Changes in the partial pressure of carbon dioxide in the Mauritanian-Cap Vert upwelling region between 2005 and 2012

Biogeosciences, 14, 3859–3871, 2017 https://doi.org/10.5194/bg-14-3859-2017 © Author(s) 2017. This work is distributed under the Creative Commons Attribution 3.0 License. Changes in the partial pressure of carbon dioxide in the Mauritanian–Cap Vert upwelling region between 2005 and 2012 Melchor González-Dávila1, J. Magdalena Santana Casiano1, and Francisco Machín1,2 1Instituto de Oceanografía y Cambio Global, Grupo QUIMA, Universidad de Las Palmas de Gran Canaria, 35017, Las Palmas de Gran Canaria, Spain 2Departamento de Física, Universidad de Las Palmas de Gran Canaria, 35017, Las Palmas de Gran Canaria, Spain Correspondence to: Melchor González-Dávila (melchor[email protected]) Received: 9 March 2017 – Discussion started: 29 March 2017 Revised: 26 July 2017 – Accepted: 28 July 2017 – Published: 31 August 2017 Abstract. Coastal upwellings along the eastern margins of major ocean basins represent regions of large ecological and economic importance due to the high biological productivity. The role of these regions for the global carbon cycle makes them essential in addressing climate change. The physical forcing of upwelling processes that favor production in these areas are already being affected by global warming, which will modify the intensity of upwelling and, consequently, the carbon dioxide cycle. Here, we present monthly high-resolution surface experimental data for temperature and partial pressure of carbon dioxide in one of the four most important upwelling regions of the planet, the Mauritanian–Cap Vert upwelling region, from 2005 to 2012. This data set provides direct evidence of seasonal and interannual changes in the physical and biochemical processes. Specifically, we show an upwelling intensification and an increase of 0.6 Tg yr−1in CO2outgassing due to increased wind speed, despite increased primary productivity. This increase in CO2outgassing together with the observed decrease in sea surface temperature at the location of the Mauritanian Cap Blanc, 21◦N, produced a pH rate decrease of −0.003 ±0.001 yr−1. 1 Introduction The excess of CO2in the atmosphere, largely responsible for global climate change, has prompted research on the role of the oceans in the carbon cycle. The aim in recent decades has been to assess how the oceans act as sources or sinks within the carbon cycle. To achieve this goal, highly resolved spatial and temporal observations representative of the distribution of CO2fluxes between the ocean and atmosphere are necessary. Automated instruments on volunteer observing ships (VOSs) serve to provide as many observations throughout the global ocean as possible. This is in addition to data collected on scientific cruises and at long-term moorings (i.e., Astor et al., 2005: Lüger et al., 2004, 2006; González-Dávila et al., 2005, 2009; Schuster et al., 2009; Ullman et al., 2009; Watson et al., 2009; Padin et al., 2010; Gruber et al., 2002; Dore et al., 2003; Santana-Casiano et al., 2007; Bates et al., 2014). With the amount of data already gathered (http://www. socat.info/; Pfeil et al., 2013), climatologies that present average CO2fluxes between the atmosphere and the ocean have been developed, identifying areas acting as a source or sink (Key et al., 2004; Takahashi et al., 2009). However, the low spatial resolution of these databases limits the applicability, especially in coastal areas. Upwelling regions are particularly under-represented in such large databases. Upwelling presents a dynamic process that raises nutrient and CO2-rich water from relatively deep areas to the surface. The nutrients reaching the photic zone promote primary production, which consumes CO2. This process generates a CO2flux into the ocean. On the other hand, upwelling also brings up CO2 from deep seawater, which generates uncertainty about the actual role of upwelling areas as a source or sink of CO2 (Michaels et al., 2001). Indeed, upwelling areas may act as a source or sink of CO2depending on their location (Cai et al., 2006; Chen et al., 2013), where upwelling regions at low latitudes mainly act as a source of CO2(Feely et al., 2002; Astor et al., 2005; Friederich et al., 2008; Santana-Casiano et al., 2009; González-Dávila et al., 2009) and those at midlatitudes Published by Copernicus Publications on behalf of the European Geosciences Union. 