Volume 118, Issue 4 p. 1584-1592
Regular Article
Free Access

Variability of North Sea pH and CO2 in response to North Atlantic Oscillation forcing

Lesley A. Salt

Corresponding Author

Lesley A. Salt

Royal Netherlands Institute for Sea Research, Texel, Netherlands

Corresponding author: L. A. Salt, Royal Netherlands Institute for Sea Research, Texel, NL-1790 AB, Netherlands. ([email protected])Search for more papers by this author
Helmuth Thomas

Helmuth Thomas

Department of Oceanography, Dalhousie University, Halifax, Nova Scotia, Canada

Search for more papers by this author
A. E. Friederike Prowe

A. E. Friederike Prowe

Centre for Ocean Life, National Institute of Aquatic Resources, Technical University of Denmark, Charlottenlund, Denmark

GEOMAR Helmholtz Centre for Ocean Research, Kiel, Germany

Search for more papers by this author
Alberto V. Borges

Alberto V. Borges

Unité d'Océanographie Chimique, Institut et Physique B5, Université de Liège, Liège, Belgium

Search for more papers by this author
Yann Bozec

Yann Bozec

CNRS,UMR 7144, Equipe Chimie Marine, Station Biologique de Roscoff, Roscoff, France

Search for more papers by this author
Hein J. W. de Baar

Hein J. W. de Baar

Royal Netherlands Institute for Sea Research, Texel, Netherlands

Search for more papers by this author
First published: 06 November 2013
Citations: 28


[1] High biological activity causes a distinct seasonality of surface water pH in the North Sea, which is a strong sink for atmospheric CO2 via an effective shelf pump. The intimate connection between the North Sea and the North Atlantic Ocean suggests that the variability of the CO2 system of the North Atlantic Ocean may, in part, be responsible for the observed variability of pH and CO2 in the North Sea. In this work, we demonstrate the role of the North Atlantic Oscillation (NAO), the dominant climate mode for the North Atlantic, in governing this variability. Based on three extensive observational records covering the relevant levels of the NAO index, we provide evidence that the North Sea pH and CO2 system strongly responds to external and internal expressions of the NAO. Under positive NAO, the higher rates of inflow of water from the North Atlantic Ocean and the Baltic outflow lead to a strengthened north-south biogeochemical divide. The limited mixing between the north and south leads to a steeper gradient in pH and partial pressure of CO2 (pCO2) between the two regions in the productive period. This is exacerbated further when coinciding with higher sea surface temperature, which concentrates the net community production in the north through shallower stratification. These effects can be obscured by changing properties of the constituent North Sea water masses, which are also influenced by NAO. Our results highlight the importance of examining interannual trends in the North Sea CO2 system with consideration of the NAO state.

Key Points

  • The North Sea shelf pump is affected by the North Atlantic Oscillation
  • Variability in the water mass composition of the North Sea impacts pH
  • The North Sea carbon system responds to forcing on different time scales

1 Introduction

[2] Coastal and marginal seas play an important role in the atmosphere-ocean carbon exchange, responsible for a disproportionately large amount of primary production relative to their surface area [Gattuso et al., 1998], which is triggered by large inputs of nutrients and organic carbon from the adjacent ocean, land, and atmosphere [Wollast, 1998; Thomas et al., 2005a; Thomas et al., 2008b]. The export of this carbon into the adjacent open ocean, thus sequestering large quantities of anthropogenic CO2, is known as the continental shelf pump [Tsunogai et al., 1999; Thomas et al., 2004]. The effectiveness of this pump is related to the physical and biological conditions governing the CO2 disequilibrium between the atmosphere and the sea surface, which in turn is thermodynamically responsible for the CO2 uptake and the subsequent variation in pH and partial pressure of CO2 (pCO2). The codependence of this variability on physical and biological factors makes it difficult to discern the increase in CO2 solely attributable to atmospheric pCO2 increases [Santana-Casiano et al., 2007].

