On Anomalous Ocean Heat Transport toward the Arctic and Associated1

Climate Predictability2

Marius Arthun∗ and Tor Eldevik3

Geophysical Institute, University of Bergen, and Bjerknes Centre for Climate Research, Bergen,

Norway.

J. Climate, In press

4

5

6

7

8

∗Corresponding author address: Marius Arthun, Geophysical Institute, University of Bergen,

Allegaten 70, 5007 Bergen, Norway.

9

10

E-mail: marius.arthun@gfi.uib.no11

Generated using v4.3.2 of the AMS LATEX template 1

ABSTRACT

A potential for climate predictability is rooted in anomalous ocean heat

transport and its consequent influence on the atmosphere above. Here we

assess the propagation, drivers, and atmospheric impact of heat anomalies

within the northernmost limb of the Atlantic meridional overturning cir-

culation using a multi-century climate model simulation. Consistent with

observation-based inferences, simulated heat anomalies propagate from the

eastern subpolar North Atlantic, into, and through the Nordic Seas. The dom-

inant time scale of associated climate variability in the northern seas is 14

years, including that of observed sea surface temperature and modeled ocean

heat content, air–sea heat flux, and surface air temperature. A heat budget

analysis reveals that simulated ocean heat content anomalies are driven by

poleward ocean heat transport, primarily related to variable volume transport.

The ocean’s influence on the atmosphere, and hence regional climate, is mani-

fested in the model by anomalous ocean heat convergence driving subsequent

changes in surface heat fluxes and surface air temperature. The documented

northward propagation of thermohaline anomalies in the northern seas and

their consequent imprint on the regional atmosphere – including the existence

of a common decadal time scale of variability – detail a key aspect of eventual

climate predictability.

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

2

1. Introduction31

The Atlantic region exhibits distinct interannual to multidecadal variability (Deser and Black-32

mon 1993; Kushnir 1994; Frankcombe et al. 2010; Williams et al. 2014), reflected in upper-ocean33

thermohaline anomalies that propagate persistently through the North Atlantic Ocean and Nordic34

Seas toward the Arctic (Sutton and Allen 1997; Polyakov et al. 2005; Holliday et al. 2008).35

Decadal variability in ocean temperature plays an important role in the marine climate system (e.g.,36

Drinkwater et al. 2014), influencing marine life from primary production to cod stocks (Helland-37

Hansen and Nansen 1909; Hatun et al. 2009). Ocean heat anomalies also play an important role38

in Arctic sea ice variability (Francis and Hunter 2007; Arthun et al. 2012; Onarheim et al. 2014;39

Carmack et al. 2015), which in turn could influence weather conditions and climate (e.g., Screen40

et al. 2013; Vihma 2014).41

Anomalous ocean heat can extend its influence beyond the marine climate by being imprinted42

on the atmosphere (Rhines et al. 2008; Farneti and Vallis 2011; Gulev et al. 2013; Schlichtholz43

2013), acting to increase the persistence of atmospheric circulation anomalies and, hence, provide44

predictability of atmospheric variability and continental climate (e.g., Sutton and Hodson 2005).45

This, however, requires that oceanic variability is communicated to the atmosphere through surface46

heat fluxes. Understanding the mechanisms and time scales involved in the propagation of ocean47

heat anomalies and how they interact with the atmosphere is thus a prerequisite for skillful climate48

predictions for the North Atlantic/Arctic sector (Latif and Keenlyside 2011; Meehl et al. 2014).49

The flow of warm, saline Atlantic waters toward higher latitudes takes place with the North50

Atlantic Current and its poleward extension, the Norwegian Atlantic Current (NwAC; Fig. 1a). In51

the Nordic Seas, the NwAC consists of two branches; a western branch enters the Nordic Seas52

over the Faroe–Iceland Ridge and is topographically guided northward along the front between53

3

the Arctic and Atlantic waters, while a warmer and more saline eastern branch inflows through54

the Faroe–Shetland Channel and continues north as a near-barotropic shelf-edge current (Orvik55

et al. 2001). Upon reaching the western boundary of the Barents Sea the eastern branch of the56

NwAC bifurcates flowing eastward into the Barents Sea, while the northward flow converges with57

the western NwAC and continues toward the Fram Strait as the West Spitsbergen Current. Part of58

the West Spitsbergen Current continues north into the Arctic Ocean (Rudels et al. 1999), but the59

majority of the current recirculates westward in the Fram Strait (Bourke et al. 1988) and joins the60

southward flowing deeper branch of the East Greenland Current en route to the Denmark Strait,61

thus forming a cyclonic loop within the Nordic Seas. While traversing the periphery of the Nordic62

Seas and the Arctic Ocean, the warm and saline Atlantic water is gradually transformed into a63

colder and fresher outflow as a result of oceanic heat loss and freshwater input (Mauritzen 1996;64

Rudels et al. 1999). Following Eldevik et al. (2014), the three regions connected by the NwAC65

(Fig. 1a) – the northern North Atlantic, the Nordic Seas, and the Arctic Ocean – will hereafter be66

jointly referred to as the northern seas.67

Temperature anomalies have been observed to propagate northwards from the eastern subpolar68

North Atlantic along the path of the North Atlantic Current and NwAC (Furevik 2000; Holliday69

et al. 2008; Chepurin and Carton 2012; Yashayaev and Seidov 2015). In the northern seas, anoma-70

lies travel from the Greenland–Scotland Ridge to the west coast of Svalbard in approximately 1–371

years (Dickson et al. 1988; Eldevik et al. 2009). This corresponds to a propagation speed of 2–572

cm s−1, which is an order of magnitude less than the typical current speed of the NwAC (Orvik73

et al. 2001). Both anomalous air–sea heat fluxes due to changes in the large-scale atmospheric cir-74

culation (Furevik and Nilsen 2005) and changing composition and strength of ocean currents have75

been suggested to generate temperature anomalies in the northern seas (e.g., Furevik 2001; Carton76

et al. 2011; Mork et al. 2014). Specifically, Mork et al. (2014) found that heat fluxes explain about77

4

half of the observed interannual heat content variability, but that the fraction varies considerably78

in time. Carton et al. (2011), on the other hand, found that surface heat flux variations in some79

cases act to reinforce anomalies, but that the contribution was too small to explain the concomitant80

changes in ocean heat storage. It is in most cases, however, not possible to construct a closed81

observation-based heat budget because of sparse data coverage, especially in terms of ocean cur-82

rent measurements. The relative importance of ocean and atmosphere in modifying ocean heat83

anomalies can therefore not be fully distinguished from observations. Heat flux reanalysis prod-84

ucts also partly disagree, making it sometimes problematic to compare with changes in observed85

hydrography (Carton et al. 2011).86

The purpose of this paper is twofold: To disentangle the contributions from ocean circulation87

and air–sea exchange in the propagation of ocean heat anomalies from the North Atlantic toward88

the Arctic, and to assess the potentially predictable relation between anomalous ocean heat and89

climate in the northern seas region. To this end, a 500-year control simulation with the Bergen90

Climate Model is used (Ottera et al. 2009). The model analysis is aided by historical sea surface91

temperatures (HadISST; Rayner et al. 2003).92

The model and the observations are introduced in section 2. In section 3 we evaluate the model93

performance for the northern seas. The propagation and drivers of anomalies are then analyzed94

in section 4 and section 5, while the link to upstream variability in the subpolar North Atlantic95

is discussed in section 6. The atmospheric imprint of ocean heat anomalies and the identified96

characteristic time scale of oceanic and atmospheric variability are discussed in section 7. Finally,97

the main conclusions and implications are presented in section 8.98

5

2. Data and Methods99

a. Bergen Climate Model100

This study uses a 500-year pre-industrial control simulation from the Bergen Climate Model101

(BCM), a fully coupled atmosphere-ocean-ice general circulation model. A general description of102

the model is given by Furevik et al. (2003), while the model run used is described in Ottera et al.103

(2009). Only a short summary is given here.104

The ocean component of BCM is a modified version of the Miami Isopycnic Coordinate Ocean105

Model (MICOM; Bleck et al. 1992). The version used for this model run uses potential density106

with reference pressure at 2000 dbar as vertical coordinates (σ2 coordinates). The ocean model107

consists of 34 isopycnic layers, ranging from σ2 = 30.119 kgm−3 to σ2 = 37.800 kgm−3, be-108

low a non-isopycnic mixed layer. The mixed layer depth is calculated from the turbulent kinetic109

energy balance of a Kraus-Turner type one-dimensional mixed layer model (Gaspar 1988), with110

modifications detailed in Medhaug et al. (2012). The horizontal grid resolution is 2.4◦ longitude111

× 0.8◦ latitude at the equator, becoming more isotropic with increasing latitude. In the northern112

seas the horizontal resolution ranges from 70–100 km. MICOM is coupled to a multi-category113

dynamic-thermodynamic sea ice model, GELATO (Salas-Melia 2002). The atmospheric com-114

ponent is ARPEGE-CLIMAT3 (Deque et al. 1994), a low-top spectral model with a horizontal115

resolution of ∼2.8◦ and 31 vertical levels. Fluxes of mass, energy, and/or momentum are calcu-116

lated in ARPEGE and communicated to the ocean via the OASIS (Terray and Thual 1995) coupler.117

