Main

The observed freshwater—or, equivalently, low salinity—anomalies in the Nordic Seas have two possible sources when disregarding local precipitation, ice melt and river run-off: the relatively fresh Arctic Ocean to the north, and the saline North Atlantic Ocean to the south (Fig. 1).

Figure 1: Climatology of the Nordic Seas.
figure 1

a, Sea surface salinity, schematic upper-ocean circulation and domains. The Norwegian Sea domain is extended to the Fram Strait for water mass consistency. b, The θ–S (potential temperature-salinity) scatter of the gridded Nordic Seas hydrography. Water mass definitions are simplified12,30. Labels refer to polar surface water (PSW), warm polar surface water (PSWw), deep water (DW), Atlantic- and Arctic-origin intermediate, Atlantic inflow and Norwegian Coastal Current (NwCC) waters. The black lines are the 0 °C isotherm, the 35 isohaline and the σθ = 27.70 kg m−3 and σ0.5 = 30.44 kg m−3 isopycnals.

Full size image

The variable magnitude of the freshwater anomalies entering the Nordic Seas from either source is uncertain, as observation-based time series of volume transport generally do not extend far into the past. Annual freshwater input to the Arctic Ocean is estimated to about 8,500 km3, with an imbalance between input and export—that is, a change in freshwater flow—of roughly 700 km3 yr−1 (ref. 5). Freshwater anomalies entering the Nordic Seas across the Greenland–Scotland Ridge can be estimated using a mean inflow of 8 Sv (1 Sv = 106 m3s−1) (ref. 6), carrying a representative salinity anomaly of 0.1 (refs 7, 8). This yields an anomalous freshwater input of about 700 km3 yr−1 (with respect to any reference salinity in the range 33–36). It is likely that these two sources differ in their impact on the entire Nordic Seas freshwater budget, based both on their amount and in terms of their pathways and transit times. The inflow across the Greenland–Scotland Ridge directly affects the entire Norwegian Sea via the Norwegian Atlantic Current (Fig. 1). By contrast, the inflow from the Arctic primarily takes place within the East Greenland Current, which has purportedly limited impact on the basin interior9, potentially allowing freshwater anomalies to pass from the Arctic Ocean to the Atlantic Ocean without significantly influencing the Nordic Seas freshwater budget. However, there is evidence for exchange across the shelf break as well as a bifurcation of the East Greenland Current north of the Denmark Strait, diverting freshwater into the interior10,11.

The subpolar North Atlantic is changing in pentadal tandem with the Nordic Seas freshwater content at a 4:1 ratio1. This large reservoir has thus both the capacity and the timing to explain Nordic Seas change downstream. Nevertheless, the Arctic Ocean is commonly understood to be the main provider of anomalous freshwater both to the Nordic Seas and the subpolar Basins12,13,14. The cause of change beyond the Arctic and Atlantic gateways to the Nordic Seas is not pursued here, but we note that the Arctic’s general capacity for matching the subpolar freshwater budget as well as the specific concept of great salinity anomalies are currently debated4,15,16. Recent observation-based assessments of subpolar freshwater content attribute its variability to local sources such as net precipitation17 and changing westerlies affecting air–sea interactions and ocean circulation18.

We use a comprehensive observational database of the Nordic Seas19 to assess the origin of the freshwater anomalies (see Methods for details). As salinity is practically conserved in the ocean, the anomalous freshwater content can be calculated, relative to a mean field, from the observed salinity changes. In agreement with previous results1, we find that the estimated freshwater content of the Nordic Seas increased substantially between 1965 and 1995. Our extended observational record demonstrates that the freshening trend ceased after 1995 (Fig. 2).

Figure 2: Observed freshwater (FW) storage of the Nordic Seas.
figure 2

a, Total storage anomaly relative to the mean over the observational period. b, Height of the equivalent freshwater layer covering the basin corresponding to the observed salinity changes. c, Freshwater storage with 5- and 10-year memory, resulting from changes in Atlantic inflow salinity assuming a constant volume transport of 8 Sv.

