The North Atlantic Ocean is a major sink region for atmospheric
Since the start of the industrial era and the concomitant rise of
atmospheric
The North Atlantic Ocean is a key region for Cant uptake and storage (Sabine
et al., 2004; Mikaloff-Fletcher et al., 2006; Gruber et al., 2009; Khatiwala
et al., 2013). In this region, storage of Cant results from the combination
of two processes: (1) the northward transport of warm and Cant-laden
tropical waters by the upper limb of the meridional overturning circulation
(MOC; Álvarez et al., 2004; Mikaloff-Fletcher., 2006; Gruber et al.,
2009; Pérez et al., 2013) and (2) deep winter convection in the Labrador
and Irminger seas, which efficiently transfers Cant from surface waters to
the deep ocean (Körtzinger et al., 1999; Sabine et al., 2004; Pérez et
al., 2008). Both processes are characterized by high temporal variability in
response to the leading mode of atmospheric variability in the North
Atlantic, the North Atlantic Oscillation (NAO). Hurrell (1995) defined the
NAO index as the normalized sea-level pressure difference in winter between
the Azores and Iceland. A positive (negative) NAO phase is characterized by
a high (low) pressure gradient between these two systems corresponding to
strong (weak) westerly winds in the subpolar region. Between the mid-1960s
and the mid-1990s, the NAO changed from a negative to a positive phase. The
change in wind conditions induced an acceleration of the North Atlantic
Current (NAC), as well as increased heat loss and vertical mixing in the
subpolar gyre (e.g., Dickson et al., 1996; Curry and McCartney, 2001;
Sarafanov, 2009; Delworth and Zeng, 2016). Concomitant enhanced deep
convection led to the formation of large volumes of Labrador Sea Water (LSW)
with a high load of Cant (Lazier et al., 2002; Pickart et al., 2003;
Pérez et al., 2008, 2013). Between 1997 and the early 2010s, the NAO index
declined, causing a reduction in LSW formation (Yashayaev, 2007; Rhein et
al., 2011) and a slowdown of the northward transport of subtropical waters
by the NAC (Häkkinen and Rhines, 2004; Bryden et al., 2005; Pérez et
al., 2013). As a result, the increase in the subpolar Cant inventory was
below values expected solely from rising anthropogenic
Based on the analysis of time series of physical and biogeochemical
properties between 1997 and 2006, Pérez et al. (2013) proposed that Cant
storage rates in the subpolar gyre were primarily controlled by the
intensity of the MOC. A weakening of the MOC would lead to a decrease in
Cant storage and would give rise to a positive climate–carbon feedback. The
importance of the MOC in modulating the North Atlantic Cant inventory was
previously suggested by model studies, which projected a decrease in the
North Atlantic Cant inventory over the 21st century in response to a
MOC slowdown under climate warming (e.g., Maier-Reimer et al., 1996; Crueger
et al., 2008; Schwinger et al., 2014). Zunino et al. (2014) extended the
time window of analysis of Pérez et al. (2013) to 1997–2010. They
proposed a novel proxy for Cant transport defined as the difference of Cant
concentration between the upper and the lower limbs of the overturning
circulation times the MOC intensity (please refer to the Supplement (Sect. S1) for a model-based discussion of the proxy and for the MOC
intensity definition). The authors concluded that while the interannual
variability of Cant transport across the OVIDE section was controlled by the
variability of the MOC intensity, its long-term change depended on the
increase in Cant concentration in the upper limb of the MOC. The latter
reflects the uptake of Cant through gas exchange at the atmosphere–ocean
boundary and questions the dominant role attributed to ocean dynamics in
controlling Cant storage in the subpolar gyre at decadal and longer timescales (Pérez et al., 2013). Were the storage rate of Cant in the
subpolar gyre indeed controlled at first order by the load of Cant in the
upper limb of the MOC, the increase in the subpolar Cant inventory would
follow the increase in atmospheric
The objective of this study is to evaluate the variability of transport,
air–sea flux and storage rate of Cant in the North Atlantic Subpolar Ocean and its
drivers over the past 53 years (1959–2011). It relies on the combination of
a multi-annual data set representative of the area gathered from
25
NEMO–PISCES and the in situ data are introduced in Sect. 2 and compared in Sect. 3 to evaluate model performance. An analysis of mechanisms controlling the interannual to decadal variability of the regional Cant fluxes and storage rate is presented in Sect. 4, and results are discussed in Sect. 5.
Column inventory (molC m
This study is based on a global configuration of the ocean model system NEMO
(Nucleus For European Modelling of the Ocean) version 3.2 (Madec, 2008). The
quasi-isotropic tripolar grid ORCA (Madec and Imbard, 1996) has a resolution
of 0.5
Locations of the 24.5
References of cruises used in this study.
