ICES Journal of Marine Science, 53: 887–898. 1996
Sea surface temperature changes in the North Sea and their
causes
Gerd A. Becker and Manfred Pauly
Becker, G. A., and Pauly, M. 1996. Sea surface temperature changes in the North Sea
and their causes. – ICES Journal of Marine Science, 53: 887–898.
On the basis of 25 years of digitized weekly sea surface temperature (SST) maps of the
North Sea, regional patterns, decadal changes, and trends are investigated with a view
to re-examining the assumption of a general warming trend, as suggested by the
observation that the years 1989–1994 represent the mildest winters observed in the past
50, perhaps even 130, years. A regional classification on the basis of an analysis of the
patterns of SST anomaly fluctuations suggests eight different regions resembling those
identified for the distribution of the North Sea water mass. The oceanic influence on
the western side and the continental influence on the eastern side are reflected in the
strikingly different behaviour of the western and eastern North Sea SST. Positive
anomalies usually start in the Southern Bight and move, with a delay of several weeks,
into the central and northern North Sea. None of the regions shows a temperature
trend over the last 25 years. However, spectral analysis of long time series and the
SST maps clearly indicate climatic fluctuations with a significant peak at a period
of about 8 years, a periodicity also found in the North Atlantic Oscillation index
(NAO). Salinity at the entrances to the North Sea is high when the NAO indicates an
increased zonal atmospheric circulation, whereas the correlation between SST
anomaly and NAO is quite low in these areas. By contrast, the correlation is high in
the central North Sea. Changes in the air–sea exchange processes and advection of
heat and salt from the North Atlantic appear to dominate climatic fluctuations in the
North Sea.
? 1996 International Council for the Exploration of the Sea
Key words: climatic fluctuations, North Atlantic Oscillation index, sea surface
temperature.
G. A. Becker: Federal Maritime and Hydrographic Agency, Postfach 311220, D-20305
Hamburg, Germany. M. Pauly: Universität Trier, Fachbereich IV – Geomathematik,
D-54286, Trier, Germany.
Introduction
In a contribution to the 1975 Århus Symposium, Becker
and Kohnke (1978) concluded that short-term (2–10
years) temperature fluctuations in the German Bight are
larger than the trend over the entire time series from
around the beginning of the 20th century to 1973.
Filtered time series are therefore more appropriate than
linear trends for assessing environmental influences on
biological processes. In a follow-up paper, Becker and
Kohnke (1977) presented sea surface temperature (SST)
time-series data for the Bay of Biscay, a station at the
Faroes, two southern North Sea stations (Noordhinder,
Helgoland) and an area off the Scottish east coast (F1).
All five low-pass-filtered SST series indicate that temperature has not increased significantly over more than
100 years. However, there is a pronounced inter-annual
and inter-decadal variability. The correlation between
the southern stations (Bay of Biscay, Noordhinder, and
1054–3139/96/060887+12 $18.00/0
Helgoland) was higher than between the northern series
(F1 and Faroes), but the southern and northern time
series were poorly correlated. The question of whether
air–sea exchanges or advective processes are dominant
for long-term climatic changes within the North Sea
could not be answered (Becker and Kohnke, 1978).
Unfortunately, some of the longer time series have later
been disrupted.
In a comprehensive compilation of surface data,
with energy fluxes derived for the entire North
Atlantic, Isemer and Hasse (1987) show that the net
annual air–sea exchange for the North Sea is close to
zero. In an investigation of the North Sea heat budget,
Becker (1981) derived an average heat gain of
8 W m "2 due to advective transport from the North
Atlantic. Any change, for instance in the position and
strength of the North Atlantic Current system, will
therefore change oceanographic conditions in the
North Sea.
? 1996 International Council for the Exploration of the Sea
888
G. A. Becker and M. Pauly
Figure 1. Rockall Channel (1948–1992). (a) Five-year running mean temperature (– · · –) and salinity (——) 1948–1992 (data
provided by Ellett and Jones, 1994). (b) Cross-correlation between temperature (5-year running mean) and salinity vs time-lag.