[3] The North Sea is a shelf sea on the northwest European continent with links to the North Atlantic Ocean in the south and the north. The majority of water exchange with the North Atlantic occurs in the northern North Sea, where inflowing waters enter through the Orkneys-Shetland shelf, Shetland shelf, and the Norwegian channel, with the Norwegian Trench providing the main exit pathway of circulated water out of the North Sea [Otto et al., 1990; Winther and Johannessen, 2006]. The total net carbon export to the North Atlantic via the Norwegian Trench has been estimated to be 6±1 × 1012 mol C yr−1 [Wakelin et al., 2012], which includes more than 90% of the CO2 drawn down from the atmosphere in the North Sea [Thomas et al., 2005a]. The effectiveness of the North Sea CO2 pump is determined principally by water mass exchange between the North Sea and the North Atlantic Ocean in combination with the export of carbon out of the surface layer, predominantly as sinking particulate organic matter [Thomas et al., 2004; Bozec et al., 2005]. The latter applies most significantly to the deeper (>50 m), seasonally stratified northern part of the North Sea (>56°N), where seasonality in pH is controlled by production of organic matter and its export [Thomas et al., 2009]. The seasonality of pH and pCO2 in the southern North Sea is also closely coupled to primary production; however, the export of organic matter is reduced by the rapid remineralization, which takes place in the shallow (<50 m) and well-mixed water column [Borges and Frankignoulle, 1999, 2002; Schiettecatte et al., 2006, 2007].

[4] Over the North Atlantic Ocean, a number of atmospheric teleconnection patterns influence climate variability of which the North Atlantic Oscillation (NAO) is the most prominent. The NAO Index (NAOI) is defined as the difference of atmospheric sea level pressure (SLP) between the Icelandic low and the Azores high and accounts for the greatest proportion (>30%) of the observed SLP variance in the region from December to March [Hurrell, 2003]. The effects of the NAO control a number of large-scale processes at different timescales [e.g. Hurrell, 1995; Hurrell and van Loon, 1997; Greatbatch, 2000]. Since the atmospheric pressure anomalies are most pronounced during northern hemisphere winter [Greatbatch, 2000] and the ratio of signal to noise is the highest [Hurrell and van Loon, 1997], commonly (but not exclusively) the NAOI recorded during December, January, and February (DJF), has been referred to in the literature yielding the most accentuated NAOI variability. Thus, although the NAOI is commonly established for DJF consequences of the NAO have been identified at various timescales. While the atmospheric realm responds at immediate, shorter timescales, for example via variability of trajectory, direction and strength of winds, the oceanic system responds at times scales from seasons to decades, for example via altered circulation patterns at various spatial scales [e.g., Hurrell, 1995; Hurrell and van Loon, 1997; Greatbatch, 2000; Thomas et al., 2008a]. Modeling studies [Thomas et al., 2008b; Levine et al., 2011; McKinley et al., 2011] and long-term observations [Santana-Casiano et al., 2007; Pérez et al., 2010; Bates, 2001] suggest that NAO-driven changes exert significant control over the interannual variability of hydrographic properties and in turn the uptake of CO2.

[5] Recently, observations from the North Sea [Thomas et al., 2007] and the North Atlantic Ocean [Watson et al., 2009] indicate that the surface water pCO2 has risen faster than the atmospheric pCO2, which has been linked to the effects of the NAO, with varying time scales of effect across the region [Santana-Casiano et al., 2007; Thomas et al., 2008a]. In-depth discussions of the role of the NAO in regulating the climate and weather go beyond the scope of the present paper, and can be found elsewhere, for example in Hurrell [1995], or Greatbatch [2000]. However, within the North Sea, many processes have demonstrated significant correlations with the wintertime (DJF) NAOI, which, as we later show, impact the carbonate system. The strength of the water mass exchange between the North Atlantic and the North Sea is regulated by the NAO, which in turn affects physical and chemical characteristics of the North Sea water column for the annual cycle. As a consequence of enhanced water mass exchange between North Atlantic and North Sea during years of positive NAOI (NAO+)[Winther and Johannessen, 2006; Kühn et al., 2010], the corresponding increase of the North Sea's nutrient inventory leads to higher productivity throughout the productive season from spring until the end of summer [Pätsch and Kühn, 2008]. Characteristic in the North Sea's response to NAO forcing can also be a hysteresis between cause and effect: In the North Sea, NAO+ has further been associated with higher precipitation across Scandinavia with drier conditions over central Europe [Ionita et al., 2011]. Changes in precipitation patterns over the drainage area of the Baltic Sea during winter will affect the runoff from the Baltic Sea into the North Sea over the relevant (runoff) seasons. Additionally, stronger westerly winds during winter, correlated with a positive wintertime NAOI, push North Sea water into the Baltic Sea, a process that in turn leads to an enhanced outflow from the Baltic Sea into the North Sea during the subsequent spring and summer [Hordoir and Meier, 2010]. These patterns are generally reversed during a NAO negative (NAO−) state. From these few examples, it is evident that despite the fact that the NAOI is commonly established for winter (DJF), the consequences for the North Sea are complex, not restricted to the winter season, and can be (partly) masked or even overridden by local or regional weather. One of the primary aims of this paper is to unravel this complex situation and to understand the variability of the North Sea CO2 system in front of this background.

