Interactive comment on “ Ocean acidification in the North Atlantic : controlling mechanisms ”

This manuscript deals with the identification and quantification of the main drivers of pH changes in the Iceland and Irminger basins between 1981 and 2015. To do so, highquality data of 13 research cruises were combined, quality-checked and statistically analysed. Moreover, the contribution of Cant to changes in DIC was calculated and the change in pH was decomposed into five factors that were numerically estimated.


INTRODUCTION
The oceanic uptake of a fraction of the anthropogenic CO 2 (i.e., C ant ; CO 2 released from humankind's industrial and agricultural activities) has resulted in long-term changes in ocean CO 2 chemistry, commonly referred to as ocean acidification, OA (e.g., Caldeira andWickett, 2003, 2005;Raven et al., 2005;Doney et al., 2009;Feely et al., 2009).The changes in the ocean CO 2 chemistry result in declining pH and reduced saturation states for CaCO 3 minerals (e.g., Bates et al., 2014).The average pH (-log 10 [H + ]) of ocean surface waters has decreased by about 0.1 pH units since the beginning of the industrial revolution (1750), and based on model projections we expect an additional drop of 0.1-0.4 by the end of this century, even under conservative CO 2 emission scenarios (Caldeira and Wickett, 2005;Orr, 2011;Ciais et al., 2013).The rate of change in pH is at determined by developing either a full titration curve (Millero et al., 1993;Dickson and Goyet, 1994;Ono et al., 1998) or from single point titration (Pérez and Fraga, 1987;Mintrop et al., 2000), with an overall accuracy of 4 μmol•kg -1 .For samples without direct A T measurements, it was estimated using a 3D moving window multilinear regression algorithm (3DwMLR), using potential temperature (θ), salinity, nitrate, phosphate, silicate and oxygen as predictor parameters (Velo et al., 2013).The total dissolved inorganic carbon (DIC) samples were analysed with coulometric titration techniques (Johnson et al., 1993), and were calibrated with Certified Reference Materials (CRMs), achieving an overall accuracy of 2 μmol•kg -1 .The exception to the use of this analytical technique was the 1981 TTO-NAS (Transient Tracer in the Ocean-North Atlantic Survey) cruise, where DIC was determined potentiometrically (Bradshaw et al., 1981) and no CRMs were used.The TTO-NAS DIC measurements were deemed unreliable (Brewer et al., 1986), therefore, the DIC values compiled in the GLODAPv2 merged data product are those calculated from pCO 2 and revised A T reported by Tanhua and Wallace (2005).pH was determined either potentiometrically (Dickson, 1993a, b) using pH electrodes or, more commonly, with a spectrophotometric method (Clayton and Byrne, 1993) using either scanning or diode array spectrophotometers and m-cresol purple as an indicator.The spectrophotometric pH determination has a typical precision of 0.0002-0.0004pH units (Clayton and Byrne, 1993;Liu et al., 2011).However, Carter et al. (2013) reported an inaccuracy of the spectrophotometric pH determination of 0.0055 pH units.When direct pH measurements were not performed, it was computed from A T and DIC using the thermodynamic equations of the seawater CO 2 system (Dickson et al., 2007) and the CO 2 dissociation constants of Mehrbach et al. (1973) refitted by Dickson and Millero (1987).For these calculated pH values, we estimated an uncertainty of 0.006 pH units by random propagation of the reported A T and DIC accuracies.The exception to the latter is the 1981 TTO-NAS cruise, whose DIC problems caused the estimated uncertainty for calculated pH values to be slightly higher (0.008 pH units).A T data from the 1981 TTO-NAS cruise were checked against A T values generated by the 3DwMLR (Velo et al., 2013).A T values differing by more than two times the standard deviation (confidence interval; 7 µmol•kg -1 ) of the difference between measured A T and 3DwMLR predicted A T were replaced with the predicted A T value.However, for leg 6 of the 1981 TTO-NAS cruise (which was not analysed by Tanhua and Wallace ( 2005)) the limit of substitution for the predicted A T value was lowered to 4 µmol•kg -1 .Note that the effect of A T corrections on pH trends is negligible, since A T corrections of 4 µmol•kg -1 lead to pH changes lower than a thousandth.The pH values reported here are at in situ conditions and on the total scale (pH Tis ).

