Seasonal variability of coastal pH and CO 2 using an oceanographic buoy in the Canary Islands

Ocean acidi ﬁ cation, caused by the absorption of carbon dioxide (CO 2 ) from the atmosphere into the ocean, ranks among the most critical consequences of climate change for marine ecosystems. Most studies have examined pH and CO 2 trends in the open ocean through oceanic time-series research. The analysis in coastal waters, particularly in island environments, remains relatively underexplored. This gap in our understanding is particularly important given the profound implications of these changes for coastal ecosystems and the blue economy. The present study focuses on the ongoing monitoring effort that started in March 2020 along the east coast of Gran Canaria, within the Gando Bay, by the CanOA-1 buoy. This monitoring initiative focuses on the systematic collection of multiple variables within the CO 2 system, such as CO 2 fugacity ( f CO 2 ), pH (in total scale, pH T ), total inorganic carbon (C T ), and other hydrographic variables including sea surface salinity (SSS), sea surface temperature (SST) and wind intensity and direction. Accordingly, the study allows the computation of the CO 2 ﬂ ux (FCO 2 ) between the surface waters and the atmosphere. During the study period, stational (warm and cold periods) behavior was found for all the variables. The lowest SST values were recorded in March, with a range of 18.8-19.3°C, while the highest SST were observed in September and October, ranging from 24.5-24.8°C. SST exhibited an annual increase with a rate of 0.007°C yr -1 . Warmer months increased SSS, while colder periods, in ﬂ uenced by extreme events like tropical storms, led to lower salinity (SSS=34.02). The predominant Trade Winds facilitated the arrival of deeper water, replenishing seawater. The study provided insights into atmospheric CO 2 . Atmospheric f CO 2 averaged 415 ± 4 µatm (2020-2023). Surface water f CO 2sw presented variability, with the highest values recorded in September and October, peaking at 437 µatm in September 2021. The lowest values for f CO 2sw were found in February 2021 (368 µatm). From 2020 to 2023, surface water f CO 2sw values displayed an increasing rate of 1.9 µatm yr -1 in the study area. The assessment of f CO 2sw decomposition into thermal and non-thermal processes revealed the importance of SST on the f CO 2sw . Nevertheless, in the present study, it is crucial to remark the impact of non-thermal factors on near-shallow coastal regions. Our ﬁ ndings highlight the in ﬂ uence of physical factors such as tides, and wind effect to horizontal mixing in these areas. The C T showed a mean concentration of 2113 ± 8 m mol kg -1 and pH at in-situ temperature (pH T,


Introduction
Over the past two centuries, there has been an exponential increase in atmospheric CO 2 concentrations as a result of anthropogenic activities (Denman et al., 2007;Takahashi et al., 2009;Lynas et al., 2021;Friedlingstein et al., 2022), also indicated in the 6 th IPCC Report (IPCC, 2022; IPCC is the Intergovernmental Panel on Climate Change).A substantial portion of this anthropogenic CO 2 is directly transferred to the ocean, accounting for about 26% of the total anthropogenic CO 2 emissions (Friedlingstein et al., 2022).The ocean's capacity to absorb CO 2 , has exhibited an increase from 1.0 ± 0.3 gigatons of carbon per year (Gt C yr -1 ) in 1960 to 2.5 ± 0.6 Gt C yr -1 in the 2010 to 2019 period (Friedlingstein et al., 2020).This transfer of CO 2 from the atmosphere to the ocean has profound repercussions on the marine chemistry and ecosystems (Wollast and Mackenzie, 1989;Walsh, 1991;Falkowski and Wilson, 1992), for example influencing the potential acidification of coastal marine waters (Borges and Gypensb, 2010;Wallace et al., 2014;Carstensen and Duarte, 2019).The oceanic pH has decreased by 0.1 units since the onset of the Industrial Revolution, representing a 26% increase in ocean acidification over the past two centuries (Doney et al., 2009).Projections suggest that the global CO 2 concentration will increase by more than 500 parts per million (ppm) by the end of this century, leading to a pH decrease of 0.4 units from the preindustrial values (Orr et al., 2005;Jiang et al., 2023).
To comprehend the evolution of any variable, such as temperature, atmospheric and oceanic CO 2 , pH, sea level, etc., and their relationship to climate change, the establishment of long-term time series is essential.It is widely acknowledged that observing stations, particularly fixed stations, constitute the most reliable data source for investigating and estimating CO 2 fluxes between the atmosphere and the ocean (Takahashi et al., 2014;Bates and Johnson, 2020;Skjelvan et al., 2022).An important development in this regard is the Global Ocean Acidification Network (GOA-ON; http://www.goa-on.org/),which aims to coordinate, promote, and sustain long-term observations of the carbonate system at both local and national scales.Measurements of the CO 2 system have predominantly focused on open waters, while coastal regions are underrepresented in the Global Carbon Budget (Friedlingstein et al., 2022) due to limited observational data, insufficient high-frequency monitoring, and the complexity of modelling these diverse environments (Takahashi et al., 2002;Gonzaĺez Davila et al., 2005;Gonzaĺez-Davila et al., 2007;Santana-Casiano et al., 2007;Bates, 2012;Bates et al., 2014;Gonzaĺez-Davila and Santana-Casiano, 2023).
While these CO 2 trends have been studied in the open ocean, there is a lack of extensive information on coastal zones, which, despite covering only 7-10% of the total ocean surface area and less than 0.5% of the ocean volume (Laruelle et al., 2013), serve as a critical interface between land, atmosphere, and ocean (Bauer et al., 2013).Coastal zones concentrate up to 30% of the primary production and organic matter remineralization in coastal shelf areas (Walsh et al., 1988;de Haas et al., 2002;Bauer et al., 2013).Consequently, they exhibit high uptake and release of dissolved inorganic carbon and partial pressure of CO 2 (pCO 2 or fCO 2 ) (Thomas et al., 2005).
The behavior of coastal zones with respect to CO 2 exchange is complex and depends on several factors (Walsh and Dieterle, 1994;Chen, 2004;Borges, 2005;Borges et al., 2005Borges et al., , 2006;;Cai et al., 2006;McNeil, 2010;Shaw and McNeil, 2014;Terlouw et al., 2019;Gac et al., 2020).Existing studies highlight the need for long-term coastal time series data, as many estimates have extrapolated values from specific coastal regions to a global scale (Borges, 2005;Borges et al., 2005;Cai et al., 2006;Chen et al., 2013).There are latitudinal variations in coastal regions, with mid-and high-latitude shelf systems generally functioning as net CO 2 sinks (-0.33 Pg C yr -1 ), while low-latitude shelf systems tend to act as net CO 2 sources (0.11 Pg C yr -1 ) (Borges et al., 2005;Cai et al., 2006;Chen et al., 2013).In broad terms, the global continental shelves exhibit a net CO 2 uptake, with estimates ranging from approximately -0.25 (Cai, 2011) to -0.4 Pg C yr -1 (Chen et al., 2013).Shallow near-shore coastal areas including estuaries, salt marshes, coral reefs, coastal upwelling systems, and mangroves, act as sources of CO 2 to the atmosphere (Bouillon et al., 2008;Chen and Borges, 2009), with estuaries being the major contributors to this ocean-atmosphere CO 2 flux.Coastal ecosystems are particularly characterized by substantial and variable inputs of nutrients discharged by rivers.These inputs trigger strong seasonal and interannual variability in the carbonate system (Gypens et al., 2009(Gypens et al., , 2011)).
Further research is needed to acquire additional CO 2 data for scaling air-water CO 2 fluxes in outer estuaries, which may exert a substantial influence on the overall flux of estuarine systems (Borges and Frankignoulle, 2002;Borges, 2005).The same is true for nearshallow coastal areas on islands, where data is lacking and where the CO 2 system is intricately linked to biological activities, physical processes, wind regimes, precipitation patterns and the significant input of nutrients and carbon from the land via rivers and runoff.Moreover, the coastal ocean, which extends from the open ocean to the continental margins, is one of the most biogeochemically active domains within the biosphere (Gattuso et al., 1998).
Despite all the previously reported studies, information on CO 2 monitoring in islands are scarce.This study represents the first scientific effort dedicated to monitoring the CO 2 system within coastal areas of the Canary Islands, employing a time series approach.While global studies on coastal areas exist, islands such as the Canary Islands offer natural laboratories conducive to monitoring the transfer of CO 2 between the atmosphere and the ocean.Moreover, according to the 6 th IPCC report, islands are one of the most vulnerable regions to the impact of climate change.The main objective of this study was to quantify variations in CO 2 fugacity (fCO 2 ), pH (at total scale), Total Inorganic Carbon (C T ), and atmosphere-ocean CO 2 flux (FCO 2 ) in Gando Bay, a coastal region located to the east of the island of Gran Canaria.This study covers the first three years of observations, with a particular focus on elucidating the diverse processes governing the atmosphereocean CO 2 transfer and studying the first seasonal variability of the CO 2 system in the coastal waters of the Gando Bay to have a preliminary trend of each variable.