[6] The intra-annual variability of pCO2 and dissolved inorganic carbon (DIC) has been well documented in the North Sea [Frankignoulle and Borges, 2001; Thomas et al., 2005b; Prowe et al., 2009; Bozec et al., 2006; Omar et al., 2010; Artioli et al., 2012]; however, on long time scales little work has been done [Thomas et al., 2005a; Thomas et al., 2007; Borges and Gypens, 2010] and despite the proximity of the North Atlantic, the drivers for interannual variability of the CO2 system in relation to the North Atlantic variability have not yet been investigated from field observations. Further, the full scope of variations in pH and pCO2 is still difficult to constrain when attempting to reproduce them in models [Prowe et al., 2009; Gypens et al., 2011; Lorkowski et al., 2012; Artioli et al., 2012]. Relying on a unique data set covering three basin-wide occupations of the North Sea during all relevant NAO phases, we are now able to examine the influences of NAO forcing on the North Sea carbonate system.

2 Materials and Methods

[7] The North Sea was sampled during August/September 2001, 2005, and 2008 using a station grid of approximately 90 identical stations each time [Bozec et al., 2005, 2006]. These years experienced NAO (DJF) indices of −1.9, 0.12, and 2.1, respectively (http://www.cgd.ucar.edu/cas/jhurrell/indices.html, 2012). As we examine the influence of the NAO in the North Sea on different time scales, we assume that the wintertime NAO forcing will be responsible for producing the most prominent signal in the data on a basin-wide scale.

[8] Due to greater spatial coverage of DIC and pCO2 data, in all three years, compared to that of total alkalinity (AT) and pH, the latter two were calculated from the former two, as previously done in intercomparison studies [Thomas et al., 2009]. Internal consistency studies from cruises with full carbonate parameter coverage in late summer indicate that AT and pH can be predicted, using DIC and pCO2, with an accuracy of ±9 μmol kg−1 and ±0.008, respectively (L.A. Salt, manuscript in preparation, 2013). Only stations with valid values for DIC and pCO2 for all of the three years were used, resulting in a total of 85 stations worth of data for comparison.

2.1 DIC and pCO2

[9] Samples for the carbonate parameters were obtained following the operating procedures outlined in DOE [2007]. Carbonate system parameters, DIC, AT, and pH (in 2005) were determined at 8–15 depths per station, yielding approximately 700 samples per cruise. All samples were analyzed within 12 h of sampling, and were verified for quality control using certified reference material (CRM) supplied by Prof. Andrew Dickson (Scripps Institute of Oceanography, USA). A single sample was obtained for both DIC and AT and these were determined by coulometric and potentiometric titrations, respectively. For further details, please see Thomas et al. [2007].

[10] Surface water pCO2 was measured every minute using a flow-through system with continuous equilibration and infrared detection [Körtzinger et al., 1996], yielding approximately 20,000 measurements per cruise with an accuracy of ±1 μatm. A temperature normalization was applied to the pCO2 data to obtain pCO2@16°C, which is independent of temperature differences between the years [Takahashi et al., 1993].

2.2 Calculations

2.2.1 Water Mass Analysis

[11] The North Sea surface waters (5 m depth) were separated into three simpler constituents using a mixing analysis of the dominant water masses in the North Sea [Kempe and Pegler, 1991; Shadwick et al., 2011], with North Atlantic water, Baltic Sea water, and German Bight water as end-members. The latter two consist of a fraction of North Atlantic water; however, the large freshwater contribution makes both very distinct from North Atlantic water and allows us to track them into the central North Sea. In order to differentiate water mass fractions from n number of contributors, n − 1 end-member variables are required. As this decomposition was done for the surface waters, temperature cannot be used, because due to seasonal heating and cooling it is not a conservative tracer as used in traditional multiparametric optimizations for analysis of deep waters. Salinity, DIC, and AT all offer sufficiently distinct end-member concentrations to be used; however, the codependence of AT on salinity makes DIC favorable. The DIC is nonconservative due to biological uptake/release of CO2; however, the end-members used (Table 1) are sufficiently distinct that changes in DIC concentration affected by mixing are much greater than the potential interference of primary production/respiration. This was confirmed by similar results being obtained using salinity and AT, which is more conservative with respect to primary production.

