• Antonov, J. I., 1993: Linear trends of temperature intermediate and deep layers of the North Atlantic and the North Pacific Oceans:1957–1981. J. Climate,6, 1928–1942.

  • Barnett, T. P., 1983: Interaction of the monsoon and Pacific trade wind system at interannual time scales. Part I: The equatorial zone. Mon. Wea. Rev.,111, 756–773.

  • Cayan, D. R., 1992: Latent and sensible heat flux anomalies over the north oceans: Driving the sea surface temperature. J. Phys. Oceanogr.,22, 859–881.

  • Delworth, T., S. Manabe, and R. J. Stouffer, 1993: Interdecadal variations of the thermohaline circulation in a coupled-atmosphere model. J. Climate,6, 1993–2011.

  • ——, ——, and ——, 1997: Multidecadal climate variability in the Greenland Sea and surrounding regions: A coupled model simulation. Geophys. Res. Lett.,24, 257–260.

  • Deser, C., and M. L. Blackmon, 1993: Surface climate variations over the North Atlantic Ocean during winter: 1900–1989. J. Climate,6, 1743–1353.

  • Dickson, R. R., J. Meincke, S.-A. Malmberg, and A. J. Lee, 1988: The “Great Salinity Anomaly” in the northern North Atlantic 1968–1982. Progress in Oceanography, Vol. 20, Pergamon, 103–151.

  • Greatbatch, R. J., and S. Zhang, 1995: An interdecadal oscillation in an idealized ocean basin forced by constant heat flux. J. Climate,8, 81–91.

  • ——, and K. A. Peterson, 1996: Interdecadal variability and oceanic thermohaline adjustment. J. Geophys. Res.,101 (C9), 20 467–20 482.

  • Griffies, S. M., and E. Tziperman, 1995: A linear thermohaline oscillator driven by stochastic atmospheric forcing. J. Climate,8, 2440–2453.

  • Grötzner, A., M. Latif, and T. P. Barnett, 1998: A decadal climate cycle in the North Atlantic Ocean as simulated by the ECHO coupled GCM. J. Climate,11, 831–847.

  • Häkkinen, S., 1993: An Arctic source for the Great Salinity Anomaly:A simulation of the Arctic Ice-Ocean System for 1955–1975. J. Geophys. Res.,98 (C9), 16 397–16 410.

  • Hansen, D. V., and H. F. Bezdek, 1996: On the nature of decadal anomalies in North Atlantic sea surface temperature. J. Geophys. Res.,101 (C4), 8749–8758.

  • Horel, J., 1984: Complex principal component analysis: Theory and examples. J. Climate Appl. Meteor.,23, 1660–1673.

  • Houghton, R. W., 1996: Surface quasi-decadal fluctuations in the North Atlantic. J. Climate,9, 1363–1373.

  • Kushnir, Y., 1994: Interdecadal variations in North Atlantic sea surface temperature and associated atmospheric conditions. J. Climate,7, 141–157.

  • Latif, M., and T. P. Barnett, 1994: Cause of decadal climate variability over the North Pacific and North America. Science,266, 634–637.

  • Levitus, S., J. I. Antonov, and T. P. Boyer, 1994: Interannual variability of temperature at a depth of 125 meters in the North Atlantic Ocean. Science,266, 96–99.

  • Lozier, M. S., W. B. Owens, and R. G. Curry, 1995: The climatology of the North Atlantic. Progress in Oceanography, Vol. 36, Pergamon, 1–44.

  • Molinari, R. L., D. A. Mayer, J. F. Festa, and H. F. Bezdek, 1997: Multiyear variability in the near-surface temperature structure of the midlatitude western North Atlantic Ocean. J. Geophys. Res.,102 (C2), 3267–3278.

  • Sharp, R. J., 1996: Complex empirical orthogonal function analysis in two dimensional data sets: Theory and examples. Thesis, The Florida State University, 35 pp. [Available as Honors 1519, from Strozier Library Subbasement Archives, The Florida State University, Tallahassee, FL 32306-1005.].

  • Shriver, J. F., M. A. Johnson, and J. J. O’Brien, 1991: Analysis of remotely forced oceanic Rossby waves off California. J. Geophys. Res.,96 (C1), 749–757.