At the end of the spin-up cycle, two 143-year long simulations were started
in 1870 and run in parallel. The first one, the historical simulation, was
forced with spatially uniform and temporally increasing atmospheric
The model simulates a global ocean inventory of Cant in 2010 of 126 PgC. It
is at the lower end of the uncertainty range of the estimate by Khatiwala et al. (2013)
of 155
Observations used to evaluate the transport of Cant in ORCA05–PISCES were
collected along the Greenland–Portugal OVIDE section and at 24.5
The OVIDE program aims to document and understand the origin of the
interannual to decadal variability in circulation and properties of water
masses in the North Atlantic Subpolar Ocean in the context of climate change
(
Data were collected along 24.5
For both data sets, C
The gridded sea surface
The simulated transport of Cant (
The budget of Cant was computed for several North Atlantic subregions
(boxes) defined later on. A budget was defined for each box as the balance
between (i) the time rate of change in vertically and horizontally
integrated Cant, (ii) the incoming and outgoing transport of Cant across
boundaries of each region and (iii) the spatially integrated air–sea flux of
anthropogenic
Heat transport across a section was computed from horizontal velocity orthogonal to the section times the heat term estimated from temperature and salinity using the international thermodynamic equations of seawater (TEOS 2010). Heat transport is used in Sect. 4.1 to evaluate model performance to reproduce the well-known mechanism controlling its interannual variability correctly and to compare to results for Cant transport.
Anthropogenic C budget of the North
Atlantic Subtropical and Subpolar regions over the period 2003–2011. Average values and their
standard deviations were estimated from smoothed time series. Horizontal
arrows show total Cant transport in PgC yr
Figure 3 summarizes the budget of Cant in the North Atlantic simulated by
the model over the period 2003–2011. In order to enable the comparison of
the model-derived budget to previous estimates (e.g., Jeansson et al., 2011;
Pérez et al. 2013; Zunino et al., 2014, 2015a, b; Guallart et al., 2015),
we defined two boxes separated by the Greenland–Portugal OVIDE section. The
first box extends from 25
In the model, over one-third of Cant entering in the southern box at
25
Volume transport (Sv) across the OVIDE section as simulated by the
model for the month of June (continuous line for mean value; shaded band for
confidence interval) and compared to the observation-based assessments
(dashed line) over the period 2002–2010. In panel
Model–data comparison over the period covered by the OVIDE cruises (2002–2010). Average and standard deviation (SD) for observation-based estimates (column 2) and model output (columns 3 to 4). Model output: (1) June average, with SD being a measure of interannual variability, and (2) yearly average, with SD corresponding to the average interannual variability.
Volume transport (Sv) integrated zonally at 24.5
Figure 4 shows the accumulated volume transport simulated by ORCA05–PISCES
along the Greenland–Portugal section compared to assessments based on
observations from OVIDE. The simulated intensity of the MOC (see Sect. S1
for details of its estimation) underestimates the observational estimate of
15.5
Model–data comparison along 25
At 25
Water column distribution of anthropogenic C concentrations
(
The simulated spatial distribution of Cant between 25
At 25
Water column distribution of anthropogenic C concentrations
(
Interannual variability of air–sea flux of total
To summarize Sect. 3.1, the underestimation of Cant transport in
ORCA05–PISCES is likely due to the combination of weak volume transports of
NAC and Nordic overflows and low Cant concentrations. The latter is partly
explained by the preindustrial condition for atmospheric
Simulated air–sea fluxes of total and anthropogenic
Over 88 % of the simulated Cant flux entering the North Atlantic between
25
Correlation coefficient (
Next, the contribution of air–sea uptake and transport of Cant to the
variability of the North Atlantic Cant inventory is derived for each box
from the analysis of multi-annual time series of air–sea fluxes of
anthropogenic
In this section, we present the analysis of the full period covered by our
simulations (1958–2012). The objective is to better understand the
interannual to decadal variability of the North Atlantic Cant storage rate
and to identify the driving processes. The study area is now divided into
three boxes: the first box extends from 25 to 36
Annual time series of MOC intensity (Sv), heat transport (PW) and
Cant transport (PgC yr
Figure 10 presents annual time series (1958–2012) of the MOC intensity and
the transports of heat and Cant across 25
Annual time series of contributions to the anthropogenic carbon
(Cant) budget (Pg yr
Summary of
The transport of Cant across all sections increased
continuously over the period of study (Fig. 10, Table 5c). Neither heat
transport, nor MOC intensity, nor the net volume of water transported across
the sections display a similar increase (Table 5c). Zunino et al. (2014)
attributed essentially the increase in the northward transport of Cant since
1958 to its accumulation in the northward flow of the MOC upper limb. In
order to isolate the circulation effect, we removed the positive trend from
the time series of Cant transport. The correlation (
Correlation coefficient (
Figure 11 shows the budget of Cant from 1959 to 2011 for the three boxes.
Each budget is composed of the storage rate of Cant, the air–sea flux of
anthropogenic
Distribution of volume transport integrated into density (
Correlation coefficient (
The trend in response to increasing atmospheric
In this section, we identify major water masses making up the upper and
lower limb of the MOC to evaluate their contributions to the regional Cant
storage rate over the period 1959–2011. The North Atlantic circulation is
well documented. Based on previous studies (e.g., Arhan, 1990; McCartney,
1992; Hernández-Guerra et al., 2015; Daniault et al., 2016) and on the
vertical distributions of volume transports integrated zonally at
25
NACW (Class 1) is transported by upper ocean circulation, either northward
(Class 1N) by the Gulf Stream and the NAC, or southward (Class 1S) by the
subtropical gyre recirculation in the western European basin. The southward
recirculation is composed of colder and denser waters (Talley et al., 2008),
allowing the distinction of Class 1S from Class 1N in our study (Fig. 12).