3860 M. González-Dávila et al.: Changes in the partial pressure of carbon dioxide mainly act as a sink of CO2(Frankignoulle and Borges, 2001; Hales et al., 2005; Borges and Frankignoulle, 2002; Borges et al., 2005; Santana-Casiano et al., 2009; González-Dávila et al., 2009). Several anthropogenic interactive effects strongly influence eastern boundary upwelling systems (EBUSs), including upper ocean warming, ocean acidification, and ocean deoxygenation (Gruber, 2011; Feely et al., 2008; Keeling et al., 2010). Moreover, evidence of increasing wind speed that would favor upwelling (Bakun, 1990; Demarcq, 2009; Oerder et al., 2015) supports the possibility of a change in the dynamics of these highly productive areas. Recently, eddyresolving regional ocean models have shown how upwelling intensification can cause a major impact on the system’s biological productivity and CO2outgassing (Lachkar and Gruber, 2013; Oerder et al., 2015). Wind observations and reanalysis products are controversial regarding the Bakun intensification hypothesis (Bakun, 1990). Using different wind databases for the Canary region, Barton et al. (2013) concluded that there was no evidence for a general increase in the upwelling intensity off northwest Africa. Marcello et al. (2011) found an intensification of the upwelling system in the same area during a 20-year period, while the alongshore wind stress remained almost stable. Cropper et al. (2014) found that coastal summer wind speed increased, resulting in an increase in upwelling-favorable wind speeds north of 20◦N and an increase in downwelling-favorable winds south of 20◦N. Santos et al. (2005, 2012) showed that sea surface temperature (SST) was not homogeneous either along latitude or longitude and depended on the upwelling index (UI) intensity. Varela et al. (2015) demonstrated opposite results worldwide depending on the length of data, season evaluated, and selected area within the same wind data set or between data sets. For the Mauritanian region, when wind stress data were used (Varela et al., 2015), a more persistent increasing trend in upwelling-favorable winds north of 21◦N and a decreasing trend south of 19◦N was determined. Starting in June 2005, the QUIMA-VOS line visited the Mauritanian–Cap Vert upwelling region northwest of Africa on a monthly basis (Fig. 1 and Table S1 in the Supplement) producing for the first time a high-resolution database of SST and partial pressure of CO2expressed as fugacity fCO2. This database shows the variations in the CO2system under changes in the upwelling conditions in the Canary ecosystem from 27 to 10◦N for the period 2005 to 2012. More data for the region from other surveys exist (http://www.socat.info/; Pfeil et al., 2013) but they were not considered in this study as they do not follow the same track as the QUIMA-VOS line. Those data are strongly influenced by the distance to the upwelling cells with the corresponding physical effects in the partial pressure of CO2. Figure 1. Ship track (black line) in the area from 28◦N (Gran Canaria, the Canary Islands) to 10◦N. The locations of Cap Blanc and Cap Vert are indicated. Monthly OceanColor Web (https:// oceancolor.gsfc.nasa.gov/) data for average chlorophyll aconcentration (mgm−3)were included in a MATLAB routine and annually averaged. The map has been generated using MATLAB 7.12 R2011a. 2 Experimental 2.1 Study region The VOS line crosses the east Atlantic Ocean from the north of Europe (English Channel) to South Africa, calling at Gran Canaria, the Canary Islands, with a periodicity of 2 months, which provides monthly data (southward or northward sections). In this work, the area between Gran Canaria at 27 and 10◦N has been selected in order to study the Mauritanian– Cap Vert upwelling region. On its route south (Fig. 1), the ship leaves Gran Canaria and goes straight to 100 km off Cap Blanc at 21◦N, 17◦450W. It then follows this longitude, passing at 100km off Cap Vert until 12◦N, where it changes direction to Cape Town, reaching 10◦N, 17◦W at 330km off the coast of Guinea. Between 22 and 20◦N, the ship reaches the 500m isobath. South of 15◦N, the ship moves between the 1000 and 500 m isobaths. On its route north, the ship follows the reverse track. 2.2 Experimental data Experimental data were obtained under the EU projects CARBOOCEAN and CARBOCHANGE (www.CarboOcean.org and https://carbochange.b.uib.no/) and now also available at http://www.socat.info/ (Pfeil et al., Biogeosciences, 14, 3859–3871, 2017 www.biogeosciences.net/14/3859/2017/ M. González-Dávila et al.: Changes in the partial pressure of carbon dioxide 3861 2013). An autonomous instrument for the determination of the partial pressure of CO2developed by Craig Neill following NOAA recommendations was installed on a VOS line. This was operated by the Mediterranean Shipping Company S.A. from 2005 to 2008 and Maersk from 2010 to 2012. This VOS line (QUIMA-VOS) ran between the UK and Cape Town from July 2005 to January 2013 (Table S1 in the Supplement). Temperature was measured at three positions along the sampling circuit: in