The 1980s was a period characterized by strong
westerlies and by increased heat advection into the
north-east Atlantic (Rodionow and Krovnin, 1992)
and probably also to the North Sea. This period is
also characterized by increased climatic variability on
both annual and inter-annual time scales in the North
Atlantic region. In the late 1980s and early 1990s,
salinity levels in the northern North Sea, the English
Channel, the southern North Sea and western German
Bight were extremely high with some reported values
exceeding 35.5 (Heath et al., 1991). The salinity increase
(Ellett and Turrell, 1992) affected a large area in the
North Sea as well as the adjacent Atlantic, which was
the source of these highly saline waters (Becker and
Dooley, 1995; Pollard et al., 1995). Ellett and Jones
(1994) compiled time series of temperature and salinity
for the central Rockall Channel for the period 1948–
1992, which are reproduced here as 5-year running mean
monthly values (Fig. 1a). The cross-correlation between
the two time series indicates that rmax =0.46 at a phase
lag of 7 months (Fig. 1b).
Relative to the historic data from the early 1900s
onwards, the 1990 salinity levels were not that exceptional. However, the magnitude of the 1990 high salinity
anomaly approached that of the Great Salinity Anomaly
(GSA) event (Dickson et al., 1988). The GSA could be
traced around the Atlantic subpolar gyre for over 14
years, from its origin north of Iceland in the 1960s until
its return to the Greenland Sea in 1981–1982. The
associated freshening of the upper 500–800 m of the
northern North Atlantic was observed around 1976
above the Rockall Channel (Fig. 1a) and appeared in the
northern and central North Sea in 1977/1979 (ICES,
1990). During the GSA period, the North Sea annual
mean SST anomaly was markedly negative (Becker and
Dooley, 1995). Turrell et al. (1996) demonstrated that
the seasonality of salinity in the northern North Sea may
be used as an index for changes in the Atlantic transport
Sea surface temperature changes in the North Sea
889
Figure 2. Global radiation at the Isle of Norderney (German Bight), monthly averaged means 1967–1994.
into the North Sea. In an earlier investigation, Ellett and
Turrell (1992) showed that salinity anomalies in the
northern North Sea can be related to changes in
the transport of Eastern North Atlantic Water within
the Slope Current.
The 1989/1990 high salinity anomaly was connected
with a large positive temperature anomaly in the North
Sea. The period 1989–1994 seems to represent the mildest North Sea winter climate years of the last 50 years,
while 1977 to 1979 and, perhaps, 1942 and 1962 were
probably the coldest. In a simulation of the North Sea
heat content, Pohlmann (1995) showed that interaction
between windstress and thermal forcing is the main
cause of heat content variability. Pohlmann noted a
change in the heat content during the summer months of
the 1980s. The period 1988–1992 (end of simulation)
exhibits a significantly higher summer heat content. The
summer of 1994 again showed very high SST anomalies.
The series of mild winters and relatively warm summers
therefore leads to the assumption of a more general
warming trend of the North Sea in the second half of
the 1980s. However, strong anomalies of this type also
occurred in the 1970s and are documented in the
temperature records since the beginning of regular
observations.
This contribution addresses mainly two questions: (1)
are there SST and/or salinity trends which could explain
changes in the North Sea ecosystem? (2) what are the
main causes of North Sea temperature and salinity
variations on climatic time scales?
Data
In autumn 1968, the Deutsches Hydrographisches
Institut started distributing weekly sea surface temperature (SST) maps of the North Sea (Mittelstaedt,
1969), based on near-surface temperature observations
taken on board research, commercial, and light vessels,
and from buoys and coastal stations. In recent years,
IR remote sensing data (NOAA satellites) have been
used more intensively. On average, about 200 to 600
observations were available per week. The accuracy of
the emerging weekly SST pattern was estimated to be
better than 0.5)C. Beginning in 1969, the weekly analyses were subjectively digitized on a 20 nautical mile
grid. Because of limited computer memory capacity at
the time, the North Sea was initially represented by
30#24 grid points, 420 of which were ‘‘wet’’. The
total data set now includes about 1400 digitized weekly
maps, representing about 600 000 data. The SST data
have been used to provide weekly and monthly mean
values, standard deviations, minima and maxima, and
long-term changes for the decades 1971–1980 and
1981–1990 (Becker et al., 1986; Becker and Wegner,
1993) and to force stratification in an advanced
numerical prognostic North Sea model (Pohlmann,
1995). The data are also available as a computer
animation.