Anthropogenic CO 2 (i.e., C ant ) estimation
C ant concentrations were estimated using the back-calculation method φC T 0 (Pérez et al., 2008;Vázquez-Rodríguez, 2009a) that has previously been applied for the entire Atlantic Ocean (Vázquez-Rodríguez et al., 2009b).Back-calculation methods determine C ant for any sample in the water column as the difference between DIC concentration at the time of the measurement and the DIC concentration it would have had in preindustrial times.This is represented as the difference in preformed DIC between the time of observation and the preindustrial as: The φC T 0 method presents two main advantages.First, the spatiotemporal variability of A T 0 is taken into account.And second, C ant estimation needs no "zero-C ant " reference, since the parameterizations of A T 0 and ΔC diseq are determined using the subsurface layer as reference for water mass formation conditions (Vázquez-Rodríguez et al., 2012a).The overall uncertainty of the method has been estimated at 5.2 μmol•kg -1 (Pérez et al., 2008;Vázquez-Rodríguez, 2009a).
The reproducibilities and uncertainties of the main variables were determined from the deep waters sampled at Iberian Abyssal Plain during the seven repeats of the OVIDE line, since these waters are expected to be in nearsteady state.The confidence intervals of those samples for each cruise (Table 2) were taken as an estimate of the uncertainty of the methodologies.The uncertainties of the Apparent Oxygen Utilization (AOU; the difference between the saturated concentrations of oxygen calculated using the equations of Benson and Krause (1984) and the measured concentrations of oxygen), A T and pH on the total scale at 25ºC (pH T25 ) for the seven cruises were similar.The confidence intervals of C ant (2.4-3.2 μmol•kg -1 ) and pH T25 (0.004-0.006 pH units) across the seven cruises are lower than the inherent uncertainty of the φC T 0 estimates (5.2 μmol•kg -1 ) and the accuracy of the spectrophotometric pH measurements (0.0055 pH units), which provides confidence that these data are suitable for trend determination.The confidence intervals of the C ant estimates are rather similar than in other regions where C ant has been compared across many cruises (i.e., 2.4 μmol•kg -1 in the South Atlantic Ocean, Ríos et al. (2003); 2.7 μmol•kg -1 in the Equatorial Atlantic Ocean, 24ºN, Guallart et al. (2015); and 2.7 μmol•kg -1 reported from a transect along the western boundary of the Atlantic Ocean from 50ºS to 36ºN, Ríos et al. (2015)).The confidence interval of the mean values of the Iberian Abyssal Plain samples across the seven cruises (last row of Table 2) was taken as an estimate of the reproducibility of the methodologies.The high reproducibilities, an order of magnitude lower than the uncertainties, render confidence to the estimated trends.
To better determine the interfaces between layers and the average value of each variable in each layer, cruise bottle data were linearly interpolated onto each dbar before determining average variable values, an improvement with respect to the previous approaches of Pérez et al. (2008Pérez et al. ( , 2010) ) and Vázquez-Rodríguez et al. (2012b).Upper layer data (pressure ≤ 100 dbar) were replaced with the mean value in the pressure range 50-100 dbar to reduce the influence of seasonal differences in sampling on the inter-annual trends (Vázquez-Rodríguez et al., 2012a).Then, the interpolated profiles were divided into the different water mass density intervals (Fig. 1b).Next, the variables were averaged over each density layer on a station by station basis for each cruise.The average values of the variables for each layer and their confidence intervals can be found in the Supplementary Table S1.

pH deconvolution
Changes in ocean pH may be brought about by changes in in situ temperature (T is ), salinity (S), A T , and/or DIC, of which changes in the latter may be brought about by C ant uptake or by natural processes (C nat ), such as remineralisation.C nat is determined as the difference between measured DIC and estimated C ant .To estimate how much each of these altogether five factors contributed to the observed change in pH, we assumed linearity and decomposed the observed pH changes into these potential drivers according to: To estimate

𝜕𝑣𝑎𝑟
(where  refers to each of the drivers: T is , S, A T and DIC) we calculated the mean pH Tis for each layer and cruise using the real average value of  but keeping the values of the other three drivers constant and equal to the mean value for the layer over all the cruises.To estimate each   term we performed a linear regression between  and time for each layer.
Trends of all variables involved in Eq. (2) were calculated using the annual interpolation of the observed values to avoid the bias due to the reduced availability of cruises during the 80's and 90's with respect to the 2000's.