  • Sutton, R. T., and M. R. Allen, 1997: Decadal predictability of North Atlantic sea surface temperature and climate. Nature,388, 563–567.

  • Talley, L. D., 1984: Meridional heat transport in the Pacific Ocean. J. Phys. Oceanogr.,14, 231–241.

  • Weaver, A. J., and E. S. Sarachik, 1991: Evidence for decadal variability in an ocean general circulation model: An advective mechanism. Atmos.–Ocean,29, 197–231.

  • Winton, M., 1996: The role of horizontal boundaries in parameter sensitivity and decadal-scale variability of coarse-resolution ocean general circulation models. J. Phys. Oceanogr.,26, 289–304.

  • Woodruff, S. D., R. J. Slutz, R. L. Jenne, and P. M. Steurer, 1987: A comprehensive ocean–atmosphere data set. Bull. Amer. Meteor. Soc.,68, 1239–1250.

  • View in gallery

    (a) January SST from climatology. Isobath intervals are 2°C for positive values and 1°C for negative values. (b) Eigenvalue spectrum (in percent of explained variance) of SST anomaly.

  • View in gallery

    (a) Spatial amplitude function obtained from the CEOF mode 1 multiplied by 10. (b) SSTA trend of averaged warm event (1950–64) minus cold event (1970–84). The difference fields are based upon the CEOF reconstructions. (c) Spatial phase function obtained from CEOF mode 1. Only the phase from 0° to 80° is shown in order to highlight the region of interest.

  • View in gallery

    (a) Temporal functions obtained from CEOF mode 1. Solid line represents the time series of the real part and dotted line is the imaginary part. (b) Dotted line is the temporal phase function obtained from the CEOF mode 1. Solid lines is the best linear fit to the functions; a is a y intersect and b is the slope of the best fit function; y = a + bx represents the linear function. The x axis is a time axis from 1949 to 1990 and the y axis is an argument of the phase functions in degrees. Reciprocal of the slope is a period of one cycle.

  • View in gallery

    Chronological sequence of January reconstructed SST anomalies obtained from the CEOF mode 1 (4-yr low-pass filtered). The values are multiplied by 10. Units are in degrees Celsius.

  • View in gallery

    Time–latitude plot of SSTA obtained from the CEOF mode 1 and zonally averaged across the North Atlantic between 20° and 80°N. The values are multiplied by 10. Units are in degrees Celsius.

  • View in gallery

    Spatial amplitude functions obtained from CEOF mode 2 (a) and mode 3 (b). Spatial phase functions obtained from CEOF mode 2 (c) and mode 3 (d).

  • View in gallery

    As in Fig. 3a except for CEOF mode 2 (a) and mode 3 (b). As in Fig. 3b except for CEOF mode 2 (c) and mode 3 (d).

  • View in gallery

    Chronological sequence of January reconstructed SST anomalies obtained from CEOF mode 2 + 3 (4-yr low-pass filtered). The values are multiplied by 10. Units are in degrees Celsius.

  • View in gallery

    Solid line is the SSTA time series obtained from CEOF mode 2 + 3 averaged over the Labrador Sea region from 48° to 60°N, 50° to 60°W. Superimposed dotted line is the SSTA time series from Houghton (1996).

All Time Past Year Past 30 Days
Abstract Views 0 0 0
Full Text Views 82 78 1
PDF Downloads 7 6 0

Multi- and Quasi-Decadal Variations of Sea Surface Temperature in the North Atlantic

View More View Less
  • 1 Center for Ocean–Atmospheric Prediction Studies, The Florida State University, Tallahassee, Florida
  • | 2 Space Applications Centre, Ahmedabad, India
  • | 3 Center for Ocean–Atmospheric Prediction Studies, The Florida State University, Tallahassee, Florida
© Get Permissions
Full access

Abstract

The multi- and quasi-decadal variabilities of the sea surface temperature (SST) anomaly over the North Atlantic are investigated by means of two-dimensional propagating CEOF. Forty-six years of the COADS SST dataset from 1947 to 1992 are used. After removing the monthly climatology and four-year low-pass filtering the SST anomaly is submitted to CEOF. The total variance of the three largest modes accounts for 87.0% of the total variance.

