pCO2 variability in the surface waters of the eastern Gulf of Cádiz (SW Iberian Peninsula)

Spatiotemporal variations of the partial pressure of CO2 (pCO2) were studied during 8 oceanographic cruises conducted between March 2014 and February 2016 in surface waters of the eastern shelf of the Gulf of Cádiz (SW Iberian Peninsula) between the Guadalquivir River and Cape Trafalgar. pCO2 presents a range of variation between 320.6 and 513.6 μatm, with 15 highest values during summer and autumn and lowest during spring and winter. For the whole study, pCO2 shows a linear dependence with temperature, and spatially there is a general decrease from coastal to offshore stations associated with continental inputs and an increase in the zones deeper than 400 m related to the influence of the eastward branch of the Azores Current. The study area acts as source of CO2 to the atmosphere during summer and autumn and as a sink in spring and winter, with a mean value for the study period of -0.18 ± 1.32 mmol m d. In the Guadalquivir and Sancti Petri 20 transects, the CO2 fluxes decrease towards offshore, whereas in the Trafalgar transect fluxes increase due to the presence of an upwelling. The annual uptake capacity of CO2 in the Gulf of Cádiz is 4.1 Gg C year.

thermal effect from the observed pCO2, the data were normalized to a constant temperature, the mean in situ SST depending on the focus considered, according to Eq. (1). pCO2 at SSTmean = (pCO2)obs·exp[0.0423·(SSTmean -SSTobs)] (1) where the subscripts "mean" and "obs" indicate the average and observed SST values, respectively.

155
The effect of thermal changes on pCO2 has been computed by perturbing the mean pCO2 with the difference between the mean and observed temperature. The pCO2 value at a given observed temperature (SSTobs) was calculated based on Eq. (2).

165
(ΔpCO2)T = (pCO2, SSTobs)max -(pCO2, SSTobs)min (4) The relative importance of each effect can be expressed in terms of the ratio between the thermal effects (T) and non-thermal effects (B): A T/B ratio greater than 1 implies the dominance of thermal effects over non-thermal on the pCO2 dynamics. However, a T/B 170 lower than 1 reveals a greater influence of non-thermal processes. This method was originally designed for open oceanic systems, but it has been widely used by other authors in coastal areas (e.g. Schiettecatte et al., 2007;Ribas-Ribas et al., 2011;Qu et al., 2014;Burgos et al., 2018).
In addition, Olsen et al. (2008) propose a method in which the seasonal signal of pCO2 data is decomposed into individual components due to variations in SST, in air-sea CO2 exchange, in SSS, and in combined mixing and biological processes.

175
pCO 2 sw,i = SST pCO 2 sw,i + AS pCO 2 sw,i + SSS pCO 2 sw,i + MB pCO 2 sw,i where the superscript "i" refers to the mean value between consecutives cruises for all variables; pCO 2 sw,i is the observed change in pCO2; SST pCO 2 sw,i is the change due to SST changes; AS pCO 2 sw,i is the change due to air-sea exchange; SSS pCO 2 sw,i is the change due to salinity variations; and MB pCO 2 sw,i is the change due to mixing plus biology. At the same time, each process is calculated with the following equations (Olsen et al., 2008): 180 SST pCO 2 sw,i = pCO 2 sw,i ·e 0.0423(ΔSST) -pCO 2 sw,i where ΔSST is the SST difference between two cruises AS pCO 2 sw,i = -(d·F i ) / MLD i where d is the number of days passed between two cruises (90 days approximately); F i is the mean flux of CO2; and MLD i is the mean mixed layer depth. SSS pCO 2 sw,i is not determined in this study, since data on the variations of total alkalinity and dissolved inorganic carbon are not available, and the spatial SSS changes are only significant near the Guadalquivir River mouth. MB pCO 2 sw,i is calculated as a residual, that is, as the change in pCO2 that is not explained by other processes. Additionally, as this study includes both coastal areas and deeper areas, the analysis is divided, in function of the system depth, between coastal (< 50 m) and distal (> 50 m) areas. Thus, MLD i in distal areas (Table 3) was calculated derived from the thermocline position that separates the SAW 190 and the ENACW (71. 3 -96.8 m), while the coastal areas correspond to the depth of these areas (15 -50 m).