External forcing from, e.g., solar insolation and greenhouse gases, is set to constant pre-industrial118

values. No flux corrections are applied, allowing the model to freely develop its own climatology.119

The initial conditions for the pre-industrial control simulation are obtained from the end of a 500-120

6

year spin-up integration (detailed in Ottera et al. 2009), after which the model is run for 600 years.121

Here we use the last 500 years of the simulation.122

b. Methods123

Ocean heat anomalies are assessed as follows. The heat content is calculated as:124

H(x,y, t) = ρwcp! η

−D[T (x,y, t)−Tref]dz, (1)

where T is temperature, ρw is density of seawater, and cp is the specific heat capacity of seawater.125

The heat content is calculated between the free surface η and the full depth of the ocean, D, using126

a reference temperature Tref =−2◦C, although we note that heat content anomalies (relative to the127

local mean) presented herein are practically insensitive to the specific reference temperature (not128

shown). We also note that an ocean heat budget calculated from the heat convergence of a closed129

mass budget (section 5) is independent of Tref. Monthly anomalies at each grid point are then130

obtained by subtracting the respective mean seasonal cycle. Unless stated otherwise time series131

are then low-pass filtered with a third-order Butterworth filter with a cut-off period of 3 years to132

emphasize interannual to decadal variability.133

Statistical significance is assessed using a two-tailed Student t-test, adjusted for serial autocor-134

relation (Chelton 1983). All correlations given in the text are significant at the 95% confidence135

level.136

c. Complex principal component analysis137

Propagating phenomena can be identified and objectively analyzed from a complex principal138

component (CPC) analysis which detects traveling waves in the input time series. A full descrip-139

tion of the procedure is given by Horel (1984), and only a short summary is presented here. First, a140

complex dataset f (x, t) is formed from the original data (by rotating its Fourier components π/2).141

7

Complex eigenvectors are then computed from the cross-covariance matrix derived from the com-142

plex dataset. From the covariance matrix, the CPCs [Pn(t)] and the complex empirical orthogonal143

functions [CEOFs; Fn(x)] are calculated. The (complex) dataset can thus be represented as a sum144

of the contribution from the N principal components:145

f (x, t) =N

∑n=1

Pn(t)F∗n (x), (2)

where x denotes spatial position and t is time. The asterisk denotes complex conjugation. The146

elements of the CPC time series can furthermore be written in the form of an amplitude an and a147

phase φn; Pn(t) = an(t)eiφn(t). The importance of each CEOF (mode) is defined as the proportion148

of variance explained by each principal component.149

d. Observed SST150

To corroborate the model analysis we use sea surface temperature (SST) data from the Hadley151

Centre (HadISST; Rayner et al. 2003) covering the period between 1870 and 2013. These data152

have a spatial resolution of 1◦ longitude by 1◦ latitude and monthly temporal resolution. We will153

only consider winter (December–April) SST as it represents the upper-ocean heat content as a154

result of a deep winter mixed layer (Nilsen and Falck 2006). The observation-based analysis is155

furthermore restricted to the southern Norwegian Sea to avoid the potential influence of sea ice. A156

good agreement in terms of interannual to decadal variability has been found between HadISST157

and data from standard hydrographic sections in the northern seas (Hughes et al. 2009).158

3. Model performance in the northern seas159

To assess the propagation of ocean heat anomalies in the northern seas it is essential that the160

model of choice is able to adequately represent the northward flow of Atlantic water and the161

gradual transformation into dense overflow water as it circulates the periphery of the Nordic Seas.162

8

Previous applications of the Bergen Climate Model in the northern seas (e.g., Ottera et al. 2010;163

Langehaug et al. 2012a,b; Medhaug et al. 2012; Lohmann et al. 2014) have found that the model164

realistically simulates the structure and the mean poleward heat transport of the NwAC, as well165

as the dense overflow and fresh surface waters in the Denmark Strait. Notably, the thermohaline166

contrast between these three water masses occupying the Greenland–Scotland ridge is consistent167

with observations, although the model hydrography is skewed toward warmer and more saline168

properties. For the Atlantic inflow specifically, the model is ∼2◦C warmer and ∼0.5 saltier than169

observations (cf. Fig. 7 in Langehaug et al. 2012a). The associated modeled volume transport into170

the Nordic Seas is 7.4 Sv for the NwAC, 2.1 Sv in the East Greenland Current, and 5.7 Sv of171

overflow water, which is in good agreement with observational estimates of 8.5 Sv, 0.4–2.1 Sv,172

and 6.4 Sv respectively (see Langehaug et al. 2012a and references therein). The model also173

captures the bifurcation at the western boundary of the Barents Sea (Fig. 1b), with a heat transport174

into the Barents Sea (61 TW; 1 TW = 1012 J s−1; Medhaug et al. 2012) which is close to that175

observed (Arthun et al. 2012). The realistic model transports and properties of inflowing and176

outflowing waters point to an accurate modification of water masses within the Nordic Seas. This177

is corroborated by Langehaug et al. (2012b) who found a realistic structure of surface-forced water178

mass transformation (diagnosed from surface buoyancy fluxes) along the path of the North Atlantic179

Current in the model.180

The observed and simulated winter sea ice extent is shown in Figure 1. The model ice cover181

is in good agreement with observations in the Norwegian Sea and western Barents Sea, while182

it is generally larger than the observed in the Greenland Sea. However, as noted by Smedsrud183

et al. (2013), a more extensive ice cover in the model is reasonable as the simulation uses constant184

pre-industrial external forcing, and therefore does not include recent sea ice decline in the Arctic.185

The simulated minimum and maximum winter sea ice edge is furthermore in agreement with the186

9

observed minimum and maximum (Smedsrud et al. 2013), i.e., the model variability is within the187

observed range.188

Admittedly, the Bergen Climate Model and other global climate models are unable to resolve189

the smaller scale features of ocean circulation, e.g., mesoscale ocean eddies and narrow boundary190

currents. However, multidecadal variability of the coupled atmosphere–ocean system can only be191

studied using relatively coarse climate models, as multicentury simulations with eddy-resolving192

grid resolution are generally not available. Fully coupled climate models are nevertheless valuable193

and necessary tools for assessing climate variability on decadal time scales and beyond.194

4. Propagation of simulated ocean heat anomalies195

To assess the properties and modification of modeled ocean heat anomalies in the northern seas196

we first need to determine the path of propagation. Based on the mean circulation (illustrated by197

the barotropic stream lines in Figure 2) and extent of the simulated NwAC (Fig. 1b), 11 stations198

(St1–11) have been defined that capture the mean propagation. This includes one station in the199

North Atlantic, corresponding to the Rockall Trough, one at the Greenland–Scotland ridge, and200

nine stations downstream within the Nordic Seas. The grid points included in each station are201

shown as circles in Figure 2. The lagged correlation between adjacent stations is generally high202

(>0.7; lags vary, but are typically around 6 months) for both heat (Fig. 2) and salt content (not203

shown), except for lower values between St2 and St3, and St9–11. The former could be a result of204

both variable communication between the North Atlantic and Nordic Seas (e.g., Hatun et al. 2005)205

and hydrographic variability internal to the Nordic Seas (e.g., Mork et al. 2014), while the less206

coherent signal between St9 and St11 could be a result of the branch of Atlantic water entering the207

Nordic Seas west of Iceland (Fig. 1b; Langehaug et al. 2012a) or branching of the southward flow208

in the south Greenland Sea (Mauritzen 1996).209

10

The time evolution of ocean heat anomalies along the defined path is shown in Figure 3a. Prop-210

agating warm and cold anomalies are evident throughout the time period. Heat anomalies with211

a depth-averaged standard deviation of 1.4·1019 J (corresponding to 0.5◦C) are associated with212

salinity anomalies of 0.05, warm conditions being accompanied by higher salinities (Fig. 3a), i.e.,213

anomalies being largely density compensated. Elevated variability is found in the southern end214

of the Norwegian Sea (St2–3) and in the area between Norway and Svalbard (St6–7). This spa-215

tial pattern is consistent with that inferred from observed Norwegian Sea heat content variability216

(Skagseth and Mork 2012). The observed anomalies discussed by Furevik (2001) also showed217

the largest amplitude in the Sørkapp section at 76.5◦N (approximate position of St7). The along-218

path evolution (relative to the local mean) is determined by the concomitant anomalous forcing219

(ocean and atmosphere) within the northern seas. A northward strengthening of an anomaly can220

for instance be explained by anomalously low surface heat loss or by an increased advection speed221

(Furevik 2001).222

In the North Atlantic and southern Norwegian Sea (St1–5) the magnitude of ocean heat content223

variability is largest between 300 m and 500 m depth (Fig. 4). The stations further downstream224

(St6–9) show maximum variability deeper in the water column (although less pronounced at St8–225

9), reflecting the along path modification and deepening of the Atlantic water. The marked change226

between St7 and St8 relates to the boundary between the ice-free Norwegian Sea and the seasonally227

ice covered Fram Strait and Greenland Sea (Fig. 1b), i.e., the northward extent of the Atlantic228

domain. The associated temperature variability is approximately 0.6–1.0◦C in the Norwegian229