Table 1. Dissolved Inorganic Carbon and Salinity End-Membersa
DIC (μmol kg−1) and Salinity End-Members 2001 2005 2008
North Atlantic 2047 (35.06) 2078 (35.30) 2065 (35.13)
Baltic 1530 (8) 1530 (8) 1530 (8)
German Bight 2090 (30.36) 2012 (30.77) 2149 (32.28)
  • a The DIC end-members used for the water mass fraction calculation for the three years, with the corresponding salinity values in parentheses.

[12] The end-members for the North Atlantic and German Bight were determined individually for each year, with only the Baltic end-member remaining constant. As we lacked observations in the Baltic Proper, DIC and AT values were taken from literature for a representative salinity of 8 [Thomas and Schneider, 1999; Hjalmarsson et al., 2008] (Table 1). The North Atlantic end-member was determined by the maximum surface salinity found in the northwest North Sea (Latitudes > 58°N and Longitudes < 0°E) and its corresponding DIC value. The German Bight end-member was determined by the salinity and DIC value found at the station closest to the Elbe river mouth (54.75°N, 8.25°E). By using annually determined end-members for the three years, we can rule out any change in water mass fractions occurring due to changes in the chemical signal of end-members, which we later show does occur in the North Sea. Statistics describing the calculated water mass fractions are shown in Table 2.

Table 2. Statistics of the Different Water Mass Fractions Present in the North Sea in 2001, 2005, and 2008a
2001 2005 2008
Water Mass Range Average Median Range Average Median Range Average Median
North Atlantic 0–100 83 91 0–100 81 87 0–100 84 91
Baltic 0–16 2 0 0–15 3 1 0-20 3 7
German Bight 0-100 15 4 0–100 16 10 0–100 13 1
  • a The basin-wide range of values, average, and median are given with the units representing the % of the total water present in the North Sea.

2.2.2 DIC Inventory Calculation of the North Sea

[13] To quantify the changes in DIC between years, differences in biological activity and remineralization must be accounted for. Here we use the apparent oxygen utilization (AOU = [O2]sat − [O2]obs) to account for production and remineralization, applying the Redfield ratio [Anderson and Sarmiento, 1994] in the following equation:

[14] These values were integrated throughout the water column to a common maximum depth per station, and then station totals were extrapolated over the entire North Sea basin. To help us better understand the changes, the International Council for the Exploration of the Sea (ICES) defined boxes [ICES, 1983] were used to examine the regional changes in DIC and salinity. Unless specified otherwise, boxes 1–5 refer to the entire water column, not just the upper 30 m, whereas boxes 11–15 refer to the deep (>30 m) portion of these boxes, respectively.

2.2.3 Calculation of Brunt-Väisälä Frequency Squared

[15] To assess the vertical stability of the water column in the northern North Sea (>56°N), we utilized the Brunt-Väisälä frequency squared:
where z is the depth (m), ρ is density (kg m−3) computed following the density equation of Fofonoff and Millard [1983], and g is the gravitational acceleration (9.807 m s−2). The temperature and salinity data from the conductivity-temperature-depth (CTD) casts with a 1 m resolution were smoothed using a cubic spline. The N2 was calculated from the smoothed dataset and the depth at which the maximum N2 occurs we defined as the mixed-layer depth.

