We present a binned product of sea surface temperature, sea surface
salinity, and sea surface density
data in the North Atlantic subpolar gyre from 1993 to 2017 that resolves
seasonal variability along specific ship routes (
Francis Bringas' and Gustavo Goni's copyright for this publication is transferred to the National Oceanic and Atmospheric Administration (NOAA)..
The North Atlantic subpolar gyre (NASPG) has been extensively studied and observed during the last 25 years. This time span presents the succession of a cold period in the early 1990s associated with strong North Atlantic Oscillation (NAO) forcing, a warmer period from 2000 to 2009, followed by a cooling (Robson et al., 2016), and strong NAO forcing in 2014 and 2015 (Josey et al., 2017). These conditions were associated with strong variability in intermediate water formed in the Labrador Sea, southwestern Irminger Sea, and south of Greenland, with strong formation years following strong atmospheric and NAO forcing years (Yashayaev and Loder, 2016; Fröb et al., 2016; De Jong and de Steur, 2016; Piron et al., 2017). There has also been extensive variability in mode waters and their thickness in the northern and northeastern subpolar gyre, such as the Reykjanes mode water (Thierry et al., 2008) or the Rockall Trough mode water (Holliday et al., 2015). The changes in these subsurface water properties and distributions drive ocean circulation and in particular play a likely role in the Atlantic Meridional Overturning Circulation (AMOC) variability (Robson et al., 2016; Rahmstorf et al., 2015). The surface layer provides the link between the ocean interior and the atmosphere.
Surface variability in oceanic properties responds to atmospheric forcing and
ocean circulation changes. In particular, the NAO and the East Atlantic Pattern are known to strongly influence heat
and freshwater fluxes in this region (Cayan, 1992; Hurrell et al., 2013;
Bojariu and Reverdin, 2002) and thus sea surface temperature (SST) and sea
surface salinity (SSS) (Josey and Marsh, 2005). Changes in freshwater fluxes
from continental run-off and ice melt are also expected to change surface
properties in the NASG (Böning et al., 2016). Net run-off from Greenland
has considerably increased during the last decades (van der Broeke et al.,
2016). The role of changes in ocean circulation has also been identified. For
instance, the proportion of inflowing subtropical water was found to have
increased in the 1995–2005 period compared to the previous 2 decades
(Häkkinen et al., 2011, 2013), followed by a net reduction of this input
(Robson et al., 2016), which could have contributed to the more recent
decadal cooling/freshening (see also Piecuch et al., 2017). Decadal changes
in the strength of the gyre circulation have been associated with zonal
displacements of the subpolar front (Hatun et al., 2005; Sarafonov, 2009).
This has been disputed (Foukal and Lozier, 2017), and has not been clearly
identified in subsets of in situ current measurements along 59
The strong changes in thermocline and water masses associated with
displacements of the subpolar front and in the southern part of the NASPG
have been used by Stendardo et al. (2016) to reconstruct surface temperature and salinity based on satellite
altimetry data. However, this method does not work well in the interior of
the NASPG, nor on the slopes and shelves. To complement these analyses and
the very valuable coverage of the NASPG at low resolution by Argo floats, we
construct monthly time series of temperature (
The data and methods used are first described in Sect. 2, and then the time series are presented in Sect. 3. Some statistical properties are described to illustrate the potential of the binned data to characterize interannual variability: the pattern of standard deviations, and empirical orthogonal function (EOF) analysis. Descriptions of the data validation performed and comparisons with other products are presented in the Appendices.
A large part of the data presented here are from SBE21 and SBE45
thermosalinographs (TSGs) installed on cargo ships running along the AX01
transect between Denmark and western Greenland and along the AX02 transect
between Iceland, Newfoundland, and the northeastern USA (Fig. 1). Along AX01,
TSG data were collected on M/V
Map of the bins along B-AX01 (red), B-AX02 (black), G-AX01 (blue), and N-AX01 (green). A typical example of a ship track is shown along B-AX02.