the intake (Sea-Bird SBE38L), in the equilibrator (Sea-Bird thermosalinograph SBE21 and internal PT100 thermometer), and in the oxygen sensor (Optode 3835, Aanderaa™). After the seawater pump, the intake is divided into two lines, one feeding the CO2system and the other feeding the oxygen sensor, the fluorometer, and the Sea-Bird thermosalinometer. Differences between equilibrator and intake temperatures were constant in time due to the high seawater flow but varied among ships due to the different locations of the equipment. Values varied between 0.06 ◦C when the equipment was placed close to the intake and 0.35◦C when the equipment was one floor above and inside the engine room. The SST was also obtained from the NOAA_OI_SST-V2 data provided by the NOAA/OAR/ESRL PSD from Boulder, Colorado, USA (http://www.esrl.noaa.gov/psd). These data had a spatial resolution of 1◦latitude and 1◦longitude and monthly averages were used. The correlation between our experimental SST data and satellite data was better than ±1◦C, and improved to ±0.4 ◦C after removing the most affected upwelling regions (19–22 and 14–16◦N), which related to the high variability imposed by the upwelling. The CO2molar fraction, xCO2, was obtained every 150s in seawater, while atmospheric xCO2data were obtained every 180min. The seawater intake was located at a 10m depth. The system was calibrated every 3 hours by measuring four different standard gases with mixing ratios in the ranges of 0.0, 250–290, 380–410, and 490–530 ppm of CO2in the air, provided by NOAA and traceable to the World Meteorological Organization scale. The precision of the system is greater than 0.5µatm and the accuracy estimated with respect to the standard gases is of 1µatm inside the standards’ range. For xCO2values higher than the highest standard (532.04ppm), the accuracy will be reduced, even when linearity was observed in all cases inside the standards range. The fugacity of CO2(fCO2, µatm) was calculated from xCO2after correcting for temperature differences between intake and equilibrator, according to the expressions for seawater given by DOE (1994). Normalized fCO2(NfCO2) derived from the mean SST for the area (Tmean)was computed following Takahashi et al. (1993) as (NfCO2)=fCO2·exp[0.0423(Tmean −SST)].(1) In order to compute a second carbonate system variable, the surface total alkalinity (AT) was computed from sea surface salinity (SSS) and SST (Lee et al., 2006). pHTat the in situ temperature was computed from fCO2and ATand with average annual surface ocean total phosphate and total silicate concentrations of 0.5 and 4.8 µmolkg−1, respectively, from the World Ocean Atlas 2009, using the carbonic acid acidity constants by Mehrbach et al. (1973) refitted by Dickson and Millero (1987). Air–sea CO2fluxes (FCO2, mmol m−2d−1) were evaluated as FCO2=0.24 ·k·s·(f COsw 2−fCOatm 2), (2) where 0.24 is the scale factor, kis the gas transfer velocity, sis the CO2solubility, fCOsw 2is the seawater fugacity of CO2, and fCOatm 2is the atmospheric fugacity of CO2. In order to evaluate (fCOsw 2−fCOatm 2),fCOatm 2data were linearly interpolated to the fCOsw 2time vector. A positive value for FCO2corresponds to CO2outgassing from the ocean. k (cm h−1)was evaluated with the following parameterization (Nightingale et al., 2000): k=(0.222 ·W2+0.333 ·w) ·(Sc/660)−1/2,(3) where Wis the wind speed at 10m above the sea surface (m s−1)and Sc is the Schmidt number. The variables involved in estimating FCO2data (i.e., fCOsw 2,fCOatm 2, SST, and SSS) were fitted to sinusoidal expressions (Lüger et al., 2004) for a given latitude as follows: X(lat)∗=a0+a1(t−2005)+a2sin(2πt)+a3cos(2πt) +a4sin(4πt)+a5cos(4πt),(4) where aiare the fitting coefficients, tis the sampling time expressed as year fraction, and X(lat)∗represents any of the four fitted variables. This procedure allowed us to reconstruct the series of experimental data for periods without monthly data. The variables were decomposed into an interannual term X(lat)∗ t=a0+a1(t−2005)plus a periodical term X(lat)∗ p=a2sin(2πt)+a3cos(2πt)+a4sin(4πt)+ a5cos(4πt), that is, X(lat)∗=X(lat)∗ t+X(lat)∗ p. The periodical term accounts for the high-frequency seasonal variability, while the interannual term marks the year-to-year trend. First, observations were grouped in a natural year for a given latitude, as if they had been taken in a single year (no correction was done for interannual variability). The mean seasonal climatology data associated with the periodic coefficients (i.e., a2,a3,a4, and a5)throughout the sampling period were determined. Next, the interannual coefficient a1 was calculated by fitting the residuals resulting from subtracting the periodical component, X(lat)∗ p, from the original variable X(lat). By fixing