Mode 1 is a very slow oscillation with an approximate 42-yr period and has basinwide spatial evolution, with alternate warm and cold anomalies appearing off Newfoundland and migrating northward until they disappear south of Greenland. The center of action of the anomalies both occur in the vicinity of the Labrador Sea, Newfoundland, and south of Greenland where deep-water formation takes place.

Mode 2 + 3 is a quasi-decadal fluctuation with an approximate 14-yr period. Mode 2 + 3 contains quasi-decadal timescales with alternating warm and cold anomalies propagating from the Labrador Sea eastward following the North Atlantic Current and the subpolar gyre.

Corresponding author address: Dr. Ken-ichi Mizoguchi, Center for Ocean–Atmospheric Prediction Studies, The Florida State University, Tallahassee, FL 32306-2840.

Email: ken@arcticocean.coaps.fsu.edu

Abstract

The multi- and quasi-decadal variabilities of the sea surface temperature (SST) anomaly over the North Atlantic are investigated by means of two-dimensional propagating CEOF. Forty-six years of the COADS SST dataset from 1947 to 1992 are used. After removing the monthly climatology and four-year low-pass filtering the SST anomaly is submitted to CEOF. The total variance of the three largest modes accounts for 87.0% of the total variance.

Mode 1 is a very slow oscillation with an approximate 42-yr period and has basinwide spatial evolution, with alternate warm and cold anomalies appearing off Newfoundland and migrating northward until they disappear south of Greenland. The center of action of the anomalies both occur in the vicinity of the Labrador Sea, Newfoundland, and south of Greenland where deep-water formation takes place.

Mode 2 + 3 is a quasi-decadal fluctuation with an approximate 14-yr period. Mode 2 + 3 contains quasi-decadal timescales with alternating warm and cold anomalies propagating from the Labrador Sea eastward following the North Atlantic Current and the subpolar gyre.

Corresponding author address: Dr. Ken-ichi Mizoguchi, Center for Ocean–Atmospheric Prediction Studies, The Florida State University, Tallahassee, FL 32306-2840.

Email: ken@arcticocean.coaps.fsu.edu

1. Introduction

The North Atlantic is the only place in the Northern Hemisphere where the atmosphere communicates with deep oceanic water masses through convective overturning (Talley 1984). The deep water in this region communicates with other oceanic waters at timescales longer than decadal. The predictability of the atmosphere by itself is merely on the order of weeks whereas that of atmospheric climate variations may potentially be over decadal scales due to coupling of the atmosphere with the ocean. The ocean retains integrated atmospheric memory at the long-term (decadal and longer) timescales due to the former’s relatively large heat capacity and inertia. If sea surface temperature (SST) variations in the North Atlantic at the climatic timescale are predicted it may be possible to improve the forecast of the global climate system.

Interdecadal fluctuations in North Atlantic SST anomalies have been investigated in several studies. Kushnir (1994) partitioned SST anomalies into two regimes of basinwide warm (1950–64) and cold (1970–84) events and concluded that, according to the nonlocal spatial distribution, the ocean plays an active role driving the atmosphere at longer than decadal timescales. Such an active role of the ocean is considered to be forced by the thermohaline circulation driven by the meridional overturning cell (MOC) associated with the deep-water formation of the North Atlantic. Delworth et al. (1993) found an approximately 50-yr thermohaline oscillation mode in their coupled ocean–atmosphere model and explained that this mode is basically produced by the ocean only and does not depend on the coupling with the atmosphere. This oceanic mode is physically explained as a linear thermohaline oscillator driven by stochastic atmospheric forcing (Griffies and Tziperman 1995).

Using a real empirical orthogonal function (EOF) analysis technique, Deser and Blackmon (1993) examined the two major patterns of SST anomalies in the Atlantic. The leading mode was characterized by a uniform polarity over the entire basin with its largest value along the Gulf Stream, whereas the second had a dipole pattern with anomalies of one sign east of Newfoundland and anomalies of the opposite polarity off the southeast coast of the United States. Their results indicated after 1945 about a 25-yr (multi-decadal) period of fluctuation in SST anomalies in their first mode, and a 12–14 yr (quasi-decadal) fluctuation in their second mode, of which the first mode seems to correspond to the Kushnir’s mode described above. The quasi-decadal mode has a spatial structure similar to the interannual variability of surface heat flux anomalies (Cayan 1992), which may come from the feedback mechanism of the ocean–atmosphere coupled system. Deser and Blackmon (1993) also suggested a linkage between SST anomalies in the North Atlantic and sea ice concentration anomaly at quasi-decadal period and found a significant correlation between the timeseries of the second EOF SST anomalies and that of ice concentration at a 1–2-yr lag, which they speculated was due to the advective nature of the SST anomalies in the Labrador Sea region.