3 Results

[16] The main distribution pattern of the carbonate parameters is relatively constant between years (Figure 1). The brackish Baltic outflow around the Norwegian headland has low AT and DIC signals. The German Bight, in the southwest, is distinguished by its high AT and high DIC content. The Shetland shelf represents the main North Atlantic inflow site, and is clearly identified by notably higher AT values than the basin-wide average (Figure 1b). The anticorrelated pH and pCO2 both show a strong gradient at the 50 m depth contour, representing the boundary at which the dominant control on pCO2 changes from temperature, in the well-mixed south, to biology, in the stratified north [Prowe et al., 2009; Thomas et al., 2005b]. The gradient between the north and south is steepest in 2008, closely followed by 2001, and weakest in 2005. This pattern reflects the measured surface water temperatures in the North Sea, 2001 and 2008 being the warmest (mean surface temperatures of 16.2°C (±1.3) and 16.1°C (±1.1), respectively), and 2005 the coldest (15.4°C (±1.0)). The highest northern pH values and lowest pCO2 values are observed in the northern North Sea in 2001, with progressively decreasing mean pH and increasing mean pCO2 trends over time. The average surface DIC increased by 21 μmol kg−1, although not uniformly, and the average surface AT remained relatively constant, ± 4 μmol kg−1, throughout all three years.

Details are in the caption following the image
Surface layer distribution of carbonate parameters with 50 and 100 m depth contours. (a–c) Total alkalinity (μmol kg−1) for the years 2001, 2005, and 2008, (d–f) dissolved inorganic carbon (μmol kg−1) for the years 2001, 2005, and 2008, (g–i) pH for the years 2001, 2005, and 2008, and (j–l) partial pressure of CO2 (pCO2; μatm) for the years 2001, 2005, and 2008. The DIC and pCO2 are observations, AT and pH are calculated parameters from DIC and pCO2, using carbonic acid dissociation constants of Mehrbach et al., 1973, refit by Dickson and Millero [1987], and pH is given on the Total scale. Average values for 2001, 2005, and 2008 are 2299, 2298, and 2291 μmol kg−1 for AT, respectively, 2034, 2052, and 2055 μmol kg−1 for DIC, 8.129, 8.105, and 8.079 for pH, and 323, 344, and 369 μatm for pCO2.

[17] The North Sea CO2 system is largely governed by the relative contribution of different water masses composing the North Sea water as well as the rate at which these are circulated within the North Sea shelf. The mixing analysis (Figure 2a) clearly identifies the North Atlantic Ocean water as the dominant water mass, constituting an average 83% fraction throughout all three years (Table 2). The average, basin-wide fraction of Baltic water increases from 2% in 2001 to 3% in 2005 and 2008 (Table 2). The dominant North Atlantic inflow follows an anticlockwise circulation from the north, mixing in the south with German Bight water. In the northeastern areas, the Baltic Sea outflow plays an additional role where it is introduced to the circulation on its way out of the North Sea. This leads to the presence of two main mixing regimes, one consisting of North Atlantic and German Bight water, and the other of North Atlantic and Baltic water (Figure 2a). It can be seen that in 2008 these two regimes are most distinguishable, which corresponds to the year in which the maximum Baltic fraction was recorded (20%; Table 2). The formation of the two mixing environments is reflected in the formation of a dichotomy in pH measurements in 2008, which is absent in 2001 and 2005 (Figures 2b–2d). The dichotomy divides the North Sea, with more acidic waters in the south and higher pH values in the north. The divide is formed at, approximately, the 50 m depth contour accounting for the strong gradient visible in Figures 1i and 1l.

Details are in the caption following the image
Water type contributions to the North Sea. (a) The fractions of North Atlantic water, Baltic water, and German Bight water in the North Sea in 2001 (blue), 2005 (green), and 2008 (red). (b–d) Histograms of the frequency of calculated pH values binned in 0.01 pH unit intervals with the mean value per year marked by the red line.

[18] Examining the mixing relationship between the North Atlantic and Baltic water further reveals a change in the water mass end-members from 2001 to 2005. A linear regression was applied, for each year, to data with a Baltic water mass fraction greater than the three-year average of 3% (Figure 3) revealing three statistically significant relationships. The years 2005 and 2008 show a remarkably similar mixing line with the most notable difference being a general decrease in the salinities observed in the Skagerrak in 2008 compared to 2005. Such a shift is reflective of different proportions of the two water masses but indicates that the water masses remain constant in their properties. The difference between these two years and 2001, however, is much greater. The slope of the regression line is shallower in 2001, indicative of a higher DIC Baltic end-member and a lower DIC North Atlantic end-member. The same relationship is reflected, between 2005 and 2008, in the AT mixing diagram (Figure 3b).

