ISSN 1068-3739, Russian Meteorology and Hydrology, 2016, Vol. 41, No. 8, pp. 544–558. Ó Allerton Press, Inc., 2016. Original Russian Text Ó G.V. Alekseev, N.I. Glok, A.V. Smirnov, A.E. Vyazilova, 2016, published in Meteorologiya i Gidrologiya, 2016, No. 8, pp. 38–56.

The Influence of the North Atlantic on Climate Variations in the Barents Sea and Their Predictability G. V. Alekseev, N. I. Glok, A. V. Smirnov, and A. E. Vyazilova Arctic and Antarctic Research Institute, ul. Beringa 38, St. Petersburg, 199397 Russia, e-mail: [email protected] Received July 20, 2015 Abstract—The effects of Atlantic water inflow on the climate variability in the Barents Sea are studied. Initial data are the series of water temperature at the Kola meridian cross-section, monthly values of ice extent, air temperature at the stations, sea level pressure from the reanalysis data, and sea surface temperature. The methods of multivariate correlation, spectral, and factor analysis and EOF decomposition are used. It was found that variations in the Atlantic water inflow define the main part of interannual variability of sea ice extent, water temperature, and air temperature in the Barents Sea in the cold season. The influence of regional atmospheric circulation on the interannual variability of these parameters is small. The effects that water temperature anomalies in the area of Newfoundland and in the equatorial part of the North Atlantic have on climate parameters in the Barents Sea are discovered. The response of these parameters lags behind the respective anomalies by 9–58 months. The high correlation between them makes it possible to develop the method of statistical forecasting of sea ice extent and water temperature in the Barents Sea with the lead time up to 4 years.

DOI: 10.3103/S1068373916080045 Keywords: Climate, Barents Sea, sea ice, North Atlantic

INTRODUCTION The Atlantic Arctic including Svalbard, the Barents and Kara seas, and the adjoining part of the Arctic basin is under the warming influence of the North Atlantic. The warm salt water flows through the Faroe-Shetland Channel into the Norwegian and Greenland seas and further to the Arctic basin and the Barents Sea. Warm and moist air moves over this water surface to the east and northeast to the Arctic seas and Arctic basin. The water with positive sea surface temperature in the Barents Sea is a powerful “heater” of the Barents Sea region in the cold season. However, its strength varies due to variations in the atmospheric and oceanic circulation in the North Atlantic which provide heat transfer from the low to high latitudes. As a result of multiyear oceanographic observations in the Barents Sea that started in the late 19th century, variations in its thermohaline conditions are well documented and their scales are determined. Especially valuable data on variations in the Atlantic water inflow are found in the observations at the Kola meridian cross-section (33°30¢ E) which started in the 1900s and have continued till now. Based on the routine oceanographic observations at the cross-section along the Kola meridian, the specialists of Knipovich Polar Research Institute of Marine Fisheries and Oceanography (PINRO) created and maintain the unique datasets of average temperature and salinity in different water layers and in the different parts of the cross-section [6, 11]. Using the long-term data for 1900–2006, paper [26] considered the variability of average water temperature in the layer of 100–150 m in the zone of 69°–78° N and 11°–57° E for every month and revealed multidecadal oscillation with the amplitude up to 4°C corresponding to the Atlantic multidecadal oscillation (AMO). Earlier the authors of [38] discovered the oscillation periods of 74, 18.6, and 9.3 years in the 100-year series of water temperature along the Kola meridian and found the two-year lag between the temperature variations in the Faroe-Shetland Channel and along the Kola meridian.

544

THE INFLUENCE OF THE NORTH ATLANTIC ON CLIMATE VARIATIONS

545

In the recent three years the series of papers was published which studied the variability of climate parameters in the Barents Sea and their causes. Special attention is given to the variability of sea ice parameters caused both by the atmosphere and ocean. Each factor and its role as a potential root cause is discussed. Paper [33] revealed the relationship between the Atlantic water temperature in the Barents Sea and atmospheric circulation. At the same time, the effects that heat fluxes to the atmosphere make on sea surface temperature (SST) anomalies in the Norwegian Sea were noted. Later the authors of [32] used the regression analysis of the relationship between summer SST anomalies at the Barents Sea entrance and subsequent wintertime atmospheric fields in 1982–2006 and made the conclusion that the ocean is of key importance for the formation of variability of meteorological parameters in the troposphere over the Nordic seas in winter. The opposite conclusion was made by the authors of [23] who discovered that the anomalies of heat fluxes in winter are by four times more important for the variability of sea ice edge than SST anomalies transported to the Greenland and Barents seas. The increase in the localization of cyclone centers over the Barents Sea in 1991–2007 was noted in [10] that, in the authors’ opinion, favors the reduction of sea ice extent. The attempts to find stable relations between the large-scale modes of atmospheric circulation such as the North Atlantic Oscillation (NAO) or Arctic Oscillation (AO) and the climate parameters in the Barents Sea, were not successful [34, 35]. Attempts have been made to find regional modes and circulation indices [19, 21] related to variability in the Barents Sea. Some papers investigate inverse effects that sea ice variations in the Barents Sea make on the atmospheric circulation over the sea and beyond it [18, 22, 27, 29, 34]. It is noted that when the Barents Sea is ice-free at the beginning of winter, cyclonicity prevails in the atmosphere which is accompanied by the western displacement of the ridge of the Siberian high with cold air inflows along its western periphery. However, the contribution of local conditions to the formation of cyclonicity over the Barents Sea is estimated in different ways. The authors of [18, 29, 34] consider that the displacement of sea ice boundaries considerably contribute to the formation of atmospheric circulation over the sea. This point of view is partly approved by the authors of review [35] who concluded that if sea ice extent increases, the ocean effects on sea ice variability prevail, but the contribution of variations in heat fluxes “controlled” by the atmosphere increases as sea ice concentration decreases. They suggest that the heat flux from the Barents Sea to the atmosphere perturbs the large-scale circulation and is able to cause the cooling over Eurasia in winter; it is also of great importance for the warming intensification in the Arctic and even for the formation of the Northern Hemisphere climate in the recent 2500 years. At the same time, the authors of [35] conclude that the warming in the 1930s was closely related to the increase in the Atlantic water inflow to the Barents Sea. The validity of this statement was proved as early as in the 1930s by V.Yu. Vize [7] who studied the warming exactly during the period of its development and revealed its relation to the global warming in the hemisphere. A close relation between the Atlantic water inflow and sea ice extent at the end of winter in the Greenland and Barents seas was noted in [8, 9]. The recent experiments with atmospheric general circulation models (AGCM) and with global climate models have not corroborated the significant influence of the Barents Sea on the climate at the high and middle latitudes. The experiments with AGCM [5, 13] investigated the response of the atmosphere at the high and middle latitudes to the variations in sea ice extent in the Arctic and in SST in the Northern Hemisphere. The model revealed no relationship between sea ice extent reduction and cold winters in Europe (in the authors’ opinion, they are more likely related to the variations in atmospheric circulation under the influence of SST rise and increase in meridional transfer of heat). The results of experiments with global coupled climate models aimed at the study of sensitivity to external forcing are more credible than experiments with AGCM because they do not artificially break variations in SST and sea ice. The experiments with the Norwegian global climate model aimed at the assessment of the effects of ocean heat transport on the Arctic sea ice [31] revealed that the increase in the Atlantic water inflow to the Barents Sea exerts considerable influence on sea ice extent as a result of the decrease in ice formation. Also, they established that the ocean affects ice mass variations more significantly than the atmosphere, both the average value and the variability. The complexity of the processes forming the climate variability in the Barents Sea is manifested in significant differences between the simulated and observed climate that are revealed from the comparison of the results of global coupled atmosphere–ocean models. In [20] the maximum difference was found between the simulated and observed climate over the Barents Sea for the period of 1981–2000: the simulated air temperature turned out to be by 6–8°C lower, and the sea ice extent was much larger than the obRUSSIAN METEOROLOGY AND HYDROLOGY