The hypothesis that North Sea SST changes are
taking place as a result of local atmosphere–ocean
processes was checked using daily total global radiation
data (J cm "2) from the island of Norderney provided by
the German Weather Service. The time series starts in
1967. In this article, we use monthly data (Fig. 2) and
the standardized anomalies in respect of 1969–1993
mean value.
The NOA index is defined as the difference between
the normalized pressure anomalies in winter at Punta
Delgada (Azores) and Akureyri (Iceland). The 30-year
mean period is 1961–1990. A high index (>1) is
associated with strong westerlies, and a low index
(< "1) represents weak westerlies. A ‘‘normal’’ index
890
G. A. Becker and M. Pauly
covers the mid-range "1 to +1 and stands for a zonal
circulation of average strength (Koslowski and Löwe,
1993).
The International Young Fish Survey salinity data set
(ICES, 1990) provides a nearly synoptic coverage of the
North Sea during February each year since 1972.
Methods
To describe and explain time-dependent SST fluctuations in the entire North Sea, a spatial differentiation is
necessary. The 420 grid points were therefore classified
to form homogeneous regions based upon computations
of time series of annual SST anomalies (1969–1993) for
each grid point xi (i=1,2, . . . 420) by standardization of
the annual mean values:
zij =(xij "xi)/si
where zij is the SST anomaly value and xij the mean
value of SST at grid point i in year j (j=1,25), and xi is
the mean value and si the standard deviation of the mean
values at grid point i for the period investigated.
Because of the standardization, the anomaly variances
at all grid points have the value 1. In this way, the
typical ranges of fluctuation in the annual mean temperatures at the grid points can be taken into account
when evaluating extreme deviations. Assuming a
Gaussian distribution, the probability of z-values >P2P is
less than 5%. Therefore, anomalies >P2P can be regarded
as extreme values.
Cluster analysis for grouping grid points with similar
anomaly fluctuations was preceded by a principal components analysis in order to reduce the number of
variables (annual anomaly values), some of which are
highly correlated, to a few independent components.
References to literature describing numerous similar
applications in the field of climatological regionalization
are given in Pauly (1993). The multivariate methods
referred to in this article have been described in detail by
Johnson (1988). The principal components analysis
showed that approximately 80% of the total variance of
the annual anomaly series of all 420 grid points could be
reduced to six components, the values of which were
computed for each grid point by means of regression
analysis.
Classification, or regionalization, was carried out in
two steps. First, a hierarchical cluster analysis was performed to determine a suitable number of clusters. However, the number of clusters cannot be singularly
determined because it depends, among other factors,
on the particular choice of distance and of cluster algo-
Figure 3. Result of hierarchical cluster analysis: number of
clusters vs distance within clusters.
rithm, for which hardly any objective criteria exist. To
estimate the effects of methods used, different combinations of distances and algorithms were used. In all runs,
the maximum distance increase occurred at a cluster
number between five and eight. The results of clustering
using the Euclidean distance is shown in Figure 3.
In a second, iterative step, the grid points were
allocated to the proposed number of clusters by
means of partitioning cluster analysis. The quality
criterion applied in the final selection of a cluster was the
degree of spatial contingency, a decisive criterion in
regionalizing classification procedures (Pauly, 1993).
Since the spatial contingency among the 420 points was
already optimal for a number of eight clusters (Fig. 4a),
clustering was not repeated with a smaller number.
The assumption of cluster homogeneity postulates
similarities in the anomaly tendencies within the regions
in individual years. To evaluate the differences among
clusters, a one-way analysis of variance (ANOVA),
another quality criterion in regionalization, was carried
out for each year. The cluster class of each grid point
was used as the categorizing variable. Significant
(p<0.001) differences were found in all years between the
‘‘regional’’ anomalies. By comparing the variances
within regions and the variance of all grid points,
quantitative proof was furnished of the homogeneity of
all regions. However, striking differences were observed
between regions. Regions 8, 7, 4, and 2 had the lowest
variance quotient (variance within a region:variance of
all points), with values ranging from 0.22 to 0.44, while
the regions to the north and west had the highest
variance quotient, with values around 0.7.