Sea, while it is smaller (<0.4◦C) in the Greenland Sea and in the Rockall Trough. The surface230

intensified variability at St6 and St7 is most likely related to large surface heat loss [both modeled231

and observation-based; Langehaug et al. (2012b)], whereas the large surface variations at St11 is232

caused by the Atlantic inflow west of Iceland. The depth of maximum heat content variability at233

11

St1–6 corresponds the average depth of the winter mixed layer (Fig. 4). Noting that the depth of234

the winter mixed layer in the Norwegian Sea reflects the base of the Atlantic layer (Nilsen and235

Falck 2006), this suggests that heat content changes are largely driven by changes in the layer236

thickness and, hence, volume of Atlantic water (Sandø et al. 2012). This will be elucidated further237

in the next section.238

Propagating phenomena can be objectively assessed from a complex principal component anal-239

ysis, which detects traveling waves in the input time series and orders the dataset into modes of240

phase propagation in space and time according to the variance explained (Section 2). The leading241

mode of phase propagation of along-path heat anomalies (Fig. 3b) explains 50% of the total vari-242

ance in the full dataset (Fig. 3a) and is well separated from the second mode (accounting for 21%243

of the variance). The phase angle of the leading mode increases with increasing station number244

(Fig. 3c), which implies that the simulated heat anomalies predominantly travel along the rim of245

the basin in the direction of the mean current, consistent with observation-based inferences (Holli-246

day et al. 2008; Eldevik et al. 2009). The 500-year time period consists of 31 complete cycles with247

a a phase propagation that is rather constant in time (Fig. 3d). This yields a period of 16 years.248

The circulation from St1 to St11 constitutes just over half a cycle which, with a travel distance of249

about 4800 km, implies that the speed of modeled anomalies is on average 2 cm s−1, an estimate250

which is in reasonable agreement with observations (Furevik 2000; Polyakov et al. 2005; Holliday251

et al. 2008; Chepurin and Carton 2012).252

The representativeness of the time scale associated with propagating anomalies obtained from253

the complex principal component analysis compared with the full variance can be evaluated by254

a frequency analysis of the anomalous heat content at individual stations (note that no filter is255

applied in the frequency analysis). Heat anomalies in the northern North Atlantic (St1) and the256

Norwegian Sea (St3) both have a significant (95% confidence level) characteristic time scale of257

12

14 years (Fig. 5a). The 14-year time scale is also clearly identifiable for salinity (Fig. 5b). All258

Atlantic-dominated stations (St1–7; cf. Fig. 4) as well as downstream St8 and St9 show significant259

power on this time scale. As mentioned above, the weaker signature of a propagating signal at St10260

and St11 is most likely a result of thermohaline anomalies exiting the Nordic Seas both west (St11)261

and east of Iceland (Mauritzen 1996; Eldevik et al. 2009). Significant interdecadal variability262

is also found in the observation-based HadISST winter sea surface temperatures between 1870263

and 2013 for the same region (0.5–17.5◦E, 60.5–71.5◦N; Fig. 5c), increasing the confidence in264

the model’s ability to simulate climate variability in the northern seas. The modeled ocean heat265

content also displays significant multidecadal variations (∼40–50 years). Variability on this time266

scale will, however, not be addressed here and the reader is referred to e.g., Frankcombe et al.267

(2010) and references therein for a discussion on mechanisms for multidecadal variability in the268

North Atlantic.269

5. Heat budget for the Norwegian Atlantic Current270

To assess the relative roles of ocean advection and air–sea fluxes in driving ocean heat anomalies,271

and how anomalous ocean heat might imprint on the atmosphere, the depth-integrated heat budget272

for the Norwegian Sea (Fig. 6) is now assessed in particular. The chosen area corresponds to273

the Atlantic dominated eastern Nordic Seas (Fig. 1) where the heat content variability is highest274

(Fig. 6) and interaction with the atmosphere is strongest (Langehaug et al. 2012b). The heat budget275

area is also similar to that used in the observation-based heat content analysis by Carton et al.276

(2011) and Mork et al. (2014). Although heat content variability in the whole water column is277

considered, changes in Norwegian Sea heat content predominantly reflect variability within the278

well-mixed Atlantic layer (r = 0.72; Nilsen and Falck 2006; Chepurin and Carton 2012) which is279

in contact with the atmosphere.280

13

Changes in heat content within a control volume occur as a result of the imbalance between the281

surface heat flux (area A) and advective and diffusive heat transports through the vertical bound-282

aries (area S):283

∂H∂ t"#$%

Ht

= ρwcp!

SvTdS

" #$ %

Qadv

−!

AqdA

" #$ %

Qs

+Qres, (3)

whereHt is the heat content tendency, v is the cross-sectional velocity into the control volume used284

to calculate the heat transport convergence (Qadv), q is the net ocean–atmosphere heat flux (the sum285

of both turbulent and radiative components) which integrated over an area A yields Qs, and Qres is286

a residual term representing the ocean heat transport into the domain resulting from parameterized287

diapycnal mixing and lateral turbulent mixing (Ottera et al. 2009). Note that positive surface heat288

fluxes indicate ocean heat loss. In Figure 6a the time-varying heat budget components are plotted289

as anomalies (for presentation purposes only the first 100 years are plotted, while all calculations290

are based on the full 500-year time series), calculated as previously described in relation to Eq. (1).291

The heat content rate of change is strongly associated with ocean heat transport convergence, both292

in terms of variability (r = 0.70) and amplitude. The larger contribution from oceanic variability293

in driving heat content change is generally the case within the northern seas (Fig. 7) and especially294

along the path of the NwAC, whereas air–sea fluxes are relatively more important in parts of the295

Greenland Sea and in the northwestern Barents Sea.296

The advective heat budget for the Norwegian Sea is in turn dominated by anomalous heat trans-297

port between Iceland and Scotland (r = 0.82), i.e., the Atlantic inflow (HTaw; Fig. 6b). The heat298

transport anomalies associated with the Atlantic inflow lead the northern (HTwsc) and eastern299

(HTbs) outflow by about 2.5 years, consistent with the calculated speed of anomalies. Heat trans-300

port anomalies (HT ′ = ρwcp&

S (vT )′dS) can occur as a result of changes in advection speed (v′T )301

or temperature (vT ′), or through eddy fluxes (v′T ′). The respective contributions of these terms to302

14

the Iceland–Scotland heat transport are shown in Figure 8. The variability of HTaw is dominated303

by v′T , i.e., velocity and thus volume transport fluctuations rather than temperature fluctuations304

drive heat transport anomalies into the area. The dominant role of ocean advection on heat storage305

rates on interannual to decadal time scales agrees with recent findings from the subpolar North306

Atlantic (Buckley et al. 2014; Williams et al. 2014).307

Heat anomalies in the Norwegian Sea (Fig. 6), rooted in the ocean, are furthermore found to force308

changes in air–sea fluxes. The correlation between ocean heat transport convergence and surface309

heat loss is 0.66, i.e., anomalously high ocean heat transport corresponds to enhanced heat loss to310

the atmosphere. Consistent with variable air–sea exchange associated with ocean heat anomalies,311

the surface heat flux and surface air temperature (SAT) within the Norwegian Sea (evaluated at312

grid points shown in Fig. 6) also have the same decadal-scale oscillation of 14 years (Fig. 9a).313

The atmospheric temperature anomalies covary with ocean heat content (r = 0.73), leading to a314

reduced thermal contrast between ocean and atmosphere and thus contributing to weaker air–sea315

fluxes and reduced damping of ocean heat anomalies [r(SAT,Qs) = −0.47]. Changes in surface316

air temperature lag variations in the Atlantic inflow (HTaw) by 2–3 years (r = 0.57). The potential317

predictability of atmospheric variability from ocean heat anomalies is further discussed in section318

7.319

6. Source of northern seas heat anomalies320

The temporal development of heat content anomalies in the northern seas (Fig. 3) implies a prop-321

agating signal. The poleward progression of thermohaline anomalies – heat and freshwater – is322

also a robust finding in observations and in other ocean and climate models (e.g., Dickson et al.323

1988; Hansen and Bezdek 1996; Krahmann et al. 2001; Holliday et al. 2008; Chepurin and Car-324

ton 2012; Glessmer et al. 2014). There is nevertheless no complete mechanistic understanding of325

15

the driving mechanisms of these anomalies, including the roles of ocean dynamics and stochastic326

atmospheric forcing (see review by Liu 2012). In support of the latter, a number of studies have327

related low-frequency temperature variability in the North Atlantic to atmospheric variability as-328

sociated with the winter North Atlantic Oscillation (NAO; Hurrell 1995). Visbeck et al. (1998) and329

Krahmann et al. (2001) demonstrated how the formation and propagation of temperature anomalies330

along the pathway of the North Atlantic Current can be obtained by temporal changes in NAO-like331

wind forcing. The propagation was found to be a result of both advection of existing temperature332

anomalies by the mean ocean currents and locally generated anomalies from spatial variations in333

the external forcing. Simple advection of coherent temperature anomalies through the North At-334

lantic is also not supported by recent drifter studies (e.g., Burkholder and Lozier 2014), showing335

no direct advective pathway of anomalous heat between the subtropical and subpolar region.336

