2 0 2 0 p r o j e c t s R E S -CO A S T M A C 3 / 3 . 5 b / 3 1 4 a n d PLANCLIMAC MAC2/3.5b/244. The participation of DC-H was funded by the PhD grant PIFULPGC-2020-2 ARTHUM-2.