For each grid point within each region, the distance
to the cluster centroids in the component space was
Figure 4. Subdivision of the North Sea in areas. (a) Based on SST patterns; position of 10 selected representative points is indicated
(in Figs 7 and 9, the points 1–8 are indicated by a preceding C; NW is representative of the northern advection area, SW of the
southern advection area; SST grid marked by crosses). (b) Based on hydrographic and biological information (ICES boxes).
892
G. A. Becker and M. Pauly
Figure 5. Thermo-isopleth diagrams ()C) for points (a) C3, (b) C4, (c) NW, and (d) SW.
calculated and the point of minimum distance was
selected as the typical grid point. A more detailed
description of the statistical technique applied is given
by Pauly (1993). The SST grid and the representative
points selected for each box (numbers 1 to 8) as well
as for the main two advection areas, the North-west
Atlantic inflow area (NW) and the South-west Channel
inflow area (SW) are given in Figure 4a.
Results and discussion
A remarkable feature is the spatial contingency of the
eight regions defined by clustering, which are solely
based on the temporal and spatial fluctuations of the
annual mean SST anomaly pattern. The result is very
similar to the subdivision of the North Sea into the
so-called ICES boxes (ICES, 1983), which were defined
on the basis of hydrographic and biological data (Fig.
4b). Region 1 represents the Northern Atlantic inflow
(ICES box 2 and parts of 1+3). Region 2 extends from
the German Bight into the Skagerrak, following the
Jutland current, and along the Norwegian coastal area
(ICES boxes 5 and 6). Region 3 represents part of the
Norwegian coastal current flowing into the Norwegian
Sea inflow (part of ICES boxes 6 and 1). Region 4
largely coincides with ICES box 7 (central North Sea).
Region 5 marks the area of advection from the English
Channel (Southern Bight; ICES box 4). Region 6 follows
the UK east coast (ICES box 3). Region 7 represents the
area around the tidal amphidromic point south-west of
Norway, where SST appears to be mainly influenced by
air–sea exchanges and less by advection. This area does
not appear as a separate ICES box, but includes parts of
boxes 1, 2, 6, and 7. Region 8 identifies the inner
Sea surface temperature changes in the North Sea
893
Figure 6. Standardized annual SST anomaly pattern in year (a) 1976, (b) 1979, and (c) 1990.
German Bight and Wadden Sea, where there is strong
continental influence (subset of ICES box 8).
For four selected representative points (NW, SW, C3,
and C4), the original, untreated, SST time series
were plotted as thermo-isopleth diagrams for the
25-year period (Fig. 5), showing clearly the different
ranges of the annual temperature variation. Close to the
continental coast (SW), the seasonal temperature range
is in the order of 16 deg C, whereas the variation is
limited to about 4 deg C in the Atlantic inflow region
(NW). The corresponding values for the central North
Sea (C4) and the Norwegian coastal current area (C3)
are in between (10 deg C and 8 deg C, respectively). In
general, the isopleths follow a horizontal course, indicating no significant climatic trend in the SST system, as
also observed by Becker and Kohnke (1977) based on a
894
G. A. Becker and M. Pauly
Figure 7. Standardized annual SST anomaly time series C1 (—.—) and C8 (—/—), 1969–1993. (Mean=——.)
longer time series. Colder and warmer periods are
reflected throughout the North Sea (e.g. cold winters
from 1978 to 1980; warm summers of 1975 and 1976;
warm winters of 1989 and 1990), suggesting a high
correlation between surface temperature in different
areas.
From the time series of standardized yearly SST
anomalies (z-values), the years 1976 (warm), 1979 (cold),
and 1990 (warm) are presented here (Fig. 6a–c).
Although these years have quite different anomaly
patterns, the distribution within a particular year
provides a consistent picture. When compared with the
corresponding February salinity distributions (ICES,
1990; not shown here), the spatial extent of the Northern
Atlantic inflow characterized by the 35 psu isohaline
appears to be limited to the Northern North Sea during
years with negative temperature anomalies. In contrast,
Atlantic Water covers large areas of the North Sea
during years with positive anomalies.