Full size image

We compare the temporal changes in the integrated freshwater content of each basin relative to the temporal development of the total freshwater content (Fig. 2a). The freshwater anomaly time series of the Norwegian Sea corresponds most closely with that of the Nordic Seas. This suggests that the Norwegian Sea is the basin downstream of the main source of Nordic Seas freshwater anomalies, which would imply an Atlantic origin for most of the freshwater anomalies. The Norwegian Sea is responsible for more than half the observed Nordic Seas freshwater anomaly. The relation is corroborated by the equivalent freshwater columns of the Norwegian and Nordic seas, calculated by normalizing the freshwater anomaly with the respective surface areas, being practically identical (Fig. 2b). Furthermore, the normalized freshwater columns in all three basins are of comparable height. A simple explanation for this may be that Atlantic inflow anomalies propagate through the three basins within a pentad, as previously inferred from the same database8. A rapid communication of anomalies is also required if oceanic freshwater transport is to explain the above-mentioned co-variance between the Nordic Seas and the North Atlantic. On the basis of the general agreement, the most prominent outlier is the relatively fresh, though still saline compared to the mean, Iceland Sea in 1965–1969. The timing corresponds to the initiation of the so-called Great Salinity Anomaly which was observed in the Iceland Sea13,14.

The hydrography of Atlantic inflow is well observed7. A time series of inflow salinity across the Greenland–Scotland Ridge is readily converted to an equivalent freshwater flux entering the Nordic Seas from the south (Fig. 2c). Assuming that the mean volume transport of 8 Sv carries a given salinity anomaly for a pentad, the result equals the subsequent Nordic Seas freshwater content if this constant inflow is the only source of change, and associated with a residence time of five years. Discrepancies between the anomalous freshwater thus contributed and the Nordic Seas freshwater content could be caused by changes in volume transport across the Greenland–Scotland Ridge, a different or varying residence time, or other freshwater sources such as the Arctic. Increasing the residence time to ten years improves the correspondence with the observed Nordic Seas freshwater anomaly (Fig. 2c).

In order better to link the variability of the anomalous freshwater content to source water masses, the observations are shown in potential temperature–salinity (θ–S) space (Fig. 1b). The warm and saline extreme is the Atlantic inflow. It is cooled as it progresses northwards through the Norwegian Sea20, visible as a predominantly vertical line in θ–S space. When the cooled inflow subducts in the vicinity of Svalbard, it has attained overflow density. The bulk of the volume of the Nordic Seas (74%) is located in the cold and saline part of θ–S space (Fig. 3a) associated with the densest waters (waters denser than σ0.5 = 30.44 kg m−3; the deep water (DW) of Fig. 1b). The transformation of Atlantic-origin water into DW is completed by mixing with colder and fresher water masses. Two additional lines of transformation are visible in Fig. 1b. The warm and fresh line is associated with the Norwegian Coastal Current and contributes to the slight freshening during the surface cooling of the inflow. The ‘baseline’ is associated with the cold and fresh polar water10 contributing to DW through mixing with Atlantic-origin waters.

Figure 3: Observed hydrographic distribution of freshwater anomalies.
figure 3

a, Reference climatology for 1950–2005 shown as a volumetric (km3) θ–S plot. b–l, Freshwater anomalies (km3) in θ–S space for individual pentads, calculated relative to the reference climatology plot in a.