Mean distribution of water mass properties
The Irminger and Iceland basins in the North Atlantic are characterized by warm and saline surface waters, and cold and less saline intermediate and deep waters (Fig. 2a,b).The central waters (here represented by the SPMW layer), which dominates the upper ~700 m, are warmer and saltier in the Iceland basin than in the Irminger basin, reflecting the water mass transformation that takes place along the path of the North Atlantic Current (NAC) (Brambilla and Talley, 2008).In particular, the mixing of the SPMW layer with the surrounding waters while flowing around the Reykjanes Ridge (evident in the salinity distribution; see also García-Ibáñez et al. (2015)), in conjunction with the air−sea heat loss, results in a colder and fresher SPMW layer in the Irminger basin.The uLSW and cLSW layers, below the SPMW layer, are warmer and saltier in the Iceland basin due to their mixing with the surrounding waters during their journey from their formation regions (Bersch et al., 1999;Pickart et al., 2003;García-Ibáñez et al., 2015).The ISOW layer dominates at depths beneath the cLSW layer.
This layer is warmer and saltier in the Iceland basin, reflecting its circulation.ISOW comes from the Iceland-Scotland sill and flows southwards into the Iceland basin, where it mixes with the older North Atlantic Deep Water (NADW).Then, it crosses the Reykjanes Ridge through the Charlie−Gibbs Fracture Zone (Fig. 1a), where it mixes with the recently ventilated cLSW and DSOW, becoming colder and fresher.In the bottom of the Irminger basin, a fifth layer is distinguished, DSOW, being the coldest and freshest layer of the section.coinciding with relatively high AOU and DIC values (Fig. 2e,f).This layer could be associated to an area of slower circulation where the products of the remineralization of the organic matter accumulate.This thermocline layer could also been influenced by waters of southern origin (Sarafanov et al., 2008), which are advected into the region by the NAC, whose arrival is closely related with the North Atlantic Oscillation (Desbruyères et al., 2013).The presence of this low pH layer lowers the average pH of our SPMW layer in the Iceland basin compared to the Irminger basin (Fig. 3).An opposite pattern is found in the uLSW layer.The water mass formation occurring in the Irminger basin (Pickart et al., 2003;García-Ibáñez et al., 2015;Fröb et al., 2016;Piron et al., 2016) transfers recently ventilated low DIC and high pH waters to depth, which causes the mean pH of uLSW in the Irminger basin to be higher than in the Iceland basin.Finally, the layers that contain the overflow waters have the lowest pH values.The presence of the older NADW in the ISOW layer in the Iceland basin decreases the mean pH of this layer here, making it lower than in the Irminger basin.
The surface waters of the section have low DIC values, which rapidly increase when increasing depth (Fig. 2f).
The low DIC values in the uppermost ~200 m are also related to the photosynthetic activity that withdraws DIC near the surface and decrease with depth (Fig. 2h), because Cant enters the ocean from the atmosphere.The C nat distribution has an opposite pattern, similar to that of the AOU distribution (Fig. 2e), with low surface values and high bottom values (Fig. 2g), for reasons discussed above.
The A T distribution along the section resembles the salinity distribution, with high values associated with the relatively saline central waters and low and almost homogeneous values in the rest of the section (Fig. 2d).The exception comes with the deep waters of the Iceland basin, which have among the highest A T values while salinity is not extraordinarily high.This reflects the influence of NADW, which contains relatively large amounts of silicate related to the influence of the Antarctic Bottom Water.