Estimation of CO2 fluxes
Fluxes of CO2 across the sea-air interface were estimated using the relationship: FCO2 = α · k · (ΔpCO2)sea-air (9) where k (cm h -1 ) is the gas transfer velocity; α is the solubility coefficient of CO2 (Weiss, 1974); and ΔpCO2 is the difference 195 between the sea and air values of pCO2. The atmospheric pCO2 (pCO₂ atm ) values were obtained from the monthly atmospheric data of xCO2 (xCO₂ atm ) at the Izaña Atmospheric Station in Spain (Earth System Research Laboratory; https://www.esrl.noaa.gov/gmd/dv/data/index.php, last access: 9 January 2019). The xCO₂ atm was converted to pCO₂ atm as described in DOE (2007).
The gas transfer velocity, k, was calculated using the parameterization formulated by Wanninkhof (2014): where u (m s -1 ) is the mean wind speed at 10 m height on each cruise, obtained from the Shipboard Weather Station; Sc is the Schmidt number of CO2 in seawater; and 660 is the Sc in seawater at 20 ºC.

Statistical analysis
Statistical analyses were performed with IBM SPSS Statistics software (Version 20.0. Armonk, NY). The dataset was analysed 205 using one-way analysis of variance test (ANOVA) for analysing significant differences between cruises for discrete and continuous surface data on hydrological and biogeochemical characteristics. The threshold value for statistical significance was taken as p < 0.05. Moreover, all reported linear correlations are type I and they are statistically significant with p-values smaller than 0.05 in the entire manuscript unless indicated otherwise. Table 1 gives the ranges of variation and the mean and standard deviation of SST, SSS and pCO2 during the 8 sampling cruises  (Table 1). For the whole period, the averaged values for both seasons were highest during summer (21.0 ± 0.8 ºC) and autumn

Underway variables
Since the cruises were carried out at the beginning of each meteorological season, it is appropriate to analyse how representative is the range of temperatures that has been obtained. Figure 3 shows the mean value over the last 10 years of the 18.0 ± 11.4 µatm, respectively) and an oversaturation in summer and autumn (-20.4 ± 24.6 and -8.0 ± 15.3 µatm, respectively).
In Fig. 2 a sharp variation of SST and pCO2 can be observed in some zones that coincides with the stations where discrete water samples were taken. This may be due to the different sampling time at these stations, which varied between 2 and 8 240 hours in function of the depth of the system.
The database of this study includes the transition from coastal zones with depths of the order of 20 m to distal shelf waters with depths greater than 800 m. Figure 4 shows the general trend of the mean values of pCO2 and SST for different intervals of depth of the water column based on the information obtained in the 8 cruises.  for those of December 2014; that exception may have been due to the exceptional mixing of the water column caused by the storms. No general trend in the spatial variations of pH and AOU was found.

Discrete surface variables
Chlorophyll-a values presented significant differences among the cruises and between the same seasons of each year. This 260 parameter varied from 0.02 to 2.37 µg L -1 , with the highest mean value measured in March 2015 (0.76 ± 0.55 µg L -1 ), which coincides with the lowest (negative) mean value of AOU (Table 2). The lowest mean value was in June 2014 (0.18 ± 0.14 µg L -1 ). With reference to the seasons of both years, the highest value was in spring (0.71 ± 0.46 µg L -1 ), followed by winter (0.58 ± 0.33 µg L -1 ), autumn (0.26 ± 0.30 µg L -1 ) and the lowest value in summer (0.23 ± 0.25 µg L -1 ). The SP transect presented the lowest mean value of the whole study (0.33 ± 0.31 µg L -1 ), and the TF zone the highest (0.49 ± 0.37 µg L -1 ).

