Global Biogeochemical Cycles

Interannual variability of chlorophyll and the influence of low-frequency climate modes in the North Atlantic subtropical gyre



[1] The impact of low-frequency climate modes on the large-scale variability of chlorophyll-a (a proxy for phytoplankton biomass) concentration in the subtropics is little known. Here we examined the concurrent monthly chlorophyll-a and hydrographical records obtained at two subtropical time series stations (Bermuda Atlantic Time series Study, BATS and European Station for Time series in the Ocean Canary Islands, ESTOC) from the beginning of the time series (1989 at BATS and 1994 at ESTOC) to 2003, in order to analyze its seasonal and interannual variability and investigate the potential correlation with large-scale atmospheric oscillations. At BATS, Chl-a variations are mainly caused by differences in the convective mixing and mesoscale phenomena. Variability in winter mixing is a significant factor at ESTOC as exemplified by years with anomalously deep mixed layer depths. An additional nutrient source causing Chl-a variability at this station likely occurs due to nutrient advection driven by the baroclinic flow caused by the Trade Winds during summertime. We found that interannual variability in mean integrated total Chl-a (TChl-a) is significantly correlated with temperature and salinity anomalies at BATS. Chl-a also covaried with changes in temperature although the correlation was not significant at ESTOC. We could not find any direct correlation between TChl-a and NAO at BATS; the correlation improved between TChl-a and NAO +1 year but was still insignificant. However, significant correlations were found for ESTOC between TChl-a anomalies and ENSO (El Niño Southern Oscillation) +3 years and NAO +1 year time lag.

1. Introduction

[2] The subtropical gyres of the world oceans are critical regions for understanding biogochemical processes in the global ocean. The controversies surrounding new production estimates derived from in situ measurements [Jenkins and Wallace, 1992; Sarmiento et al., 1990] or model estimates [McGillicuddy et al., 2003; Oschlies, 2002], for example, point to the need to improve our knowledge about these regions. The analysis of time series data reveals the importance of interannual variability [Letelier et al., 1996; Neuer et al., 2007; Steinberg et al., 2001] and of mesoscale processes [Cianca et al., 2007; McGillicuddy et al., 2006; Sweeney et al., 2003] in the biogeochemistry of these vast biomes. In addition, low-frequency and large-scale climate modes such as El Niño Southern Oscillation (ENSO) and North Atlantic Oscillation (NAO), may exert a substantial influence on the variability of biogeochemical processes in the North Atlantic subtropics [Bates, 2007; González-Dávila et al., 2007; Santana-Casiano et al., 2007].

[3] The North Atlantic Oscillation (NAO) is the dominant mode of the winter climate variability over the upper North Atlantic Ocean on monthly to decadal scales [Hurrell, 1995; Hurrell and Deser, 2009; Hurrell et al., 2003]. In addition, the ENSO phenomenon is known to cause climate variability on interannual and decadal time-scales in both the North Pacific and North Atlantic Oceans.Lau and Nath [2001] suggested the existence of an “atmospheric bridge,” and Gallego and Cessi [2001] illustrated in a modeling study the likely connection between the decadal variability in both the North Atlantic and North Pacific. Lee et al. [2008]studied El Niño- induced warming of the tropical North Atlantic (TNA) and found that the Atlantic internal variability (such as NAO) could greatly reduce ENSO effects or vice versa. As shown byGrossmann and Klotzbach [2009], NAO can also be influenced by the Atlantic Multidecadal Oscillation (AMO) [Enfield et al., 2001].

[4] Open ocean time series stations have been used as platforms for several of the studies investigating large-scale variability of oceanic biogeochemical processes and the low-frequency atmospheric modes. Interannual anomalies of hydrography and ocean biogeochemistry in the western North Atlantic were partially linked to NAO and ENSO, generally co-occurring with the cold phase of the NAO or lagging ENSO by 6–12 months [Bates, 2001, 2007]. More recently, Lomas et al. [2010]found that chlorophyll-a variability was directly linked with the winter phase of the NAO in the Sargasso Sea. The authors hypothesized that a shift in the NAO in 1996 from positive to neutral but variable caused an increase in winter convective mixing and subsequent increase in productivity.

[5] On the eastern side of the North Atlantic, significant correlations between hydrography or ocean bioeochemistry and NAO were inferred with a delay in the oceanic response of around three years [Santana-Casiano et al., 2007]. Siedler et al. [2005] hypothesized that a change in the sign of the NAO index produced changes both in the baroclinic transport and in the spatial pattern of the North Atlantic subtropical gyre (NASTG) with a similar three year delay in the oceanic response. Palter et al. [2005] linked the variability of the nutrient budget and primary production in the Sargasso Sea to NAO induced changes in the NASTMW (North Atlantic Subtropical Mode Water) stability. The analysis consisted of a comparison of two different periods where cold winters and intense convective mixing (negative NAO) produced dense and thick layers of NASTMW whereas warm winters (positive NAO) decreased NASTMW formation. The apparent difference in the correlation of vertical mixing with NAO between the eastern and western parts of subtropical North Atlantic gyre was furthermore confirmed by [Oschlies, 2001]. Recently, Boyce et al. [2010]observed both a decline in global phytoplankton biomass over the past century as well as fluctuations which were clearly correlated with basin-scale climate indices.