these five coefficients (a1–a5), new distributions for fCOsw∗ 2,fCOatm∗ 2, SST∗, and SSS∗were constructed with a daily resolution based on the curve fits given for each variable as in Eq. (4), providing the coefficient a0. The accuracy of this fitting procedure was checked by both computing the correlation between experimental and reconstructed values and by determining the mean residuals. The Pearson coefficients were always over 0.87 for www.biogeosciences.net/14/3859/2017/ Biogeosciences, 14, 3859–3871, 2017 3862 M. González-Dávila et al.: Changes in the partial pressure of carbon dioxide SST (average 0.94 ±0.03), over 0.69 for both fCOsw 2and fCOatm 2(average of 0.79 ±0.07 and 0.82 ±0.04, respectively), and over 0.67 for SSS (average 0.79 ±0.07). The mean residual on the determination of those four variables were ±3.7 µatm, ±1.5 µatm, ±0.22 ◦C, and ±0.05 for fCOsw∗ 2,fCOatm∗ 2, SST∗, and SSS∗, respectively. When the monthly satellite SST values were considered, the new SST∗ function averaged for each month produced values within ±0.47 ◦C, confirming that this procedure was able to fit nonsampled periods. It was assumed that the same procedure was valid for non-sampled fCO2.Finally, daily FCO∗ 2time series between 10 and 27◦N with a latitudinal resolution of 0.5◦were calculated with a standard error of estimation of 0.5 mmol m−2d−1(15 % of error) that produced mean residuals (experimental FCO2–FCO∗ 2)of 0.4 mmol m−2d−1and Pearson correlation coefficients between experimental and computed FCO∗ 2of r > 0.6, p < 0.01. Chlorophyll awas calculated from measurements made by the Moderate-resolution Imaging Spectroradiometer (MODIS) aboard NASA’s Aqua satellite. We used monthly averages with a spatial resolution of 9 km supplied by OceanColor Web (https://oceancolor.gsfc.nasa.gov). Wind data were downloaded from the NCEP CFSR database at http://rda.ucar.edu/pub/cfsr.html developed by NOAA and retrieved from the NOAA National Operational Model Archive and Distribution System and maintained by the NOAA National Climatic Data Center. The spatial resolution is approximately 0.3×0.3◦and the temporal resolution is 6 h. The reference height for the wind data is 10m. Rainfall data were collected by the precipitation radar installed on the Tropical Rainfall Measuring Mission (TRMM) satellite (http://precip.gsfc.nasa.gov). Monthly averages with a spatial resolution of 0.5 ×0.5◦(product 3A12, version 07) were used (Fig. S1 in the Supplement) in order to explain changes in seasonal surface salinity distributions. 3 Results and discussion 3.1 Physical properties The variability of the Mauritanian–Cap Vert upwelling was analyzed in terms of the upwelling index (Nykjaer and Van Camp, 1994) (Fig. 2) using satellite wind data. Negative UI values correspond to upwelling-favorable conditions and positive values to downwelling-favorable conditions. The lowest negative values of the index correspond to more intense upwelling. Results clearly distinguish two main subareas in the upwelling system: (1) north of 20◦N, the upwelling conditions were favorable throughout the year, although the highest upwellings were observed from March to September with a northward shift from 20 to 22◦N. (2) South of 20◦N, a marked seasonality was observed with favorable upwelling conditions during autumn and winter, with the maximum intensity observed during January and February. In this Figure 2. Time series of upwelling index (UI, ×10−3m2s−1)in the Mauritanian–Cap Vert upwelling region along the ship track computed following Nykjaer and Van Camp (1994). Blue colors are related to upwelling events and red colors to downwelling events. region, a downwelling regime is present between May and November when the summer trade winds are replaced by the monsoonal winds advecting warm water (Fig. 3a) northward along the shore (Nykjaer and Van Camp, 1994). Our results (Fig. 2) are quite consistent with previous research (Nykjaer and Van Camp, 1994; Marcello et al., 2011; Santos et al., 2005, 2012; Cropper et al., 2014) but include the years 2010 to 2012, when the UI at around 20–21◦N presented a shift of the upwelling intensity from high (−2000 m2s−1)to strong (−2800 m2s−1). The analysis of upwelling trends along this route has been controversial since it is highly dependent on the selected region (Santos et al., 2012). The interannual evolution of the UI over the period 2005 to 2012 (Fig. 4, green line) for each degree in latitude indicates an increase in the UI (mean confidence interval of 9m2s−1)as showed by Santos et al. (2012). North of 15◦N, the upwelling index confirmed the stronger upwelling observed since 1995–1996 in this region after more than a 10-year (from at least 1982 to 1995) period of weaker upwelling (Santos et al., 2012). Local