Hansen and Bezdek (1996, hereafter HB) presented evidence of advection of SST anomalies along the path of the North Atlantic Current (NAC), an extension of the Gulf Stream. Warm and cold anomalies propagated along the subpolar and subtropical gyres. The speed of these anomalies were in general 1–3 km d−1 (1.2–3.6 cm s−1), which is less than the near-surface ocean speed. Sutton and Allen (1997) showed the alternate propagating SST anomalies at quasi-decadal timescale, embedded in the prominent warm and cold SST anomalies at multidecadal timescale. In their analysis a multidecadal timescale was not separated from amplitude modulation of the quasi-decadal signals. A series of warm and cold anomalies propagated along the Gulf Stream and NAC eastward and poleward. An average velocity of the anomalies was estimated as ∼1.7 cm s−1 and the alongpath length scale as ∼2000 km with their amplitude in the range 0.5°–1.0°C. The SST power spectrum averaged along the path exhibited a quasi-decadal spectral peak at 12–14 years.

The advective aspects of the observation in SST seems to be manifested in numerical ocean modeling experiments that exhibited decadal oscillations under steady forcing (Weaver and Sarachik 1991). The salinity and warm temperature anomalies were advected in between the subtropical and subpolar gyres eastward and poleward. The speed of the propagation was estimated as the order 1 to 2 km d−1 (1.2 to 2.3 cm s−1). On the other hand, propagating anomalies along the western boundary current and its extension were also identified in a ECHO coupled GCM (Grötzner et al. 1998) with an approximate 17-yr period. An ocean–atmosphere coupled theory of the quasi-decadal oscillation was first proposed by Latif and Barnett (1994) as a delayed feedback and adjustment to the variation of wind stress anomalies.

Many of the studies analyzing SST anomaly data, including those mentioned above, have been done using conventional real EOF analysis, which decomposes the data fields into real orthogonal pairs of spatial and temporal patterns. The product of the resulting time functions (TF) and spatial functions (SF) reproduces the major patterns of the analyzed datasets. However, the shortcoming of the EOF is that it tends to capture only the stationary patterns. That is, the extracted patterns are spatially fixed. Whereas, complex EOF (CEOF) analysis extracts nonstationary propagating features retaining physical information related to phase and wavenumber frequencies.

As described above, SST anomalies in the subpolar North Atlantic have propagating features. Therefore, it is necessary to understand the spatiotemporal evolution of the SST anomaly field at the decadal timescales. Latif and Barnett (1994) investigated the characteristic evolution of the integrated upper 500-m heat content in the Pacific Ocean from the ECHO coupled GCM, by using CEOF analysis with a 3-yr low-pass filter. They found that the leading mode of CEOF had a period of 20 yr. Studies using CEOF in other applications include Barnett (1983), Horel (1984), Shriver et al. (1991), and Sharp (1996).

In this paper the investigation focuses on the SST anomaly mainly in the subpolar gyre at multi- and quasi-decadal timescales. The monthly SST anomaly (SSTA) is defined as the departure from the monthly climatology. The mathematical method to analyze the dataset is CEOFs. The SST dataset is from the Comprehensive Ocean–Atmosphere Data Set (COADS) covering the North Atlantic. The data are described in section 2, the method in section 3. The results are shown in the context of other observations in section 4. The discussion is in section 5.

2. Data description

A 46-yr subset of COADS (Woodruff et al. 1987) monthly sea surface temperature covering the North Atlantic from 20°N to 80°N, 70°W to 30°E from 1947 through 1992 with a grid size of 2° × 2° is examined. COADS has inherent uncertaintities due to historical changes in instrumentation, observation techniques, coding methods, data density, and ship tracking. The historical datasets after 1947 were chosen due to their high spatial coverage. This is after the large shift in SST anomaly observed by Deser and Blackmon (1993). The Mediterranean Sea and Baltic Sea are ignored for this analysis.