[6] The main purpose of this study is to improve our current understanding of the impact of low-frequency climate modes on the basin-scale variability of chlorophyll-a concentration (a proxy for phytoplankton biomass) in the subtropical North Atlantic. We examined the concurrent monthly chlorophyll-a concentrations analyzed by the Turner fluorometer and hydrographic records at the two time series stations Bermuda Atlantic Time series Study (BATS) and European Station for Time series in the Ocean Canary Islands (ESTOC) from January 1989 and February 1994, respectively, until 2003. Logistical issues impeded the continued time series observations at ESTOC and difficulty with the Turner based chlorophyll record at BATS curtailed a longer time series. WhileCianca et al. [2007] compared the in situ nutrient budgets and also showed the chlorophyll time series at both stations, this present comparison gives us the opportunity to study possible climate connections over a basin scale.

[7] Changes in temperature and salinity have been associated with long-term atmospheric variations in the North Atlantic and, in turn, these cause changes in circulation [Curry and McCartney, 2001; Siedler et al., 2005] and nutrient supply. All of these would impact phytoplankton populations both in terms of standing stock, primary production and possibly export production. We therefore examined the impact of the hydrography (mixed layer depth and summer stratification) and nutrient availability (NO3+NO2) on the chlorophyll concentration in the upper 200 m as well as in the Deep Chlorophyll Maxima (DCM) which is present at both sites during most of the year. The DCM reflects layers of increased chlorophyll concentration at minimal light and nutrient replete conditions [Huisman et al., 2006] and thus is sensitive to changes in nutrient supply and light conditions.

[8] In order to understand the connection between Chl-a fluctuations and climate variability at both sides of the North Atlantic subtropical gyre, we estimated the correlations between the main low-frequency modes of climate and Chl-a variability in the water column. We investigated the indexes of the ENSO (the monthly Southern Oscillation Index, SOI; the Oceanic Niño Index, ONI; and Multivariate ENSO index, MEI), the NAO and the Atlantic Multidecadal Oscillation (AMO).

[9] The connection between phytoplankton dynamics and the hydrographic variability caused by atmospheric climate modes on regional scales is a topic fundamentally important for our understanding of marine ecosystems, biogeochemical cycling and fisheries.

2. Characteristics of the Time Series Sites and Data Analysis

[10] BATS (31°45′N, 64°10′W) and ESTOC (29°10′N, 15°30′W) are located in the western and eastern subtropical North Atlantic Gyre, respectively (Figure 1). ESTOC is situated within the weak southward return flows on the eastern side of the North Atlantic subtropical gyre [Neuer et al., 2007] while BATS is located within the stronger western gyre recirculation [Steinberg et al., 2001]. In addition, meso-scale patterns show less energetic variability in the eastern boundary of the gyre (ESTOC) compared to the west (BATS) [Cianca et al., 2007]. BATS and ESTOC exhibit seasonal differences in nutrient supply between winter and summer [Cianca et al., 2007; Neuer et al., 2002]. Winter mixing can vary substantially between years at BATS (around 125 to 350 m), while winter mixing at ESTOC is shallower and varies from 100 to 200 m. Despite deeper winter mixing and higher eddy activity, the overall new nutrient input at BATS is only 25% greater than at ESTOC, mainly because of the steeper and shallower nitracline at ESTOC compared to BATS [Cianca et al., 2007]. Higher values in the annual rates of primary production and phytoplankton biomass would be expected at BATS; however, the phytoplankton biomass and productivity are similar at the two stations. In contrast, export production varies considerably, thought to be caused partly by differences in nutrient input to the mixed layer, both from eddy induced mixing and dinitrogen fixation, as well as differences in the community composition [Neuer et al., 2002, Helmke et al., 2010].

Figure 1.

Location of the three subtropical time series stations overlaid over a multiyear average of MODIS Aqua derived sea surface chlorophyll from 2002 to 2007.

[11] BATS and ESTOC have been sampled monthly. Additionally, during the bloom season (February to June), BATS is sampled twice a month. Data collection at BATS began in October 1988 and at ESTOC in February 1994. The BATS data were obtained from the Web ( No significant gaps were noted in the BATS sampling program and only one at ESTOC (from October 2001 to February 2002). Sampling methodologies and results obtained from the long-term data sets have been reported for BATS byMichaels et al. [1994], Michaels and Knap [1996] and Steinberg et al. [2001] and for ESTOC by González-Dávila et al. [2003], Llinás et al. [1999] and Neuer et al. [2007].