Water mass acidification and drivers
Trends of pH Tis in each layer and basin are presented in Table 3 and Fig. 3.The pH Tis has decreased in all layers of the Irminger and Iceland basins during the time period of more than 30 years  that is covered by our data.The trends are stronger in the Irminger basin due to the presence of younger waters.The rate of OA decreases with depth, except for the DSOW layer that has acidification rates close to those found in the cLSW layer.This indicates that DSOW is a newly formed water that has recently been in contact with the atmosphere.Moreover, the acidification rate in the ISOW layer in the Irminger basin is relatively low, which could be related to the increasing importance on this layer of the relatively old NADW with the diminution in volume of cLSW since mid-90s (Lazier et al., 2002;Yashayaev, 2007).The normalized pH values (pH N ) for each layer was obtained using multiple linear regressions between the observed mean pH SWS25 (pH at seawater scale and 25ºC) and the observed mean values of θ, salinity, silicate and AOU, referred to the mean climatological values of θ, salinity, silicate and AOU compiled in WOA05 (http://www.nodc.noaa.gov/OC5/WOA05/pr_woa05.html).This normalization, combined with the lower temporal coverage  and the fact that they evaluated trends in pH at 25ºC and not at in situ conditions renders direct comparisons between their and our derived trends difficult.
To infer the causes of the acidification trends reported here, we decomposed the pH trends into their individual components as described in Sect.2.2.The results are presented in Table 3.The sum of the pH changes caused by the individual drivers (in situ temperature, salinity, A T and DIC) matches the observed pH trends, which renders confidence to the method.
The temperature changes (Fig. 4a,b) have generally resulted in small to negligible pH declines (Table 3).
Specifically, warming corresponds to a pH decrease of more than 0.0001 pH units•yr -1 in the SPMW layer of both basins and in the LSW layers of the Irminger basin, while the effect of temperature changes on pH in the other layers is negligible.Temperature driven pH change is larger in the LSW layers in the Irminger than in the Iceland basin.In the case of the uLSW layer, this is possibly explained by the deep convection occurring in the Irminger basin (Pickart et al., 2003;García-Ibáñez et al., 2015;Fröb et al., 2016;Piron et al., 2016).In the case of the cLSW layer, the higher pH changes driven by temperature changes in the Irminger basin could be explained by the rapid advection of this water mass from the Labrador Sea to this basin (Yashayaev et al., 2007).
The temperature effect on pH evaluated here is mostly thermodynamic.The same applies to the salinity effect, which however is small to negligible, reflecting that salinity changes in the region (Fig. 4c,d The A T has increased in all layers (Fig. 5a,b), corresponding to increasing pH (Table 3), which counteracts the acidification from the CO 2 absorption.The contribution from A T to reduce ocean acidification is significant for all the layers, except for ISOW of the Irminger basin and uLSW of the Iceland basin (in which A T trends over time are not significant; Fig. 5a,b).The A T increasing trends observed in SPMW may indicate the increasing presence of waters of subtropical origin (with higher A T ) as the subpolar gyre was shrinking since mid-90s (e.g., Flatau et al., 2003;Häkkinen and Rhines, 2004;Böning et al., 2006).The A T effect is evident in the ISOW layer of the Iceland basin, which can be explained by the circulation and mixing of this layer.As ISOW flows downstream along the Reykjanes Ridge, it mixes with cLSW and NADW (van Aken and de Boer, 1995;Fogelqvist et al., 2003).The reduced volume of cLSW since mid-90s (Lazier et al., 2002;Yashayaev, 2007) has increased the importance of NADW (with high A T ; Fig. 2h) in the ISOW layer, making the pH decrease of the ISOW layer of the Iceland basin lower than in the Irminger basin.
The DIC increase (Fig. 5c,d) is the main cause of the observed pH decreases, and corresponds to pH drops between -0.00085 and -0.00134 pH units•yr -1 (Table 3).The waters in both the Irminger and Iceland basins gained DIC in response to the increase in atmospheric CO 2 ; the convection processes occurring in these basins (Pickart et al., 2003;Thierry et al., 2008;de Boisséson et al., 2010;García-Ibáñez et al., 2015;Fröb et al., 2016;Piron et al., 2016) and in the surrounding ones (i.e., Labrador and Nordic Seas) provide an important pathway for DIC to pass from the surface mixed layer to the intermediate and deep layers.The effect of the DIC increase on pH is generally dominated by the anthropogenic component (Table 3).The exception comes with the cLSW layer of the Irminger basin, where dominates the natural component resulting from the aging of the layer.All layers have higher C ant increase rates in the Irminger basin than in the Iceland basin (Fig. 6a,b), and therefore larger pH declines, presumably a result of the proximity of the Irminger basin to the regions of deep water formation.The highest C ant increase rates are found in the SPMW layer, owing to its direct contact with the atmosphere, and result in the highest rates of pH decrease.The higher pH drops related to C ant increase found in the SPMW layer in the Irminger basin compared to those found in the Iceland basin layer, can be related to the differences in the rise in C ant levels in both basins.In the Irminger basin, the rise in C ant levels of the SPMW layer correspond to about 85% of the rate expected from a surface ocean maintaining its degree of saturation with the atmospheric CO 2 rise (computed using as reference the measurements of Mauna Loa), while in the Iceland basin, this rate is about 73% of the expected rate.The lower fraction in the Iceland basin compared to the Irminger basin is a consequence of the inclusion of the aforementioned poorly ventilated thermocline waters in our SPMW layer (Fig. 2e,h).Note than none of the C ant trends of the SPMW layers correspond to 100% of the rate expected from assuming saturation with the atmospheric CO 2 rise.This can be explained by the fact that surface waters CO 2 concentration rise lags that of the atmosphere by between two to five years in this region (Biastoch et al., 2007;Jones et al., 2014).We also note that the temperature and A T changes impact the pH of SPMW, decreasing and increasing it, respectively.This could indicate the increasing presence of warmer and more saline (with higher A T ) waters of subtropical origin, which, because A T effects dominate, in last instance counteracts the effects of increasing DIC values.Overall this change can be explained as the result of the contraction of the subpolar gyre that took place since mid-90s (e.g., Flatau et al., 2003;Häkkinen and Rhines, 2004;Böning et al., 2006).Wakita et al. (2013)  The greater influence of C nat in the cLSW layer is the result of the aging of this water mass after its last formation event, in the mid-90s (eg., Lazier et al., 2002;Azetsu-Scott et al., 2003;Kieke et al., 2007;Yashayaev, 2007).C nat also contributes to pH changes in the ISOW layer of the Iceland basin, which is related to the increasing influence of the relatively old NADW over time due to the decreasing contribution of LSW (Sy et al., 1997;Yashayaev, 2007;Sarafanov et al., 2010;García-Ibáñez et al., 2015).