Study area
The CanOA-1 buoy is located within the Canary Islands, on the eastern side of the island of Gran Canaria (Figure 1), off the northwest coast of Africa and in shallow coastal waters (27.930°N ; 15.365°W; at 12 m depth), within a military area that prevents vandalism with controlled access.This geographical location places the islands between two important oceanic features: the African upwelling to the east and the oligotrophic waters of the North Atlantic Subtropical Gyre.The Canary Islands are also influenced by the Canary Current, which delineates the eastern boundary of the subtropical gyre (Knoll et al., 2002).The prevailing winds are the Trade Winds.

Data collection
The CanOA-1 buoy structure is equipped with an array of seven sensors, including sea surface temperature and salinity (SST and SSS respectively; SBE 37-SI/SIP Thermosalinometer MicroCAT sensor manufactured by Sea-Bird Scientific -accuracy for SST is ±0.002°C and for SSS is ±0.01 units), fluorescence (Cyclops-7F from Turner Designs with a detection limit of 0.03 μg L -1 ), dissolved oxygen concentration (Optode 4835 Oxymeter, manufactured by Aanderaa with an accuracy<0.1 μmol L -1 ), photometric pH (SAMI-pH meter from Sunburst, precision<0.003ppm and accuracy ±0.01 units), pCO 2 (partial pressure of CO 2 , here expressed as fugacity of CO 2 -fCO 2 ; measured with a CO 2 -pro CV sensor from PrOceanus -precision ±0.01 ppm and accuracy ±0.5%, using a Non-Dispersive Infrared Detector -NDIR) operating on a three-hour schedule, measuring the molar fraction of CO 2 (xCO 2 ) and converted internally to pCO 2 in seawater using a CO 2 -permeable membrane.An external Sea-Bird Scientific pump (SBE-5) supplies seawater from outside the buoy body (60 cm), including copper-intake tubing to reduce biofouling effects.An internal zero determination is made every 24 hours to eliminate any signal drift.The pCO 2 was also measured using a Battelle system (model 635108H1010), which assesses the xCO 2 in both seawater and the atmosphere every three hours using equilibration-CO 2 Infrared detection.Atmospheric xCO 2 data from the Battelle system at 2.5 m above sea level, were calculated to 10 m (Hsu et al., 1994) and compared with those obtained at the ICOS Izaña Atmospheric Research Station (Tenerife, Canary Islands) and provided by the Agencia Estatal de Meteorologıá (AEMET).The agreement was better than ± 3 ppm.After the first year of work, the Batelle system had to be repaired so the Izaña data was used.Various meteorological variables (wind speed, wind direction, air temperature, humidity, atmospheric pressure, precipitation, solar radiation, and GPS coordinates -Gill MaxiMet GMX 501 GPS) were measured.All the sensors were installed at a depth of 1.5 m depth except for the meteorological station that was located 2.5 m above sea level.All data are free and the last 2000 data are available in real-time on the free Telegram app under Boya Morgan (@QUIMAbot).
Despite all the sensors installed on the buoy, the pH, chlorophyll and oxygen data have not been used due to their low stability and biofouling problems.In this sense, the pH sensor showed high variability and the pH data used in the manuscript were computed from total alkalinity to salinity relationship.
The buoy was visited every 2-3 months for inspection and maintenance.In addition, to determine the sensitivity and accuracy of the sensors, 23 surface water samples (with duplicates) were collected throughout the observation period.These samples were analyzed in the laboratory for total alkalinity (A T ), total dissolved inorganic carbon (C T ), and oxygen concentration.A T and C T were determined using the VINDTA 3C system (Mintrop et al., 2000) with Certified Reference Material (CRM) from batches 108, 122, 163, 177 and 196, provided by A. Dickson (Scripps Institute of Oceanography, University of South California, San Diego, United States) with allowed accuracy of ±1.5 mmol kg -1 for both C T and A T .Oxygen was measured using the Winkler method (Graneĺi and Graneĺi, 1991).These laboratory measurements were subsequently used to verify the response of the various sensors (see below) and to establish relationships with continuous salinity data.Additionally, each measurement of CO 2 obtained by the Battelle sensor was calibrated every 3-hours prior to analysis using a zero and an external CO 2 gas cylinder with a known concentration of 553.35 ± 0.02 ppm traceable to the World Meteorological Organization.