[12] In this study, we examined the concurrent monthly extracted Chlorophyll-a analyzed by Turner fluorometry as mentioned above.Saba et al. [2010] in a model comparison concluded that the majority of ocean color models produced in situ net primary production (NPP) trends that were closer to the observed data when derived from chlorophyll determined by High Performance Liquid Chromatography (HPLC) instead of Turner fluorometry. However, HPLC measurements are not available at ESTOC to compare with BATS during the survey period and we are restricted to the comparison with the available Turner chlorophyll records.

[13] The original sampling records were interpolated linearly to the largest common resolution to facilitate a direct comparison. The original data sets were made up of eleven levels (surface, 10, 20, 40, 60, 80, 100, 120, 140, 160 and 200 m) from surface to 200 m at BATS (36 levels in the water column) and nine levels (surface, 10, 25, 50, 75, 100, 125, 150, 200 m) at ESTOC (24 levels in total). A linear interpolation was used to re-locate the values at a common depth. Thus, the resolution of the interpolated profiles are the same or lower than the originals and the maximum error estimate, which is proportional to the square of the distance between points, is lower than 0.02 mg m−3. To make sure that the shapes of the interpolated profiles were not affected, we visually reviewed the original and interpolated profiles and confirmed that there were no significant differences in the shapes (mainly in the maxima in Chl-a distribution). Some estimates used in this study such as the depth of Deep Chlorophyll Maximum (DCM) are limited by the resolution of the data set (discrete measurements in lieu of continuous values in the vertical with high resolution); however we consider this approach most suited for the comparative analysis presented here.

[14] In order to determine the derived variables for use in the statistical analyses, estimates from the individual profiles were made prior to calculating any monthly or seasonal averages. Thus, estimates such as mixed layer depth (MLD) or temperature gradient maxima (GRD) were calculated in the individual CTD profiles (two dbar pressure resolutions) and subsequently averaged to obtain monthly or seasonal means. A threshold criterion based on a temperature difference of 0.5°C from the level of 10 m depth was used to assess the mixed layer depth (MLD). This criterion is the most widely used in the literature [de Boyer Montégut et al., 2004] and is the same as that used by Cianca et al. [2007]. At ESTOC, temperature from January 1994 to August 1995 was recorded by deep-sea reversing thermometers (as no CTD, Conductivity, Temperature, Depth probe, was available during this time period). This is the only period where discrete temperature measurements were used to estimate the MLD and GRD. The pressure/depth measurements were estimated by calculation of thermometric depth that is deduced from the difference between the paired protected and unprotected reversing thermometer readings; the unprotected reversing thermometer indicates higher temperature due to pressure effects on the instrument. To estimate depths in the bottle sampling, pressure sensors attached in the Niskin bottle were also used. The MLD assessments obtained with the CTD during the latter months was then compared with temperature from reversing thermometers to ensure the temporal coherence between the results [Cianca et al., 2007]. The GRD was estimated from the maximum temperature gradient in the upper 100 m depth and filtered with a median filter to exclude any erratic data.

[15] With the aim of facilitating the statistical analysis, we defined two time periods, characterized by water column mixing (January to April) and stratification (August to November). We selected these periods based on the deepest MLD and steepest GRD observed during those months. We could include December in lieu of April or July for November but based on tests which were carried out ahead of the analysis this does not change the results significantly. The standardized anomalies used to estimate the interannual variability of the properties were assessed by the difference of the actual value minus the mean and divided by the standard deviation. The correlation between seawater properties and Chl-a and the correlation between the atmospheric climate modes and the interannual chlorophyll were estimated using the Pearson coefficient (r). For instance, the monthly Southern Oscillation Index (SOI), the Oceanic Niño Index (ONI) and Multivariate ENSO index (MEI), which were obtained from the Climate Prediction Center (, were used to correlate the ENSO phenomenon with the oceanic variables. SOI is based on a traditional definition of the fluctuation in the air pressure differences between Tahiti and Darwin. ONI and MEI are new definitions based on sea surface temperature departures from the average in the Niño 3.4 region and the analysis of the six main observed variables over the tropical Pacific (sea level pressure, zonal and meridional components of the surface wind, surface sea temperature, surface air temperature and total cloudiness fraction of the sky) [Wolter, 1987]. SOI shows a contrary sign compared to ONI and MEI indexes (negative/ positive respectively), in order to mark an El NIÑO event.

3. Results and Discussion

3.1. Seasonal Characteristics of the Chlorophyll Distribution and the Nitracline

[16] There are several studies on the seasonality of biogeochemical variables at BATS [Bates et al., 1998, Michaels et al., 1994, Steinberg et al., 2001] and ESTOC [González- Dávila et al., 2003, Neuer et al., 2007, Santana-Casiano et al., 2007]. However, we consider the comparative analysis as described below helpful for the reader to be able to understand similarities and differences between both sites.