The general trend in the z-value time series for C1 and
C8 (Fig. 7) is similar over the 25 years but the anomaly
structure appears to have changed. From 1969 to about
1977, the annual mean standardized SST anomalies
fluctuated with relatively small, irregular oscillations.
Thereafter, around the time when the GSA reached the
North Sea, amplitudes become larger and periods
longer. However, the time series is still too short for a
definite conclusion that a change in climate variability
has taken place.
The correlation matrix (Table 1) between the 10 mean
annual anomaly time series shows that the anomalies
are in general significantly correlated. The correlation
coefficient is lowest for the NW-inflow and SW-inflow,
suggesting that the inflow processes at the two entrances
to the North Sea are largely independent. When comparing the salinity time series from the Rockall Channel
with salinity data from the western part of the English
Channel (data provided by MAST-NOWESP), a highly
significant correlation was found with a 74-month lag
(M. Visser, pers. comm.). This is apparently the result of
at least two different pathways or mean transports of the
North Atlantic Current west of the European continent.
The salinity signal carried in a southerly North Atlantic
Current branch arrives at the southern entrance of
the Rockall Channel at about 51)N off the Porcupine
Table 1. Correlation coefficients between selected representative
points
C1
C2
C3
C4
C5
C6
C7
C8
NW
C2
0.76
1.00
C3
0.86
0.82
1.00
C4
0.86
0.92
0.84
1.00
C5
0.70
0.92
0.79
0.87
1.00
C6
0.78
0.82
0.75
0.87
0.88
1.00
C7
0.85
0.95
0.87
0.95
0.88
0.87
1.00
C8
0.71
0.94
0.81
0.87
0.97
0.84
0.90
1.00
NW
0.81
0.58
0.75
0.66
0.53
0.57
0.70
0.53
1.00
SW
0.41
0.69
0.57
0.56
0.85
0.72
0.59
0.82
0.31
Sea surface temperature changes in the North Sea
895
Figure 8. Correlation between NAO index and standardized annual SST anomalies, 1969–1993.
continental slope (Pingree, 1993). There, this branch
splits into a faster northward moving branch and a
weaker and slower branch moving toward the Bay of
Biscay and eventually to the area adjacent to the English
Channel. Whether the apparent time-lag of 74 months
between the salinity time series can be explained by this
current pattern is still open to discussion, but the
time-lag would explain the low correlation (in lag 0)
between the annual mean SST anomalies at the North
Sea entrances.
Warm and cold periods seem to follow each other
with a certain periodicity (Fig. 7). A rather simple
time-series analysis of the SST anomalies revealed cycles
of around 80 months, but the series are too short to
obtain statistical confirmation. Because the wellestablished NAO time series (Löwe and Koslowski,
1993) reveals prominent quasicycles with periods of 2.4
and 7.7 years, the possible correlation between SST
anomalies and NAO index (1879–1992) was investigated, revealing a very interesting pattern (Fig. 8).
Although coefficients greater than 0.505 are significant
at p=0.01 (df=23), this criterion should be interpreted
as an informal one because both time series have
marked autocorrelation structures and thus fail to meet
896
G. A. Becker and M. Pauly
Figure 9. Comparison of standardized global radiation anomaly (——) at Norderney and (a) SST Anomalies at Norderney; (– – –)
(b) NAO index. (– – –).
important requirements for the application of significance tests. However, since the autocorrelations of the
SST anomaly series do not show any systematic space
variant differences, the spatial pattern of correlations
with NAO index can be interpreted independently of
autocorrelation structures.