Data treatment
The Battelle sensor provides xCO 2 values while the PrOceanus sensors provide pCO 2 .The xCO 2 data were converted to pCO 2 (pCO 2,equ ; Dickson et al., 2007) (Equation 1).The pCO 2,equ is the partial pressure of CO 2 in the equilibrator, P atm (atm) is the atmospheric pressure and the expression for water vapor (pH 2 O) is given in Equation 2. SST (K) is the sea surface temperature and SSS is the measured salinity.Once pCO 2 was obtained, the fCO 2 was calculated (Equation 3).The coefficients B CO2 (cm 3 mol -1 ) and d CO2 are given by Equations 4 and 5, respectively.(2) To determine the CO 2 flux (FCO 2 ) between the atmosphere and the sea surface water, Equations 6 and 7 were used: where 0.24 is a conversion factor to have the flux in mmol m -2 d -1 , S is the CO 2 solubility in mol dm -3 atm -1 (Weiss, 1970), and DfCO 2 is the difference between fCO 2 in seawater and atmosphere (fCO 2 SW -fCO 2 atm ).k is the gas transfer velocity given by Wanninkhof (2014), where w is the wind velocity in m s -1 (at 10 m height), Sc is the Schmidt number which considers the kinematic viscosity of seawater divided by the gas diffusion coefficient (Wanninkhof, 2014).The FCO 2 flux depends on the difference between fCO 2 in the seawater and the atmosphere, the temperature, and the wind speed.If the flux is negative, the ocean acts as a sink, and if it is positive, it acts as a source.Wind speeds were averaged from two hours before and two hours after each study point.
The fCO 2Tmean was calculated at the approximate annual mean temperature (21°C) (Takahashi, 1993) to obtain the temperatureindependent fCO 2 (Equation 8).Hence, the annual non-thermal effect is obtained by the difference of the minimum and the maximum values obtained in Equation 9.The same process was followed to know the thermal effect (Equation 10) on the average observed fCO 2 .The annual thermal effect was determined by the difference of the minimum and maximum values (Equation 11).
The thermal to non-thermal ratio (T/NT) shows the importance of both (physical and biological) effects.It was calculated by dividing the terms in Equations 9 and 11 (D(fCO 2 ) therm =D(fCO 2 ) non−therm indicating that when the ratio is greater than 1, the temperature effect dominates over the other effects.
Total dissolved inorganic carbon (C T ), pH at in situ temperature (pH T,IS ) and normalized to a mean temperature of 21°C (pH T,T=21°C ) were calculated with the Excel program CO 2 sys (Pierrot et al., 2021) using total alkalinity (computed from salinity) and the measured pCO 2 in seawater.The carbonic acid dissociation constants of Lueker et al. (2000), the HSO 4 -dissociation constant of Dickson (1990) and the value of [B] I determined by Lee et al. (2010) were used.Alkalinity concentrations used in the calculations (data not shown) were obtained from in situ samples (n=23, collected every 2-3 months) and normalized to a salinity (SSS) of 35 (NA T = A T / SSS•35).A constant value for NA T = 2292.3± 2.8 mmol kg -1 was obtained (similar to that obtained at the ESTOC site, located 60 miles north of the buoy site, Gonzaĺez- Davila et al., 2010), confirming that alkalinity is controlled by salinity variability and is not affected by atmospheric CO 2 increase or spring-summer primary productivity.In addition, since the SAMI sensor failed due to bubbles in the tubing, the pH was calculated in the total scale (pH T ) with alkalinity determined from salinity and pCO 2 variables.
The A T -C T pair of discrete data were used to test pCO 2 sensor values (23 pairs) and other carbonate system variables.The computed pCO 2 values and those provided by the sensor for the same day and time of the day were within ± 6 matm.Calculated pH T values from the A T -C T discrete values pair and those from A T from salinity and sensor pCO 2 data were within ± 0.01 pH units.Moreover, discrete C T concentrations (n = 23) and those determined from A T from salinity and sensor pCO 2 data were within ± 3 mmol kg -1 .
The surface water displacement was calculated using the Ekman Equations 12 and 13, which include variables such as DE (Ekman depth), z (depth of interest, in this case, 8 meters), f (Coriolis parameter), and Az (turbulent viscosity coefficient).The solution to this equation yields a displacement angle of 45°when z is 0, meaning that the surface current flows at an angle of 45°to the right of the wind direction (Pond and Pickard, 1983).This calculated value was then added to the measured wind direction for further analysis.