CONCLUSIONS
The progressive acidification of the North Atlantic waters has been assessed from direct observations obtained over the last three decades , with the greatest pH decreases observed in surface and intermediate waters.By separating the observed pH change into its main drivers, we corroborate that the observed pH decreases are a consequence of the oceanic C ant uptake and in addition we find that they have been partially offset by A T increases.However, while the C ant concentration of the upper layer roughly keeps up with that expected from rising atmospheric CO 2 , the pH decreases at a lower rate than expected from C ant increase.The where the preformed DIC for the time of observation is represented as the measured DIC (DIC meas ) less any DIC added to the water due to organic matter remineralisation and calcium carbonate dissolution (ΔC bio ), and the preindustrial preformed concentration is represented by the DIC concentration the water would have if in Biogeosciences Discuss., doi:10.5194/bg-2016-66,2016 Manuscript under review for journal Biogeosciences Published: 29 February 2016 c Author(s) 2016.CC-BY 3.0 License.equilibrium with the preindustrial atmosphere (DIC preind ) less any offset from such an equilibrium value, known as the disequilibrium term (ΔC diseq ).The procedure requires DIC and A T as input parameters, and the empirical parameterization of the preformed A T (A T 0 ) for the computation of the calcium carbonate dissolution and of the ΔC diseq term.