[17] The Chlorophyll-a and nitrate plus nitrite (here after nitrate) time series (Figures 2a and 2b) illustrate the seasonal and interannual variability in these parameters. The DCM is observed as a typical characteristic of the Chl-a vertical distribution of the subtropical gyres [Eppley et al., 1988, Macedo et al., 2000; Pérez et al., 2003] and was centered near 110 m from spring to fall, while it extended to the surface during winter at BATS and ESTOC (Figure 2a). The thickness of the DCM layer was roughly constant and around 50 m at BATS (based on the 0.15 mg m−3Chl-a isoline to define the layer, see black lines inFigure 2a), while it was thicker than 50 m at ESTOC. Interestingly, the DCM layer displayed several discontinuities at BATS but not at ESTOC.

Figure 2.

(a) Chl-a time series at the two stations from the beginning to 2004. Black lines define the DCM layer based on the 0.15 mg m−3Chl-a isoline. (b) Same for nitrate time series.

[18] Two factors control the DCM: light supplied from above and nutrient input from below [Huisman et al., 2006]. Euphotic zone depths (based on the photosyntetically available radiation; 1% PAR) range between 80 and 120 m at BATS and ESTOC [Siegel et al., 1995; Zielinski et al., 2002]. Concerning the second factor, the annual cycle of nitrate is linked to physical forcing (winter convective overturn and summer stratification) at BATS and ESTOC. Normally, nitrate is not detectable in the upper 100 m using the colorimetric methodology of nitrate detection, but it is known that nitrate can show variability at nanomolar levels during the year when using chemiluminescent detection methods [Dore and Karl, 1996]. During summer stratification, nutrients were depleted in the upper layers at both stations. The nitracline (considered as the depth where the nitrate reaches 0.5 μmol kg−1) was located between 100 and 125 m (mean of 115 m, StDev ± 40 m) at BATS and between 75 and 100 m (mean of 92 m, StDev ± 22 m) at ESTOC [Cianca et al., 2007]. The means were significantly different (t-test, p = 0.001). The mean depth of the nitracline was not statistically different between the ‘mixing’ period from January to April and the ‘stratification’ period from August to November at either station. At ESTOC the mean depth of the nitracline was at 100 m (93 ± 27 m during mixing and 90 ± 23 m during stratification). At BATS, the mean nitracline depth varied more, being closer to the 100 m depth level during the mixing period (106 ± 42 m) and closer to 125 m during the stratification period (117 ± 20 m). Below this layer, the nutrient concentrations increased more quickly at ESTOC compared to BATS. The mean nitrate concentration at 200 m depth, for example, was 5.8 ± 0.9 μmol kg−1 at ESTOC compared to only 2.1 ± 0.8 μmol kg−1 at BATS [Cianca et al., 2007]. In addition, vertical movements of the nitracline depth could be observed at BATS and ESTOC (Figure 2b). At ESTOC, the nitracline moved along with the displacements of the upper layers of the main thermocline. At BATS, the nitrate concentrations near the nitracline, below 150 m, were higher from 1989 to 1996 compared to the period from 1997 to 2004 (end of the study period presented here). Lomas et al. [2009] also observed the latter differences and separated the nearly two decades of data at BATS between 1995 and 1996 due to the NAO phase change. The periods are likely influenced by spatial and temporal nutrient reservoir changes in the NASTMW as shown by Palter et al. [2005]. This Eighteen degree mode water layer which is normally nutrient depleted due to biological utilization downstream from the formation region injects a wedge of low nutrient water into the upper layers of the ocean. Counter intuitively, cold winter events that support deep convective mixings and favor mode water formation could reduce primary productivity in the area influenced by the mode water [Palter et al., 2005] (see section 3.3).

[19] The seasonal average profiles of the chlorophyll concentrations (Figures 3a and 3c) were slightly higher at ESTOC compared to BATS. The average Chl-a concentration maximum was 0.288 ± 0.120 mg m−3 at ESTOC, whereas it was 0.236 ± 0.128 mg m−3at BATS. The means are significantly different (t-test, p = 0.005). The winter maxima found at BATS and ESTOC had high standard deviations due to the yearly MLD variability (see the white lines in the upper plots of Figures 2 and 3 inCianca et al. [2007]). These deviations were clearly significant compared to the error estimate in the data interpolation (0.02 mg m−3, see section 2). The seasonal peaks of Chl-a were between 75 and 100 m at both sites. An obvious connection existed between the Chl-a peaks and the upper nitracline at the two sites. No seasonal differences existed in the slope of the nitracline at BATS, while a steeper nitracline could be observed, including a higher standard deviation, during summer and fall at ESTOC (Figures 3b and 3d). The deep Chl-a maxima in summer are just above the nitracline at ESTOC, whereas the maximum nitrate gradient was below the eighteen degree mode water layer at BATS (not shown). This difference makes ESTOC more susceptible to physical nutrient input to the euphotic zone compared to BATS.