The monthly SST climatology fields are calculated by averaging all the months in time at each location. In the early years of the time series, there is a lack of data in the North Atlantic from 60° to 80°N and in the Labrador Sea due to severe weather conditions and ice coverage. The area off the African continent from 20° to 30°N, 30° to 20°W lacks data in late 1940s and 1950s regardless of the seasons. The area from 60° to 80°N of the historical dataset is replaced by the climatology when there is a missing value. Missing values in other regions are filled by interpolating from surrounding data values. Each month of the SSTA and the 46-yr time series of SST at each location then has an E–W–E and a N–S–N 1–2–1 Hanning filter applied three times. A new climatology is calculated from the filled, interpolated, and smoothed fields and the original climatology is discarded. The data used in the calculation are the differences from the new climatology.

In January, all the isotherms are positive and run roughly east–west (Fig. 1a). The gradient of the isotherms is strongest off Newfoundland at 45°N, 55°W where the Gulf Stream flows. The minimum and maximum values are 2° and 24°C, near the northern and southern boundaries, respectively.

The domain is on a 51 × 31 grid, where there are 950 spatial points in the ocean and 631 land points set. The monthly anomaly time series at each ocean point are low-pass filtered using a 4-yr running mean. The time series of anomalies are not detrended. As will be shown below, the low-pass filtered SSTA has a very slow variation, which would be suppressed by linear detrending.

3. Method

a. Two-dimensional propagating complex empirical orthogonal function

Two-dimensional propagating complex empirical orthogonal function analysis is used in this study to extract physical information on propagating features from the SSTA at climatic timescales. The variable u(xm, t) = SST(rm, t), where rm is the position vector and t is the time index, can be written
i1520-0485-29-12-3133-e1
where am(ω), bm(ω) are the Fourier coefficients. Define the complex representation of a variable as
i1520-0485-29-12-3133-e2
where cm(ω) = am(ω) + ibm(ω).
The expansion of (2) gives
i1520-0485-29-12-3133-e3
The imaginary part ûm(t) is the Hilbert transform of the real part, which is in fact a π/2 shifted function of the real part in time with the same magnitude of the Fourier transform.

The variable Um(t) is represented as a complex-valued spatiotemporal matrix. Here Um,t is formed so that each row of the matrix is a temporal array, and each column is a spatial array, where m = 1, 2, · · · , M, t = 1, 2, · · · , N, and M = 950 (ocean points), N = 504 (months). The covariance array is calculated by multiplying the M × N matrix by its transpose to form an N × N Hermitian matrix, which is symmetric with real-valued diagonal elements. By solving the eigensystem equation of the covariance matrix, the eigenvalue vector λi, where i = 1, 2, · · · , N, and the spatial and temporal functions (eigenvectors) are obtained. Normalized eigenvalues represent a relative variance of the original dataset [λi(Σni=1λi)−1].

The data Um(t) are reconstructed by the summation of SF and temporal TF as follows:
i1520-0485-29-12-3133-e4
where SF = Sl(m)el(m) and TF = Rl(t)el(t) and m and * denote a spatial index and a complex conjugate, respectively.
The spatial amplitude function Sl(m) is defined as
Slm1/2
The spatial phase function θl(m) is defined as
θlm
The wavenumber k is obtained by
i1520-0485-29-12-3133-e7
where x = (x, y).
The temporal amplitude function Rl(t) is
Rlt1/2
The temporal phase function ϕl(t) is
ϕlt
and the frequency ω is calculated from the temporal phase function as
i1520-0485-29-12-3133-e10

4. Results

CEOF analysis was done on 4-yr low-pass filtered data. The first three modes of SSTA CEOF account for more than 87.0% of the total variance, of which the first mode accounts for 58.3%, the second and the third 16.1% and 12.7%, respectively. The first three modes appear to be distinctively above the noise level (Fig. 1b) and, therefore, may have some physical meaning. Only the first three modes are discussed hereafter.

a. CEOF mode 1

The spatial amplitude function for CEOF1 (Fig. 2a) is maximized in the western portion of the subpolar gyre, with local extrema south of Greenland and east of Newfoundland. This is the region where the largest warm anomalies appear in the reconstructed SSTA (Fig. 4) in the early 1950s and the cold anomalies appear in the mid-1970s.