Seasonal detrending of data
The trend analysis performed on the observed data includes inherent seasonal variability, which is influenced by sampling irregularities throughout the study period.To mitigate this seasonality, a seasonal detrending approach was implemented, in line with methodologies used in other time series analyses (e.g., Bates, 2012;Gonzaĺez-Davila and Santana-Casiano, 2023).The data were organized into corresponding monthly bins, spanning the time series from 2020 to 2023.Within each month, the mean and standard deviation were computed.Anomalies were then determined by subtracting the monthly mean from each data point within the dataset.This procedure effectively mitigated the temporal non-uniformity present in the data.Furthermore, a harmonic fitting technique was applied, similar to methods previously used in studies such as those conducted at the ESTOC site (Gonzaĺez- Davila et al., 2010).This fitting allows the determination, in a single step, of both seasonal effects (terms c, d, e, f) and interannual trends (b, Equation 14) as a function of time for the variable considered (y), expressed as an annual fraction (x).The results of this harmonic fitting analysis gave trends that closely aligned with those obtained within the estimated error margins for each considered parameter.3 Results

Hydrographic variability
The study area is characterized by a seasonal amplitude of SST as depicted in Figure 2A, with a range of approximately 5-6°C, fluctuating between 19°C and 24.8°C.The lowest SST were recorded in March, ranging from 18.8 -19.3°C, while the highest SST were observed in September-October, ranging from 24.5 -24.8°C.It is noteworthy that the minimum SST was registered in the year 2022, with SST reaching 18.8°C.During the period from 2020 to 2023, the average SST was 21.2 ± 1.6°C.Even if three years of data are not enough to obtain a significant trend analysis, it is observed that the surface waters in the study area exhibit a warming trend at a rate of 0.007°C yr -1 (Table 1; SupplementaryFigure SI-1), a value that is consistent with solid trends shown by 25 years of ESTOC data (i.e.Gonzaĺez- Davila and Santana-Casiano, 2023).It is worth noting that there is a seasonal shift in SST, with the warmest temperatures typically occurring in July and August, comparable to the usually warmer months of September and October (Curbelo-Hernańdez et al., 2021;Gonzaĺez-Davila and Santana-Casiano, 2023), indicating that factors other than warming are acting on these coastal waters.
The SSS exhibited distinct characteristics during the 2020-2023 period (Figure 2B).The SSS reached its maximum salinities, ranging from 36.92 to 37.04, during the months of September and October.At certain time points, the SSS dropped to as low as 34 (e.g., SSS = 34.02 in September 2022), coinciding with periods of heavy rainfall (Figure 2B).The mean SSS observed throughout the study period was 36.71 ± 0.14.It was also observed that coinciding with the dominance of the Trade Winds from May to September (Figure 2B, green data, and Figure 2C), when the wind speed was above 7 m s -1 , low anomalous salinities were registered with respect to those described by the harmonic fit.The salinity has increased by 0.02 ± 0.001 during the observed period (Table 1; Figure 2B; Supplementary Figure SI-1).

Seasonality of the carbon system
Figure 3A shows that both CO 2 sensors provided highly consistent data (linear relationship with r 2 = 0.962, where the root mean square deviation (RMSD) was 2.9 μatm, n = 4600).The temporal overview of the fCO 2 in both the atmosphere and surface water within the study area is shown in Figure 3A.The average atmospheric fCO 2,atm was 415 ± 4 μatm.Surface water fCO 2sw showed variability, with the highest values occurring in September and October, reaching a maximum of 444 μatm (in September 2021).Conversely, the lowest fCO 2sw values were observed during the coldest months, particularly in February and March, when they hovered around 368 μatm.Consequently, a more pronounced seasonal fCO 2sw amplitude of about 55 -60 μatm, was observed in the surface water compared to the atmosphere (about 12 μatm).Remarkably, in March 2023, the fCO 2sw values did not reach the lowest values observed in previous years, only decreasing to 384 μatm.This trend suggests that the surface water is undersaturated with CO 2 relative to the atmosphere in cold months, whereas it becomes oversaturated in warm months.As Sea Surface Temperature (SST, °C) (A), Sea Surface Salinity, rain (mm) and wind speed (m s -1 ) for the stronger winds > 7 m s -1 (B), and all the wind data [(C), m s -1 ] recorded at the CanOA-1 site.The lines in (A) and (B) correspond to the harmonic fit of the observed data.
mentioned above, three years of data provide only a first estimate of any trend, including the fact that more local events are acting on coastal areas than in open ocean waters making more complex the calculation of a definitive trend.Nevertheless, the fCO 2sw values for the period 2020-2023 increase with an annual rate of 1.9 ± 0.1 μatm yr -1 (Table 1) in the study area, considering both the harmonic fitting (Equation 14; Figure 3A; Table 1) and the detrended calculation (Supplementary Figure SI-1).This value is similar to that observed at the ESTOC site (Gonzaĺez-Davila and Santana-Casiano, 2023), which supports the values observed during these three years in Gando Bay.
Figure 3B shows the decomposition of fCO 2sw together with the observed values (black dots) to assess the influence of thermal and non-thermal processes.In this study area, SST appears to control fCO 2sw , although the contributions of other physical mixing processes and biological factors should not be overlooked.
Figure 3C shows C T measured with discrete samples and the C T estimated from fCO 2 and A T derived from SSS, in the study area.Concentrations decreased from colder to warmer months, with an average concentration of 2113 ± 8 mmol kg -1 (Figure 3C).Maximum C T typically occurred from mid-March to April, averaging 2123 ± 7 mmol kg -1 .Minimum concentrations were observed at the end of October, with an average of 2101 ± 3 mmol kg -1 .These observations suggest a seasonal amplitude of about 20 mmol kg -1 .During the observation period, C T showed an increase of 2.2 ± 0.2 mmol kg -1 yr -1 (Table 1).However, when the C T data were normalized to a constant salinity of 36.8 (NC T = C T /SSS•36.8)(Figure 3C), which removes precipitation and evaporation effects, the rate of increase was reduced to 1.0 ± 0.1 mmol kg -1 yr -1 , a value similar to the oceanic ESTOC site (Gonzaĺez- Davila and Santana-Casiano, 2023).
The computed pH measurements, both in situ (pH T,IS ) and calculated at a constant temperature of 21°C (pH T,21 ), are shown in Evolution of fCO 2 in the atmosphere and ocean (A), the components of the fCO 2sw , considering the thermal (fCO 2,therm ) and non-thermal processes (fCO 2,non−therm ) (B), the total inorganic carbon (C T ) (estimated in red and measured in black) and seawater salinity-normalized inorganic carbon (NC T ) at SSS= 36.8 (C), and the pH at total scale, both at in situ temperature (pH T,IS ) and at constant temperature of 21°C (pH T,21 ) (D), in the CanOA-1 site.
The lines correspond to the harmonic fit of the data.  .The pH T,IS displayed a decreasing trend from winter to summer, with a mean value of 8.05 ± 0.02, for the studied period.The maximum pH T,IS was typically recorded between February and March, with a mean of 8.07 ± 0.01, while minimum pH T,IS occurred between September and October, averaging 8.03 ± 0.01.This variation represents a decrease of about 0.04 units from winter to summer.Characterizing a rate of change in a variable such as pH will need an extended time series of data.However, when detrended pH data at ESTOC site (period 1995-2023, Gonzalez-Davila and Santana-Casiano, 2023) and those at the CanOA-1 site are plotted together (Figure 4), it is shown the coastal pH values are following the same pH trend than that at oceanic waters, this last one decreasing at 0.002 ± 0.0002 units yr -1 (Figure 4; Supplementary Figure SI-1).It is also consistent with other oceanic carbon time series (Bates et al., 2014).
Examination of the variability of fCO 2sw and the corresponding CO 2 flux (Figure 5) revealed periods of oversaturation (higher levels of fCO 2sw than those in fCO 2atm ) from May and June to November and December, approximately.During these months, the coastal zone acted as a source of CO 2 , releasing it from the ocean into the atmosphere.Conversely, during the rest of the year, the coastal zone acted as a carbon sink.It is noteworthy that on average DfCO 2 is lower during the period June to November than during the period December to June (Figure 3A).However, the highest FCO 2 coincided with the June to November period, indicating the effect of the increased wind intensity.In this region, the prevailing Trade Winds exhibit their maximum strength in July and August, with wind speeds reaching 16 m s -1 (Figure 2C).During this period, it is common for wind speeds to consistently exceed 7 m s -1 each year (Santana-Casiano et al., 2007).
Year 2021 acted as a slight sink of 79 mmol m -2 yr -1 , related to lower SST in the area during winter months (Figure 2A).When extrapolated to the entire 6 km 2 area of Gando Bay (Figure 1), the region acted as a slight CO 2 source for the entire 2020-2022 period, quantified at 33 ± 3 tons of CO 2 yr -1 .