Figure 3.

Comparison of the seasonally averaged depth profiles of chlorophyll and nitrate for the two sites. Error bars = standard deviations.

[20] Chl-a concentrations in the DCM are comparable to those found during winter at ESTOC, whereas Chl-a maxima are higher at BATS during winter compared to summer. This difference could be associated with the reinforcement of the baroclinic flow caused by the Trade Winds at ESTOC during summertime, favoring nutrient supply to the euphotic zone by epipycnal advection [Pelegrí et al., 2006]. This flow is affected by the seasonal pattern of the northeasterly winds, with maximum values during summer. This unstable flow also produces mesoscale eddy events which originate in the prominent areas of the African coastline and could reach the ESTOC region [Cianca et al., 2007].

3.2. Interannual Variability of the Chlorophyll Distribution

[21] We investigated the interannual variability by comparing the annual average of TChl-a during the two selected periods, mixing, from January to April and stratification, from August to November. In order to establish similar patterns of variation between Chl-a and hydrography, we compared the yearly Chl-a anomalies for the selected periods with anomalies in temperature and salinity and MLD for the mixing period, whereas the temperature gradient maxima (GRD) were used as indicators of stratification. The yearly TChl-a, MLD, Temperature and Salinity anomalies during the mixing period (January to April) are shown inFigures 4a and 4c. TChl-a anomalies which correspond to the stratification period (August to November) with GRD, temperature and salinity yearly anomalies are shown inFigures 4b and 4d. During the mixing period, BATS and ESTOC showed decreasing linear tendencies in MLD, however the TChl-a linear trends hardly changed in spite of some punctual large variations along the time series. No significant correlations existed between anomalies of TChl-a and MLD, however, the interannual variations were similar at the two sites and were only interrupted by anomalous years (those years with more than a 100% increase in TChl-a compared to the rest of the time series;Figure 2a). The main anomalous Chl-a maximum value occurred at ESTOC during 1999, higher than 200% of the Chl-a mean maximum value, whereas the main anomalous Chl-a maximum value occurred during 1995 at BATS. The latter was interpreted as the responses of the autotrophic community to the passage of a Mode water eddy in July 1995 [McGillicuddy et al., 2007, 1999; McNeil et al., 1999; Mouriño-Carballido and McGillicuddy, 2006]. Mesoscale variability is also a factor at ESTOC, albeit not as profoundly [Cianca et al., 2007]. A significant difference is distinguishable between the anomalous Chl-a events at both sites. The Chl-a concentration maxima measured were higher than 0.6 mg m−3 at ESTOC during 1999 compared to 0.29 mg m−3(mean value in the time series). The TChl-a also drastically increased from 34.1 mg m−2 (mean value for convective mixing events at the time series) to values higher than 70 mg m−2. The number of profiles with anomalous values during winter 1999 was more than 50% of the profiles used in the average (up to 4 profiles from 7). McClain et al. [2004]showed using SeaWIFS ocean color observations that 1999 was an elevated chlorophyll year in the entire subtropical gyre. At BATS, the peak of Chl-a concentration measured during July1995 was 1.12 mg m−3 compared to 0.25 mg m−3(mean value of the time series). However, the TChl-a measured in this profile (51 mg m−2) is similar to those found during the mixing period, suggesting that the mode water eddy passage during summer 1995 at BATS triggered a phytoplankton response similar to deep convective episodes. However, the extraordinary TChl-a values measured at ESTOC were likely caused by unusual hydrographic conditions linked to long-term phenomena. We will discuss this in depth insection 3.3.

Figure 4.

(a and c) Yearly anomalies of temperature, salinity and MLD computed from the annual maximum MLD value and TChl-a anomaly computed from the 4-monthly averages. Black lines represent the variable linear trends of the time series. Correlations between TChl-a and the variables are estimated by the Pearson coefficients “r” with p < 0.05 (in boldface). Negative r values mean the variables inversely co-varied and vice versa. (b and d) Same as in Figures 4a and 4c for the stratification period.