Vol. 41

No. 8

2016

546

ALEKSEEV et al.

served value. The new generation of global models CMIP5 better simulated the observed climate in the Barents Sea but the discrepancies between the model and observations still remain ones of the most significant in the Arctic [16, 34]. A problem of simulation of the observed climate with global models causes the problems of construction of climate change projections in the Barents Sea region. The alternative approach to the regional forecasts based on statistical models and observational data [14] also requires the understanding of basic mechanisms and influencing factors. First of all, the role of the effects of the ocean and atmosphere and their interrelation should be revealed. The present paper deals with solving these problems and with searching for predictors to construct the climatic forecast of variations in water temperature and sea ice in the Barents Sea based on the data of observations and reanalysis. DATA AND METHODS The initial data for the study are the long-term series of water temperature measured at the cross-section along the Kola meridian; the series are represented by average annual values for 1900–2013 and by monthly mean values for 1951–2013. The series were prepared in PINRO [6, 11] and are available at http://www/pinro.ru/n22/index/phpstructure/labs/labhidro/. The present study uses the average water temperature in the layer of 0–200 m at stations 3–7 at the cross-section which are located in the mainstream of the Atlantic water. Besides, the data were used on sea surface temperature in the Atlantic Ocean and Norwegian, Greenland, and Barents seas from HadISST dataset [30] with the spatial resolution of 1° ´ 1° for the period of 1951–2013. To assess the variability of atmospheric parameters over the Barents Sea, the monthly mean data on surface air temperature and sea level pressure from NCEP/NCAR reanalysis are used (http://www.esrl.noaa.gov/psd) [24] as well as the monthly mean data on surface air temperature from nine stations in the Barents Sea region from the meteorological dataset of Arctic and Antarctic Research Institute (AARI) [1]. The data on monthly mean sea ice extent in the Barents Sea were collected in AARI for 1928–2013 [15] and are available at http://wdc.aari.ru/datasets. The research methods included the cross-correlation and cross-spectral analysis of time series, the multivariate correlation and factor analysis, and the empirical orthogonal function decomposition of sea level pressure fields. The cross-spectral analysis discriminated between the climatic signal and the weather noise in the long-term series of climate parameters that led to the significant increase incorrelation between climate variables. The research had two stages. At the first stage the effects of the Atlantic water inflow on the variability of climate parameters in the Barents Sea and their predictability were analyzed and assessed. At the second stage the search and estimation of teleconnections between the climate parameters in the Barents Sea and the variability of sea surface temperature in the North Atlantic as well as of their predictability were carried out. THE EFFECTS OF THE ATLANTIC WATER INFLOW ON SEA ICE AND AIR TEMPERATURE IN THE BARENTS SEA The clear evidence of the effects of the Atlantic water inflow on sea ice extent in the Atlantic Arctic is the composite maps of salinity distribution on the surface of the Barents, Norwegian, and Greenland seas and the sea ice edge position (Fig. 1). It is clear that in the years with the increased presence of saltier Atlantic water, sea ice extent decreases and, on the contrary, ice occupies the larger area if the area of the Atlantic water inflow is reduced. June was selected because this month is characterized by the maximum volume of the data of oceanographic observations and, at the same time, the summer melting of sea ice begins in June. To obtain the quantitative estimate of the effects that the Atlantic water inflow makes on sea ice extent variations in the Barents Sea, the long-term series of average annual water temperature TKM in the layer of 0–200 m at the Kola meridian cross-section are used as an indicator of the Atlantic water inflow to the Barents Sea. Table 1 presents the results of such estimation based on the correlation analysis. The table data demonstrate that the highest correlation between variations in water temperature and sea ice extent is registered in May when sea ice extent stops increasing and the melting starts. The effects of variations in the Atlantic water inflow on the interannual displacement of the sea ice edge were assessed by computing the correlation between the principal component of the set of the ice edge positions and water temperature at the cross-section in May in 1951–2012. The sea ice edge position RUSSIAN METEOROLOGY AND HYDROLOGY

Vol. 41

No. 8

2016

THE INFLUENCE OF THE NORTH ATLANTIC ON CLIMATE VARIATIONS

547

Fig. 1. Surface water salinity in the Barents, Norwegian, and Greenland seas from the Atlas data [25] and the sea ice edge position from the HadISST data [30] in June in the cases of (a, c, e) large and (b, d, f) small sea ice extent in the Barents Sea. (a) 1969; (b) 1976; (c) 1972; (d) 1987; (e) 1979; (f) 1990.