Details are in the caption following the image
North Sea - Skagerrak mixing diagram. Linear fitted regressions for the mixing relationship between the Skagerrak and North Atlantic inflow in the northern North Sea. Upper graph shows DIC and lower graph shows AT. Points were selected based on percentage of Baltic water mass fraction present (>3%). Dashed lines represent one standard deviation (±1σ) of the residuals for each fit, also given in parentheses.

[19] The inorganic carbon, after correction for metabolic DIC (calculations in section 2.2.2) and respective salinity inventories were used to quantify the changes (∆S and ∆DIC) between the two time periods. The change in each ICES box for the two time periods showed a significant correlation coefficient of 0.903 and 0.839 (excluding Box 8) for the period 2001–2005 and 2005–2008, respectively. Box 8, heavily influenced by the German Bight, recorded very low salinities for 2008, coinciding with a larger Elbe discharge [http://coast.gkss.de/staff/kappenberg/elbe/abfluss/elbe.abfluss]. This box is likely to be more representative of local effects thus skewing the general pattern for the rest of the North Sea (including box 8 gives an R2 = 0.49). Although the increase in DIC was basin-wide from 2001 to 2005, it did not occur evenly across the basin or throughout the water column.

[20] The greatest increases in average DIC concentration from 2001 to 2005 occurred in the central and northwestern North Sea. The latter region is significantly deeper than the southern North Sea; thus, the ICES boxes were further divided into a surface (30 m) and deep box. The surface boxes showed an increase of the same order of magnitude as the deep boxes; however, due to the greater DIC concentrations at depth, this constitutes a lower percentage of total increase. From 2005 to 2008, the changes in DIC were much smaller and demonstrated no clear pattern. The North Sea DIC inventory was 8.11 Pg C (1 Pg = 1015 g) in 2001, 8.17 Pg C in 2005, and 8.18 Pg C in 2008. Hence, it increased by approximately 0.8% (+ 6 × 10−2 Pg C) from 2001 to 2005, while the inventory remained almost stable (+ 5 × 10−3 Pg C, <0.1%) between 2005 and 2008. The deep northern boxes demonstrated a greater inventory increase of 1.1% from 2001 to 2005, and a 0.0% change from 2005 to 2008. Of the entire observed change in these boxes, 97% occurred from 2001 to 2005.

[21] The changes in the DIC concentrations in the surface also manifest in the change in temperature-normalized pCO2 (pCO2@16°C) distribution (Figure 4). The increase in the average North Sea pCO2@16°C from 2001 to 2005 is 26 μatm, compared to 4 μatm from 2005 to 2008 (the increases for 2001 to 2005 and 2005 to 2008 for nontemperature normalized pCO2 were 21 μatm and 25 μatm, respectively). Using an average basin-wide pCO2@16°C from all three years (346 μatm), these increases correspond to 7.5% and 1.0% increases, which are proportional to the calculated increase in the North Sea DIC inventory. This agreement implies that the surface expression of pCO2@16°C is, to a large extent, representative of the entire water column, which, due to the shallow depths of the majority of the North Sea is not surprising. Similar to the distribution of DIC increase, the majority of the increase in pCO2@16°C occurs in the northern North Sea, namely, where the main inflow site is for North Atlantic water. In the north (>56°N), the average surface increase from 2001 to 2005 is 44 μatm, and from 2005 to 2008 the increase is 2 μatm. The southern North Sea shows much more stable increases, of 22 μatm and 18 μatm. At this time of year in the northern North Sea, pCO2 without the effect of temperature is predominantly controlled by biology [Prowe et al., 2009]. Using this parameter pCO2@16°C to investigate the differences in production between the three years would indicate a drop in primary productivity from 2001 to 2005 and 2008. However, the AOU in subsurface waters in the northern North Sea show comparable values between all three years (36.4 μmol kg−1 in 2001, 30.1 μmol kg−1 in 2005, and 31.3 μmol kg−1 in 2008), indicating that net community production (NCP) was similar.

Details are in the caption following the image
Distribution of temperature-normalized (16°C) pCO2 in surface waters. In (a) 2001, (b) 2005, and (c) 2008.