zonal differences between ocean and coastal SST trends determined with satellite data confirmed the intensification of the upwelling regime along the African coast for the period 1982 to 2000 (Santos et al., 2005) and extended by Santos et al. (2012) until 2010 and further extended in this study until 2012 (data not shown). This has been described as a decadal-scale shift of the upwelling regime intensity (Marcello et al., 2011; Santos et al., 2012). South of 15◦N, the annual UI values and trends (Figs. 2 and 4) both for the upwelling (values close to −2800 m2s−1 in January) and downwelling (values reaching 1850 m2s−1 in July) periods are becoming stronger. At 11–12◦N, where downwelling is becoming stronger, this results in negative annual temperature rates that approach zero. The UI serves as an indication of decadal variability of the summer monsoon winds and associated northward advection of warm water along the coast (Santos et al., 2012). The highest upwelling intensity along the VOS line was located at the capes, Cap Blanc and Cap Vert. From satellite chlorophyll adata, especially off Cap Blanc, giant filaBiogeosciences, 14, 3859–3871, 2017 www.biogeosciences.net/14/3859/2017/ M. González-Dávila et al.: Changes in the partial pressure of carbon dioxide 3863 10 15 20 25 18 20 22 24 26 28 30 Latitude Temperature ( C) (a) W 10 15 20 25 32 33 34 35 36 37 Latitude Salinity W (b) 10 15 20 25 18 20 22 24 26 28 30 Latitude Temperature ( C) Sp 10 15 20 25 32 33 34 35 36 37 Latitude Salinity Sp 10 15 20 25 18 20 22 24 26 28 30 Latitude Temperature ( C) Sm 10 15 20 25 32 33 34 35 36 37 Latitude Salinity Sm 10 15 20 25 18 20 22 24 26 28 30 Latitude Temperature ( C) Au 10 15 20 25 32 33 34 35 36 37 Latitude Salinity Au Figure 3. In situ data of column (a) SST and column (b) SSS in the Mauritanian–Cap Vert coastal region grouped by seasons: winter (W; December, January, and February), spring (Sp; March, April, and May), summer (Sm; June, July, and August), and autumn (Au; September, October, and November). The averaged values for all cruises in Table S1 are shown in black for each season including the 95% confidence limits. The color code for each cruise is indicated in Table S1. ments with chlorophyll concentrations above 1mg m−3persist year-round, spreading from the coast to several hundred kilometers offshore (Fig. 1). North of Cap Blanc the upwelled water originates from the North Atlantic Central Water, and mixes with South Atlantic Central Water (SACW) towards the south (Mittelstaedt, 1983). South of Cap Blanc, the upwelling of nutrient-rich SACW (Mittelstaedt, 1983) promotes phytoplankton growth between Cap Blanc and Cap Vert. Towards 12◦N, upwelling is also fed by the North Equatorial Undercurrent (Hagen and Schemainda, 1984). Moreover, the entire northwest African coast is also influenced by the African desert dust transport by the midtropospheric Harmattan winds originating from the central Sahara, which supplements the levels of micronutrients (such as iron) to the adjacent marine ecosystem (Mittelstaedt, 1983; Neuer et al., 2004). The study area is also affected by the migration of the Intertropical Convergence Zone (ITCZ), related to maximum precipitation rates (Hastenrath, 1995). To have a significant satellite precipitation record in our region of interest, precipitation data were integrated longitudinally between 25.25 and 9.75◦W. Time series for the latitudinal distribution of integrated precipitation (Fig. S1 in the Supplement) identified the average position of the ITCZ related to maximum precipitation rates. The ITCZ was located at its southernmost position (2◦N) during winter, reaching its northernmost position www.biogeosciences.net/14/3859/2017/ Biogeosciences, 14, 3859–3871, 2017 3864 M. González-Dávila et al.: Changes in the partial pressure of carbon dioxide −0.2 −0.1 0 0.1 0.2 0.3 10 12 14 16 18 20 22 24 26 28 Change per year Latitude (°N) (a) Trends in T Trends in S Trends in UI 0 1 2 3 4 5 10 12 14 16 18 20 22 24 26 28 Change per year (b) Trends in fCO2 sw Trends in fCO2 atm Figure 4. Latitudinal distribution of the interannual trends for the upwelling index (UI) and for the four experimental variables along the QUIMA-VOS line integrated over every degree between 2005 and 2012. Panel (a) presents the trends for upwelling index (UI, ×10−3m2s−1, mean confidence interval of 9m2s−1), SST (◦C yr−1, confidence interval 0.13◦C), and SSS (yr−1, confidence interval 0.06) and (b) the trends for fCOsw 2and fCOatm 2(confidence intervals 4.23 and 0.44µatm). (14–16◦N) around summer. The ITCZ reached our area of interest (>10◦N) from late spring to late summer. The latitudinal distributions of measured SST and SSS along the vessel track are shown in Fig. 3, grouped by seasons (labeled W, Sp, Sm, and Au). The temperature generally decreased from 10 ◦N to about 20–21◦N, where the ship meets the Mauritanian upwelling. From there to the north, the temperature rises as the ship leaves the upwelling area on its way to the Canary Islands. In situ temperature at 27◦N shows temperatures in the range of 18 to 24 ◦C with the minimum in winter and maximum in late summer to early autumn. The annual temperature range was somewhat higher at 20◦N, with a summer maximum of around 26 ◦C and minimum in spring of about 17 ◦C. At 10◦N, temperatures were the highest throughout the year (>25 ◦C), with minimum values in winter and maximum in late spring and late autumn. The low values observed during the end of summer are related to the arrival of the ITCZ (Fig. S1 in the Supplement) at those latitudes. The thermal distribution shows a temperature increase as we move to the Equator and a notable cooling at the upwelled waters off Mauritania. The upwelling of cold water from the Cap Vert area was only detected during winter time and the beginning of spring. Salinity minimum values were normally located at 10◦N, increasing to maximum values at the Canaries’ latitude. The minimum values of salinity were exceptionally low during autumn from 10 to 16◦N by both the freshwater input from rivers that increase their outflow during this season (Nicholson, 1981) and by the northward shift of the ITCZ during this time of the year. Anomaly fields for temperature and salinity (data not shown) were calculated as the difference between the observations and the mean values at each season for individual latitudes. For temperature, the largest anomalies in winter and spring were located south of 18◦N, with values of ±2◦C, related to the seasonal cycle of the Cap Vert upwelling. During summer the pattern changed and the largest anomalies were detected in the upwelling area at 18–22◦N, with values of ±5◦C when the upwelling index for the Mauritanian area was highest (Fig. 2). In autumn the temperature anomalies were shifted slightly to the north, 20–24◦N, with values of ±3◦C related to the observed pulses in upwelling-favorable winds that affected the surface seawater properties. On the other hand, salinity anomalies showed a very homogeneous pattern in all latitudes for winter, spring, and summer, with values generally within ±0.5. However, during autumn important anomalies south of 18◦N were observed, with values in the range of ±1.5. In this region, the upwelling development, the river discharge, and the rainy season controlled the observed distribution (Yoo and Carton, 1990). To conclude, the data show a permanent annual upwelling regime observed north of 20◦N and a seasonal regime across 10–19◦N, in accordance with the climatology of previous studies. The data also confirm an increase in upwelling conditions north of 20◦N and an increase in downwelling conditions south of 20◦N. 3.2 Carbon dioxide variability The latitudinal distribution of the seasonal fCOsw 2data (Fig. 5a) showed the highest values between 18 and 23◦N for all seasons due to the variability imposed by the upwelling off Mauritania. fCOsw 2was consistently greater than the fCOatm 2. During winter, when the Cap Vert upwelling develops (Fig. 2), the 12–15◦N region also presented higher fCOsw 2values than those in the atmosphere. fCOsw 2data showed a latitudinal shift between the seasons following the shift observed in the upwelling index: in winter, the largest values were located between 19 and 24◦N; in spring, they were located between 16 and 22◦N; and during summer and autumn, the largest fCOsw 2values were recorded in the range 20 to 23◦N. The difference between fCOsw 2normalized to the mean SST of 22◦C for the region (NfCOsw 2)and fCOsw 2 (1f CO2=NfCOsw 2−fCOsw 2, Fig. 5b) reinforced the variability at 20–23◦N all year around and at 12–17◦N during winter and spring, indicating that upwelling is the major factor contributing to the fCO2variability. According to Takahashi et al. (1993), fCOsw 2increases with temperature at a rate of 4.3% µatm ◦C−1(between 15 and 26 µatm ◦C−1in this area) in a thermodynamically controlled system. At 27◦N, as SST increases, the rate was only 7.45 µatm ◦C−1due mainly to biological uptake and also to CO2outflux. At 20◦N the rate became negative with a value Biogeosciences, 14, 3859–3871, 2017 www.biogeosciences.net/14/3859/2017/ M. González-Dávila et al.: Changes in the partial pressure of carbon dioxide 3865 10 15 20 25 200 300 400 500 600 700 800 Latitude fCO2 ( atm) (a) W 10 15 20 25 −150 −100 −50 0 50 100 150 Latitude fCO2 ( atm) W (b) 10 15 20 25 200 300 400 500 600 700 800 Latitude fCO2 ( atm) Sp 10 15 20 25 −150 −100 −50 0 50 100 150 Latitude fCO2 ( atm) Sp 10 15 20 25 200 300 400 500 600 700 800 