Discussion
Long-term time series data of CO 2 variables in coastal ecosystems are a rare but essential resource for assessing their response to climate change.In particular, time-series data that include at least two carbonate system variables are essential for understanding the underlying processes governing observed trends.In regions close to the coastline, such as the studied area, which has a depth of approximately 10 meters, variables such as wind intensity and direction, SST, and primary production exert certain influence.
In the context of this specific study, it is evident that forcing factors such as strong winds and precipitation events, such as tropical storms Theta (November 2020), Filomena (January 2021), and Hermine (September 2022), have a discernible impact on sea surface salinity (SSS, Figure 2B).Wu et al. (2021) previously demonstrated the influence of extreme events, such as typhoons, on the net atmosphere-ocean CO 2 exchange in the East China Sea, due to the water mixing and biological drawdown.In addition, the increase in SST, and also the atmospheric temperature due to extreme events such as heat waves occurring in the Canary Islands (Suaŕez-Molina and Sanz, 2022) also affects the CO 2 system through the dependence of fCO 2 on SST.In this sense, the Canary Islands have experienced several heat waves (AEMET, 2023): three in 2021 (11 days, 5 in August and 3 + 3 in September), two in 2022 (3 + 3 in July) and 26 days in 2023 (5 + 5 days in August and 16 days in October).
The study area, a coastal region with a natural barrier in the form of Mountain Gando (Figure 1), approximately 100 meters in height (IDE Canarias -https://visor.grafcan.es/),experiences fluctuations in wind intensity and direction, particularly within a few meters (approximately 50 meters) above the surface.These wind-related variations influence the water circulation patterns within the Gando Bay, resulting in lower SSS (Figure 2B) during periods of strong winds or winds blowing from certain directions.In this regard, a comprehensive analysis of the prevailing winds in the region and the surface water direction (Supplementary Figure SI-2) shows maximum wind speeds of 15 m s -1 and an average of 8.2 ± 2.1 m s -1 throughout the study period.The Trade Winds, which blow from the northeast direction (Van Camp et al., 1991), shift to the north-northwest in the study area due to the sheltering effect of the prominent Gando Mountain, forming a small peninsula that changes the wind direction (Supplementary Figure SI-2).These prevailing winds lead to a surface water column displacement of up to 8 meters in the W-SW direction at times.The calculation of this surface water displacement was achieved using the Ekman Equations (see experimental section).
Consistent findings, as established by Pond and Pickard (1983), confirm that surface currents in the Northern Hemisphere flow at an angle of 45°to the right of the prevailing wind direction.This displacement of surface water instigates the upward movement of deep water to compensate for the lost volume, a phenomenon elucidated by Kämpf and Chapman (2016).The replenishing seawater from deeper areas outside of the bay (about 150 m water depth), exhibits specific characteristics, with an average temperature of 22°C, a mean salinity of 36.4,a substantial C T concentration of 2117 mmol kg -1 , and a mean pH T,IS of 8.03 (Curbelo-Hernańdez et al., 2023).Notably, these values differ from the typical surface conditions in the bay during June-August, which include SST of 23.5°C, SSS reaching up to 36.7,C T concentrations of 2105 mmol kg -1 , and a pH T, IS of 8.04.In this sense, it was observed every year that there was no increase in SST from June to September, but the SST was relatively constant or even decreased by July-August consisting with the strongest predominant Trade winds blowing in the area that favored the entrance of deeper water from outside of the bay.Depending on the wind strength and the moment when that force is exerted, it is observed that the increase in SST does not consistently rise but rather slows down due to the arrival of colder water and increased surface mixing.During these periods of decreasing SST, an increase in SSS variability was observed, also related to the arrival of deeper water with lower salinity and pH and higher C T (Figure 2B).
Efforts were made to discern any potential correlation with tidal intensity throughout the lunar period (SupplementaryFigure SI-3).Although some instances of alignment between full and new moon phases and variations in SSS were identified, these occurrences lacked temporal consistency.Consequently, while tidal effects may exert some influence, they do not appear to constitute a primary or determinative factor within this particular environment.Nonetheless, it is worth noting that tidal height plays a crucial role in carbon exchange within estuaries and river mouth regions, as highlighted in previous studies (Ortega et al., 2005;Bauer et al., 2013).
The hydrographic features of Gando Bay are manifested in the seasonal and interannual variability of the CO 2 variables.The interannual increase of fCO 2sw , quantified at 1.9 μatm yr -1 (Table 1; Figures 3A, B), considering only three years of data is close to the observed value at the oceanic station ESTOC of 2.1 ± 0.1 μatm yr -1 (Gonzaĺez- Davila and Santana-Casiano, 2023), indicating the important control of the increased atmospheric CO 2 concentrations in the seawater concentration.Further years of observation are required to confirm this trend in Gando Bay.During the year, thermodynamic effects control the observed variability (Figure 3B) with a T/NT ratio of 2.0 ± 0.1, but other physical and biological effects should not be ignored.From February to November (Figure 3B, black dots), the increase in SST led to an increase in measured fCO 2sw (a slope of 10 ± 0.2 matm °C-1 was calculated).According to Takahashi (1993), the theoretical change should be 17 matm °C-1 (indicated by the red dots in Figure 3B).The observed lower slope in Gando Bay includes both the effects of the arrival of deeper, less saline water with highly variable fCO 2sw content and changes in the productivity of the area (blue dots in Figure 3B).Oxygen data from the sensor (data not shown) were strongly affected by biofouling during the early years of study until a flow system with a copper intake located far from the buoy body was included.However, there is not enough oxygen data to decompose the nonthermal component between biological and water mixing processes.Moreover, there is a seagrass bed (locally known as sebadales) in the vicinity of the buoy which is known for its robust biological activity characteristics (Duarte and Krause-Jensen, 2017;Serrano et al., 2021).As mentioned above, the study area is impacted by the influence of the Trade Winds (wind direction and intensity), and the presence of the nearby mountain, all of which contribute to its unique characteristics, especially during the summer months period.The physical factors should also consider the influence of tides, wind, and horizontal mixing (Xue et al., 2016) on the impact of mixing processes between coastal and open ocean waters, mainly driven by horizontal advection, leading to changes in SSS and productivity, a phenomenon that should also affect the present study.
This outcome differs from the prevailing pattern in coastal regions, where changes in C T due to non-thermal forcings are expected to dominate, especially in mid-latitudes (Cao et al., 2020;Torres et al., 2021).This unexpected result can be attributed to the shallow nature and low productivity of the coastal area studied (Arıśtegui et al., 2001), which differs from the conditions reported by these authors.Moreover, in the northern region of Gran Canaria (station 16.5°W), research by Curbelo-Hernańdez et al. ( 2021) revealed a T/NT ratio of 2.1, comparable to the results obtained in Gando Bay.Therefore, it can be inferred that in Gando Bay, despite the impact of non-thermal processes including encompassing biological activity and advective mixing, the thermal component controls the observed seasonality in the fCO 2sw .These climatological results parallel findings from other coastal studies, such as those conducted in Hawaii and Australian regions, where temperature dominantly controls fCO 2sw (Shaw and McNeil, 2014;Terlouw et al., 2019).In the northern coastal regions of the Atlantic Ocean, non-thermal processes control the fCO 2sw (Gac et al., 2020).