Full size image

Anomalies can be shown in θ–S space using mean potential temperature and salinity as coordinates. The consequent distribution of anomalous freshwater reveals the water masses carrying the anomalies at different pentads (Fig. 3). The largest freshwater anomalies in θ–S space occur along the transformation line between Atlantic inflow and DW. Anomalies in DW can persist for a decade or more after anomalies in the inflow have changed sign, because the volume of DW is large compared to the volume transport in and out of the Nordic Seas. Anomalies along the transformation line associated with polar water contributed most prominently to the 1975–1979 pentad (Fig. 3g). In the context of the Great Salinity Anomaly, however, the 1975–1979 pentad has been associated with the Atlantic inflow13,14, which is also anomalously fresh (Fig. 3g). The variance of the pentadal freshwater anomaly provides a climatological measure of variability in θ–S space. The variance is largest along the inflow–DW transformation line, reaffirming the prominent Atlantic role in the Nordic Seas freshwater budget (Fig. 4).

Figure 4: Nordic Seas freshwater variance and modelled freshwater budget.
figure 4

a, Observed pentadal variance in freshwater storage. b, Modelled pentadal variance in freshwater storage. c, The model-equivalent to Fig. 2c, with the possible accumulated contribution from sea-ice import added (five-year running mean). The observed storage anomaly is included for reference. d, The annual time series of model freshwater transports into the Nordic Seas compared with model storage shown at a three-year lag. The thick lines are five-year running means.

Full size image

We complement our analysis of the observed hydrography with a state-of-the-art global eddy-permitting ocean–sea-ice hindcast simulation21, spanning the period 1948–2009 (see also Methods). The simulated freshwater content is in overall agreement with the observed content, and the temporal evolution corresponds closely (Fig. 4c). Initially more saline than the observations, the modelled Nordic Seas freshwater content increases by almost 8,000 km3 during 1950–1984, stays high until 1999 and then decreases towards the mean state.

As opposed to the observations, the model provides time series of freshwater transports through the source gateways. These are compared with simulated storage in Fig. 4d. The sea-ice contribution through Fram Strait has the most variable flux, whereas the liquid contribution here is relatively negligible. The Atlantic inflow and the sea-ice flux both seem to influence the long-term storage pattern. The sea ice is more aligned with the 1960s freshening, whereas the Atlantic inflow is more aligned with the overall three-phased pattern; a saline first period, then a fresh state continuing into the 1990s, followed by the gradual return to a more saline state. In more general and quantitative terms, Atlantic inflow linearly explains about 60% of the interannual variance in freshwater at three-year lag, while sea-ice flux explains 20%.

The model is consistent with the observed relationship between inflow and freshwater content (compare Figs 4c and 2c). The freshwater content contributed by sea-ice is of similar magnitude to that of the inflow, but it is more variable and accordingly less correlated with the total storage.

Although water mass properties differ slightly, the modelled variance of freshwater anomalies is consistent with the observations (Fig. 4). As in the observations, the variance is high along the ‘spine’ connecting Atlantic inflow with DW, the latter being the most variant. The relatively high but more disconnected model variance at S ≈ 34 (Fig. 4b) is geographically confined to the East Greenland Coast, a relatively undersampled region with respect to observations.

Observations and a model simulation show that the Atlantic inflow is a dominant contributor to the Nordic Seas freshwater budget. Arctic freshwater anomalies seem to be less influential, particularly for the observational record. The model simulation indicates an Arctic contribution; however, less pronounced than that of the Atlantic inflow. The present study does not attempt to identify the cause of change in the reservoirs providing anomalous freshwater to the gateways and thus the Nordic Seas. Recent literature on the Atlantic inflow and the subpolar North Atlantic predominantly attributes change to the Atlantic domain17,18,22,23, although an Arctic influence cannot be excluded13,14, even if debated15,16.

It has been suggested that increased freshwater input to the Nordic Seas could lead to diminished density contrasts across the Greenland–Scotland Ridge, with serious ramifications for the density-driven ocean exchanges across the ridge1. This presupposes that freshwater input to the Nordic Seas is rooted in the Arctic Ocean, and that this freshwater reservoir, probably replenished by external forcing, is sufficiently large. If, on the other hand, the North Atlantic provides the freshwater anomalies in the Nordic Seas via the Atlantic inflow, densities on either side of the ridge would co-vary beyond interannual timescales as actually manifested by the co-varying freshwater contents1. Hence, a predominantly Atlantic origin of freshwater anomalies implies that the Nordic Seas freshwater content has limited impact on the Atlantic meridional overturning circulation.