Variable (NAO-like) atmospheric forcing can also induce upper-ocean temperature anomalies337

through modulation of the North Atlantic Ocean circulation and subpolar gyre (SPG) strength,338

driving changes in poleward heat transport (Czaja and Marshall 2001; Eden and Jung 2001;339

Lohmann et al. 2009). This has also been shown for the Bergen Climate Model (Langehaug340

et al. 2012a; Medhaug et al. 2012). It has previously been demonstrated both from observations341

(e.g., Hatun et al. 2005; Yashayaev and Seidov 2015) and modeling studies (e.g., Jungclaus et al.342

2014) that variability in the amount and temperature of Atlantic water flowing northward across343

the Greenland–Scotland ridge are driven, in part, by the strength of the SPG. In the Bergen Cli-344

mate Model this is resonated in enhanced spectral power at the same frequencies for the modeled345

heat transport by the NwAC (HTaw) and the SPG strength (calculated from the barotropic stream-346

function in the subpolar region), including the 14-year periodicity found in ocean heat anomalies347

within the northern seas (Fig. 9b). A strengthening of the subpolar gyre precede heat and salt348

content changes in the Norwegian Sea by 1 year, r = 0.46 and r = 0.60, respectively. Our results349

16

thus support a close coupling between the subpolar North Atlantic and climate variability in the350

northern seas through variations in poleward heat transport.351

7. Oceanic forcing of atmospheric variability352

The identified persistent northward advection of anomalous ocean heat and the consequent353

decadal changes in surface heat fluxes and surface air temperature (Fig. 6; Fig. 9a) suggest po-354

tential climate predictability in the northern seas region. To further demonstrate the predictive355

capability associated with a variable ocean heat transport, Figure 10 shows the two-year lagged356

response in northern seas surface heat fluxes (Fig. 10a) and surface air temperature (Fig. 10c) to357

changes in the Atlantic inflow based on linear regression. Inflow-driven heat fluxes of 5–20Wm−2358

(per standard deviation of heat transport) occur offshore of the Norwegian coast and further north359

in the marginal ice zone around Svalbard and in the Barents Sea, reflecting fluctuations in the zonal360

and meridional extent of the Atlantic domain, and hence sea ice extent (Fig. 10b), respectively. The361

air temperature response is, on the other hand, pronounced over large parts of the northern seas.362

The magnitude of the atmospheric response to ocean heat anomalies varies from >0.5◦C in the363

marginal ice zone to 0.1–0.3◦C in the southern Norwegian Sea (Fig. 10c), values being similar to364

temperature anomalies in the ocean (Fig. 4). There is also a significant atmospheric response over365

land (0.1–0.4◦C), in agreement with Norwegian climate (air temperature) reflecting decadal tem-366

perature variability in the Norwegian Sea (e.g., Eldevik et al. 2014), and, more broadly, European367

continental climate reflecting North Atlantic SSTs (e.g., Sutton and Hodson 2005).368

Understanding the interaction between the ocean and the atmosphere is a prerequisite for un-369

derstanding and predicting climate variability. To what extent anomalous ocean heat leads to370

atmospheric circulation changes is not addressed here. Modeling studies have generally consid-371

ered that the amplitude of the atmospheric response to extratropical large-scale SST anomalies372

17

is modest compared with internal atmospheric variability (Kushnir et al. 2002), but this is still a373

matter of debate and might be both model (Omrani et al. 2014; Smirnov et al. 2015) and time scale374

dependent (Gulev et al. 2013; Sheldon and Czaja 2014). Ocean heat anomalies in the northern seas375

can in any case yield a significant regional atmospheric response. In line with our results van der376

Swaluw et al. (2007), also using data from a pre-industrial control run, showed that anomalous heat377

transport by the NwAC forces the atmosphere by increased heat fluxes in the marginal ice zone378

(>70◦N; cf. our Fig. 10b). The atmospheric response to increased heat transport was associated379

with a cyclonic pressure anomaly and decreased atmospheric heat transport by baroclinic eddies380

as a result of a decreased poleward temperature gradient in the atmosphere. Similar mechanisms381

and impact were also identified by Schlichtholz (2013) based on observational data from the north-382

ern seas. The regionally confined anomalous atmospheric circulation in response to decadal-scale383

ocean-driven sea ice variability in the northern seas, and in particular in the Barents Sea, can also384

drive larger-scale surface climate variability (e.g., Semenov et al. 2010; Liptak and Strong 2014).385

In support of decadal ice–ocean interaction in the Barents Sea, the modeled winter (December–386

April) sea ice cover in the Barents Sea (15–60◦E, 70–81◦N) has a similar spectrum to that of387

climate variability in the northern seas (Fig. 5; Fig. 9b), including a dominant time scale of 14388

years (not shown). The correlation between the low-pass filtered heat transport into the Barents Sea389

(HTbs) and sea ice extent in the Barents Sea is -0.77, with a time lag of 1–2 years in agreement with390

observations (Arthun et al. 2012; Onarheim et al. 2015). The decadal-scale oscillation furthermore391

agrees with Venegas and Mysak (2000) and Vinje (2001) who found fluctuations in the observed392

Barents Sea ice extent with a time scale of 16–20 years and 12–14 years, respectively, related to a393

variable NwAC.394

Identifying a time scale associated with northward propagating ocean heat anomalies from the395

subpolar North Atlantic toward the Arctic (Fig. 5; Fig. 9b) and their consequent interaction with396

18

the atmosphere (Fig. 9a; Fig. 10) is essential in terms of potential climate prediction. A recent397

model study by Escudier et al. (2013) found that a 20-year coupled mode of atmosphere–ice–398

ocean variability may exist in the subpolar North Atlantic in which the propagation of thermohaline399

anomalies from the subpolar gyre interacts with the atmosphere in the northern seas to eventually400

produce anomalies of the opposite sign in the Labrador Sea. Results demonstrated herein further401

supports a coupled mode of variability in the northern seas associated with the propagation and402

atmospheric imprint of ocean heat anomalies. The different time scale in the two models is likely403

related to the stronger northward heat transport in the Bergen Climate Model (Langehaug et al.404

2012b) as the time scale of the cycle is set by the propagation of anomalies along the rim of the405

Nordic Seas. The robustness of the time scale identified herein needs to be further assessed, but406

we reiterate that the characteristic 14-year time scale of climate variability in the northern seas is407

supported by observed SST fluctuations in the Norwegian Sea (Fig. 5c).408

8. Conclusions409

Interannual to decadal-scale ocean heat anomalies associated with the northern limb of the At-410

lantic meridional overturning circulation propagate persistently toward the Arctic (e.g., Holliday411

et al. 2008). This poleward propagation of anomalous ocean heat is commonly understood to be412

a primary source for climate predictability (e.g., Latif and Keenlyside 2011). Here, we have used413

a 500-year control simulation from the fully coupled Bergen Climate Model (BCM), aided by ob-414

served sea surface temperatures (HadISST), to assess the propagation and drivers of ocean heat415

anomalies in the northern seas (northern North Atlantic, Nordic Seas, and Arctic Ocean), and to416

what extent these anomalies imprint on the atmosphere.417

Ocean heat anomalies are found to propagate from the eastern subpolar North Atlantic, into,418

and along the rim of the Nordic Seas with a speed of 2 cm s−1 (Fig. 3). The characteristic time419

19

scale of variability is 14 years, which is also that of observed sea surface temperature variability420

in the Norwegian Sea during the last century (Fig. 5). The relative roles of ocean and atmosphere421

in driving ocean heat anomalies are assessed by constructing a depth integrated heat budget for an422

area covering the Atlantic domain of the Nordic Seas, i.e., the Norwegian Sea. Changes in heat423

content are found to be caused mainly by anomalous ocean heat transport convergence (Fig. 6).424

Variations in ocean heat convergence largely originate in the inflow from the Atlantic proper, and425

a temporal decomposition of the Atlantic heat transport shows that volume transport anomalies426

dominate (Fig. 8). Simulated ocean heat anomalies in the northern seas are thus driven mainly by427

changes in the strength of the northward flowing Atlantic water. A similar decadal-scale oscillation428

in the strength of the subpolar gyre (Fig. 9b) further supports the close coupling, observed and429

modeled, between the subpolar North Atlantic and Nordic Seas–Arctic Ocean (e.g., Hatun et al.430

2005; Glessmer et al. 2014; Jungclaus et al. 2014).431

A potentially predictable relation between anomalous ocean heat and climate in the northern432

seas region is furthermore identified. Ocean heat anomalies in the northern seas are reflected in433

regional sea ice extent and found to influence the atmosphere by driving changes in surface air434

temperatures through anomalous air–sea fluxes (Fig. 10). The interaction with the atmosphere435

is also most pronounced on a 14-year time scale. The identified time scale of climate variabil-436

ity, manifested both in anomalous ocean heat transport and its consequent atmospheric response,437

provides encouraging evidence for climate predictability rooted in the northern seas.438

Acknowledgments. This research was supported by the Centre for Climate Dynamics at the439

Bjerknes Centre for Climate Research through the project PRACTICE and the Research Coun-440

cil of Norway projects EPOCASA and NORTH. We thank Odd Helge Ottera for providing the441