According to the observations, the decrease in pH T,IS at the CanOA-1 site followed the same behavior as that at the oceanic ESTOC site, (Figure 4), and was comparable to other coastal areas such as SOMLIT-Brest, where pH decreased by -0.0026 ± 0.0004 yr -1 .The observed diurnal variation is linked with the diel biological cycle, such as in the Bay of Brest and the tidal cycles in Roscoff (Gac et al., 2020(Gac et al., , 2021)), and was of a similar magnitude to the seasonal variability.These results are consistent with previous studies of coastal seas in NW Europe that estimated ocean acidification based on seasonal cruises or voluntary observing ship surveys (Clargo et al., 2015;Ostle et al., 2016;Omar et al., 2019).Ocean acidification rates in North Sea surface waters ranged from -0.0022 yr -1 (period 2001-2011;Clargo et al., 2015) to -0.0035 yr -1 (period 1984-2014;Ostle et al., 2016), with a recent estimate of -0.0024 yr -1 in the northern North Sea (Omar et al., 2019).When the temperature effect is eliminated, pH T,21 shows an average of 8.05 ± 0.02 and follows an inverse pattern compared to pH T,IS .pH T,21 increases from February to September (mean of 8.07 ± 0.01) related to the increase of biological activity and decreases from September to February (mean of 8.03 ± 0.01) due to vertical mixing with deeper seawater from out of the bay and possible with higher nutrient concentrations.Unfortunately, we did not measure the biological activity, the nutrient concentration in the buoy location and oxygen data are not enough accurate to estimate this component.Despite its status as a coastal zone, the biodiversity in the area appears to be insufficient to absorb excess atmospheric CO 2 due to the results of T/NT ratio.Consequently, this leads to acidification levels in the region that are similar to those observed at the ESTOC oceanic station in the Northeast Atlantic, characterized by an interannual variability of -0.002 pH units yr -1 (Santana-Casiano et al., 2007;Gonzaĺez-Davila et al., 2010;Gonzaĺez-Davila and Santana-Casiano, 2023).
The recorded C T with an average of 2113 ± 8 mmol kg -1 (for the study period) were close to the observed concentrations at the ESTOC station for the years 1995-2020 with an average of 2109.5 ± 9.6 mmol kg -1 .These concentrations converge if the observed annual C T increase at ESTOC of 1.09 ± 0.10 mmol kg -1 yr -1 is applied (Bates et al., 2014).However, the observed trend of increase in C T over the three-year period studied (2.2 ± 0.4 mmol kg -1 yr -1 ) is twice that observed in open oceanic waters at ESTOC.When the NC T data are considered (Figure 3C, blue open circles), the trend is reduced to 1 ± 0.1, as that observed in ESTOC.Therefore, we assume that most of the anomalies with respect to the harmonic function are related to the arrival of deeper waters into the area, which could bring more remineralized (positive anomalies) or more productive (negative anomalies) waters.As a result, the NC T does not decrease after the end of March, but keeps relatively constant concentrations until July, when the productivity of the area compensates the physical processes.The detectable seasonal amplitude of 15 mmol kg -1 in NC T after July (Figure 3C) is associated with CO 2 consumption by organisms for the production of organic matter and the exchange of CO 2 between the atmosphere and the surface layer.In this coastal area, the influence of the Trade Winds seems to be more important due to horizontal advection and seawater renewal in the bay.
The variability of fCO 2 in both the atmosphere and seawater describes the studied system as an overall CO 2 source.The surface water was undersaturated from late October to mid-June.For the rest of the year, the surface water was supersaturated and the system acted as a source.The calculated mean FCO 2 over the study period is 0.35 ± 0.04 mmol m -2 d -1 (126 ± 13 mmol m -2 yr -1 ), with peak outgassing occurring between August and November, and maximum ingassing observed between February and March.This shift from a sink to a source role coincides temporally with the onset of Trade Winds, which typically blow at high and constant velocity in the Canary Islands from mid-June.This climatology is consistent with previous coastal studies, such as those conducted in the Hawaiian region, where the influence of Trade Winds is prominent and affects the CO 2 system in coastal waters (Terlouw et al., 2019).The occurrence and strength of the dominant northeast and east Trade Winds between 1973 and 2009 have been previously studied (Garza et al., 2012) and reported velocities ranging from 0.8 to 8.2 m s -1 , with summer being the period of higher intensity.Furthermore, the significance of winds and mixing processes becomes clear when comparing these results with coastal studies along the East Australian coast, where mixing processes are relatively low (McNeil, 2010;Shaw and McNeil, 2014), or along the Northwest coast of the North Atlantic Ocean, where coastal systems act as CO 2 sinks primarily due to SST effects at higher latitudes (Boehme et al., 1998).
The FCO 2 results here are comparable to other coastal stations such as in Brest (France, Gac et al., 2020) and Hawaii (Terlouw et al., 2019).In Brest, at higher latitudes than the Canary Islands, the authors estimated fluxes of 0.18 ± 0.10, 0.11 ± 0.12, and 0.39 ± 0.08 mol m -2 yr -1 in three coastal stations.In Hawaii, it is also important to highlight how the coastal areas could be a strong source of CO 2 from the ocean with 1.24 ± 0.33 mol m −2 yr −1 , and close to the equilibrium with 0.05 ± 0.02 and 0.00 ± 0.03 mol m −2 yr −1 at the other two coastal stations.In the coastal waters of the Mediterranean Sea, the coastal waters of the Gulf of Trieste act as a CO 2 sink in winter, especially in the presence of strong wind events (FCO 2 up to −11.9 mmol m −2 d −1 ; Cantoni et al., 2012;Ingrosso et al., 2016).In the Gando Bay region, with a net annual outgassing flux of 0.26 mol m -2 yr -1 (period 2020-2022), with values of 0.20 mol m -2 yr -1 for the years 2020 and 2022, but ingassing CO 2 at -0.08 mol m -2 yr -1 for the year 2021 due to lower winter SST values, the influence of Trade Winds is responsible for a temporary increase in fluxes, causing the mean values to remain positive (0.13 ± 0.01 mol m -2 yr -1 ), a phenomenon also observed at the ESTOC station (Gonzaĺez- Davila et al., 2003;Santana-Casiano et al., 2007;Gonzaĺez-Davila and Santana-Casiano, 2023).Additionally, Curbelo-Hernańdez et al. (2021) reported an average FCO 2 of -0.25 ± 0.04 mol m -2 yr -1 for the oceanic waters near Gran Canaria, which is consistent with the behavior observed in the Northeast Atlantic Ocean (-0.6 mol m -2 yr -1 ) (Takahashi et al., 2009).
Considering the mean flux determined at the coastal buoy site within Gando Bay (6 km 2 ), a net annual outgassing flux of 33 ± 4 Tons of CO 2 per year through the bay is calculated, which is close to equilibrium.If we consider the values determined in the open ocean waters of the Canary Islands (Curbelo-Hernańdez et al., 2021), where a CO 2 sink of -0.24 ± 0.04 mol m -2 yr -1 was calculated, that is, -170 Tons of CO 2 yr -1 for the entire coastal area of the Canary Islands.Coastal areas located in the easternmost part of the archipelago are affected by the arrival of Northwest African coastal waters (Curbelo-Hernańdez et al., 2021).The Gando Bay, with an area of only 6 km 2 , causes the amount of CO 2 absorbed by the Canary region to decrease by 9%.
Coastal regions at lower latitudes typically act as sources of CO 2 to the atmosphere, while high latitudes tend to act as sinks (Chen, 2004;Cao et al., 2020).Notably, the boundary for this behavior typically occurs around 30°N (Cai et al., 2006), placing the study area in the transition zone between low and high latitudes.Consequently, it is consistent with the result that this coastal area of the Canary Islands acts as a source while it is almost in equilibrium.The flux depends on several factors such as SST, DfCO 2 , and wind speed.This makes different mixing processes and biological activity in both the coastal and open ocean environments crucial influencers of the carbonate system parameters and controllers of the CO 2 air-sea exchange.In the case of this study area, it remains relatively unaffected by rivers, agriculture, or other anthropogenic activities that could disrupt the biological activity and the physical processes that control the carbonate system in the region.
The findings of this research clearly emphasize the need to explore additional coastal areas, since the hydrodynamic and CO 2 system-altering phenomena exhibit pronounced locality.Their effects vary from one area to another.In addition, studies should also extend the observations for at least 10 years to allow more accurate estimates of rates of change.