265
Nitrate concentration did not show significant differences among the cruises, ranging between 0.00 and 1.93 µmol L -1 . The highest mean value was recorded in spring (0.82 ± 1.09 µmol L -1 ) and the lowest in summer (0.25 ± 0.35 µmol L -1 ) of both years. The TF transect presented the highest mean concentration for the whole study (0.77 ± 0.76 µmol L -1 ). Phosphate concentration showed significant differences among all the cruises. By season, the highest mean value was obtained during autumn (0.31 ± 0.30 µmol L -1 ), although the average data in October 2014 (0.09 ± 0.03 µmol L -1 ) was lower than that of 2015

270
(0.50 ± 0.55 µmol L -1 ) ( Table 2). The lowest mean value was observed during summer (0.10 ± 0.05 µmol L -1 ). The GD transect presented the highest mean value of the whole study (0.28 ± 0.39 µmol L -1 ), and the lowest values were found in the TF and SP transects, with a similar value in each, 0.15 ± 0.07 µmol L -1 and 0.14 ± 0.09 µmol L -1 , respectively. The mean N/P ratio in surface waters for the whole study was 3.5 ± 2.0, similar to that estimated by Anfuso et al. (2010) in the northeast continental shelf of the Gulf of Cádiz, which indicates a relative phosphate deficit with respect to the Redfield ratio (Redfield et al., 1963). Table 3 summarizes the mean values and standard deviation for atmospheric pCO2, wind speed, gas transfer velocity and the air-sea CO2 fluxes measured in this study. The mean wind speeds were relatively similar for the whole study period, ranging between 5.5 ± 2.8 m s -1 (March 2015) and 7.7 ± 4.2 m s -1 (December 2014). The gas transfer velocity varied between 6.9 ± 0.1 cm h -1 in March 2015 and 14.4 ± 0.3 cm h -1 in June 2015, since it is very sensitive to changes in wind speed. There was a clear 280 seasonal variability in the dataset of CO2 fluxes. The study area acted as source of CO2 to the atmosphere during summer and autumn (0.7 ± 1.5 mmol m −2 d −1 and 1.2 ± 0.9 mmol m −2 d −1 , respectively) and as a sink in spring and winter (-1.3 ± 1.6 mmol m −2 d −1 and -1.3 ± 1.6 mmol m −2 d −1 , respectively).

Thermal influence in pCO2
285 Numerous research studies have determined that temperature is one of the most important factors that control the variability of pCO2 in the ocean (e.g. Millero, 1995;Bates et al., 2000;Takahashi et al., 2002;Carvalho et al., 2017), as a consequence of the dependence of the solubility of CO2 with the temperature (Weiss, 1974;Woolf et al., 2016). When pCO2 is affected only by the temperature, Takahashi et al. (1993) determined a relative variation of pCO2 of 0.0423 ºC -1 , equivalent to 16.9 µatm ºC -1 for experimental pCO2 of 400 µatm. In our study a seasonal variation was observed with a linear increase of the values of 290 pCO2 with SST for the entire database (r 2 = 0.37, Fig. 5A). This relationship becomes more significant when it is obtained from the mean values of pCO2 and SST of each cruise (r 2 = 0.71, Fig. 5B). The slope, 4.80 µatm ºC -1 , is lower than the thermal effect on pCO2 described by Takahashi et al. (1993), and indicates the influence of other non-thermal processes on the distribution of pCO2 in this zone of the Gulf of Cádiz. m) are described (Table 4). Ribas-Ribas et al. (2011) found in the north eastern shelf during June 2006 and May 2007 a dependence of pCO2 with temperature similar to that found in this study (5.03 µatm ºC -1 , r 2 = 0.42), and a pCO2 that ranged between 338 and 502 µatm. In 2003, Huertas et al. (2006)  temperature has also been determined in other studies of continental shelves, such as in the east China Sea (Wang et al., 2000), in the northern east China Sea (Shim et al., 2007) and in the northern Yellow Sea (Xue et al., 2012).
Comparing the data given in previous studies of the Gulf of Cádiz with the mean value found in this study (398.9 ± 15.5 µatm), it is evident that there has been an increase of pCO2 during the last decade, even taking into account the uncertainty associated 310 https://www.esrl.noaa.gov/gmd/dv/data/index.php, last access: 9 January 2019)). This suggests a possible increase of the anthropogenic nutrient and C inputs from land (Mackenzie et al., 2004) since the direction and magnitude of estuarine and continental shelf CO2 exchange with the atmosphere is highly dependent on the terrestrial organic budget and nutrient supplies to the coastal ocean (Borges and Abril, 2011;Cai, 2011).