[22] When we search for correlations of each variable between sites, no significant correlations existed between TChl-a at BATS and TChl-a at ESTOC (r = 0.2; not shown), however the correlation was significant for MLD anomalies at BATS and at ESTOC (r = 0.61; not shown). Colder and fresher anomalies were mainly associated with higher TChl-a at both stations, however only the correlation of TChl-a with temperature and salinity was significant at BATS during the mixing period (Figure 4a). This confirms the obvious agreement between winter convection and TChl-a in spite of not finding a significant correlation between TChl-a and MLD at either site and between TChl-a and temperature and salinity at ESTOC.Cianca et al. [2007]showed that the nutrient transport to the euphotic zone by winter convection constituted 32% of the total nutrient supply at BATS and 50% at ESTOC. The causes of not finding a significant correlation between TChl-a and temperature or salinity at ESTOC could be explained by sampling gaps in February 1995 and 2001 at ESTOC, which could affect the correlations. February is the month when the deepest MLD and highest Chl-a values are most likely to occur at this site. Another likely cause for the lack of correlation could be light limitation.Lomas et al. [2009]suggested that slow and steady Chl-a increases during deep convective events could be limited by light, although they could not confirm this hypothesis due to lack of light data. At BATS, mesoscale anticyclonic passages could also negatively affect the correlation. For instance, the anticyclonic eddy that occurred during winter 1994 constituted the deepest MLD found during the time series but did not coincide with a high Chl-a concentration.

[23] During the stratification period, the TChl-a correlation between sites was inversely correlated (r = −0.61, p < 0.05; not shown), whereas the GRD correlation between both sites was also inverse but not significant (r = −0.27, p > 0.1; not shown). BATS showed decreasing linear tendency in the yearly potential temperature gradient anomalies while they increased at ESTOC. These inverse tendencies between BATS and ESTOC regarding GRD and TChl-a during the stratification period could be a consequence of highest (lowest) opposite anomalies occurring during 1994 and 1996 at BATS and ESTOC. During these years large variations in TChl-a were observed during the stratification period at both sites. If warmer summers were the reason for an increase in GRD, it would imply similar correlations at both sites. However, nutrient supply to the euphotic zone by eddy-pumping which represents the 50% of the total annual flux at BATS [Cianca et al., 2007] would affect the relationship between GRD and Chl-a. At ESTOC, the seasonal contribution of the northeasterly winds, with maximum values during summer as mentioned insection 3.1, produces mixing in the upper layer and tempers the temperature increase at ESTOC. In addition to hydrography, other factors could affect nutrient availability and consequently TChl-a, such as nitrogen fixation which was hypothesized to play a greater role at BATS compared to ESTOC [Neuer et al., 2002].

3.3. Influence of Low-Frequency Climate Modes on the Chlorophyll Variability

[24] NAO and ENSO are usually correlated with variability in sea surface temperature and salinity or water column stratification in the North Atlantic through changes in the wind-forcing (westerly and trade wind variability), as well as variability in the oceanic circulation (e.g., latitudinal or longitudinal displacements of Gulf, Azores or Canary Currents) [Curry and McCartney, 2001; Siedler et al., 2005]. These processes can in turn affect the thermohaline distribution and can cause anomalies in the temperature and salinity field.

[25] Oschlies [2001] linked the winter variability with the NAO index (negative index / intense convective mixing; positive index / moderate convective mixing). The authors modeled nutrient supply over the subtropical North Atlantic for the observed swing in the NAO between the early 1960s and 1990s and found a decreased nutrient supply by more than 30% in the western side, whereas it increased by about 60% in the upwelling region off West Africa. Palter et al. [2005] linked the variability of the nutrient budget and primary production in the Sargasso Sea to NAO induced changes in the NASTMW (North Atlantic Subtropical Mode Water) stability. Other investigations such as by Santana-Casiano et al. [2007] proposed two thermohaline variability scenarios to explain changes in the carbon dioxide concentrations at ESTOC. The first scenario depicted periods of cooler and fresher waters. In addition, these periods presented changes in the vertical mixing (intense convective mixing) and changes in the water mass composition (annual class of the mode water). The second scenario represents the periods where temperature and salinity exhibit inverse relations. These changes in the thermohaline behavior presumably were related to variations in the regional evaporation/precipitation balance, as well as produced by north/south displacements of the frontal areas and current axes (Gulf, Azores or Canary Currents). These displacements in the oceanic circulation were previously observed and correlated with the NAO index by Curry and McCartney [2001].

[26] In the Bermuda region, air-sea CO2fluxes were correlated with NAO in summer and fall, whereas wintertime air-sea CO2 influxes were poorly correlated with NAO, although they increased during El Niño years, both with a + six months time lag [Bates, 2001, 2007]. In the ESTOC region, air-sea CO2 fluxes were correlated with NAO with a + three year time lag and with the Eastern Anomaly (EA) during wintertime (without time lag) [González-Dávila et al., 2007; Santana-Casiano et al., 2007]. These latter authors found a SST increase and a SSS decrease for the time series and hypothesized that these changes were caused by a weakening of the overturning circulation in the West Atlantic Ocean, changing heat and freshwater fluxes, increasing precipitation in the temperate Atlantic, and a shift in the Azores Current axis toward the southwest.