High correlation coefficients had been expected in the
advection areas (regions 1 and 5) but the opposite
was found. In the northern inflow in particular, but also
in the Southern Bight, SST anomalies seem to be
independent of the zonality and strength of the North
Atlantic atmospheric circulation. Advection into the
North Sea depends on the large-scale North Atlantic
Current system. A band of lower correlation coefficients
extends from the Scottish coast through the southern
central North Sea to the outer German Bight, following
the mean North Sea circulation system. Lower correlations are also found above the Norwegian Trench with
Sea surface temperature changes in the North Sea
its high dynamics. The highest correlation (r>0.7) is
found in the central North Sea (cluster area 7). This
region is less dynamic and is characterized by relatively
small advection and variable and weak surface currents. It seems likely that air–sea exchange is the main
cause of the local SST anomalies, because this process
closely depends on atmospheric forcing and prevailing
weather systems. Higher correlation coefficients (r>0.6)
are also observed close to the English, North Frisian,
and Danish coasts. These shallow waters are much
influenced by local ocean–atmosphere processes and
temperature variability, short-term (days) up to the
entire annual cycle, is large (Becker and Wegner,
1993). On the other hand, the influence of oceanic
advection is very weak, river run-off and the
larger atmospheric variability above land being the
dominating factors.
SST anomalies within the entire North Sea are
closely (r>0.6) related to the global radiation
anomalies at Norderney in the period from May to
September. It is only during these months that global
radiation dominates the heat flux between the atmosphere and the ocean. From October to April, longwave back radiation, latent and sensible heat fluxes
determine air–sea exchange (Becker, 1981). A remarkable event was an abrupt change of the global radiation anomaly sign. During 1976/1977, which is about
the time the GSA arrived in the eastern North
Atlantic, the radiation anomaly became strongly
negative (Fig. 9a). This period lasted until 1980/1981.
The cause of the large positive SST anomaly in the
German Bight at the end of the 1980s is probably not
just a result of the positive radiation anomaly; the high
salinities observed during this period point to some
advection of saline and warmer water from the English
Channel to the German Bight (Becker and Dooley,
1995). There is no clear correlation between global
radiation anomaly at Norderney and NAO index.
There are periods in which the two parameters fluctuate in phase but during others they appear to be in
opposite phase (Fig. 9b).
The spatial correlation patterns (Fig. 8) and the close
correlation between SST anomalies and global radiation
anomaly (Fig. 9a) point to a dominance of local ocean–
atmosphere processes within the North Sea. The general
conclusion that zonality and strength of the mid-latitude
Atlantic atmospheric circulation determine the air–sea
exchanges, and hence SST anomaly fluctuations in the
North Sea, is consistent with the hypothesis put forward
by Bjerknes (1964) and Kushnir (1994) on SST variability in the North Atlantic. The lower correlation
coefficients in the Atlantic inflow regions indicate that
factors other than air–sea interaction processes are
responsible for the inter-decadal SST variability, of
which changes in the North Atlantic large-scale oceanic
circulation are the most likely causes.
897
Conclusions
(1) The North Sea does not show a significant trend in
the mean annual SST over the past 25 years.
(2) Climatic variability occurs mainly on time scales
from 5 to 10 years.
(3) Larger SST anomalies (positive or negative) are
related to salinity anomalies in the eastern North
Atlantic and in the North Sea.
(4) There is an indication of a change in the structure
of the SST oscillation. This indication of a less
stable system is also observed in global radiation at
Norderney. However, both time series are too short
to substantiate this feature statistically.
(5) The correlation between SST anomalies at the
southern and northern inflow areas is low, suggesting different transport processes from the North
Atlantic into the North Sea.
(6) SST changes are explained largely by local air–sea
exchange processes, which depend on the North
Atlantic atmospheric circulation. However, advective transports from the North Atlantic do play a
role and probably explain part of the observed
changes, especially at the entrances to the North
Sea.
References
Becker, G. A. 1981. Beiträge zur Hydrographie und
Wärmebilanz der Nordsee. Deutsche Hydrographische
Zeitung, 34: 167–262.
Becker, G. A., and Dooley, H. D. 1995. The 1989/91 high
salinity anomaly in the North Sea and adjacent areas. Ocean
Challenge, 6 (1).
Becker, G. A., Frey, H., and Wegner, G. 1986. Atlas der
Temperatur an der Oberfläche der Nordsee. Deutsche
Hydrographische Zeitung Ergänzungsheft Reihe B, Nr. 17.
Becker, G., and Kohnke, D. 1977. Comments on long-term
variations of sea surface temperature of the European shelf
seas. ICES CM 1977/C: 37.
Becker, G., and Kohnke, D. 1978. Long-term variations of
temperature and salinity in the inner German Bight.