Table 1. The coefficients of correlation between the series of monthly mean Tm and average annual Ty values of water temperature in the layer of 0–200 m along the Kola meridian and monthly mean values of sea ice extent and surface air temperature in the Barents Sea in 1951–2013 TKM, °C

January

February

March

April

May

June

August

September

October

November

December

–0.53 –0.57

–0.47 –0.51

–0.38 –0.45

–0.46 –0.41

–0.59 –0.60

0.49 0.55

0.44 0.41

0.35 0.20

0.51 0.35

July

Sea ice extent, 103 km2 Tm Ty

–0.72 –0.59

–0.74 –0.69

–0.70 –0.65

–0.77 –0.77

–0.88 –0.85

–0.80 –0.79

–0.65 –0.69

Surface air temperature, °C Tm Ty

0.69 0.60

0.58 0.61

0.37 0.43

0.60 0.62

0.58 0.57

0.58 0.65

0.60 0.58

0.59 0.54

Note: The 95% significance level of correlation coefficients is 0.25.

was identified from the information of the HadISST dataset on the 15% ice concentration in the zone from 20° to 60° E and was specified in the degrees of latitude counted from the North Pole. The maximum propagation of the sea ice edge to the south was limited by 70° N. The principal component represents 60% of the interannual variability of the sea ice edge position, and its correlation with average annual water temperature and water temperature in May is –0.65 and –0.70, respectively. The variations in the sea ice edge position and sea ice extent in May are characterized by the correlation coefficient equal to 0.78. The atmosphere over the Barents Sea exerts the warming influence on the Atlantic water inflow, and air temperature varies from year to year as a result of its variations. However, the correlation between water temperature along the Kola meridian cross-section and average temperature over the Barents Sea is lower due to the effects of atmospheric circulation and summertime heating of water by solar radiation (Table 1). RUSSIAN METEOROLOGY AND HYDROLOGY

Vol. 41

No. 8

2016

548

ALEKSEEV et al.

Fig. 2. The first vector of EOF decomposition of the field of monthly mean pressure in (a) March and (b) August over the Barents Sea. The dots are the weather stations, the triangles are the points between which the pressure differences were computed, the crosses are the grid points where the values of pressure were taken.

For example, the coefficients of correlation r between average seasonal values of TKM and surface air temperature are the following: Season r

Winter 0.75

Spring 0.69

Summer 0.69

Autumn 0.42

Year 0.77

Average air temperature over the Barents Sea was determined from the data of weather stations whose location is presented in Fig. 2. It follows from the above data that the correlation with water temperature weakens in autumn and is much higher in winter and for the year as a whole. In summer from June to August high (relative to autumn) correlation is also kept as a result of the synchronous heating of water and air by solar radiation. Variations in average air temperature over the Barents Sea are also connected with variations in sea ice extent. The correlation between average air temperature in winter and sea ice extent in March is –0.56 and between temperature in summer and sea ice extent in September is –0.57 for 1934–2013. As air temperature in winter is associated with water temperature, the correlation between sea ice extent and air temperature in winter is caused to a significant degree by the effects of variations in water temperature which, in turn, defines interannual variations in sea ice extent (see Table 1). Air temperature in summer is an indicator of the joint effects of heat coming with the Atlantic water and heat from solar radiation. Using summer air temperature as an indicator of the integrated effects of many factors on sea ice extent in September is efficient as demonstrated in [2, 4] where the correlation between variations in summertime air temperature and sea ice extent in the Arctic in September was –0.93 for 1980–2014. THE EFFECTS OF ATMOSPHERIC CIRCULATION ON WATER TEMPERATURE, AIR TEMPERATURE, AND SEA ICE IN THE BARENTS SEA The introduction noted the debatable nature of the problem concerning the degree of the influence that the Atlantic water inflow, atmospheric circulation, and their interaction produce on the interannual variability of water temperature, air temperature, and sea ice extent in the Barents Sea. Let us examine this problem based on the statistical analysis of the time series of the difference in sea level pressure and principal components of pressure field as the indicators of variability of atmospheric circulation and the above series of sea ice extent, water temperature, and air temperature. Differences in sea level pressure are computed from monthly mean data of NCEP/NCAR reanalysis between Norway and Svalbard Dp1 (between the values of pressure at 70° and 77.5° N averaged for three points: 15°, 17.5°, and 20° E) and between Svalbard and Franz Josef Land Dp2 (between the values of pressure at 20° and 55° E at 80° N). The pressure difference between Norway and Svalbard was used before in RUSSIAN METEOROLOGY AND HYDROLOGY

Vol. 41

No. 8

2016

THE INFLUENCE OF THE NORTH ATLANTIC ON CLIMATE VARIATIONS

549

Table 2. The coefficients of correlation between the parameters of atmospheric circulation, water temperature TKM at the Kola meridian cross-section, and sea ice extent in the Barents Sea Parameter

Period

Janu- Febary ruary

March

April

May

June

July

August

September

October

November

December

Year

0.26 –0.41 –0.32 –0.33 –0.35 –0.31

0.28 –0.41 –0.34 –0.34 –0.36 –0.32

0.23 –0.39 –0.35 –0.36 –0.40 –0.30

0.28 –0.34 –0.28 –0.37 –0.34 –0.30

0.27 –0.37 –0.28 –0.28 –0.22 –0.19

0.27 –0.42 –0.24 –0.21 –0.08 –0.18

0.25 –0.45 –0.35 –0.27 –0.35 –0.14

TKM, °C Dp 1 Dp 2 PC

I–III I I–III II VIII I–III

0.15 –0.41 –0.29 0.02 –0.27 –0.06

0.15 –0.41 –0.34 –0.06 –0.30 –0.13

0.18 –0.42 –0.34 –0.15 –0.33 –0.23

0.23 –0.43 –0.35 –0.25 –0.35 –0.28

0.24 –0.44 –0.36 –0.29 –0.38 –0.30

0.23 –0.42 –0.32 –0.29 –0.36 –0.29

Sea ice extent, 103 km2 Dp 1 Dp 2 PC

I–III I I–III II VIII I–III

–0.14 –0.24 0.29 0.37 0.14 0.27 0.17 0.26 0.25 0.32 0.05 0.24

–0.28 0.27 0.32 0.42 0.38 0.33

–0.19 –0.21 –0.20 –0.08 –0.13 –0.23 –0.13 –0.26 –0.31 0.34 0.40 0.42 0.44 0.38 0.32 0.19 0.19 0.23 0.26 0.20 0.27 0.27 0.35 0.30 0.27 0.13 0.14 0.31 0.25 0.18 0.13 0.09 0.09 –0.06 0.04 0.09 0.39 0.32 0.33 0.28 0.35 0.22 0.21 –0.03 0.09 0.44 0.33 0.22 0.06 0.02 0.04 –0.12 –0.07 0.05

– – – – – –

Note: Dp1 is the pressure difference between the coast of Norway and Svalbard; Dp2 is the pressure difference between Svalbard and Franz Josef Land; PC is the first coefficient of EOF decomposition of the air pressure field over the Barents Sea. The period is indicated as follows: I–III is January–March, I is January, II is February, VIII is August. Correlation coefficients being significant at the level of 95% are bolded.