The spatial phase from 50° to 70°N, where deep-water formation occurs, is highlighted in Fig. 2c. The phase in this region increases in a nearly linear fashion to the north, indicating nearly steady meridional propagation of SST anomalies.

Both real and imaginary parts of the time series (Fig. 3a) show slow variability that does not finish a single cycle. A half-cycle occurs approximately from about 1950 to 1970, which is when the warm event occurred from Newfoundland to southeast of Greenland (Fig. 4).

Since the argument of the phase function is defined from 0° to 360° the time series of the function cannot be plotted as a continuous function. Therefore, it is defined from 0° to 1500° to make it continuous (Fig. 3b). The y intersect has the value of 332.7°. The linear fit is applied to this function so that the fundamental period ω is estimated. The period of mode 1 is about 43.7 yr and is about double the half period of about 21 yr estimated from the series in Fig. 3a.

The time evolution of the spatial field of each mode is obtained by multiplying the spatial and temporal function of that mode [as in Eq. (4)] and taking the real component. The reconstructed spatial patterns for SSTA for CEOF1 are shown in Fig. 4. Several years of variation are represented by each month due to the 4-yr low-pass filtering. Following HB, January is chosen as the representative month of each year.

In January 1949, a warm anomaly is located off Newfoundland at around 45°N, 60°W. Contemporaneous with the warm anomaly is the cold anomaly south of Greenland at around 35°W, 60°N. As the warm anomaly grows, the cold anomaly weakens and slowly dissipates. By 1953 most of the ocean midbasin from 20° to 70°N is covered with the warm anomaly. After 1954 this anomaly reduces its size and migrates northward until eventually it disappears at around 45°N, 60°W in 1973, the same location where the cold anomaly disappeared in 1951.

In 1958 a new premonitory cold event appears off New York at around 40°N, 70°W while the opposite anomaly occurs to the north. During 1967–68 another cold patch emerges from the eastern side of the basin off the African continent, merging with the one from the west. As the cold anomaly strengthens it follows the same path of the preceding warm event, and the warm event weakens and disappears south of Greenland. In 1985 another warm anomaly appears off New York and begins to follow the same path until 1990, the end of the time series. Over the course of the entire 42 years, only the warm event completes nearly one cycle. Rough estimation of the warm anomaly speed is about 0.3 km d−1 (0.35 cm s−1) assuming the distance that the warm anomaly travels and the duration that it survives are 20° (from 40° to 60°N) and 21 yr (from 1953 to 1973), respectively.

b. Multi-decadal change

Multidecadal variability in North Atlantic SSTA is clearly seen in a time–latitude presentation of zonally averaged SSTA (Fig. 5). The timing of these variations changes with latitude. Warming begins before 1950 and lasts until around 1964, the beginning of the cold event that lasts until around 1985. The timing of these events agrees with those found in Kushnir (1994). The transition from warm to cold anomalies is also seen in Fig. 5, in which the two clearly recognized temperature events become delineated and the tilting of these figures exhibits a northward propagation. Sutton and Allen (1997) showed a similar figure (their Fig. 2a), displaying the prominence of warm and cold anomalies that began in the mid-1940s and in the late 1960s respectively, embedded in a quasi-decadal fluctuation. However, they could not discriminate between a multidecadal and quasi-decadal timescale due to their choice of mathematical technique.

The difference field of the CEOF mode 1 between the warm interval (1950 to 1964) and the cold interval (1970 to 1984) is shown in Fig. 2b. The resulting pattern is similar to the spatial amplitude function (Fig. 2a). Only one strong warm patch is situated at 50°N, 45°W. A similar analysis by Kushnir (1994) shows the three temperature extrema occurring along the northern boundary of the domain, around Iceland, and in the Labrador Sea. The result in Fig. 2b is also very similar to SST changes found by Delworth et al. (1993, their Fig. 6a) associated with a 40–50 yr periodicity in the thermohaline circulation (THC) of a fully coupled ocean–atmosphere model. The oscillation may be driven by density anomalies in the region of deep-water formation (52° to 72°N).

c. CEOF mode 2 and mode 3

The second and the third modes have comparable variance of 16.1% and 12.7%, respectively. These modes are first examined separately, then in combination as a rotating mode. The properties of interest in modes 2 and 3 are mainly found in the subpolar region; therefore, the subtropical region is excluded from the detailed analysis.