20

model data and Helene R. Langehaug for providing the SPG index. We also thank Tore Furevik442

and three anonymous reviewers for valuable comments which improved the manuscript.443

References444

Arthun, M., T. Eldevik, L. H. Smedsrud, Ø. Skagseth, and R. B. Ingvaldsen, 2012: Quantifying445

the Influence of Atlantic Heat on Barents Sea Ice Variability and Retreat. J. Climate, 25 (13),446

4736–4743, doi:10.1175/JCLI-D-11-00466.1.447

Bleck, R., C. Rooth, D. M. Hu, and L. T. Smith, 1992: Salinity-driven thermocline transients in a448

wind-forced and thermohaline-forced isopycnic coordinate model of the North Atlantic. J. Phys.449

Oceanogr., 22 (12), 1486–1505, doi:10.1175/1520-0485(1992)022⟨1486:SDTTIA⟩2.0.CO;2.450

Bourke, R. H., A. M. Weigel, and R. G. Paquette, 1988: The westward turning branch451

of the West Spitsbergen Current. J. Geophys. Res., 93 (C11), 14 065–14 077, doi:10.1029/452

JC093iC11p14065.453

Buckley, M. W., R. M. Ponte, G. Forget, and P. Heimbach, 2014: Low-Frequency SST and Upper-454

Ocean Heat Content Variability in the North Atlantic. J. Climate, 27 (13), 4996–5018, doi:455

10.1175/JCLI-D-13-00316.1.456

Burkholder, K. C., and M. S. Lozier, 2014: Tracing the pathways of the upper limb of the North457

Atlantic Meridional Overturning Circulation. Geophys. Res. Lett., 41 (12), 4254–4260, doi:458

10.1002/2014GL060226.459

Carmack, E., and Coauthors, 2015: Towards quantifying the increasing role of oceanic heat in sea460

ice loss in the new Arctic. Bull. Amer. Meteor. Soc, in press, doi:10.1175/BAMS-D-13-00177.1.461

21

Carton, J. A., G. A. Chepurin, J. Reagan, and S. Hakkinen, 2011: Interannual to decadal variability462

of Atlantic Water in the Nordic and adjacent seas. J. Geophys. Res., 116, C11 035, doi:10.1029/463

2011JC007102.464

Chelton, D. B., 1983: Effects of sampling errors in statistical estimation.Deep-Sea Res. A, 30 (10),465

1083–1103, doi:10.1016/0198-0149(83)90062-6.466

Chepurin, G. A., and J. A. Carton, 2012: Subarctic and Arctic sea surface temperature and467

its relation to ocean heat content 1982–2010. J. Geophys. Res., 117, C06 019, doi:10.1029/468

2011JC007770.469

Czaja, A., and J. Marshall, 2001: Observations of atmosphere-ocean coupling in the North At-470

lantic. Q. J. R. Meteorol. Soc., 127 (576), 1893–1916, doi:10.1002/qj.49712757603.471

Deque, M., C. Dreveton, A. Braun, and D. Cariolle, 1994: The ARPEGE/IFS atmosphere model472

- A contribution to the French community climate modeling. Clim. Dyn., 10 (4-5), 249–266,473

doi:10.1007/BF00208992.474

Deser, C., and M. L. Blackmon, 1993: Surface climate variations over the North Atantic475

Ocean during winter - 1900–1989. J. Climate, 6 (9), 1743–1753, doi:10.1175/1520-0442(1993)476

006⟨1743:SCVOTN⟩2.0.CO;2.477

Dickson, R. R., J. Meincke, S.-A. Malmberg, and A. J. Lee, 1988: The ”great salinity anomaly”478

in the northern North Atlantic 1968–1982. Prog. Oceanogr., 20 (2), 103–151, doi:10.1016/479

0079-6611(88)90049-3.480

Drinkwater, K. F., M. Miles, I. Medhaug, O. H. Ottera, T. Kristiansen, S. Sundby, and Y. Gao,481

2014: The Atlantic Multidecadal Oscillation: Its manifestations and impacts with special em-482

phasis on the Atlantic region north of 60◦N. J. Mar. Sys., 133, 117–130.483

22

Eden, C., and T. Jung, 2001: North Atlantic interdecadal variability: oceanic response to the North484

Atlantic Oscillation (1865-1997). J. Climate, 14 (5), 676–691, doi:10.1175/1520-0442(2001)485

014⟨0676:NAIVOR⟩2.0.CO;2.486

Eldevik, T., J. E. Ø. Nilsen, D. Iovino, K. A. Olsson, A. B. Sandø, and H. Drange, 2009: Observed487

sources and variability of Nordic seas overflow. Nat. Geosci., 2 (6), 405–409, doi:10.1038/488

NGEO518.489

Eldevik, T., and Coauthors, 2014: A brief history of climate–the northern seas from the Last490

Glacial Maximum to global warming. Quat. Sci. Rev, doi:10.1016/j.quascirev.2014.06.028.491

Escudier, R., J. Mignot, and D. Swingedouw, 2013: A 20-year coupled ocean-sea ice-atmosphere492

variability mode in the North Atlantic in an AOGCM. Clim. Dyn., 40 (3-4), 619–636, doi:493

10.1007/s00382-012-1402-4.494

Farneti, R., and G. K. Vallis, 2011: Mechanisms of interdecadal climate variability and the role of495

ocean–atmosphere coupling. Clim. Dyn., 36 (1-2), 289–308, doi:10.1007/s00382-009-0674-9.496

Francis, J. A., and E. Hunter, 2007: Drivers of declining sea ice in the Arctic winter: A tale of two497

seas. Geophys. Res. Lett., 34 (17), L17 503, doi:10.1029/2007GL030995.498

Frankcombe, L. M., A. von der Heydt, and H. A. Dijkstra, 2010: North Atlantic Multidecadal Cli-499

mate Variability: An Investigation of Dominant Time Scales and Processes. J. Climate, 23 (13),500

3626–3638, doi:10.1175/2010JCLI3471.1.501

Furevik, T., 2000: On anomalous sea surface temperatures in the Nordic Seas. J. Climate, 13 (5),502

1044–1053, doi:10.1175/1520-0442(2000)013⟨1044:OASSTI⟩2.0.CO;2.503

23

Furevik, T., 2001: Annual and interannual variability of Atlantic Water temperatures in the504

Norwegian and Barents Seas: 1980-1996. Deep-Sea Res. I, 48 (2), 383–404, doi:10.1016/505

S0967-0637(00)00050-9.506

Furevik, T., M. Bentsen, H. Drange, I. K. T. Kindem, N. G. Kvamstø, and A. Sorteberg, 2003:507

Description and evaluation of the Bergen Climate Model: ARPEGE coupled with MICOM.508

Clim. Dyn., 21 (1), 27–51, doi:10.1007/s00382-003-0317-5.509

Furevik, T., and J. Nilsen, 2005: Large-scale atmospheric circulation variability and its impacts510

on the Nordic Seas ocean climate–A review. The Nordic Seas: an integrated perspective,511

H. Drange, T. Dokken, T. Furevik, R. Gerdes, and W. Berger, Eds., Vol. 158, AGU Geophys.512

Monogr., 105–136.513

Gaspar, P., 1988: Modeling the seasonal cycle of the upper ocean. J. Phys. Oceanogr., 18 (2),514

161–180, doi:10.1175/1520-0485(1988)018⟨0161:MTSCOT⟩2.0.CO;2.515

Glessmer, M. S., T. Eldevik, K. Vage, J. E. Ø. Nilsen, and E. Behrens, 2014: Atlantic origin516

of observed and modelled freshwater anomalies in the Nordic Seas. Nat. Geosci., 7, 801–805,517

doi:10.1038/NGEO2259.518

Gulev, S. K., M. Latif, N. Keenlyside, W. Park, and K. P. Koltermann, 2013: North Atlantic519

Ocean control on surface heat flux on multidecadal timescales. Nature, 499 (7459), 464–467,520

doi:10.1038/nature12268.521

Hansen, D. V., and H. F. Bezdek, 1996: On the nature of decadal anomalies in North Atlantic sea522

surface temperature. J. Geophys. Res., 101 (C4), 8749–8758, doi:10.1029/95JC03841.523

24

Hatun, H., A. B. Sandø, H. Drange, B. Hansen, and H. Valdimarsson, 2005: Influence of the524

Atlantic subpolar gyre on the thermohaline circulation. Science, 309 (5742), 1841–1844, doi:525

10.1126/science.1114777.526

Hatun, H., and Coauthors, 2009: Large bio-geographical shifts in the north-eastern Atlantic Ocean:527

From the subpolar gyre, via plankton, to blue whiting and pilot whales. Prog. Oceanogr., 80 (3),528

149–162, doi:10.1016/j.pocean.2009.03.001.529

Helland-Hansen, B., and F. Nansen, 1909: The Norwegian Sea. Fiskdir. Skr. Ser. Havunders.,530

11 (2), 1–360.531

Holliday, N. P., and Coauthors, 2008: Reversal of the 1960s to 1990s freshening trend in the north-532

east North Atlantic and Nordic Seas. Geophys. Res. Lett., 35 (3), doi:10.1029/2007GL032675.533