Non-thermal factors controlling pCO2
315 Several authors have described the influence of the continental inputs on the distribution of pCO2 in surface waters. In general, the coastal zone is usually oversaturated with CO2 ( Fig. 4), whereas the continental shelf as a whole acts as a sink of atmospheric CO2 (e.g. Rabouille et al., 2001;Chen and Borges, 2009). This behaviour has been described in other systems,  remineralization of the organic matter in the surface sediments originating from the continuous deposition of organic matter through the water column (de Haas et al., 2002;Jahnke et al., 2005). The intensity of this process decreases in line with the increasing depth of the system, and the influence of the primary production and the continental supplies on the deposition of the particulate organic matter is less (Friedl et al., 1998;Burdige, 2007;Al Azhar et al., 2017). Ferrón et al. (2009) quantified the release from the sediment of DIC related to the processes of oxidation of organic matter in the coastal zone (depth < 50 m) of the Gulf of Cádiz, between the Guadalquivir and the Bay of Cádiz. These authors found a mean benthic flux of 27 ± 8 mmol C m -2 d -1 for stations with a mean depth of 23 m. This flux of DIC is equivalent to a CO2 flux of 198 ± 80 µmol C m -2 d -1 , considering a well-mixed water column, a pH = 8, in the conditions of mean temperature and salinity in the Gulf of Cádiz (18.8 ºC and 36.19, respectively) and using the K1 and K2 acidity constants proposed by Lueker et al. (2000) in the total pH scale. Moreover, this estimated CO2 benthic flux would produce an increase of pCO2 of 0.25 ± 0.10 µatm d -1 in the water 345 column.
Additionally, another factor present in the Gulf of Cádiz and that could affect the distribution of pCO2 is the vertical and lateral transport. For example, there are two upwelling systems in our study zone, one more permanent situated in the coastal zone (depth between 50 and 100 m) of the Trafalgar section (Prieto et al., 1999;Vargas-Yáñez et al., 2002)  In accordance with Olsen et al. (2008), Fig. 10 shows the decomposition of the variations of pCO2 between cruises due to changes in SST, in air-sea CO2 exchange and in combined mixing and biology, in distal and coastal areas. In general, the 375 variations are greater than those found in other works (Olsen et al., 2008;Omar et al., 2010) because this study considers seasonal changes against the monthly change analysed in previous applications. pCO 2 sw presents practically the same temporal trend in deep and coastal areas, but with a global behaviour different since the distal zones act a sink of CO2 of the system (mean pCO 2 sw = -3.4 ± 28.9 µatm) and the shallower areas as a source of CO2 (mean pCO 2 sw = 0.2 ± 22.7 µatm). In distal areas (Fig. 10), pCO2 changes are mainly brought about by SST (-58.4 -106.2 µatm) together with mixing and biology 380 processes (-90.8 -36.2 µatm). An inverse coupling is observed between SST pCO 2 sw and MB pCO 2 sw , since with the increase of the system SST (increase SST pCO 2 sw ) there is greater biological uptake of CO2 (decrease MB pCO 2 sw ). As reported in the studies of Olsen et al. (2008) and Omar et al. (2010), the change produced by the air-sea CO2 exchange is lower. Instead, in coastal areas (Fig. 10), the dominant effects on pCO2 changes are produced by air-sea CO2 exchange (-196.2 -103.4 µatm) and mixing plus biology (-101.1 -198.5 µatm). A relative inverse coupling between the two factors was also observed; 385 outgassing is produced (decrease AS pCO 2 sw ) when the system receives greater inputs/production of CO2 (increase MB pCO 2 sw ). There is a different behaviour between the transition from spring to summer of 2014 (ST1 and ST2) and 2015 (ST5 and ST6) for MB pCO 2 sw , which may be due to a greater quantity of continental inputs, as reflected in the Guadalquivir river flow rate in these periods (85.1 ± 75.4 m 3 s -1 and 25.3 ± 10.2 m 3 s -1 , respectively). A larger effect of the air-sea CO2 exchange on pCO2 variation is observed in the shallower mixed layers, as also described by Olsen et al. (2008) in the subpolar