The spatial amplitude of mode 2 has a maximum region located off Newfoundland (Fig. 6a). Three secondary maxima are found, two on either side of the southern tip of Greenland, and one north of Norway. The spatial phase (Fig. 6c) indicates propagation from north of the Labrador Strait (around 70°N) southeastward to the center of the basin at 45°N, 35°W. They then turn eastward beneath the phase nexus at 55°N, 30°W and divert to the north, up to 70°N northeast of Iceland. This phase structure suggests the presence of a phenomenon rotating about the subpolar gyre. However, the direction of rotation can only be determined in conjunction with the temporal phase, described below.

For mode 3 the spatial amplitude (Fig. 6b) has a strong maximum located northeast of Newfoundland at the entrance of the Labrador Sea. There is another maximum extending from west of Europe to east of Iceland. The corresponding spatial phase (Fig. 6d) also seems to suggest that the SST anomalies propagate southeastward from the Labrador Strait into the Atlantic basin, and turn eastward toward England. Following this path, the phase increases in a nearly linear fashion. A phase nexus is found near the same location as for mode 2, further suggesting the two modes are dynamically related.

The real part of the temporal function for mode 2 (Fig. 7a) shows a sinusoidal behavior, particularly after 1967, with a period of around 15 yr. The temporal phase (not shown) suggests a varying period, based on the time between peak values, between 12 and 18 years. A least squares linear fit to the phase function (Fig. 7c) produces an average period of 14.4 yr. The increasing phase of this temporal mode, coupled with the phase of the spatial mode, indicates cyclonic propagation about the subpolar gyre.

The temporal phase function for mode 3 (not shown) is similar to that of mode 2, though the former is relatively jagged with a burst of 1–2 yr variations in the early 1970s. The linear fit to the temporal phase yields a period of 13.9 yr. The nearly constant slope of the temporal phase (Fig. 7d) has important implications for the speed of propagating features in mode 3. In conjunction with the spatial phase [as in Eq. (4), l = 3] the temporal phase indicates roughly steady movement out of the Labrador Sea into the Atlantic and eastward toward Europe.

The similarity between mode 2 and mode 3 in their relative variance, spatial structure and temporal phase suggests that these two modes represent a single split mode. In the following subsection we will discuss the combination of CEOF mode 2 and mode 3. Sharp (1996) discussed using CEOF analysis applied to a field that has a rotating feature. CEOF analysis treats this case in a similar fashion to the way a real EOF would split a linearly propagating feature. In CEOFs, circular motion splits into two equal variance modes. Combining the two modes reconstitutes the rotating structure.

d. CEOF mode 2 + 3

The real part of the sum of modes 2 and 3 is now examined. As in the previous subsection the subtropics is again ignored. January maps of the anomalies are examined (Fig. 8) from 1949 through 1990. In 1950 a cold anomaly is situated in the midbasin around 50°N, 25°W and a weak cold anomaly is developing north of Iceland. A warm anomaly is at the entrance of the Labrador Sea (LS) at 60°N, 50°W and another east of Scandinavia. This alternating pattern of warm and cold anomalies travels around the subpolar gyre cyclonically. The center of the circular motion is located roughly in between Greenland and Iceland.

By 1955 the warm anomaly originally off Newfoundland has replaced the cold anomaly. The warm anomaly is now found in the midbasin and the cold anomaly now straddles the east and the west of southern Greenland. These features continue to travel about the gyre until the initial position of the warm and the cold anomalies is reestablished in 1965. In 1971 the 1955 pattern reestablishes, occurring again in 1983. The alternating anomalies seem to be advected by the currents in the North Atlantic, one of which is the North Atlantic Current that parallels the 50°N line.