Horel, J. D., 1984: Complex principal component analysis: Theory and examples. J. Climate Appl.534

Meteor., 23 (12), 1660–1673, doi:10.1175/1520-0450(1984)023⟨1660:CPCATA⟩2.0.CO;2.535

Hughes, S. L., N. P. Holliday, E. Colbourne, V. Ozhigin, H. Valdimarsson, S. Østerhus, and536

K. Wiltshire, 2009: Comparison of in situ time-series of temperature with gridded sea sur-537

face temperature datasets in the North Atlantic. ICES J. Mar. Sci., 66, 14671479, doi:10.1093/538

icesjms/fsp041.539

Hurrell, J. W., 1995: Decadal trends in the North Atlantic Oscillation: regional temperatures and540

precipitation. Science, 269 (5224), 676–679, doi:10.1126/science.269.5224.676.541

Jungclaus, J. H., K. Lohmann, and D. Zanchettin, 2014: Enhanced 20th-century heat transfer to542

the Arctic simulated in the context of climate variations over the last millennium. Climate of the543

Past, 10 (6), 2201–2213, doi:10.5194/cp-10-2201-2014.544

25

Krahmann, G., M. Visbeck, and G. Reverdin, 2001: Formation and propagation of temperature545

anomalies along the North Atlantic Current. J. Phys. Oceanogr., 31 (5), 1287–1303, doi:10.546

1175/1520-0485(2001)031⟨1287:FAPOTA⟩2.0.CO;2.547

Kushnir, Y., 1994: Interdecadal variations in North Atlantic sea surface temperature and associated548

atmospheric conditions. J. Climate, 7 (1), 141–157, doi:10.1175/1520-0442(1994)007⟨0141:549

IVINAS⟩2.0.CO;2.550

Kushnir, Y., W. Robinson, I. Blade, N. Hall, S. Peng, and R. Sutton, 2002: Atmospheric GCM551

response to extratropical SST anomalies: synthesis and evaluation. J. Climate, 15 (16), 2233–552

2256, doi:10.1175/1520-0442(2002)015⟨2233:AGRTES⟩2.0.CO;2.553

Langehaug, H. R., I. Medhaug, T. Eldevik, and O. H. Ottera, 2012a: Arctic/Atlantic Exchanges554

via the Subpolar Gyre. J. Climate, 25 (7), 2421–2439, doi:10.1175/JCLI-D-11-00085.1.555

Langehaug, H. R., P. B. Rhines, T. Eldevik, J. Mignot, and K. Lohmann, 2012b: Water mass556

transformation and the North Atlantic Current in three multicentury climate model simulations.557

J. Geophys. Res., 117, C11 001, doi:10.1029/2012JC008021.558

Latif, M., and N. S. Keenlyside, 2011: A perspective on decadal climate variability and predictabil-559

ity. Deep Sea Res. II, 58 (1718), 1880 – 1894, doi:10.1016/j.dsr2.2010.10.066.560

Liptak, J., and C. Strong, 2014: The winter atmospheric response to sea ice anomalies in the561

Barents Sea. J. Climate, 27 (2), 914–924, doi:10.1175/JCLI-D-13-00186.1.562

Liu, Z., 2012: Dynamics of interdecadal climate variability: A historical perspective. J. Climate,563

25 (6), 1963–1995, doi:10.1175/2011JCLI3980.1.564

26

Lohmann, K., H. Drange, and M. Bentsen, 2009: Response of the North Atlantic subpolar gyre to565

persistent North Atlantic oscillation like forcing. Clim. Dyn., 32 (2-3), 273–285, doi:10.1007/566

s00382-008-0467-6.567

Lohmann, K., and Coauthors, 2014: The role of subpolar deep water formation and Nordic Seas568

overflows in simulated multidecadal variability of the Atlantic meridional overturning circula-569

tion. Ocean Sci., 10, 227–241, doi:10.5194/osd-10-1895-2013.570

Mauritzen, C., 1996: Production of dense overflow waters feeding the North Atlantic across571

the Greenland-Scotland Ridge .1. Evidence for a revised circulation scheme. Deep-Sea Res.572

I, 43 (6), 769–806, doi:10.1016/0967-0637(96)00037-4.573

Medhaug, I., H. R. Langehaug, T. Eldevik, T. Furevik, and M. Bentsen, 2012: Mechanisms for574

decadal scale variability in a simulated Atlantic meridional overturning circulation. Clim. Dyn.,575

39 (1-2), 77–93, doi:10.1007/s00382-011-1124-z.576

Meehl, G. A., and Coauthors, 2014: Decadal Climate Prediction: An Update from the Trenches.577

Bull. Amer. Meteorol. Soc., 95 (2), 243–267, doi:10.1175/BAMS-D-12-00241.1.578

Mork, K. A., Ø. Skagseth, V. Ivshin, V. Ozhigin, S. L. Hughes, and H. Valdimarsson, 2014:579

Advective and atmospheric forced changes in heat and fresh water content in the Norwegian580

Sea, 1951-2010. Geophys. Res. Lett., doi:10.1002/2014GL061038.581

Nilsen, J. E. Ø., and E. Falck, 2006: Variations of mixed layer properties in the Norwegian Sea for582

the period 1948–1999. Prog. Oceanogr., 70 (1), 58–90, doi:10.1016/j.pocean.2006.03.014.583

Omrani, N.-E., N. S. Keenlyside, J. Bader, and E. Manzini, 2014: Stratosphere key for wintertime584

atmospheric response to warm Atlantic decadal conditions. Clim. Dyn., 42 (3-4), 649–663.585

27

Onarheim, I. H., T. Eldevik, M. Arthun, R. B. Ingvaldsen, and L. H. Smedsrud, 2015: Skillful586

prediction of Barents Sea ice cover. Geophys. Res. Lett., 42 (13), 5364–5371, doi:10.1002/587

2015GL064359.588

Onarheim, I. H., L. H. Smedsrud, R. B. Ingvaldsen, and F. Nilsen, 2014: Loss of sea ice during589

winter north of Svalbard. Tellus A, 66, 23 933, doi:10.3402/tellusa.v66.23933.590

Orvik, K. A., Ø. Skagseth, and M. Mork, 2001: Atlantic inflow to the Nordic Seas: current struc-591

ture and volume fluxes from moored current meters, VM-ADCP and SeaSoar-CTD observa-592

tions, 1995-1999. Deep-Sea Res. I, 48 (4), 937–957, doi:10.1016/S0967-0637(00)00038-8.593

Ottera, O. H., M. Bentsen, I. Bethke, and N. G. Kvamstø, 2009: Simulated pre-industrial climate594

in Bergen Climate Model (version 2): model description and large-scale circulation features.595

Geosci. Model Dev., 2 (2), 197–212, doi:10.5194/gmd-2-197-2009.596

Ottera, O. H., M. Bentsen, H. Drange, and L. Suo, 2010: External forcing as a metronome for597

Atlantic multidecadal variability. Nat. Geosci., 3 (10), 688–694, doi:10.1038/ngeo955.598

Polyakov, I. V., and Coauthors, 2005: One more step toward a warmer Arctic. Geophys. Res. Lett.,599

32 (17), doi:10.1029/2005GL023740.600

Rayner, N. A., D. E. Parker, E. B. Horton, C. K. Folland, L. V. Alexander, D. P. Rowell, E. C.601

Kent, and A. Kaplan, 2003: Global analyses of sea surface temperature, sea ice, and night602

marine air temperature since the late nineteenth century. J. Geophys. Res., 108 (D14), 4407,603

doi:10.1029/2002JD002670.604

Rhines, P., S. Hakkinen, and S. Josey, 2008: Is Oceanic Heat Transport Significant in the Cli-605

mate System? Arctic–Subarctic Ocean Fluxes, R. R. Dickson, J. Meincke, and P. Rhines, Eds.,606

Springer Netherlands, 87–109.607

28

Rudels, B., H. J. Friedrich, and D. Quadfasel, 1999: The Arctic circumpolar boundary current.608

Deep-Sea Res. II, 46 (6-7), 1023–1062, doi:10.1016/S0967-0645(99)00015-6.609

Salas-Melia, D., 2002: A global coupled sea ice-ocean model. Ocean Mod., 4 (2), 137–172, doi:610

10.1016/S1463-5003(01)00015-4.611

Sandø, A., J. Nilsen, T. Eldevik, and M. Bentsen, 2012: Mechanisms for variable North Atlantic–612

Nordic seas exchanges. J. Geophys. Res., 117 (C12), doi:10.1029/2012JC008177.613

Schlichtholz, P., 2013: Observational Evidence for Oceanic Forcing of Atmospheric Variability in614

the Nordic Seas Area. J. Climate, 26 (9), 2957–2975, doi:10.1175/JCLI-D-11-00594.1.615

Screen, J. A., C. Deser, I. Simmonds, and R. Tomas, 2013: Atmospheric impacts of Arctic sea-ice616

loss, 1979–2009: separating forced change from atmospheric internal variability. Clim. Dyn.,617

1–12, doi:10.1007/s00382-013-1830-9.618

Semenov, V. A., M. Latif, D. Dommenget, N. S. Keenlyside, A. Strehz, T. Martin, and W. Park,619