Conclusion
The coastal zones of islands require vigilant monitoring to accurately quantify the carbon balance and its consequences as an essential tool for effective governance.The economic well-being of these islands is significantly intertwined with the health of their coastal zones.
Within the Canary Islands, the CanOA-1 station, an integral part of the international GOA-ON network, is located in the eastern region of Gran Canaria, specifically in the Bay of Gando.Data from this station show a discernible seasonal pattern in the variables defining the CO 2 system.This study highlights the importance of physical processes, especially horizontal mixing, biological influences (hypothesis because no biological data are collected), and sea surface temperature, in facilitating the transfer of CO 2 from the atmosphere to the ocean.
The three-year data allow us to know that the sea surface temperature (SST) shows seasonal fluctuations, with the highest temperatures occurring in September-October and the lowest in March.The contributions of both thermal and non-thermal processes to the seasonal fCO 2 were investigated with vertical mixing, wind stress, and biological forcing as principal components.The T/NT ratio of 2.0 ± 0.1 implies that SST plays a controlling role in modulating fCO 2sw , although the influence of other factors should not be neglected.fCO 2sw increases at a rate of 1.91 μatm yr -1 , which is consistent with the observed increase in NC T in these coastal waters.The CO 2 transfer from the atmosphere to the surface waters causes a decrease of pH T,IS in the region.Furthermore, the concentration of total inorganic carbon (NC T ) showed an annual increase of 1.0 mmol kg -1 yr -1 in the surface waters of the bay.The occurrence of extreme events, such as tropical storms, and the most persistent Trade Winds direction, which are modified by the geological structure of the bay, affect the physicochemical properties of the region, renewing the bay waters with deeper seawater, affecting SSS, pH, fCO 2 , and C T .
According to the data collected at the CanOA-1 site, the Gando Bay is almost in equilibrium, with a total CO 2 release from the ocean to the atmosphere between 2020 and 2023, of 33 ± 4 Tons of CO 2 yr -1 , with the year 2021 acting as a slight sink.Small changes in SST between years, together with variability related to the prevailing strength of the Trade Winds, which affect both water exchange and CO 2 fluxes, control the Gando Bay site.This underscores the critical importance of continuous monitoring and quantification of CO 2 concentrations in seawater, especially in coastal regions and on islands worldwide, in order to estimate the CO 2 contribution of coastal regions to the global ocean.The author(s) declare that financial support was received for the research, authorship, and/or publication of this article.This study was supported by the Government of the Canary Islands the Loro Parque Foundation through the CanBIO project, CanOA subproject (2019-2023) and the CARBOCAN agreement (Consejerıá de Transicioń Ecoloǵica, Lucha contra el Cambio Climatico y Planificación Territorial, Gobierno de Canarias).Discrete sampling and maintenance were also supported by the FeRIA project (PID2021-123997NB-100), the Spanish Ministerio de Ciencia e Innovacioń.