Finally
, the average values in each density layer were determined for each cruise taking into account the Biogeosciences Discuss., doi:10.5194/bg-2016-66,2016 Manuscript under review for journal Biogeosciences Published: 29 February 2016 c Author(s) 2016.CC-BY 3.0 License.thickness of the layer and the separation between stations.Note that average values of pressure sensitive parameters, i.e. pH Tis , were referred to the mean pressure of the layer over the studied time period to avoid the effects of the heaving of the water masses due to warming and/or of the sampling strategy over the pH trends.
Biogeosciences Discuss., doi:10.5194/bg-2016-66,2016 Manuscript under review for journal Biogeosciences Published: 29 February 2016 c Author(s) 2016.CC-BY 3.0 License.The general pattern of pH Tis (Fig. 2c) follows by and large the distribution expected from the surface production of organic material and remineralisation at depth.The high surface values (> 8.05) are the result of the withdrawing of DIC by photosynthetic activity, while the values generally decrease with depth down to < 7.95 in the deepest layers, because of the DIC concentration increase resulting from remineralisation.This overall pattern is disrupted at ~500 m in the Iceland basin by a layer with relatively low pH Tis values (< 7.98), from seawater.Below ~200 m the DIC distribution is almost homogeneous, only disrupted by relatively high values in the Iceland basin at ~500 m associated with the thermocline layer, and at the bottom, associated with the old NADW.The gradients in DIC anthropogenic and natural components are much stronger.This is because the C ant and C nat distributions are anti-correlated.The C ant values are high, close to saturation (80% of saturation), BiogeosciencesDiscuss., doi:10.5194/bg-2016Discuss., doi:10.5194/bg--66, 2016     Manuscript under review for journal Biogeosciences Published: 29 February 2016 c Author(s) 2016.CC-BY 3.0 License.The observed rate of pH Tis decrease in the SPMW layer of the Iceland basin (-0.0012 ± 0.0001 pH units•yr -1 ; Table3, Fig.3b) is in agreement with that observed at the Iceland Sea time-series (68ºN, 12.66ºW;Olafsson et al. (2009Olafsson et al. ( , 2010))) for the period 1983-2014 (-0.0014 ± 0.0005 pH units yr -1 ; Bates et al. (2014)).Our rates in the SPMW layer of both basins are slightly lower than those observed at the Subtropical Atlantic time-series stations ESTOC (29.04ºN, 15.50ºW; Santana-Casiano et al. (2007), González-Dávila et al. (2010)) for the period 1995-2014 (-0.0018 ± 0.0002 pH units•yr -1 ; Bates et al. (2014)) and BATS (32ºN, 64ºW; Bates et al. (2014)) for the period 1983-2014 (-0.0017 ± 0.0001 pH units•yr -1 ; Bates et al. (2014)).However, our rate of pH Tis decrease in the SPMW layer in the Irminger basin (-0.0013 ± 0.0001 pH units•yr -1 ) is only half of that observed in the sea surface waters of the Irminger Sea time-series (64.3ºN, 28ºW; Olafsson et al. (2010)) for the period 1983-2014 (-0.0026 ± 0.0006 pH units yr -1 ; Bates et al. (2014)), which is exceptionally high compared to the other time series summarized here.Comparing with the Pacific Ocean, the OA rates in the Iceland and Irminger basins are slightly lower than those reported for the Central North Pacific based on data from the time-series station HOT (22.45ºN, 158ºW; Dore et al. (2009)) for the period 1988-2014 (-0.0016 ± 0.0001 pH units•yr -1 ; Bates et al. (2014)), but are in agreement with those found by Wakita et al. (2013) in the winter mixed layer at the Subarctic Western North Pacific (time-series stations K2 and KNOT) for the period 1997-2011 (-0.0010 ± 0.0004 pH units•yr -1 ).Vázquez-Rodríguez et al. (2012b) have previously studied the pH changes in the different water masses of the Irminger and Iceland basins.These authors carried out a pH normalization to avoid potential biases due to different ventilation stages and rates of each layer, from the different spatial coverage of the evaluated cruises.
also found lower than expected acidification rates in the surface waters of the Pacific Ocean, which they explained as being the consequence of increasing A T .Finally, the strong Biogeosciences Discuss., doi:10.5194/bg-2016-66,2016 Manuscript under review for journal Biogeosciences Published: 29 February 2016 c Author(s) 2016.CC-BY 3.0 License.influence of anthropogenic component on the pH decrease of the DSOW layer stands out, and is the main agent of the pH decline in this layer.The pH change related to C nat changes (Fig.6c,d) can be interpreted as changes related to ventilation of water masses and water mass changes (with different A T and DIC).Higher pH decreases related to C nat changes indicate lack of ventilation and accumulation of DIC from remineralised organic material.This is clearly the case for the cLSW layer, where the observed pH decrease is caused by a combination of the effects of C ant and C nat .
increasing arrival of salty and alkaline subtropical waters transported by the NAC to the study region related to the contraction of the subpolar gyre since mid-90's buffers the acidification caused by the C ant increase in the upper layer.The acidification rates in intermediate waters are similar to those in the surface waters, and are caused by a combination of anthropogenic and non-anthropogenic components.The acidification of cLSW due to the C ant uptake is reinforced by the aging of this water mass from the end of the 1990s onwards.The pH of the deep waters of the Irminger basin, DSOW, has clearly decreased in response to anthropogenic forcing.We also observe that water mass warming contributes between 13 and 18% to the pH decrease of the upper and intermediate waters of the Irminger basin, and 34% to the pH decrease of the upper waters of the Iceland basin.