2010: The impact of North Atlantic-Arctic multidecadal variability on Northern Hemisphere620

surface air temperature. J. Climate, 23 (21), 5668–5677, doi:10.1175/2010JCLI3347.1.621

Sheldon, L., and A. Czaja, 2014: Seasonal and interannual variability of an index of deep atmo-622

spheric convection over western boundary currents. Q. J. R. Meteorol. Soc., 140 (678), 22–30,623

doi:10.1002/qj.2103.624

Skagseth, Ø., and K. A. Mork, 2012: Heat content in the Norwegian Sea, 1995-2010. ICES J. Mar.625

Sci., 69 (5), 826–832, doi:10.1093/icesjms/fss026.626

Smedsrud, L. H., and Coauthors, 2013: The role of the Barents Sea in the Arctic climate system.627

Rev. Geophys., 51 (3), 415–449, doi:10.1002/rog.20017.628

29

Smirnov, D., M. Newman, M. A. Alexander, Y.-O. Kwon, and C. Frankignoul, 2015: Investigating629

the local atmospheric response to a realistic shift in the Oyashio sea surface temperature front.630

J. Climate, 28 (3), 1126–1147, doi:10.1175/JCLI-D-14-00285.1.631

Sutton, R. T., and M. R. Allen, 1997: Decadal predictability of North Atlantic sea surface temper-632

ature and climate. Nature, 388 (6642), 563–567, doi:10.1038/41523.633

Sutton, R. T., and D. L. R. Hodson, 2005: Atlantic Ocean forcing of North American and European634

summer climate. Science, 309 (5731), 115–118, doi:10.1126/science.1109496.635

Terray, L., and O. Thual, 1995: Oasis: le couplage ocean-atmosphere. LaMeteorologie, 10, 50–61.636

van der Swaluw, E., S. Drijfhout, and W. Hazeleger, 2007: Bjerknes compensation at high637

northern latitudes: the ocean forcing the atmosphere. J. Climate, 20 (24), 6023–6032, doi:638

10.1175/2007JCLI1562.1.639

Venegas, S. A., and L. A. Mysak, 2000: Is there a dominant timescale of natural climate vari-640

ability in the Arctic? J. Climate, 13 (19), 3412–3434, doi:10.1175/1520-0442(2000)013⟨3412:641

ITADTO⟩2.0.CO;2.642

Vihma, T., 2014: Effects of Arctic Sea Ice Decline on Weather and Climate: A Review. Surv.643

Geophys., 1–40, doi:10.1007/s10712-014-9284-0.644

Vinje, T., 2001: Anomalies and trends of sea-ice extent and atmospheric circulation in the Nordic645

Seas during the period 1864–1998. J. Climate, 14 (3), 255–267, doi:10.1175/1520-0442(2001)646

014⟨0255:AATOSI⟩2.0.CO;2.647

Visbeck, M., H. Cullen, G. Krahmann, and N. Naik, 1998: An ocean model’s response to North648

Atlantic Oscillation-like wind forcing. Geophys. Res. Lett., 25 (24), 4521–4524, doi:10.1029/649

1998GL900162.650

30

Williams, R. G., V. Roussenov, D. Smith, and M. S. Lozier, 2014: Decadal Evolution of Ocean651

Thermal Anomalies in the North Atlantic: The Effects of Ekman, Overturning, and Horizontal652

Transport. J. Climate, 27, 698–719, doi:10.1175/JCLI-D-12-00234.1.653

Yashayaev, I., and D. Seidov, 2015: The role of the Atlantic Water in multidecadal ocean variabil-654

ity in the Nordic and Barents Seas. Prog. Oceanogr., 132, 68–127, doi:10.1016/j.pocean.2014.655

11.009.656

31

LIST OF FIGURES657

Fig. 1. a) Observed and b) modeled winter (December–April) upper-ocean temperature (shading)658

and sea ice extent (white line; defined where the sea ice concentration is 15%) in the northern659

seas. Observations are from HadISST (Rayner et al. 2003). The arrows indicate the main660

features of the near-surface circulation (NwAC: Norwegian Atlantic Current; WSC: West661

Spitsbergen Current; EGC: East Greenland Current). Modeled temperature and velocities662

are averages within the surface mixed layer (note the different velocity scales). Isobaths are663

given every 1000 m (thin gray lines) and for 500 m depth (black line) which roughly marks664

the continental slopes. DS: Denmark Strait; RT: Rockall Trough; Sh: Shetland. . . . . . 34665

Fig. 2. Barotropic streamlines showing the mean cyclonic circulation within the northern seas (thick666

black lines; plotted for -1.5 Sv, -2 Sv, and -3 Sv; 1 Sv≡ 106 m3 s−1). Isobaths are given for667

500 m, 2000 m, and 3000 m depth (gray lines). The colored circles show the grid points668

used to define the path of ocean heat anomalies, where the associated color indicates the669

maximum lagged correlation between adjacent stations, i.e., the correlation shown for St2670

corresponds to r(St1,St2). Station numbers are assigned and used in the text. . . . . . . 35671

Fig. 3. a) Temporal development of low-pass filtered heat content (shading; units in 1019 J) and672

depth-averaged salinity anomalies (contours; plotted for 0 and ±0.05, dashed lines corre-673

sponding to a negative salinity anomaly) at defined stations (averages over grid points shown674

in Fig. 2). The thick dashed line corresponds to a propagation speed of 2 cm s−1. b) The675

dominant mode of propagation derived from complex principal component analysis (units676

in 0.5×std. deviation), explaining 50% of the total variance. c,d) Phase angle as a function677

of station number and time, respectively. Increasing phase angle with increasing distance678

(station) corresponds to a cyclonic propagation. The dashed line in c) corresponds to a679

propagation speed of 2 cm s−1, whereas in d) it corresponds to constant phase propagation. . . 36680

Fig. 4. Heat content (shading) and temperature (contours) variability (expressed as one standard681

deviation) at defined stations as a function of depth. The black line displays the average682

depth of the winter (December–April) mixed layer. . . . . . . . . . . . . 37683

Fig. 5. Power spectrum (thick lines) for a) heat content and b) salinity anomalies at St1 and St3684

based on unfiltered data together with the theoretical red noise spectrum (thin solid lines)685

computed by fitting a first order autoregressive process with a 95% confidence interval686

(thin dashed lines) around the red noise. c) Power spectrum for observed (HadISST) win-687

ter (December–April) sea surface temperatures in the Norwegian Sea (0.5–17.5◦E, 60.5–688

71.5◦N) between 1870 and 2013. Winter SST is considered because it represents the upper-689

ocean heat content as a result of a deep winter mixed layer. . . . . . . . . . . 38690

Fig. 6. a) Heat budget for the Norwegian Sea (low-pass filtered), where HCt is the heat content691

tendency, Qadv is the heat transport convergence, Qs is the net surface heat flux (positive out692

of the ocean), and Qres is a residual term. The domain (upper map) is bounded in the south693

by the Iceland–Scotland ridge and in the west and north by the extent of the Atlantic domain,694

visible by the gradient in heat content variability (expressed as one standard deviation; units695

in 108 Jm−2). The lower map displays the grid points used in the analysis. Colored circles696

and black dots present the ocean grid, while the red open squares are the atmospheric grid697

points used to analyze surface air temperature. b) Heat transport anomalies (HT ; low-pass698

filtered) through the vertical sections of the control volume. The line colors correspond to699

the section color given in the lower map and the names refer to geographic location or major700

currents. HTaw: Atlantic water; HTns: North Sea; HTgs: Greenland Sea; HTbs: Barents Sea;701

HTwsc: West Spitsbergen Current. . . . . . . . . . . . . . . . . . 39702

32

Fig. 7. Relative magnitude of oceanic (Qo = Qadv+Qres) and atmospheric (Qs) heat budget terms,703

calculated as std(Qo)/std(Qs) and presented on a log scale. Values >1 indicate that the704

oceanic contribution is the largest. . . . . . . . . . . . . . . . . . 40705

Fig. 8. Heat transport anomalies (low-pass filtered) between Iceland and Scotland (Fig. 6; HTaw)706

decomposed into a temperature component (vT ′), a velocity component (v′T ), and an eddy707

component (v′T ′). . . . . . . . . . . . . . . . . . . . . . 41708

Fig. 9. Power spectrum for a) surface heat flux (Qs) and surface air temperature (SAT) in the Nor-709

wegian Sea (cf. Fig. 6), and b) the heat transport between Iceland and Scotland (HTaw) and710

the strength of the subpolar gyre (SPG) based on unfiltered data. The red noise spectrums711

and 95% confidence intervals are also plotted (see Fig. 5 for description). The SPG strength712

is calculated as the absolute value of the minimum barotropic streamfunction in the subpolar713

region (Langehaug et al. 2012a). . . . . . . . . . . . . . . . . . 42714

Fig. 10. The influence of ocean heat transport into the Nordic Seas (HTaw; black circles) on a) surface715

heat flux to the atmosphere, b) sea ice concentration loss, and c) surface air temperature,716

based on linear regression analysis on low-pass filtered time series. The ocean heat transport717

leads by 2 years. Only regressions significant at the 95% confidence level are plotted. Units718

are given in Wm−2, %, and ◦C, respectively, per std(HTaw). . . . . . . . . . . 43719