[19] as an indicator of atmospheric circulation influence on the Atlantic water inflow and sea ice in the Barents Sea. The pressure difference between Svalbard and Franz Josef Land was used there as an indicator of the effects of the meridional component of atmospheric circulation over the northern boundary of the Barents Sea on the displacement of the sea ice edge and on air temperature. The principal component (PC) of monthly mean pressure over the Barents Sea is represented by the first coefficient of pressure field decomposition into empirical orthogonal functions (EOF) which was carried out for 42 values of pressure at the points of 2.5° ´ 10° grid for 1948–2013. The first decomposition vector is presented in Fig. 2; the figure also presents the points between which the pressure difference was computed and the location of weather stations whose data were used to compute average air temperature over the sea. Table 2 presents the results of computations of correlation between the parameters of atmospheric circulation and climate parameters which can be affected by atmospheric circulation. The low correlation should be noted between the Norway–Svalbard pressure difference, water temperature at the cross-section along the Kola meridian, and sea ice extent in the Barents Sea which is presented only for the average pressure difference for January–March. The pressure difference between Svalbard and Franz Josef Land correlates with TKM and sea ice extent more highly that indicates the greater influence of meridional flows on the Atlantic water inflow and sea ice extent. The increase in the meridional component of wind speed in January and January–March causes the decrease in the Atlantic water inflow and the increase in sea ice extent during the whole year. According to the first coefficient of PC EOF pressure decomposition, variations in atmospheric circulation over the Barents Sea affect sea ice extent in January and February and TKM in the next months. In August PC is caused by the effects of sea ice extent and TKM in the preceding months. The effects of atmospheric circulation on air temperature from the results of the correlation between monthly mean surface air temperature and circulation parameters (PC) is appreciable in the cold season from October to April. However, the absolute values of coefficients do not reach 0.50. As a whole, the considered correlations between the parameters of atmospheric circulation and climate parameters in the Barents Sea correspond to the speculations on the sign of the correlation between them; however, their absolute values are small. This means that the contribution of regional atmospheric circulaRUSSIAN METEOROLOGY AND HYDROLOGY

Vol. 41

No. 8

2016

550

ALEKSEEV et al.

Fig. 3. Frequency structure of variability and relations between (1) average annual water temperature at the Kola meridian and (2) sea ice extent in the Barents Sea in May for 1928–2013. (a) Spectral density; (b) spectral function; (c) squared coherence between the series smoothed by 3. The relationship between conditional frequency and the period is Tk = 72/k years.

tion to the variability of climate parameters in the Barents Sea is smaller than that of the Atlantic water inflow. To find a cause for this situation, let us pay attention to the frequency structure of the variability of circulation and climate parameters used in the study. CLIMATIC SIGNAL AND WEATHER NOISE IN THE VARIABILITY OF CLIMATE PARAMETERS IN THE BARENTS SEA The spectral analysis of the used series revealed the difference in the distribution of contributions that time scales make to the interannual variability of parameters of atmospheric circulation and other climate parameters. In the variability of water temperature at the Kola meridian cross-section and sea ice extent and average air temperature in the Barents Sea, the low-frequency variability prevails in which the periods of about 70, 18, and 5–6 years are separated (Fig. 3). The low-frequency variability is characterized by the highest correlation between water temperature, sea ice extent, and average air temperature estimated by the squared coherence R2. Proceeding from the type of spectral density, the boundary between climatic low-frequency variability and the rest part of variability may be chosen at the conditional frequency k = 16 (the period of 4.5 years). In this case, the signal–noise ratio is equal to 3.35 which is much lower if the real low-frequency filter, for example, the moving 5-year averaging, is used. Therefore, taking into account the maintenance of appreciable correlation at the conditional frequency that is higher than 16, let us choose the boundary between the climatic signal and noise at the conditional frequency k = 24 (the 3-year period). In this case, the ideal signal–noise ratio is equal to 7.08 for TKM and 5.99 for sea ice extent. The use of moving 3-year averaging for removing the noise reduces these ratios because such filter to 1.78 and 1.68, respectively, is not ideal. In the interannual variability of pressure difference and the first coefficient of EOF decomposition of the pressure field, the contribution of periods of more than 24 years makes up not more than 20% and that of the periods of less than 3 years makes up to 40% against 15% in the variability of marine climate parameters (TKM and sea ice extent). The signal–noise ratio in the variability of these parameters of atmospheric circulation referred to the noise of the periods of less than 3 years is within 1.5–2.3. The removal of noise by the moving 3-year averaging makes this ratio less than 1.

RUSSIAN METEOROLOGY AND HYDROLOGY

Vol. 41

No. 8

2016

THE INFLUENCE OF THE NORTH ATLANTIC ON CLIMATE VARIATIONS

551

Table 3. The coefficients of correlation between monthly mean Tm and average annual Ty values of water temperature at the Kola meridian cross-section TKM and monthly mean values of sea ice extent and surface air temperature in the Barents Sea in 1952–2012 TKM

January

February

March

April

May

June

July

August

September

October

November

December

–0.60 –0.57 –0.65 –0.64

–0.50 –0.42 –0.58 –0.51

–0.49 –0.46 –0.63 –0.60

–0.72 –0.63 –0.77 –0.72

0.71

0.54

0.46

0.60

Monthly mean sea ice extent, 103 km2 Ty Tm T y* Tm*

–0.72 –0.80 –0.79 –0.82

–0.83 –0.84 –0.87 –0.83

–0.77 –0.81 –0.81 –0.81

–0.84 –0.81 –0.83 –0.83

–0.91 –0.94 –0.90 –0.91

–0.87 –0.87 –0.89 –0.88

–0.81 –0.78 –0.85 –0.82

–0.69 –0.65 –0.68 –0.64

Monthly mean surface air temperature, °C Tm

0.79

0.71

0.49

0.70

0.73

0.83

0.72

0.83

Note: T y* and Tm* are the average annual and monthly mean values of TKM if sea ice extent variations lag behind by 1 year, respectively. All series are smoothed by the moving 3-year averaging. The maximum absolute values are bolded.