Latitude fCO2 ( atm) Sm 10 15 20 25 −150 −100 −50 0 50 100 150 Latitude fCO2 ( atm) Sm 10 15 20 25 200 300 400 500 600 700 800 Latitude fCO2 ( atm) Au 10 15 20 25 −150 −100 −50 0 50 100 150 Latitude fCO2 ( atm) Au Figure 5. Fugacity of CO2data in the Mauritanian–Cap Vert coastal region grouped by seasons: winter (W; December, January, and February), spring (Sp; March, April, and May), summer (Sm; June, July, and August), and autumn (Au; September, October, and November). Column (a) fCOsw 2latitudinal distribution. Column (b), difference between measured and fCOsw 2values normalized to a constant temperature of 22 ◦C. The averaged values for all cruises in Table S1 are shown in black for each season including the 95% confidence limits. The color code for each cruise is indicated in Table S1. of −10.9 µatm ◦C−1, clearly indicating the important injection of cool and CO2-rich seawater at the upwelling area. The injection is not being compensated for by the solubility nor by the biological carbon pumps. At 10◦N, the rate was still negative but only −4.3 µatm ◦C−1as a result of the seasonal upwelling. NfCOsw 2was related with SST (data not shown) in order to account for effects not removed during normalization. At latitudes 19 to 21◦N, in the upwelling vicinity of Cap Blanc, an inverse relationship of 70–100µatm ◦C−1 was found during winter and spring, while in summer and autumn the inverse relationship rate was reduced to 12– 18 µatm ◦C−1. While the upwelling indexes at those latitudes were quite constant throughout the year, different rates observed should be related to biological consumption of the CO2excess. However, during winter and spring the injection of CO2in the upwelling is not decreased by the biological activity in the area. But during the Chlorophyll amaximum (late spring and summer), most of the CO2was consumed and/or exported and, therefore, the rate was strongly reduced. www.biogeosciences.net/14/3859/2017/ Biogeosciences, 14, 3859–3871, 2017 3866 M. González-Dávila et al.: Changes in the partial pressure of carbon dioxide Figure 4 depicts the observed interannual trends (a1coefficient in Eq. 4) for the four experimentally recorded detrended parameters, together with the UI trend. Confidence intervals of the computed mean annual values for SST, SSS, fCOatm 2, and fCOsw 2were 0.13 ◦C, 0.06, 0.44, and 4.23 µatm, respectively. There was a clear SST trend whereby seawater along the VOS line track was getting cooler with maximum cooling rates at the location of Cap Blanc (21◦N) and Cap Vert upwellings (15◦N) with rates higher than −0.2 ◦C yr−1. Data from the first 3 years (2005 to 2008) at 21◦N showed lower temperatures with higher cooling rates that reached −0.7 ◦C yr−1, although 3 years of data are not representative. The area crossed by the VOS line along 17◦450W from 22 to 10◦N is located inside the 1000m isobath that is well inside the mean frontal activity in the Canary region, about 200km wide (Wang et al., 2015). The different changes in temperature in the coastal slope and offshore waters are related to the different origins of the waters upwelled from depths of about 100m to the surface (Mittelstaedt, 1983) that spread off the coastal area. The offshore water SST is less variable owing to longer residence time in the ocean surface. These effects and the fact that the VOS line keeps a track line that crossed the upwelling cells at a distance to the coast that varies among cells contribute to the observed spatial variability. There was no attempt to compare latitudinal and longitudinal effects on the observed values. Our experimental data, however, do not show any positive SST rates in the upwelling affected area, and only when the ship approached the Canary Islands did the trends become less negative, reaching a value of +0.02 ◦C yr−1at 27◦N, similar to those obtained for oceanic Atlantic water (Bates et al., 2014). fCOatm 2for the area showed the interannual increase of about 2 ±0.3 µatm yr−1observed in atmospheric stations, while fCOsw 2presented a heterogeneous distribution. South of 18◦N, the rate of increase was always higher than that in the atmosphere reaching a maximum value of 4.1 ±0.4 µatm yr−1at 10◦N. At 27◦N, fCOsw 2increased at a rate of 1.7 ±0.2 µatm yr−1similar to that determined at the ESTOC time series site (González-Dávila et al., 2010) located at 29◦100N, 15◦300W. In the Cap Blanc area, fCOsw 2 increased at an average rate of 2.5±0.4 µatm yr−1with the highest values in the period 2005 to 2008 (a rate of 4.6 ±0.5 µatm yr−1was computed with only those years). Around Cap Blanc, fCOsw 2always presented lower rates of increase than in the atmosphere with values well below 1 µatm yr−1. The observed decrease in SST and the trends in fCOsw 2can only be explained