Table 1 :
List of hydrographic cruises used in this study (Fig. 1a).P.I. denotes principal investigator, and #Biogeosciences Discuss., doi:10.5194/bg-2016-66,2016 Manuscript under review for journal Biogeosciences Published: 29 February 2016 c Author(s) 2016.CC-BY 3.0 License.Table 2: Mean values ± confidence interval of pressure (in dbar), potential temperature (θ, in ºC), salinity, Apparent 646 Oxygen Utilization (AOU, in μmol•kg -1 ), total alkalinity (A T , in μmol•kg -1 ), anthropogenic CO 2 (C ant , in μmol•kg -1 ) and 647 pH at total scale and 25ºC (pH T25 ) for the bottom waters of the Iberian Abyssal Plain sampled during the seven 648 OVIDE cruises."n" represents the number of data considered in each cruise.The last row represents the inter-cruise 649 confidence interval (i.e., the confidence interval of the mean values across the seven cruises).
the main water masses in the Irminger and Iceland basins for the period 1981-2015.pH changes caused 653 by the main drivers (in situ temperature, T is ; salinity, S; total alkalinity, A T ; total dissolved inorganic carbon, DIC; 654 anthropogenic CO 2 , C ant ; natural DIC, C nat ) are also shown, as well as the pH changes result of the sum of the pH 655 changes caused by the individual drivers (     ).All the trends are calculated based on the annually 656 interpolated values and are in 10 -3 pH units•yr -1 .Values in parenthesis are the percentages of the observed pH change 657 explained by each one of its drivers.Confront Fig. 1 for water mass acronyms.

Figure 2 :
Figure 2: Mean distributions along the cruise track, from Greenland (left) to the Iceland basin (right) over study 670

Figure 3 :
Figure 3: Temporal evolution of average pH at total scale and in situ conditions (pH Tis ) in the main water masses of 676

Figure 4 :
Figure 4: Temporal evolution between 1981 and 2015 of average (a and b) in situ temperature (T is , in ºC) and (c and d) salinity in the main water masses of the Irminger (a and c) and Iceland (b and c) basins.Each point represents the average property of a particular layer (SPMW (red dots), uLSW (blue dots), cLSW (black dots), ISOW (green dots) and DSOW (magenta dots)) at the time of each cruise (Table S1).The error bars are 2σ.The inset boxes give the trends (in 10 -3 units•yr -1 ) ± standard error of the estimate and the correlation coefficients (r 2 ), resulting from the annually interpolated values.* denotes that the trend is statistically significant at the 90% level (p-value < 0.1), ** at the 95% level (p-value < 0.05), and *** at 99% level (p-value < 0.01).Confront Fig. 1 for layer acronyms.

Figure 5 :
Figure 5: Temporal evolution between 1981 and 2015 of average (a and b) total alkalinity (A T , in μmol•kg -1 ) and (c and d) total dissolved inorganic carbon (DIC, in μmol•kg -1 ) in the main water masses of the Irminger (a and c) and Iceland (b and d) basins.Each point represents the average property of a particular layer (SPMW (red dots), uLSW (blue dots), cLSW (black dots), ISOW (green dots) and DSOW (magenta dots)) at the time of each cruise (Table S1).The error bars are 2σ.The inset boxes give the trends (in μmol•kg -1 •yr -1 ) ± standard error of the estimate and the correlation coefficients (r 2 ), resulting from the annually interpolated values.*** denotes that the trend is statistically significant at the 99% level (p-value < 0.01).Confront Fig. 1 for layer acronyms.