The application of moving 3-year averaging to water temperature series at the Kola meridian cross-section, to sea ice extent, and to the series of average air temperature significantly increases the correlation between them and enables revealing the lags being important for forecasting (Table 3). The removal of oscillations with the periods of less than 3 years significantly increases the correlation between variations in water temperature at the Kola meridian and in sea ice extent. The maximum (by the absolute value) of the correlation coefficient equal to –0.94 falls on May and is kept at the level of –0.91 when the variations in sea ice extent lag behind those in TKM by a year. The value of the correlation coefficient between TKM and the first coefficient of EOF decomposition of sea ice edge position in May also increases to 0.75 after the smoothing. The estimates of the effects of regional variations in atmospheric circulation on climate variations in the Barents Sea after removing the short-period variability also increase. The correlation between the variations in the principal component of air pressure field PC over the sea in August and water temperature at the Kola meridian in August and September becomes considerable (correlation coefficients are equal to 0.50 and 0.54, respectively). The effects of PC variations in winter (January–March) on water temperature in July–October are also observed. The correlation between the pressure difference between Svalbard and Franz Josef Land in winter (December–February) and sea ice extent in all months also increases with the correlation maximum in July (0.54). The correlation between the air pressure difference between Norway and Svalbard and water temperature at the Kola meridian also increases after the smoothing, especially if the pressure difference lags behind water temperature by a year. TELECONNECTIONS BETWEEN THE NORTH ATLANTIC SST AND VARIATIONS IN WATER TEMPERATURE AND SEA ICE EXTENT IN THE BARENTS SEA The influence that the Atlantic water inflow to the Barents Sea makes on the variability of regional climate parameters, allows supposing the relation of variability to the anomalies of the parameters of the Atlantic water in the areas of their formation at the middle and low latitudes of the North Atlantic. To identify the influencing areas of the North Atlantic, the factor analysis and multivariate correlation analysis were carried out of sea surface temperature fields in the North Atlantic using the HadISST dataset and water temperature at the Kola meridian cross-section. Figure 4a presents the results of the factor analysis of the fields of average annual sea surface temperature specified at the points of 1° grid for 1951–2013. Two areas with the maximum load are singled out: southward of Newfoundland and in the equatorial North Atlantic. The computation of correlation between average annual values of water temperature at the Kola meridian cross-section and sea surface temperature at the grid points covering the North Atlantic and North EuroRUSSIAN METEOROLOGY AND HYDROLOGY

Vol. 41

No. 8

2016

552

ALEKSEEV et al.

Fig. 4. The areas of the North Atlantic where SST anomalies influence water temperature and sea ice extent in the Barents Sea. (a) The areas distinguished by factor loads; (b) the areas distinguished by the distribution of the coefficients of correlation between average annual water temperature at the Kola meridian cross-section and average annual SST at the of regular grid points; (c) the areas distinguished by the distribution of coefficients of correlation between average annual water temperature at the Kola meridian cross-section and average annual SST in the North Atlantic taken 4 years ahead.

pean basin demonstrated that the maximum coefficients are concentrated in the areas adjoining the Barents Sea and in the area southward of Newfoundland where the North Atlantic Current originates (Fig. 4b). Here the correlation between SST and TKM reaches 0.67. In the rest part of the water area the correlation coefficients are small or insignificant. The absence of significant correlations between average annual values of TKM and SST in the areas situated closer to the equator is evidently associated with the lag in the influence of anomalies from these areas. The computation of correlation between average annual values of TKM and SST with different time lags enabled detecting the area in the equatorial part of the North Atlantic (Fig. 4c) where the anomalies of TKM lag behind SST anomalies by 4 years. Thus, both methods identified two areas in the North Atlantic affecting water temperature in the Barents Sea, namely, the area southward of Newfoundland with the coordinates of 32°–44° N, 38°–64° W, where the Gulf Stream effects are observed, and the equatorial area with the coordinates of 0°–20° N, 40° W–8° E. The following analysis was based on the monthly mean data on sea surface temperature averaged for the distinguished areas and on monthly mean data on TKM for 1951–2013. The correlation coefficients rmk were RUSSIAN METEOROLOGY AND HYDROLOGY

Vol. 41

No. 8

2016

THE INFLUENCE OF THE NORTH ATLANTIC ON CLIMATE VARIATIONS

553

Table 4. Characteristics of the relationship between the initial and smoothed monthly mean values of Te, TG, and TKM taking the lag into account Te(Sept.) and TG(May) Characteristic

Initial data Smoothed data Distance, km Va, cm/s

TG(Sept.) and Te(June)

Te(Sept.) and TKM(March)

Correlation coefficient

Lag, month

Correlation coefficient

Lag, month

Correlation coefficient

Lag, month

0.51 0.77

44 44

0.55 0.70

9 9

0.66 0.81

54 54

6700 6

5800 25

12500 9

Note: Va is the speed of anomaly propagation.

computed between the matrices (tables) of SST value normalized to the mean value and standard deviation in the equatorial area Te¢ and in the area of the Gulf Stream effects TG¢ and water temperature at the Kola ¢ at the lags with the step of one year: meridian TKM T