33

24oW 12oW 0o 12oE 24oE

55oN

60oN

65oN

70oN

75oN

80oN

10 cm/s5 cm/s

b)

[°C]

2 0 2 4 6 8 10

24oW 12oW 0o 12oE 24oE

55oN

60oN

65oN

70oN

75oN

80oN a)

[°C]

2 0 2 4 6 8 10

BarentsSea

Norway

Iceland

Gre

enla

nd

Svalbard

NwAC

EGC

NorthAtlantic

GreenlandSea

NorwegianSea

WSC

DS

Fram Strait

Faroe Is

Scotland

Sh

RT

FIG. 1. a) Observed and b) modeled winter (December–April) upper-ocean temperature (shading) and sea ice

extent (white line; defined where the sea ice concentration is 15%) in the northern seas. Observations are from

HadISST (Rayner et al. 2003). The arrows indicate the main features of the near-surface circulation (NwAC:

Norwegian Atlantic Current; WSC: West Spitsbergen Current; EGC: East Greenland Current). Modeled tem-

perature and velocities are averages within the surface mixed layer (note the different velocity scales). Isobaths

are given every 1000 m (thin gray lines) and for 500 m depth (black line) which roughly marks the continental

slopes. DS: Denmark Strait; RT: Rockall Trough; Sh: Shetland.

720

721

722

723

724

725

726

34

20oW 10oW 0o 10oE 20oE 30oE

55oN

60oN

65oN

70oN

75oN

80oN

−3

−1.5

−2

−1.5St1

St2 St3

St4

St5

St6

St7

St8

St9

St10

St11

0 0.2 0.4 0.6 0.8 1

FIG. 2. Barotropic streamlines showing the mean cyclonic circulation within the northern seas (thick black

lines; plotted for -1.5 Sv, -2 Sv, and -3 Sv; 1 Sv≡ 106 m3 s−1). Isobaths are given for 500 m, 2000 m, and

3000 m depth (gray lines). The colored circles show the grid points used to define the path of ocean heat

anomalies, where the associated color indicates the maximum lagged correlation between adjacent stations, i.e.,

the correlation shown for St2 corresponds to r(St1,St2). Station numbers are assigned and used in the text.

727

728

729

730

731

35

100 200 300 400 500

Dis

tanc

e [k

m]

10 20 30 40 50 60 70 80 900

1000

2000

3000

4000

12

34567891011

#Station

−4

−3

−2

−1

0

1

2

3

4

Time [yr]

Dis

tanc

e [k

m]

10 20 30 40 50 60 70 80 900

1000

2000

3000

4000

12

34567891011

#Station

100 200 300 400 500

0 0.1 0.2 0.3 0.4 0.5 0.60

1000

2000

3000

4000

5000

Phase angle [360°]

Dis

tanc

e [k

m]

c)

12

34567891011

#Station

0 100 200 300 400 500−30

−25

−20

−15

−10

−5

0

Time [yr]

Pha

se a

ngle

[360

°]

d)

a)

b)

FIG. 3. a) Temporal development of low-pass filtered heat content (shading; units in 1019 J) and depth-

averaged salinity anomalies (contours; plotted for 0 and±0.05, dashed lines corresponding to a negative salinity

anomaly) at defined stations (averages over grid points shown in Fig. 2). The thick dashed line corresponds

to a propagation speed of 2 cm s−1. b) The dominant mode of propagation derived from complex principal

component analysis (units in 0.5×std. deviation), explaining 50% of the total variance. c,d) Phase angle as a

function of station number and time, respectively. Increasing phase angle with increasing distance (station)

corresponds to a cyclonic propagation. The dashed line in c) corresponds to a propagation speed of 2 cm s−1,

whereas in d) it corresponds to constant phase propagation.

732

733

734

735

736

737

738

739

36

#Station

Dep

th [m

]

0.2

0.2

0.2

0.4

0.4

0.4

0.6

0.6

0.6

0.8

0.8

0.8

1

1

1

1 2 3 4 5 6 7 8 9 10 11

100

200

300

400

500

600

700

800

900

1000

[109 J m−2]

0 0.5 1 1.5 2 2.5 3 3.5 4

FIG. 4. Heat content (shading) and temperature (contours) variability (expressed as one standard deviation) at

defined stations as a function of depth. The black line displays the average depth of the winter (December–April)

mixed layer.

740

741

742

37

5102030500

5

10

15

20

25

Period [yr]

|Y(f)

|

St3St1

a)

5102030500

5

10

15

20

25

Period [yr]

|Y(f)

|

St3St1

b)

5102030500

1

2

3

4

5

6

7

8

9

10

Period [yr]

|Y(f)

|

HadISST

c)

FIG. 5. Power spectrum (thick lines) for a) heat content and b) salinity anomalies at St1 and St3 based on

unfiltered data together with the theoretical red noise spectrum (thin solid lines) computed by fitting a first

order autoregressive process with a 95% confidence interval (thin dashed lines) around the red noise. c) Power

spectrum for observed (HadISST) winter (December–April) sea surface temperatures in the Norwegian Sea (0.5–

17.5◦E, 60.5–71.5◦N) between 1870 and 2013. Winter SST is considered because it represents the upper-ocean

heat content as a result of a deep winter mixed layer.

743

744

745

746

747

748

38

−80

−60

−40

−20

0

20

40

60

80

Hea

t ano

mal

y [T

W]

a) HCt = Qadv − Qs + Qres

10 20 30 40 50 60 70 80 90 100−100

−80

−60

−40

−20

0

20

40

60

80

100

Time [yrs]

Hea

t ano

mal

y [T

W]

b) Qadv = HTaw + HTns + HTgs + HTbs + HTwsc

0 1 2 3 4 5

FIG. 6. a) Heat budget for the Norwegian Sea (low-pass filtered), where HCt is the heat content tendency,

Qadv is the heat transport convergence, Qs is the net surface heat flux (positive out of the ocean), and Qres is a

residual term. The domain (upper map) is bounded in the south by the Iceland–Scotland ridge and in the west

and north by the extent of the Atlantic domain, visible by the gradient in heat content variability (expressed

as one standard deviation; units in 108 Jm−2). The lower map displays the grid points used in the analysis.

Colored circles and black dots present the ocean grid, while the red open squares are the atmospheric grid points

used to analyze surface air temperature. b) Heat transport anomalies (HT ; low-pass filtered) through the vertical

sections of the control volume. The line colors correspond to the section color given in the lower map and the

names refer to geographic location or major currents. HTaw: Atlantic water; HTns: North Sea; HTgs: Greenland

Sea; HTbs: Barents Sea; HTwsc: West Spitsbergen Current.

749

750

751

752

753

754

755

756

757

758

39

24oW 12oW 0o 12oE 24oE

55oN

60oN

65oN

70oN

75oN

80oN

0.5

1

2

4

8

FIG. 7. Relative magnitude of oceanic (Qo =Qadv+Qres) and atmospheric (Qs) heat budget terms, calculated

as std(Qo)/std(Qs) and presented on a log scale. Values >1 indicate that the oceanic contribution is the largest.

759

760

40

0 50 100 150 200 250 300 350 400 450 500−100

−50

0

50

100

Time [yrs]

Hea

t ano

mal

y [T

W]

HTaw = vT′ + v

′T + v

′T

′

FIG. 8. Heat transport anomalies (low-pass filtered) between Iceland and Scotland (Fig. 6; HTaw) decomposed

into a temperature component (vT ′), a velocity component (v′T ), and an eddy component (v′T ′).

761

762

41

5102030500

5

10

15

Period [yr]

|Y(f)

|

SAT

Qs

a)

5102030500

5

10

15

Period [yr]

|Y(f)

|SPG

HTaw

b)

FIG. 9. Power spectrum for a) surface heat flux (Qs) and surface air temperature (SAT) in the Norwegian

Sea (cf. Fig. 6), and b) the heat transport between Iceland and Scotland (HTaw) and the strength of the subpolar

gyre (SPG) based on unfiltered data. The red noise spectrums and 95% confidence intervals are also plotted

(see Fig. 5 for description). The SPG strength is calculated as the absolute value of the minimum barotropic

streamfunction in the subpolar region (Langehaug et al. 2012a).

763

764

765

766

767

42

a) 20oW 0o 20oE 40oE 60oE

55oN

60oN

65oN

70oN

75oN

80oN

0

2

4

6

8

10

12

14

16

b)

55oN

60oN

65oN

70oN

75oN

80oN

0

1

2

3

4

5

6

7

8

55oN

60oN

65oN

70oN

75oN

80oN c)

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

FIG. 10. The influence of ocean heat transport into the Nordic Seas (HTaw; black circles) on a) surface heat

flux to the atmosphere, b) sea ice concentration loss, and c) surface air temperature, based on linear regression

analysis on low-pass filtered time series. The ocean heat transport leads by 2 years. Only regressions significant

at the 95% confidence level are plotted. Units are given in Wm−2, %, and ◦C, respectively, per std(HTaw).

768

769

770

771

43