NACW loses heat during its northward journey, which increases its density. As
a result, the density limits between Classes 1N and 1S and 2 change with
latitude. Based on Fig. 12, we defined Class 1N from the surface to
IW (Class 2) encompasses the densest water masses of the MOC upper limb,
such as Antarctic Intermediate Water (AAIW), Subantarctic Intermediate Water
(SAIW) or Mediterranean Water (MW). Class 2 circulates northward between
NADW (Class 3) supplies the lower limb of the MOC. It flows southward from
the subpolar gyre to the subtropical region. In the model, it is found below
Analysis of the long-term changes in the simulated transport of volume and
Cant across the four sections and for the three specified classes led to
identify two periods, before and after 1995 (Fig. S4). The distinction
between these two periods is based on Class 1N (northward NACW) at the OVIDE
section and Class 2 (IW) at 36
Before 1995, more than 50 % of Cant transported by NACW flowing northward
(Class 1N) at 25
Annual time series of the anomaly of volume transport (Sv, bar
plot) compared to the winter NAO index over the period 1959–2011 for Class 1
at 36
Annual time series of the average temperature of the mixed layer
for Box 2 (36
This cross-isopycnal transport between Class 1 and Class 2 (Fig. 13a) causes
a decrease in the volume of Class 1 waters and an increase in the volume of
Class 2 waters transported northward from 25
Figure 13a also shows that 62 % of Cant entering in Box 3 by advection of
Class 1 and Class 2 waters and by air–sea flux was converted into Class 3
inside the box and exported southward. The remainder was stored in Box 3
(18 %) or transported northward through the Greenland–Iceland–Scotland
sills as Class 2 waters (19 %). NADW was thus strongly enriched in Cant
between the OVIDE section and the Greenland–Iceland–Scotland sills by
entrainment of NACW/IW and deep convection, which is in agreement with
results from Sarafanov et al. (2012). Finally, a small fraction of Cant
entering in Box 2 within Class 3 left the area across 25
After 1995, 27 % of Cant entering within Class 1 at 25
The model–data comparison presented here highlights a large underestimation
(by 2 or 3 times) of Cant transport by the model, resulting from an
underestimation of both volume transport and Cant accumulation in the water
column. The underestimation of the NAC and Nordic overflow volume transports
was identified as the major model shortcoming. It led to an underestimation
of the intensity of the upper and lower MOC. Moreover, the underestimation
of the NAC transport resulted in a smaller transport of Cant from the
subtropical to the subpolar gyre compared to observations. The missing
southward transport of Cant associated with the Nordic overflows resulted in
a net transport of Cant to the Arctic region that was closer to observations
but for the wrong reasons (Cant transport was 3 times smaller than observations
at the OVIDE section, while it was only 2 times smaller at the sills). Our
analysis also revealed a strong overestimation of the simulated air–sea flux
of anthropogenic
Compared to the two other terms, the simulated Cant storage rate is in line with data-based estimates (Pérez et al., 2013). It reflects the compensation between the underestimation of Cant transport and the overestimation of air–sea gas exchange. However, the spatial distribution of the column inventory of Cant is well reproduced by the model, likely due to correct simulation of mechanisms controlling the interannual variability of Cant storage rate (Pérez et al. (2013); Zunino et al., 2014, 2015b) despite the underestimation of simulated Cant transport. Having assessed the strengths and limitations of the simulation, we extended the time window of analysis of interannual to multidecadal changes in the North Atlantic Cant storage rate and its driving processes to the period 1959–2011.
Over the last 4 decades, the interannual variability of the simulated
Cant storage rate in the North Atlantic Ocean was controlled by the
northward transport divergence of Cant. At the OVIDE section, the
interannual variability of Cant transport was controlled by Cant
accumulation in the MOC upper limb, whereas it was also influenced by the MOC
intensity at 25 and 36
To conclude, at the multi-decadal timescale, the long-term change in
anthropogenic
All references for the availability of in situ data sets
are indicated in the text. A simulation of the model data set can be accessed at:
The supplement related to this article is available online at:
The authors declare that they have no conflict of interest.
This article is part of the special issue “Progress in quantifying ocean biogeochemistry – in honour of Ernst Maier-Reimer”. It does not belong to a conference.
Virginie Racapé was funded through the EU FP7 project CARBOCHANGE (grant 264879).
Simulations were made using HPC resources from GENCI-IDRIS (grant
x2015010040). We are grateful to Christian Ethé, who largely contributed
to obtain Cant transport in the online mode over the period 2003–2011. We want
to acknowledge Herlé Mercier (supported by CNRS and the ATLANTOS H2020 project (GA
633211)) and colleagues for leading the OVIDE project (supported by French
research institutions IFREMER and CNRS/INSU), as well as Alonso
Hernandez-Guerra for the availability of its mass transport data at
24.5