( N - k ) -1 [ Te¢ mg ´ TKM mg + k ] = r mk

where g = 1, 2,…, N is the year; N is the series length; m = 1, 2,…, 12 is the month; k = 1, 2,…, M is the lag (years); T is the sign of matrix transposition. The use of monthly mean data allowed identifying the months, when SST in the areas under consideration exerts the maximum influence on TKM, and specifying the lag in variations in TKM relative to SST variations in each area. It turned out that the maximum influence on TKM is exerted by TG and Te in September and October, when the maximum values of SST are formed in these regions. This influence becomes more appreciable after the smoothing of initial series with the moving 3-year averaging. Table 4 presents the characteristics of the relationship between the initial and smoothed monthly mean values of SST in September and monthly mean values of TKM taking the lag into account. The table data demonstrate that the correlation between the SST anomalies and water temperature along the Kola meridian is higher when SST is specified in the equatorial North Atlantic 54 months ahead. The SST anomalies in the area southward of Newfoundland affect the anomalies of TKM to a smaller degree and are only 9 months ahead of them. The above effects of TKM on sea ice extent in the Barents Sea and the relationship between TKM and Te suppose the anticipating effects of Te on sea ice extent. The computations of correlation between monthly mean values of Te and sea ice extent using the above formula detected the maximum of correlation between Te in October and sea ice extent in July (the lag by 57 months). The correlation coefficient for the initial values is equal to 0.68 and increases up to 0.86 after applying the moving 3-year averaging. The respective coefficients for Te in September at the 58-month lag of TKM are equal to 0.66 and 0.84. RESULTS AND DISCUSSION The estimation of the effects of the Atlantic water inflow on sea ice extent in the Barents Sea using monthly mean and average annual data on water temperature at the Kola meridian cross-section demonstrated the maximum contribution of this factor (77%) to the variability of sea ice extent in May. This is explained by the fact that in the Barents Sea at the beginning of winter ice formation starts in the northern part under the significant influence of atmospheric circulation. At the same time, the warm water flowing to the southern part of the sea impedes ice propagation to the south whose maximum falls on April. So, the increase or decrease in warm water inflow indicated by the observations at the cross-section defines its propagation in the sea water area and affects the position of the southern ice edge. The effects of regional atmospheric circulation on variations in sea ice extent and air temperature in the Barents Sea taken into account using three different meteorological parameters, are insignificant as compared to the contribution of the Atlantic water inflow. The effects of the meridional component of atmospheric circulation is the most appreciable. It is assessed by the pressure difference between Svalbard and Franz Josef Land whose anomalies in winter affect water temperature and sea ice extent during almost the whole year. The influence of circulation variations over the Barents Sea indicated by the principal compoRUSSIAN METEOROLOGY AND HYDROLOGY

Vol. 41

No. 8

2016

554

ALEKSEEV et al.

nent of the pressure field is manifested in the variability of TKM from April to November and in the variability of sea ice extent from February to May. The comparison of contributions of variations in the Atlantic water inflow and atmospheric circulation using the multiple regression of TKM and circulation parameters (by turn) to the variations in sea ice extent demonstrated that 76% of sea ice extent variability in May is associated with the variations in water temperature and about 10% is associated with the parameters of regional atmospheric circulation. The basic reason for the smaller regional atmospheric influence on the interannual variability of sea ice extent is clear from the comparison of spectral structure of interannual variability of TKM, sea ice extent, and surface air temperature in the Barents Sea and the parameters of atmospheric circulation: 50% of interannual variability of the first two parameters falls on the periods of more than 10 years, and not more than 15% falls on the variations with the period of less than 3 years. In the interannual variability of the parameters of regional atmospheric circulation, up to 40% of variability falls on the variations with the period of less than 3 years which include the weather noise caused by not completely removed synoptic oscillations [3]. Besides, the computation of pressure difference between the points at the relatively small distances weakens the large-scale components of pressure variability and intensifies random variations. The variations with the periods of less than 3 years, where the correlation (coherence) between TKM and sea ice extent is low, are interpreted as noise; the rest part of variability is classified as a climatic signal. In this case, the signal–noise ratio is equal to the ratio of a part of the variance for the frequency from 0 to 1/3 year–1 to a part of the variance for the frequency from 1/3 to 1/2 year–1. To discriminate between two parts of variability and to provide such signal–noise ratio, the low-frequency filter with the rectangular frequency characteristic, the so called ideal filter, is needed. In practice, any filter is not ideal and its application worsens the signal–noise ratio. In this case, the simplest low-frequency filter is the moving averaging of the series with the time window of 3 years which reduces the ideal signal–noise ratio by about 3–4 times. Nevertheless, the application of such smoothing to the series used considerably increases the correlation between the climatic interannual variability of TKM and sea ice extent in May to –0.94 (Fig. 5a), between Te in October and TKM in January after 50 months, to 0.83 (Fig. 5b), between Te in October and sea ice extent in July after 58 months, to –0.86 (Fig. 5c). The cross-correlation functions of the last two pairs are presented in Fig. 6a. The correlation between atmospheric parameters and marine climate parameters also increased after the smoothing; however, their effects still remained much weaker than those of variations in the Atlantic water inflow. Besides, the smoothed pressure differences between Norway and Svalbard for January–March lag behind the respective variations in TKM by a year (Fig. 6b), but the correlation coefficient (0.40) does not exceed the 95% significance level. It should be noted that paper [36] as a result of experiments with the global climate model found that variations in the ocean northward of 70° N are ahead of the variations in the atmosphere. The opposite picture is observed at the middle latitudes. The high correlation between the smoothed series of TKM and sea ice extent in the Barents Sea in May (the correlation coefficient is –0.94) is maintained at the level of –0.91 at the variations in sea ice extent lagging behind by a year (Table 3) and may form a basis for the statistical climate prediction. The smoothed series of sea surface temperature in the equatorial North Atlantic in September and October may also be used as predictors to forecast climatic changes in water temperature and sea ice extent in the Barents Sea in separate months with the efficiency up to 75% and with the lead time to 58 months. The time limitation of the series used for the cross-correlation analysis (the minimum period is 1951– 2012 and the maximum period is 1928–2013) and their internal correlation (the significant autocorrelation radius) pose the question of the statistical reliability of the estimates obtained. The autocorrelation radius of the times series used varies from 3 years for the series of atmospheric parameters to 12 years for sea surface temperature in the equatorial zone. The radius is the shift at the computation of autocorrelation function at which its value is equal to the upper limit of 95% confidence interval of the zero value of autocorrelation. The smoothing increases the autocorrelation radius of the series. The following estimates were obtained for the series presented in Fig. 5. The autocorrelation radius for TKM and sea ice extent in May is 6 years after the smoothing that corresponds to the series of eight independent values (61 values in the smoothed series). Under these conditions the obtained estimate of correlation between the smoothed series (–0.94) remains statistically significant at the level of 95%. The autocorrelation radius for the series of Te in September and October after the smoothing was determined as 10 years that corresponds to the series containing six independent values. In this case, the coefficient of correlation between the smoothed series of Te in October and sea ice extent in July (–0.86) is also significant at the level of 95%. RUSSIAN METEOROLOGY AND HYDROLOGY

Vol. 41

No. 8

2016

THE INFLUENCE OF THE NORTH ATLANTIC ON CLIMATE VARIATIONS

555

Fig. 5. The examples of the graphs of interannual variations in water temperature and sea ice extent smoothed by the moving 3-year averaging. (a) Variations in (1) water temperature at the Kola meridian cross-section and (2) sea ice extent in the Barents Sea in May in 1952–2011 (the correlation coefficient is –0.94); (b) variations in (3) SST in the equatorial North Atlantic in October and (4) in water temperature at the Kola meridian cross-section in January in four years (the correlation coefficient is 0.83); (c) variations in (5) SST in the equatorial North Atlantic in October and (6) in sea ice extent in the Barents Sea in July in 4 years (the correlation coefficient is –0.86).