T/B ratio
In this study, the total T/B ratio is 1.15, which indicates that the thermal effect is an important factor controlling intra-annual variation of pCO2. This value is similar to that determined by Ribas-Ribas et al. (2011) (see date and study zone in Table 4  these two factors favour the accumulation of CO2 in this area as a convergence zone (Ríos et al., 2005). The observed variations of ΔpCO2 non-thermal between areas close to the coast and deeper areas agrees with the application of the Olsen et al. (2008) method.
The T/B ratios have also been calculated for the different transects at right angles to the coast that have been cruised for sampling in the study zone, as shown in Fig. 9. It can be appreciated that the T/B ratio increases with the distance from the coast on the three transects, and that the temperature generally has a greater influence on the distribution of pCO2 than the non-  2006). The coastal zone close to Cape Trafalgar has been characterized as a region with high autotrophic productivity and biomass associated mainly with the nutrients input due to upwelling waters (e.g. Echevarría et al., 2002;García et al., 2002).
The presence of these emerging water masses could be related to the relatively low values of ΔpCO2 thermal found in this 430 zone; in fact, the mean temperature in this area is 18.4 ± 2.3 °C, about 0.5 °C lower than in the other two zones. The Sancti Petri zone is the one that receives a smaller supply of nutrients, and presents the lowest concentrations of chlorophyll-a in this study. The high values of ΔpCO2 thermal in this part of the Gulf of Cádiz are associated with a higher mean temperature (19.0 °C) and a wider range of variation (6.8 °C).

435
In the Gulf of Cádiz, the flux of CO2 presents a range of variation from -5.6 to 14.2 mmol m -2 d -1 . These values are within the ranges observed by other authors in different areas of the Gulf of Cádiz (Table 4). As can be appreciated in Fig. 10, the fluxes of CO2 presented seasonal and spatial variations during the period studied. The Gulf of Cádiz acts as a source of CO2 to the atmosphere during the months of summer (ST2, ST6) and autumn (ST3, ST7), and as a sink in spring (ST1, ST5) and winter (ST4, ST8). Previous studies conducted in the Gulf of Cádiz are consistent with the behaviour found in this study (González-440 Dávila et al., 2003;Aït-Ameur and Goyet, 2006;Ribas-Ribas et al., 2011).
As has been observed with pCO2, temperature is one of the principal factors that control the fluxes of CO2. In fact, for each cruise, a linear and positive relationship has been found between the mean values of the CO2 fluxes and SST (r 2 = 0.72, Fig.   11). In parallel, there is a linear and negative relationship between the mean values of the CO2 fluxes and the concentration of chlorophyll-a at the discrete stations sampled (r 2 = 0.74, Fig. 11), as a consequence of the biological utilization of the CO2 445 (Qin et al., 2014). These relationships have also been found in various studies carried out in zones similar to the area studied (Zhang et al., 2010;Arnone et al., 2017;Carvalho et al., 2017).
The fluxes of CO2 in the Gulf of Cádiz tend to decrease with the distance from the coast (Fig. 10). The coastal zone (< 50 m) presents a mean CO2 flux of 0.8 ± 1.8 mmol m -2 d -1 , that reduces progressively to reach a value of -0.3 ± 1.6 mmol m -2 d -1 in open waters with bottom-depth greater than 600 m. This dependence of CO2 fluxes with distance from the coast has also been 450 reported in other systems, such as in the South Atlantic Bight of the United States (Jiang et al., 2008), in the south-western part of the Atlantic Ocean (Arruda et al., 2015), in the Patagonian Sea (Kahl et al., 2017) and on the continental shelf of on the CO2 fluxes as one moves towards the open sea. Ribas-Ribas et al. (2011) also found that in the Gulf of Cádiz the CO2 fluxes vary with the distance from the coast; the zone close to the estuary of the Guadalquivir and the Bay of Cádiz acts as a 455 source (1.39 mmol m -2 d -1 ) and the zone comprising the rest of the shelf acts as a sink (-0.44 mmol m -2 d -1 ).
In addition, on both the GD and SP transects a decrease of the CO2 flux is found towards the open ocean, due to the continental inputs associated with the estuary of the Guadalquivir and with the Bay of Cádiz, respectively. On the TF transect, in contrast, it was observed that the zone close to the coast acts as a sink of CO2 (-0.4 ± 1.2 mmol m -2 d -1 ), and the deeper zone is a weak source of CO2 to the atmosphere (0.3 ± 1.3 mmol m -2 d -1 ). This finding can be explained by the presence of an upwelling close to the coast that is likely to be causing an increase of the production (e.g. Hales et al., 2005;Borges et al., 2005). With reference to this, on the TF transect there are significant differences between the mean surface concentrations of chlorophyll-a and nitrate in the coastal zone (0.63 ± 0.43 µg L -1 and 1.09 ± 0.77 µmol L -1 , respectively) and in deeper zones (0.17 ± 0.12 µg L -1 and 0.32 ± 0.33 µmol L -1 , respectively).
The Gulf of Cádiz, during the period of this sampling, acted as a sink of CO2, with a mean rate of -0.18 ± 1.32 mmol m -2 d -1 , 465 that would give rise to an annual flux of -0.07 mol C m -2 yr -1 . With the total surface of the study area (52.8·10 2 km 2 ) and the mean annual flux during the 8 cruises, the uptake capacity estimated for the Gulf of Cádiz will be 14.9 Gg C year -1 . The findings of previous studies carried out in the Gulf of Cádiz coincide with the behaviour observed in this study González-Dávila et al., 2003;Huertas et al., 2006;de la Paz et al., 2009;Ribas-Ribas et al., 2011), with the exception of the study by Aït-Ameur and Goyet (2006) in which it was estimated that the Gulf of Cádiz acts as a source of 470 CO2 to the atmosphere, although that study only corresponds to the summer season.