FIGURE 1
FIGURE 1Study area map.The Gando Bay belongs to Gran Canaria Island.Red flag represents the location of buoy and green point, the ESTOC station.

Figure 3D
Figure3D.The pH T,IS displayed a decreasing trend from winter to summer, with a mean value of 8.05 ± 0.02, for the studied period.The maximum pH T,IS was typically recorded between February and March, with a mean of 8.07 ± 0.01, while minimum pH T,IS occurred between September and October, averaging 8.03 ± 0.01.This variation represents a decrease of about 0.04 units from winter to summer.Characterizing a rate of change in a variable such as pH will need an extended time series of data.However, when detrended pH data at ESTOC site(period 1995-2023, Gonzalez-Davila and Santana- Casiano, 2023) and those at the CanOA-1 site are plotted together (Figure4), it is shown the coastal pH values are following the same pH trend than that at oceanic waters, this last one decreasing at 0.002 ± 0.0002 units yr -1 (Figure4; Supplementary FigureSI-1).It is also consistent with other oceanic carbon time series(Bates et al., 2014).Examination of the variability of fCO 2sw and the corresponding CO 2 flux (Figure5) revealed periods of oversaturation (higher levels of fCO 2sw than those in fCO 2atm ) from May and June to November and December, approximately.During these months, the coastal zone acted as a source of CO 2 , releasing it from the ocean into the atmosphere.Conversely, during the rest of the year, the coastal zone acted as a carbon sink.It is noteworthy that on average DfCO 2 is lower during the period June to November than during the period December to June (Figure3A).However, the highest FCO 2 coincided with the June to November period, indicating the effect of the increased wind intensity.In this region, the prevailing Trade Winds exhibit their maximum strength in July and August, with wind speeds reaching 16 m s -1 (Figure2C).During this period, it is common for wind speeds to consistently exceed 7 m s -1 each year(Santana-Casiano et al., 2007).The mean annual flux of FCO 2 showed different patterns in the studied years when considered from March to March, with values of 0.70 ± 0.03 mmol m -2 d -1 (2020), -0.22 ± 0.04 mmol m -2 d -1 (in 2021), and 0.55 ± 0.04 mmol m -2 d -1 (in 2022).The average flux from 2020 to 2023 was 0.34 ± 0.04 mmol m -2 d -1 (126 ± 13 mmol m -2 yr -1 ), acting as a slight CO 2 source.The years 2020 and 2022 were registered as CO 2 sources with 255 mmol m -2 yr -1 and 202 mmol m -2 yr -1 , respectively.

FIGURE 4
FIGURE 4Detrended pH data in total scale measured at ESTOC (in black, Gonzaĺez-Davila and Santana-Casiano, 2023) and CanOA sites (in red).

FIGURE 5 CO 2
FIGURE 5CO 2 flux (FCO 2 ) in the CanOA-1 site measured with two sensors as described in the experimental section.PrOc referred to the ProOceanus sensor and Batt is to the Battle sensor.

TABLE 1
Variables for the equations of interannual trend of the different variables monitored in the Gando Bay, according with Equation 14 where x is the year fraction of each observation y.