[27] With the aim of estimating the influence of the low-frequency climate modes on the chlorophyll dynamics on both sides of the North Atlantic subtropical gyre, we obtained the anomalies of the integrated chlorophyll concentration to 200 m for both the mixing and the stratification periods. TChl-a anomalies were estimated from annual averages of 4-monthly values. This estimate was considered the best approach to average the short-term variability of the phytoplankton distribution. These anomalies were then correlated with the main climatic variability modes described for the region (e.g., NAO, ENSO or AMO;Figure 5), using several time lags. The time lags widely used are + one or + three years, as well as a + six months lag which was used to correlate oceanic parameters with NAO and ENSO at Bermuda by Bates [2001]. In our study, there is an eight month time lag between NAO (calculated using the anomalies from December to March) and the stratification period (August to November).

Figure 5.

(a–c) Yearly anomalies of low-frequency climate indices computed from the 4-monthly averages in winter (DJFM). (d–f) Yearly anomalies of summer indices computed from the 4-monthly averages (JASO).

[28] The results from our study showed a significant correlation at ESTOC between ENSO and TChl-a anomalies for the mixing period with a + one year time lag (r = −0.76 SOI; r = 0.83 MEI; r = 0.79 ONI; p < 0.01;Table 1) and smaller significant correlations with NAO + three years (r = −0.59, p < 0.05). There is also a significant correlation between MLD and NAO + three years (negative NAO, deeper MLD; Table 1). These results are in good agreement with earlier assessments using the variables of the carbon dioxide system by Santana-Casiano et al. [2007] at ESTOC. Siedler et al. [2005] hypothesized that a change in the sign of NAO index produces changes both in the baroclinic transport and in the spatial pattern of the NASTG with a delay of + three years. Regarding the relationship with ENSO, the results at ESTOC corroborate results from investigations in the African coast upwelling during 1998–1999, where an observed chlorophyll increase was thought to be related to ENSO variations in the local atmospheric fields [Pradhan et al., 2006]. This upwelling variability occurred with a few months time lag, while the adjacent areas toward the inner gyre from the coast (including ESTOC) were affected during the following winter, probably favored by a deep convective mixing in 1999 [Neuer et al., 2007]. As mentioned in section 3.2, McClain et al. [2004] showed that 1999 was an elevated chlorophyll year in the entire subtropical gyre. These conclusions are also confirmed by a study in the north of the Canary Islands region, where diatoms were observed at ESTOC in similar abundance and species composition to those near the coast [Abrantes et al., 2002]. This coastal influence which coincided with the winter bloom in the area could have been the consequence of mesoscale activity that originated off Cape Ghir [Cianca et al., 2007].

Table 1. Correlation Between Total Integrated Chl-a, MLD, GRD and the Respective Indices Regarding the Atmospheric Low-Frequency Modes, Computed From 4-Monthly Averagesa
 NAONAO +1 yrNAO +3 yrSOISOI +1 yrMEIMEI +1 yrONIONI +1 yrAMOAMO +1 yr
  • a

    Correlations with p < 0.05 are in boldface.

Mixing (January To April): BATS
Mixing (January To April): ESTOC
Stratification (August to November): BATS
Stratification (August to November): ESTOC

[29] For the BATS data, there was no correlation between TChl-a anomalies and NAO for the winter period. The correlation improved with a one year time-lag (NAO + one year) but was still insignificant (r = 0.31; p > 0.1;Table 1). However, Lomas et al. [2009]found a significant direct correlation between TChl-a variability and NAO. The difference between the results could be caused by differences in the Chl-a methodology (HPLC used byLomas et al. 2009 versus Turner fluorometry used in this study) and the length of the time series.

[30] The opposite tendency found between TChl-a at BATS and at ESTOC during stratification supports the notion that the sites show a contrary pattern regarding the response to NAO, although the correlation between NAO and TChl-a was not significant. The maximum (minimum) TChl-a anomaly occurred at BATS (ESTOC) from 1994 to 1996 (Figures 4b and 4d); 1994–1995 were positive NAO years whereas 1996 was a negative NAO index year (Figure 5a). This opposite tendency between sites is also observed in the GRD and temperature; however no significant correlations existed between the anomalies of TChl-a and the temperature gradients. Interestingly, no significant correlation was found between NAO and chlorophyll anomalies during the stratification period at either station (Table 1) for none of the climate indices and time lags tested (including the + eight months lag). In addition, our study found a significant correlation between nitrate anomalies and NAO (r = 0.79, p < 0.01), and AMO + one year (r = −0.69, p < 0.02) at ESTOC during the mixing period. At BATS, GRD anomalies and the El Niño indexes, SOI and MEI with no time lag, as well as AMO with + one year were significantly correlated, whereas MEI and ONI were correlated with GRD anomalies at ESTOC with no time lag.