Rapports et Procès-Verbaux des Réunions du Conseil
International pour l’Exploration de la Mer, 172: 335–344.
Becker, G. A., and Wegner, G. 1993. North Sea surface
temperature means 1981 to 1990. ICES CM 1993/C: 47.
Bjerknes, J. 1964. Atlantic air–sea interaction. Advances in
Geophysics. Academic Press, London. 82 pp.
Dickson, R. R., Meincke, J., Malmberg, S.-A., and Lee, A. J.
1988. The ‘‘Great Salinity Anomaly’’ in the northern North
Atlantic 1968–1982. Progress in Oceanography, 20.
Ellett, D. J., and Jones, S. R. 1994. Surface temperature and
salinity time-series from the Rockall Channel 1948–1992.
Fisheries Research Data Report No. 36, Lowestoft.
Ellett, D. J., and Turrell, W. R. 1992. Increased salinity levels in
the Northeast Atlantic. ICES CM 1992/C: 20.
Heath, M. R., Henderson, E. W., Slesser, G., and Woodward,
E. M. S. 1991. High salinity in the North Sea. Nature, 352:
116.
ICES. 1983. Flushing times of the North Sea. Cooperative
Research Report No. 123. 159 pp.
898
G. A. Becker and M. Pauly
ICES. 1990. Oceanographic distribution charts obtained from
data collected during IYFS. ICES, Copenhagen (compiled by
S. Reimert and H. Dooley).
Isemer, H.-J., and Hasse, L. 1987. The Bunker climate atlas of
the North Atlantic Ocean. Vol. II. Air–sea interactions.
Springer, Berlin, Heidelberg, New York. 252 pp.
Johnson, R. A. 1988. Applied multivariate statistical analysis.
London.
Koslowski, G., and Löwe, P. 1993. The western Baltic sea ice
season in terms of a mass-related severity index: 1879–1992.
Part I. Temporal variability and association with the North
Atlantic Oscillation. Tellus, 46A: 66–74.
Kushnir, Y. 1994. Interdecadal variations in North Atlantic sea
surface temperature and associated atmospheric conditions.
Journal of Climatology, 7: 141–157.
Löwe, P., and Koslowski, G. 1993. The western Baltic sea ice
season in terms of a mass-related severity index: 1879–1992.
Part II: Spectral characteristics and the associations with the
NAO, QBO and solar cycle. BSH – Sea Ice Branch Report
No. 2. 26 pp.
Mittelstaedt, E. 1969. Synoptische Ozeanographie in der
Nordsee. Bericht deutsche Wissenschaftliche Kommission
für Meeresforschung, N.F. 20: 1–20.
Pauly, M. 1993. Dürrefrühwarnsysteme und nationales
Dürremanagement am Beispiel der Trockenräume
Nord- und Südostkenyas – welchen Beitrag kann die statistische Analyse von Niederschlagsdaten leisten? Materialien
zur Ostafrika-Forschung, Heft 11, Trier.
Pingree, R. D. 1993. Flow of surface waters to the west of the
British Isles and in the Bay of Biscay. Deep-Sea Research II,
40: 369–388.
Pohlmann, Th. 1995. Simulating the heat storage in the North
Sea with a three-dimensional circulation model. Continental
Shelf Research (in press).
Pollard, R. T., Griffiths, M. J., Cunningham, S. A., Read, J. F.,
Perez, F. F., and Rios, A. F. 1995. Vivaldi 1991 – a study of
the formation, circulation and ventilation of Eastern North
Atlantic Central Water, pp. 8–9. Sigma, Dec. 1995.
Rodionow, S. N., and Krovnin, A. S. 1992. The 1980s in the
context of climatic changes in the North Atlantic region.
ICES Marine Science Symposia, 195: 93–102.
Turrell, W. R. 1992. The East Shetland Atlantic Inflow. ICES
Marine Science Symposia, 195: 127–143.
Turrell, W. R., Slesser, G., Payne, R., Adams, R. D., and
Gillibrand, P. A. 1996. The hydrography of the East
Shetland Basin in relation to decadal North Sea variability.
ICES Journal of Marine Science, 53: 899–916.