The first installation on
Along AX02, a succession of ships has been used, with different installations usually in the engine room at mid-ship, between 4 and 7 m below the water line. The route taken by these vessels is often roughly straight between southeastern Newfoundland and the western tip of the Reykjanes Peninsula (Fig. 1, which we will refer to as the standard route B-AX02), but with some deviations depending on sea ice or weather conditions. Due to seasonal sea ice, in particular, there were no standard TSG data on the route northeast of Newfoundland on shelf and slope in February–April 1994–1995 and 2014–2016.
The validation and correction of the TSG salinity data are mostly based on
comparison with water samples collected from a water intake at the TSG (AX01
and AX02) and using nearby upper-level Argo float data (primarily for AX02)
(Alory et al., 2015). On AX02, adjusting
B-AX02
We construct monthly binned temperature, practical salinity, and water
density time series starting in mid 1993 along two standard sections
intersecting near 59.5
First, a gridded seasonal cycle is subtracted from the data to create
anomalies that are then grouped in the bins on a monthly timescale. The
average seasonal cycle is based on 120 years of data in the NASPG provided on
a
Hovmøller diagrams of seasonal salinity anomalies are presented in Fig. 2. A rather similar variability is portrayed where the two sections B-AX01 and B-AX02 intersect (with rms differences which are less than the error estimates), although clearly B-AX01 indicates a strong longitude dependence of the signals portrayed just to the east of the intersection with B-AX02.
The B-AX02 salinity plot (Fig. 2a) suggests large spatial variations
characterized by interannual to decadal variability. In the shelf and slope
regions, in particular near Newfoundland, there seems to be more short-term
variability. However, in these regions, error estimates are also larger, to
some extent as a result of insufficient sampling, as well as due to
unresolved high-frequency variability. This results in weak correlation of
Temperature anomalies (Fig. 2c, d) tend not to be correlated with the salinity ones, although there is some suggestion that the decadal variability is correlated (except on the shelves). This is seen here as the negative SST anomalies near the beginning and end of the time series with warmer temperatures in the 2000–2009 period, roughly corresponding to SSS variability of the same sign. Variability is slightly larger along B-AX02, as expected from the known westward increase in SST variability portrayed for example in the Hadley Centre SST data set (Kennedy et al., 2011a, b). Altogether there is not a large spatial variability in the temperature signals along these transects, at least on seasonal or longer timescales, except for some differences in the southern part of B-AX02 compared to other regions.
G-AX01
Seasonal cycle of interannual variability standard deviation RMSa:
Density anomalies (Fig. 2e, f) are a result of both temperature and salinity
anomalies. Except in the southern part of B-AX02 (south of 54
To a large extent, section N-AX01 (Fig. 3) presents variability that is
coherent with what is seen on B-AX01 along 59
Finally, variability on the southwestern West Greenland Shelf (G-AX01,
Fig. 3) is rather different for
For each month of the calendar year, we evaluate the root-mean square
standard deviation of interannual variability of the anomaly time series
(referred to as RMSa). We thus estimate a seasonal cycle of RMSa (Fig. 4).
For
For
The surface density RMSa seasonal cycle (Fig. 4e, f) is a mix of
what is found on temperature and salinity. Along B-AX01 and N-AX01, density
variations are dominated by temperature variations, except west of
40
The principal components (PCs) and spatial structure (EOFs) of an EOF analysis of salinity jointly for B-AX01 and B-AX02 (July 1993–December 2017) (we applied a 15-month running mean prior to the EOF analysis). The PCs are normalized to 1, and the EOF scaling is such that 1 indicates that the EOF explains 100 % of total local variance.
To illustrate the potential of these binned data to investigate interannual variability, we perform EOF analysis. For this analysis, gaps in the time series are filled by first linearly interpolating from neighboring spatial bins, and then in time from neighboring time steps. They are then normalized to unit variance.
When performing an EOF analysis of the monthly anomalies of
EOF1 has large positive values across the two sections, except in the far
west of AX01 (close to the eastern Greenland Current), and on the Labrador
shelf. EOF2, which overall explains only 14 % of the variance, has
positive values both in the Labrador Sea (B-AX02 south of 53
This illustrates the potential of the binned data to investigate the low-frequency surface variability.