Besides the statistical significance of the obtained estimates, there is an evidence of the physical reliability of the influence of SST anomalies in the equatorial North Atlantic on the variations in TKM and sea ice extent in the Barents Sea. It is the realization (with the accuracy to 1 month) of the additive property of the time of propagation of anomalies following from Table 4. The authors of [37] revealed the effects of SST anomalies on the cyclonicity in the atmosphere in the Northwest Atlantic based on the analysis of in situ data and on the experiments with the atmospheric general circulation model. In [28] the model experiments revealed that SST anomalies at the low latitudes are responsible for the significant part of the warming in the Arctic. The authors of [5] had come to the similar conclusion before. In the pioneer paper [12] one of the centers of SST effects (the energy-active zone of the ocean) on weather and climate in Russia located in the equatorial North Atlantic was detected by solving the inverse problem using the hydrodynamic model of atmospheric circulation [17].

RUSSIAN METEOROLOGY AND HYDROLOGY

Vol. 41

No. 8

2016

556

ALEKSEEV et al.

Fig. 6. Cross-correlation functions. (a) Between (1) SST in the equatorial zone in October and temperature at the Kola meridian cross-section in January and (2) sea ice extent in July; (b) between water temperature at the Kola meridian cross-section and pressure difference between Norway and Svalbard in January–March.

CONCLUSIONS The variations in the Atlantic water inflow to the Barents Sea lead to the variations in water temperature at the cross-section along the Kola meridian and, so, define the most part of interannual variability of sea ice extent, water temperature, and air temperature over the Barents Sea in the cold season and for the year on average (for temperature). The effects of regional atmospheric circulation on the interannual variability of the mentioned climate parameters are not significant. One of the reasons is a significant part of weather noise in their variability. In particular, in May 76% of sea ice extent variability is associated with water temperature variations and about 10% is associated with the parameters of regional atmospheric circulation. The teleconnections were detected manifested in the significant effects of SST anomalies in the area southward of Newfoundland and in the equatorial North Atlantic on the climate parameters of the Barents Sea whose response lags behind the corresponding SST anomalies by 9 and 58 months, respectively. The significant correlation between SST anomalies and climate parameters of the Barents Sea builds capacity for developing the method of climatic forecasting of sea ice extent in the Barents Sea with the lead time of more than 4 years. ACKNOWLEDGMENTS The authors thank N.E. Kharlanenkova and N.E. Ivanov for their assistance in computations, E.I. Aleksandrov for providing data on air temperature from the Arctic stations, and V.F. Zakharov, V.P. Karklin, and A.V. Yulin for data on sea ice extent in the Barents Sea. Data on sea surface temperature were taken from the website of Hadley Centre for Climate Prediction and Research http://hadobs.metoffice.com.hadsst. The NCEP reanalysis data are provided by NOAA/OAR/ESRL PSD, Boulder (USA) at http://www.esrl.noaa.gov/psd/. The authors express special gratitude to the specialists of PINRO that provide data on water temperature at the Kola meridian cross-section at http://www/pinro.ru/n22/index/ phpstructure/labs/labhidro. The authors thank the reviewer for useful remarks which helped to obtain more accurate estimates. The study was supported by the Ministry of Science and Education of the Russian Federation in the part of applied scientific research and experiments on the theme “Development of new methods of monitoring of hydrometeorological and geophysical conditions at Svaldbard and in the western Arctic zone of the Russian Federation.” The grant agreement No. 14.21.0006 from October 20, 2014, the unique identifier PNIER RFMEFI61014X0006).