4 Discussion

[22] Despite the similar fundamental patterns observed in 2001, 2005, and 2008, distinct differences occurred as a consequence of different DIC mixing patterns and proportions. In 2008, the year with the most positive NAO index (+2.1), the most pronounced front between the two mixing regimes occurs, where North Atlantic and German Bight waters were on the southern side and a mix of North Atlantic and Baltic Sea waters were on the northern side. This is caused by the fact that both the inflows of North Atlantic [Winther and Johannessen, 2006] and Baltic Sea [Hordoir and Meier, 2010] water into the North Sea are strongest under NAO+ conditions due to stronger westerly winds, strengthening the anticlockwise circulation [Lenhart et al., 2004; Winther and Johannessen, 2006]. There is subsequently an increased flow from east to west along the isoclines at 50 m, (the Dooley current [Dooley, 1974]), in addition to increased flow down the west coast of the UK. When these inflows weaken during neutral (2005) or NAO− (2001) phases, there is a less pronounced mixing pattern and the contrast between the two regimes is diminished with more mixing occurring in the eastern central North Sea. The front between the north and south is thus less distinct or vanished, as for example, visible in the greater presence of Baltic Sea water (Figure 2) or in the distribution of AT (Figure 1), which is unaffected by the anthropogenic CO2 perturbation, in contrast to the remaining CO2 system parameters.

[23] The biogeochemical divide along the 50 m contour, separating the northern and southern North Sea, is a well-documented feature in CO2 distributions [Thomas et al., 2004; Bozec et al., 2005; Prowe et al., 2009; Omar et al., 2010]; however here, the manifestation of this divide is shown to be strongly influenced by the NAO. The observed mixing behavior has profound implications for the distribution of pH in the North Sea (Figures 1c–1e). The two aforementioned mixing regimes lead to a higher pH in the stratified northern region, where Baltic Sea and North Atlantic inflows mix, and a lower, more acidic pH, in the shallower southern region. When these inflows weaken during NAO− years, evidently more thorough mixing between these regions occurs and the pH histogram reveals a more transitional pattern.

[24] In addition to NAO-driven changes in mixing ratios and rates, the North Sea CO2 system is further affected by the biogeochemical properties of its composing water masses. This applies especially to the DIC concentrations of the inflowing North Atlantic Ocean waters, which have demonstrated substantial variability during the recent decade [Watson et al., 2009; Thomas et al., 2008a; Schuster and Watson, 2007]. The northern North Sea CO2 system shows a substantial increase in DIC and temperature-normalized pCO2 and decrease in pH between 2001 and 2005 (Figures 1 and 3) in comparison to that observed from 2005 to 2008. The concomitant change in salinity with the change in DIC throughout the water column reveals that the increase is water mass driven. Otto et al. [1990] state that changes in salinity in the North Sea stem from either changes in the salinity of inflowing water or changes in runoff. A shift in the salinity-DIC mixing diagram confirms that the DIC concentrations in the North Atlantic end-member have increased dramatically from 2001 to 2005, which we propose is a delayed response to NAO-driven changes in the North Atlantic [Thomas et al., 2008b]. This accelerated the decrease in pH and increase in pCO2 of the North Sea beyond that anticipated from the rise of atmospheric CO2 concentrations alone [Thomas et al., 2007]. The DIC and salinity changes between 2005 and 2008 also demonstrated a significant correlation, but were an order of magnitude smaller thus much less pronounced, yielding similar end-member DIC concentrations for 2005 and 2008, analogous to the variable behavior observed in surface pCO2 in the North Atlantic Ocean [Schuster et al., 2009].

[25] The same pattern was unsubstantial in the AT-salinity mixing line. As the increase in DIC observed in the North Atlantic, and subsequently the North Sea, is driven by air-sea pCO2 exchange, there is no concomitant change in AT with DIC. However, there is still some variation in the AT-salinity mixing lines, which could be associated with the noted change in salinity [Thomas et al., 2008a]. The NAO also exerts large influence over precipitation patterns over Europe so that NAO+ leads to drier conditions across Europe but wetter conditions over Scandinavia [Hurrell and van Loon, 1997]. It follows that in 2008 we would expect more rainfall and runoff in the northern area of the Baltic drainage basin, which is associated with granite rocks and has lower AT values compared to the south [Hjalmarsson et al., 2008]. This would lead to lower AT in 2008 compared to 2001, which is consistent with our computed AT end-members (Figure 3b). These changes in precipitation patterns over Europe and thus riverine runoff into the North Sea may additionally alter salinity and DIC patterns in the southern North Sea, where the majority of riverine input enters the North Sea. However, the riverine contribution represents just 0.5% of the water budget and 0.7% of the carbon budget in the North Sea [Thomas et al., 2005a], which is insufficient to cause the change recorded in DIC or pCO2, although it may have contributed to the observed variability.