by a reinforced upwelling. North of 18◦N, the lowest rate of increase in fCOsw 2 compared to fCOatm 2, together with a decrease in temperature, indicated that upwelling is also favoring an increase in the net community production around the Mauritanian upwelling, consuming and/or exporting the CO2-rich upwelled waters favored by the lateral transport of the Mauritanian current (Lachkar and Gruber, 2013; Varela et al., 2015). The Figure 6. pH of surface waters in total proton scale and at in situ SST computed from total alkalinity (based on regional correlations with SST and SSS; Lee et al., 2006) and fCO2at 21 ±0.25◦N. The error bars represent the standard deviation of the computed data for each cruise for the selected latitude. The black curve shows the harmonic fitting of Eq. (4) for the data and the corresponding linear trend is also shown. upwelling intensification effects observed in the trends of our experimental data support the recent wind stress trends (Cropper et al., 2014; Varela et al., 2015; Santos et al., 2012) of increased upwelling-favorable winds, at least for the period 2005–2012 in the Canary upwelling region (Figs. 2 and 4). The intensification of the upwelling results in a change in the measured upwelled water properties due to either higher upwelling velocities or deeper source upwelled waters. However, what remains unclear from these records is to what extent those changes reflect upwelling variations due to climate change forcing versus natural decadal variability in the upwelling areas occurring over interannual timescales. Because the upwelling intensity is changing, other variables will also be affected. pHT ,is at 21 ±0.25◦N was computed from fCO2and alkalinity pairs of data. Alkalinity was computed from regional correlations with SST and SSS (Lee et al., 2006), which could underrepresent seasonal and interannual variations in upwelling areas. However, pH computed from fCO2values are relatively insensitive to errors in AT, and fCO2controls the magnitude and variability of pH (a 60µmol kg−1change in ATwill affect a 0.1% in pH, that is, about 0.01 pH units). Figure 6 depicts the computed pHT ,is(AT,fCO2)data and the harmonic fitting of Eq. (4) providing the seasonal variability and interannual trend. Considering the small systematic biases in interannual dynamics, we determined a decrease in pH at a rate of −0.003 ±0.001 yr−1(Fig. 6). This decrease is one of the highest rate values determined in several time series stations (Bates et al., 2014), where oceanic SST has only slightly increased in the last decades. However, at the Mauritanian upwelling area and at the location where our VOS line approached this region, SST decreased at a rate of −0.22 ±0.06 ◦C yr−1(Fig. 4). Solely Biogeosciences, 14, 3859–3871, 2017 www.biogeosciences.net/14/3859/2017/ M. González-Dávila et al.: Changes in the partial pressure of carbon dioxide 3867 Figure 7. Latitudinal distributions of seasonal and annual CO2fluxes (FCO2, mol m−2). Fluxes of CO2were computed using Nightingale et al. (2000) parameterization and satellite winds with a resolution of 6h. (a) Integrated year to year from 2005 to 2012 and (b) latitudinally integrated for 2005 to 2012 together with annual values for the North Atlantic Oscillation (NAO) index. Latitudinal distributions of FCO2 seasonally integrated from 2005 to 2012 are depicted for (c) winter (December, January, and February), (d) spring (March, April, and May), and (e) summer (June, July, and August). this decrease in temperature would increase the pH by a rate of +0.004 yr−1and the fCO2would decrease by 4 µatm yr−1. The net effect of the increase in the amount of rich CO2and lower pH upwelled waters in the Mauritanian upwelling would be, therefore, a decrease in the pH rate of over −0.007 ±0.002 units yr−1and an increase in fCO2of +6.5 ±0.7 µatm yr−1(with periods where those rates could reach values of −0.015 yr−1in pH and +10.5 µatm yr−1in fCO2as recorded during 2005–2008). Those values are greatly compensated for by the important decrease in the SST resulting in the determined rates of −0.003 ±0.001 pH units and +2.5 ±0.4 µatm of fCO2per year. This new data set of experimental values confirmed a decrease in SST and trends in fCOsw 2that can only be explained by reinforced upwelling conditions that favor an increase in the net community production around the Mauritanian upwelling together with a more corrosive environment with pH rates that change by more than −0.007 ±0.002 yr−1 at 21◦N. However, the decrease in SST in the upwelling cell buffers this pH rate to values around −0.003 ±0.001 yr−1 and +2.5 ±0.4 µatm yr−1in fCO2, still among the highest observed in other time series. www.biogeosciences.net/14/3859/2017/ Biogeosciences, 14, 3859–3871, 2017