RUSSIAN METEOROLOGY AND HYDROLOGY

Vol. 41

No. 8

2016

THE INFLUENCE OF THE NORTH ATLANTIC ON CLIMATE VARIATIONS

557

REFERENCES 1. E. I. Aleksandrov and A. A. Dement’ev, “Database of Surface Meteorological Parameters in Polar Regions and Its Use,” in Formation of Database of Sea Ice and Hydrometeorology (Gidrometeoizdat, St. Petersburg, 1995) [in Russian]. 2. G. V. Alekseev, “Arctic Dimension of Global Warming,” Led i Sheg, No. 2 (2014) [Ice and Snow, No. 2 (2014)]. 3. G. V. Alekseev, “Weather Noise in Interannual Climate Variations,” Trudy AANII, No. 409 (1988) [in Russian]. 4. G. V. Alekseev, A. I. Danilov, V. M. Kattsov, et al., “Changes in the Climate and Sea Ice of the Northern Hemisphere in the 20th and 21st Centuries from Data of Observations and Modeling,” Izv. Akad. Nauk, Fiz. Atmos. Okeana, No. 6, 45 (2009) [Izv., Atmos. Oceanic Phys., No. 6, 45 (2009)]. 5. A. V. Baidin and V. P. Meleshko, “Response of the Atmosphere at High and Middle Latitudes to the Reduction of Sea Ice Area and the Rise of Sea Surface Temperature,” Meteorol. Gidrol., No. 6 (2014) [Russ. Meteorol. Hydrol., No. 6, 39 (2014)]. 6. Yu. A. Bochkov, “Historical Data on Water Temperature in the Layer of 0–200 m at the Kola Meridian Cross-section in the Barents Sea (1900–1981),” Trudy PINRO (1982) [in Russian]. 7. V. Yu. Vize, “A Cause for the Arctic Warming,” Sovetskaya Arktika, No. 1 (1937) [in Russian]. 8. V. F. Zakharov, The World Ocean and Pleistocene Ice Ages (Gidrometeoizdat, Leningrad, 1978) [in Russian]. 9. V. F. Zakharov, Sea Ice in the Climate System (Gidrometeoizdat, St. Petersburg, 1996) [in Russian]. 10. A. N. Zolotokrylin, T. B. Titkova, and A. Yu. Mikhailov, “Climatic Variations of the Arctic Front and the Barents Sea Ice Cover in Winter Time,” Led i Sheg, No. 1 (2014) [Ice and Snow, No. 1 (2014)]. 11. A. L. Karsakov, Oceanographic Studies at the Kola Meridian Cross-section in the Barents Sea in 1900–2008 (PINRO, Murmansk, 2009) [in Russian]. 12. G. I. Marchuk and Yu. N. Skiba, “Numerical Calculation of Conjugate Problem for the Models of Thermodynamic Interaction between the Atmosphere and the Ocean and Continent,” Izv. Akad. Nauk, Fiz. Atmos. Okeana, No. 5, 12 (1976) [in Russian]. 13. V. P. Meleshko and A. V. Baidin, “Response of Atmospheric Climate to the Arctic Sea Ice Reduction and to Other External Forcing in Recent Decades,” Trudy GGO, No. 568 (2013). 14. E. U. Mironov, Ice Conditions in the Greenland and Barents Seas and Their Long-range Forecasting (AANII, St. Petersburg, 2004) [in Russian]. 15. Review of Hydrometeorological Processes in the Arctic Ocean, Ed. by I. E. Frolov (AANII, St. Petersburg, 2014) [in Russian]. 16. T. V. Pavlova, V. M. Kattsov, V. P. Meleshko, et al., “A New Generation of Climate Models,” Trudy GGO, No. 575 (2014) [in Russian]. 17. Program of Atmospheric and Oceanic Research to Study Short-period Climate Changes (“Cross-sections” Program), Ed. by G. I. Marchuk, “Results of Science and Technology. Series: Atmosphere, Ocean, and Space —“Cross-sections” Program,” Vol. 1 (VINITI, Moscow, 1981) [in Russian]. 18. V. A. Semenov, I. I. Mokhov, and M. Latif, “Influence of the Ocean Surface Temperature and Sea Ice Concentration on Regional Climate Changes in Eurasia in Recent Decades,” Izv. Akad. Nauk, Fiz. Atmos. Okeana, No. 4, 48 (2012) [Izv., Atmos. Oceanic Phys., No. 4, 48 (2012)]. 19. L. Bengtsson, V. A. Semenov, and O. M. Johannessen, “The Early-twentieth-century Warming in the Arctic—A Possible Mechanism,” J. Climate, 17 (2004). 20. W. L. Chapman and J. E. Walsh, “Simulations of Arctic Temperature and Pressure by Global Coupled Models,” J. Climate, 20 (2007). 21. H. W. Chen, Q. Zhang, H. Körnich, and D. Chen, “The Surface Air Temperature Anomalies over the Barents Sea are Closely Associated with This Mode of Climate Variability. Arctic: The Barents Oscillation,” Geophys. Res. Lett., 40 (2013). 22. J. Inoue, M. E. Hori, and K. Takaya, “The Role of Barents Sea in the Wintertime Cyclone Track and Emergence of a Warm-Arctic and Cold-Siberian Anomaly,” J. Climate, 25 (2012). 23. D. P. Ivanova, J. L. McClean, and E. C. Hunke, “Interaction of Ocean Temperature Advection, Surface Heat Fluxes and Sea Ice in the Marginal Ice Zone during the North Atlantic Oscillation in the 1990s: A Modeling Study,” J. Geophys. Res., 117 (2012). 24. E. Kalnay, M. Kanamitsu, R. Kistler, et al., “The NCEP/NCAR 40-year Reanalysis Project,” Bull. Amer. Meteorol. Soc., 77 (1996). 25. A. Korablev, A. Smirnov, and O. K. Baranova, Climatological Atlas of the Nordic Seas and Northern North Atlantic, Ed. by D. Seidov and A. R. Parsons, NOAA Atlas NESDIS, 77 (2014). 26. S. Levitus, G. Matishov, D. Seidov, and I. Smolyar, “Barents Sea Multidecadal Variability,” Geophys. Res. Lett., 36 (2009). 27. J. Liptak and C. Strong, “The Winter Atmospheric Response to Sea Ice Anomalies in the Barents Sea,” J. Climate, 27 (2014). RUSSIAN METEOROLOGY AND HYDROLOGY

Vol. 41

No. 8

2016

558

ALEKSEEV et al.

28. J. Perlwitz, M. Hoerling, and R. Dole, “Arctic Tropospheric Warming: Causes and Linkages to Lower Latitudes,” J. Climate, 28 (2015). 29. V. Petoukhov and V. A. Semenov, “A Link between Reduced Barents-Kara Sea Ice and Cold Winter Extremes over Northern Continents,” J. Geophys. Res., 115 (2010). 30. N. A. Rayner et al., “Global Analysis of Sea Surface Temperature, Sea Ice, and Night Marine Air Temperature since the Late Nineteenth Century,” J. Geophys. Res., No. D14, 108 (2003). 31. A. B. Sando, Y. Gao, and H. R. Langehaug, “Poleward Ocean Heat Transports, Sea Ice Processes, and Arctic Sea Ice Variability in NorESM1-M Simulations,” J. Geophys. Res. Oceans, 119 (2014). 32. P. Schlichtholz, “Local Wintertime Tropospheric Response to Oceanic Heat Anomalies in the Nordic Seas Area,” J. Climate, 27 (2014). 33. P. Schlichtholz and M. N. Houssais, “Forcing of Oceanic Heat Anomalies by Air-sea Interactions in the Nordic Seas Area,” J. Geophys. Res., 116 (2011). 34. V. A. Semenov, T. Martin, L. K. Behrens, and M. Latif, “Arctic Sea Ice Area in CMIP3 and CMIP5 Climate Model Ensembles–Variability and Change,” The Cryosphere Discuss., 9 (2015). 35. L. H. Smedsrud et al., “The Role of the Barents Sea in the Arctic Climate System,” Rev. Geophys., 51 (2013). 36. E. van der Swaluw, S. S. Drijfhout, and W. Hazeleger, “Bjerknes Compensation at High Northern Latitudes: The Ocean Forcing the Atmosphere,” J. Climate, 20 (2007). 37. K. Ye, R. Wu, and Y. Liu, “Interdecadal Change of Eurasian Snow, Surface Temperature, and Atmospheric Circulation in the Late 1980s,” J. Geophys. Res. Atmos., 120 (2015). 38. H. Yndestad, W. R. Turrell, and V. Ozhigin, “Lunar Nodal Tide Effects on Variability of Sea Level, Temperature, and Salinity in the Faroe-Shetland Channel and the Barents Sea,” Deep Sea Res., Part I, 55 (2008).

RUSSIAN METEOROLOGY AND HYDROLOGY

Vol. 41

No. 8

2016