[26] The northern North Sea is sufficiently deep to develop a summer thermocline, thus facilitating phytoplankton blooms during the productive period, which are maintained in the surface mixed layer. Using the maximum water column value of a representative northern station of the Brunt-Väisäilä frequency as an indicator of the mixed layer depth, we observe a very stable, but shallower stratification with a surface mixed layer of approximately 30 m in 2001 compared to 40 m in 2005 (Figure 5). Since the residence time of the waters in the North Sea is of less than one year, there is no accumulation of metabolic DIC (previous years' primary production has no effect on the current state of the North Sea). If the NCP were the same, as it was indicated in our datasets by AOU, a change in mixed-layer depth (MLD) would impact the observed pCO2. While the biological pCO2 drawdown was on the order of 300–330 μatm in 2001 [Thomas et al., 2005b], a similar NCP would have caused a pCO2 drawdown of 200–250 μatm in 2005, characterized by a mixed layer depth of 40 m. This would lead to a difference of 50–130 μatm between the two years, which is in line with the observed increase of 51 μatm of temperature-normalized pCO2 (Figure 4) in the northern North Sea. Taking the observed salinity change into account, such an increase would correspond to a pH decrease of 0.056.

Details are in the caption following the image
Stratification at 57ºN, 2.25ºE. (a) Vertical profiles of density and temperature for 2001 (blue), 2005 (green), and 2008 (red). (b) Brunt-Väisälä frequency squared profile for 2001 (blue), 2005 (green), and 2008 (red).

[27] Kühn et al. [2010] have modeled a greater CO2 sink in the northern North Sea during years of NAO− (2001) compared to years of NAO+, which they attributed to weaker stratification facilitating greater upward transport of nutrients. The highest pH values and lowest pCO2 values recorded are observed in the northern North Sea during years of NAO−; however, we cannot ascribe this to weaker stratification. The weaker stratification in 2005, associated with cooler temperatures, demonstrates that the expression of pH and CO2 can be exaggerated or masked by physical conditions within the North Sea. The variability of the mixed layer depth in the North Sea is controlled by a combination of temperature and NAO [Lenhart et al., 2004] and we propose the latter affects it through two possible processes. First, a stronger North Atlantic inflow weakens the thermocline in years of neutral and positive NAO, leading to less stable stratification as found in 2005 and 2008. Alternatively, weaker winds may allow the thermocline to rise during more negative NAO conditions. In 2008, a stronger warming of the surface waters resulted in a more stable water column with a shallower mixed layer. This situation enhanced the NCP effect on the surface layer pCO2. Simultaneously, the warmer waters experienced in 2008 raised the pCO2, eventually leading to comparable pCO2 conditions in 2005 and 2008, with 2008 showing the warmer and shallower surface layer.

[28] In summary, our results indicate that under conditions of NAO+, as in 2008, the North Sea shelf pump is more efficient than under NAO− due to increased rates of Atlantic and Baltic inflow into the North Sea and a strengthened anticlockwise circulation. The strengthened biogeochemical divide between the northern and southern North Sea means that more carbon is exported from the northern North Sea out of the Norwegian Trench without coming into contact with the atmosphere, which limits subsequent outgassing of remineralized CO2 making the shelf pump more effective. When NAO+ occurs in conjunction with a year of higher SST, shoaling of the thermocline allows intense areas of primary production to develop. This causes lower pCO2 and higher pH in surface waters of the northern North Sea; however, this is not shown to have any effect on total carbon uptake. In addition, any changes to the biogeochemical content or quantity of the source water masses contributing to the North Sea can intensify or diminish these expressions.


[29] We thank the captains and crews of the Research Vessel Pelagia, and three anonymous reviewers for constructive comments that helped improve the paper. This program was supported by the Research Council for Earth and Life Sciences (ALW) of the Netherlands Organization for Scientific Research (NWO) and CARBOOCEAN. This work contributes to IGBP/IHDP LOICZ. AVB is a senior research associate at the FRS-FNRS. AEFP was supported by the Kiel Cluster of Excellence “The Future Ocean” and by the Centre for Ocean Life, a VKR centre of excellence supported by the Villum foundation.