[31] Thus, the relative proximity of the oligotrophic subtropical gyre station ESTOC to the eutrophic upwelling region makes that station apparently more sensitive to NAO and ENSO related hydrographic fluctuations. In this region, even small changes in the depth of winter mixing can determine whether the mixed layer penetrates the nitracline and thus allowing significant nutrient input and the immediate biological response to alterations in nutrient supply. To understand how the hydrographic variability could affect nutrient supply to the euphotic zone, we need to review studies on nutrient dynamics published during the last two decades, mainly in the Western North Atlantic Boundary Current. A summary of the discussion has been gathered by Williams et al. [2011]. Diapycnic mixing [Pelegrí and Csanady, 1991, 1994], epipycnal or isopycnal advection [Pelegrí et al., 2006; Williams et al., 2006] or imported water distribution [Palter and Lozier, 2008] appear to be the processes supplying nutrients from the upper thermocline apart from winter convection and the passage of mesoscale features already described in Cianca et al. [2007]. These processes could improve or diminish the nutrient supply in relation to changes in the oceanic circulation due to long-term variations caused by climatic variability. These processes have been widely discussed in the western subtropical Atlantic, but less so for the eastern boundary. NAO and ENSO, separately or coinciding in time, produce a strengthening of the upwelling system, which affects the baroclinic system and displaces the axis of the Canary Current toward the center of the Gyre. These changes likely produce an increase in nutrient supply by epipycnal advection. In addition, upwelled water in the African coast could be carried to the adjacent areas by mesoscale dynamics. Another potential nutrient source is atmospheric deposition as recently updated byZamora et al. [2010]. These authors suggested rates of atmospheric deposition comparable to nitrogen fixation as a cause of the excess N formation in the North Atlantic. Previous studies such as Knap et al. [1986] at BATS and Neuer et al. [2004] at ESTOC estimated a very low contribution for this nutrient source, whereas Michaels et al. [1996] suggested that interannual and decadal variations in dust deposition could play an important role in modifying the upper ocean nitrogen dynamics at the Sargasso Sea. Finally, another recent hypothesis proposed by Torres-Valdés et al. [2009] states that the input of semi labile dissolved organic nitrogen (DON) and phosphorus (DOP) produced in the upwelling zone off North Africa and transferred laterally into the inner ocean by eddies and upwelling filaments is a significant component of the total nitrogen and phosphorus pools in surface waters above the thermocline. However, the ESTOC station is not directly influenced by upwelling filaments [Neuer et al., 2007], and there is no evidence in the inorganic nutrient budget at ESTOC of possible advection and subsequent remineralization of organic nutrients from the upwelling margin.

[32] Hence, we conclude that the hydrographic variability caused by ENSO and NAO is one of the principal factors generating anomalies in the TChl-a time series at ESTOC. These results point to the need for longer regional time series to allow for a better understanding of the influence of low-frequency climate modes on the biology of the subtropical gyres.

4. Conclusions

[33] A comparative analysis of two North Atlantic time series stations (BATS; 1988–2003 and ESTOC; 1994 to 2003) provided a unique insight into their chlorophyll dynamics and sensitivity to low-frequency climate modes. BATS and ESTOC are characterized by large differences between winter and summer. Hydrography, or nitrate availability were not the only factors responsible for the variations found in Chl-a, both integrated to the DCM and to 200 m. The interannual variability and some particularly anomalous events produced several important deviations from the average. The importance of the variability in the winter convective mixing and the passage of mesoscale features at BATS appear to be responsible for these anomalies while at ESTOC, the anomalous winter convective mixing in some years and the summer strengthening of the baroclinic system are likely the main factors causing Chl-a variability.

[34] This interannual variability in hydrography and chlorophyll is likely to be connected with atmospheric variability. Correlations of TChl-a with MLD, temperature and salinity anomalies were statistically important at BATS and ESTOC during the mixing period, and significant for the relationships between TChl-a and temperature and TChl-a and salinity at BATS. The inversely significant TChl-a correlation during summertime at either station points to an offset response to NAO. Correlations with statistical significance were found between TChl-a and low-frequency climate modes for ESTOC with ENSO + one year time lag and with NAO + three years, whereas the correlation between TChl-a and NAO + one year was close but insignificant at BATS, possibly weakened by the variability of the Eighteen Degree Mode Water.


[35] We thank the U.S National Science Foundation, the Instituto Canario de Ciencias Marinas (ICCM), Gobierno de Canarias and German Ministry for Research and Education (BMBF) for financial support of the time series stations. This study has been supported by the Ministerio de Ciencia e Innovación, MOMAC project. We are especially grateful for the dedicated work of the many scientists and technical staff involved in the data collection and analysis. We thank three anonymous reviewers for helpful comments on earlier versions of the manuscript.