The gridded data set is freely available and accessible at
ARMOR3D (GLOBAL_REP_PHY_001_021) products are freely available through
the Copernicus Marine Environment Monitoring service
(
The International Argo Program is part of the Global Observing System (Argo, 2000).
The validated data presented here are able to characterize the seasonal
variability of surface temperature and salinity along two transects crossing
the North Atlantic subpolar gyre (along 59
In the interior of the subpolar gyre, the time series can be used to precisely monitor the arrival of very large freshwater salinity anomalies in recent years, and to characterize how they relate or not to temperature anomalies. They also suggest similarities to an earlier event in 1994–1996, which is unfortunately not as well sampled overall (Reverdin et al., 2002). The salinity time series are rather different on the shelves sampled here, in particular west of Greenland and near Newfoundland. This is expected, because of the different water masses with a large proportion of water advected from the Arctic or influenced by continental inputs. Sampling with the ships of opportunity is not always sufficient in these areas, due to the presence of seasonal sea ice, and would need to be complemented by other observational platforms.
In some areas, such as on the shelves or south of 54
Results and data presented here highlight the importance of repeated ocean observations from volunteer ships, and the value of complementary data to better assess and monitor the state of the ocean and its variability from seasonal to interannual timescales.
TSG observations from M/V
Statistical summary of the comparisons of
collocated Argo data (at 3–9 m) with TSG data along AX01 and AX02 (within
58–61
We checked the consistency of these T-S data of
The “adjusted” temperature reported by the TSG was also compared with the
temperature of the XBTs launched usually every 3 months from the
We carried out similar comparisons for TSG data along B-AX02 (since 1994), but
although average results are not statistically distinguishable from the ones
on AX01 (Table A1), scatter is larger (for example in comparison to Argo
data, the standard deviation of the temperature differences equals
0.51
Comparison in 1993–2015 of
Mapped analysis products of the hydrographic data sets EN4 (Good et al.,
2013; Skliris et al., 2014) and CORA (version 6.1) (Cabanes et al., 2013) are
based on objective mapping, and contain a level near the surface which is
used here. Mapped products from Armor3D are largely based on altimetric
sea-level data with
We compare the binned (B-AX01) monthly time series (59–60
For SST, the correlation of B-AX01 with all the gridded products along this
zonal section is quite large (larger than 0.80 everywhere, albeit a little
smaller for Armor3D), with rms variability of the same magnitude as the one
in B-AX01 in the different products (although slightly smaller in CORA). The
data coverage (XBTs in addition to Argo, PALACE, and CTD casts) is often
quite good, with the largest differences in 1993–1996 when data coverage is
weaker. Despite possible near-surface stratification, the large similarity in
The comparison of TSG data with Argo profile data (Appendix A) gives
confidence in
GR has contributed to the data validation and data compilation along the two
ship of opportunity lines (AX01 and AX02) since the project was initiated in
1993. HV has provided support in Iceland and contributed to the scientific
discussion on the data compilation. GA has been in charge of AX02 data
correction and validation. DD installed the TSG on M/V
The authors declare that they have no conflict of interest.
This is a contribution to the French SSS observation service, which is
supported by French agencies INSU/CNRS, IRD, CNES, and IPEV. We are very
grateful to the crews of the different vessels on lines AX01 and AX02 from
which the salinity and temperature data have been collected, in particular
under the EIMSKIP and Royal Arctic Line (RAL) managements. We acknowledge the
strong support of this operation by Lars Heilman in Nuuk, Magnus Danielsen in
Reykjavik, Hans Magnussen in Aalborg, Denis Pierrot, and Francis Bringas.
NOAA/AOML and NOAA/CPO Ocean Observing and Monitoring Division have
contributed by maintaining the TSGs along AX02 and providing XBTs on the
different ships that have operated along the AX01 and AX02 transects.
Coriolis contributed by providing XBTs to M/V