Methods

In this study we use the NISE hydrographic dataset19, which is based on data from the ICES database (www.ices.dk/) as well as data from the Marine Research Institute; the Institute of Marine Research; the Faroese Fisheries Laboratory; and the Geophysical Institute, University of Bergen. We use a total of 220,000 hydrographic profiles and bottle data that have been obtained from the Nordic Seas between 1950 and 2005. The measurement density is highest around the late 1980s on an annual timescale, and on a monthly scale it is highest in June and lowest in December and January. Spatial data coverage is best in the Norwegian Sea and around Iceland.

We construct a 1° × 0.5° gridded climatology from the NISE dataset using the HydroBase package1, which is a tool for climatological analysis of oceanographic properties. Building on isopycnal averaging techniques1,24,25, data is binned into successive, non-overlapping five-year intervals. Within each of those pentadal intervals, monthly climatologies are constructed by combining data of the respective month with those of the previous and subsequent months, as well as a so-called ‘deep climatology’. The deep climatology is the average state of the ocean below 500 m during the respective pentad, and is used to fill gaps in data coverage below the depth range influenced by the annual cycle. When constructing a given climatological month, data is weighted such that the data from that respective month is five times more important than any of the other contributing months. The baseline for all comparisons is a mean state, constructed in a similar manner, from data from the period 1950–2005.

Pentadal temporal resolution was chosen to achieve sufficient data coverage while resolving the annual cycle. Excluding the Greenland shelf (shallower than 200 m) north of 70° N, where data is extremely sparse, data coverage of the monthly climatological fields, constructed as described above, is on average 96.5%. A separate analysis performed on the dataset from the interior Nordic Seas alone results in the same outcome and shows that the sparse data coverage on the Greenland shelf does not significantly bias our conclusions.

Our methodology is based on ref. 3, but there are differences in the way that the freshwater anomalies are calculated locally and in the reference period. We calculate the freshwater anomaly relative to the full study period (1950–2005) rather than relative to the mean over 1950–19591. We also normalize by the reference salinity at each grid point rather than the reference salinity of the water column1. This effectively allows us to investigate not only the distribution of anomalies in the horizontal plane, but to also include the vertical dimension and, more importantly, assess changes in θ–S space.

A hindcast simulation with an eddy-permitting ocean–sea-ice model, ORCA025, has been used in this study21. The simulation spans the available forcing period of the interannually varying forcing fields from 1948 to 2009 (CORE2, ref. 26). The simulation was initialized from the quasi-equilibrium hydrography that resulted from spinning up the model from climatology using one cycle of 1978–2009 forcing. In a study focussed on freshwater in the Arctic Ocean, comparison with recent moorings shows that ORCA025 realistically represents the hydrography of, and advective exchanges of freshwater through, Fram Strait27, while south of the Nordic Seas the subpolar gyre has been shown to realistically be represented in both hydrography, heat content and position of the subpolar front28. Within the Nordic Seas in the Iceland Basin, the model’s heat budget closely resembles that determined from ARGO floats29. All calculations performed on the model fields were done analogously to the calculations on the observations (see above). Oceanic freshwater fluxes across a gateway (Fig. 4d) were thus first calculated locally using the mean salinity in each grid point as reference, and then added up over the section to produce the net transport.

The model is well aligned with the target observations as described in the main text (compare Figs 2c and 4c). The timing and magnitude of individual events nevertheless differ in detail from the observations, which could, for example, be due to shortcomings of the prescribed forcing fields. The model hydrography would also require a certain response time to reflect the prescribed forcing rather than the initial hydrography in order truly to hindcast the ocean state.