Conclusions
The mean value of pCO2 in the eastern part of the Gulf of Cádiz found in this study (398.9 ± 15.5 µatm) indicates that it is undersaturated in CO2 with respect to the atmosphere (402.1 ± 3.9 µatm). The spatiotemporal variation of pCO2 found responds to the influence of different factors that usually affect its distribution in the littoral oceans. In global terms, when the mean Firstly, the dominant effects in the shallower areas are also due to the continental inputs, the biological activity and the air-sea CO2 exchange. Then pCO2 values diminish progressively in line with increasing distance from the coast, out as far as an 480 approximate depth of some 400 m. There is a relative increase of SST and pCO2 as consequence of a change in the origin of the surface water, with the arrival of waters in a warm branch of the Azores current and the change produced by the biological activity.
The total T/B ratio (1.15) suggests that the distribution is principally controlled by the temperature. However, there is a different behaviour in this ratio if it is determined by bottom-depth intervals, related to the existence of non-thermal processes.

485
In the proximity of the Guadalquivir estuary the ratio takes a value of 0.93 due to the continental inputs of C and nutrients, and in the zone around the coastal upwelling off Cape Trafalgar the ratio is 1.09. Furthermore, the actual characteristics of the surface water mass that originates under the influence of a branch of the Azores current also produce a decrease of the T/B ratio in the deeper zone studied (1.05 for depths > 600 m). In contrast, the highest T/B ratio values have been found in the SP transect, where values of up to 1.54 are obtained for depths greater than 100 m.
The annual uptake capacity of CO2 by the surface waters in our study area is 14.9 Gg C year -1 . The CO2 fluxes present seasonal variation: these waters act as a source of CO2 to the atmosphere in summer and autumn and as a sink in winter and spring.
Based on the information available in the zone, there seems to have been a decrease in the capacity for CO2 capture in the zone in recent decades.

Competing interests
The authors declare that they have no conflict of interest.