From 1959 through 1966 a cold anomaly exists in the mode 2 + 3 projection of North Atlantic SSTA (Fig. 8), which has no counterpart in HB. During these years a warm anomaly appears at the same location in the mode1 projection (Fig. 4). It might be speculated this cold anomaly is a partial compensation for the warm anomaly in the mode 1 and is due to the orthogonality constraint of the CEOF analysis.

The speeds of warm and cold anomalies are both estimated as 1.3 cm s−1. The warm and cold anomalies are followed from 1951 to 1958 and from 1959 to 1966, respectively. They propagate from 50° to 20°W during this period of eight years. The scale of anomalies along the west–east path is roughly estimated as ∼2000 km (∼20°). These anomalies are selected because their signal amplitude is prominent compared to any other years.

This combined mode has a large impact on the Labrador Sea. SSTAs from CEOF mode 2 + 3 are averaged over the region from 48° to 60°N, 50° to 60°W. This is compared to the SSTA averaged over the area along the Labrador–Newfoundland continental margin from 48° to 60°N seaward from Houghton (1996, his Fig. 1) in Fig. 9. His time series was produced from January–March monthly anomalies calculated by subtracting the climatological mean. For this comparison his result is detrended and both time series are normalized by twice their respective standard deviation. The period and relative amplitude of oscillation are nearly identical in both the time series. Houghton suggested these fluctuations might be related to latent heat flux and surface freshwater anomalies due to local precipitation, river runoff, and ice melt, which are subject to modulation by variations in the anomalous wind stress.

5. Summary and discussion

The multi- and quasi-decadal variabilities of the sea surface temperature (SST) anomaly over the North Atlantic are investigated by means of two-dimensional propagating CEOF. Forty-six years of COADS SST data from 1947 to 1992 are used. After removing the monthly climatology and 4-yr low-pass filtering the SST anomaly is submitted to CEOF. The total variance of the three largest modes accounts for 87.0% of the total variance. The first three modes are assumed to be physically significant based on the small variance captured by modes four and higher (Fig. 1b). The second and third modes are found to be a split mode and are primarily considered in summation with each other. The results of this study indicate that SST anomalies in the North Atlantic have two different timescales of multi- and quasi-decadal fluctuations represented by mode 1 and mode 2 + 3, respectively.

Mode 1 is a very slow oscillation with an approximate 42-yr period and has basinwide spatial evolution, with alternate warm and cold anomalies appearing off Newfoundland and migrating northward until they disappear south of Greenland. The centers of action of the anomalies both occur in the vicinity of the Labrador Sea, Newfoundland, and south of Greenland where deep-water formation takes place. Delworth et al. (1997) emphasized the importance of SST anomalies in the Greenland Sea region associated with the 50-yr oscillation. The behavior of our model (Fig. 2b) is similar to the 50-yr oscillation of Delworth et al. (1993), suggesting they may be the same or related events. Delworth et al. (1993) explained that physical mechanism of the 50-yr oscillation in their model involved changes in both horizontal and thermohaline circulation, which were mainly controlled by upper ocean temperature anomalies and salinity anomalies in the sinking region (52°–72°N) respectively. The delayed interaction between the thermohaline and horizontal circulations determines the multidecadal timescale. Greatbatch and Zhang (1995) ran an ocean model driven by a constant, zonally uniform surface heat flux that also exhibited a 50-yr period. Their idealized ocean contained alternating warm and cold SST anomalies migrating northward (their Fig. 6). They claimed the temperature anomaly at high latitudes was controlled by the strength of the oceanic advection from lower latitudes. However, the overturning has a negative feedback from the high-latitude temperatures. An increase in the temperature anomaly in the northern ocean decreases the overturning. The opposite sense is true for the other phase of the oscillation.

Mode 2 + 3 is a quasi-decadal fluctuation with an approximate 14-yr period. Mode 2 + 3 contains quasi-decadal timescales with alternating warm and cold anomalies propagating from the Labrador Sea eastward following North Atlantic Current (NAC) and the subpolar gyre. The advective anomalies from the Labrador Sea may be linked to a winter sea ice concentration anomaly. Deser and Blackmon (1993) suggested the SST anomaly in the Labrador Sea is strongly related to the ice concentration there. Minima in SST anomaly were found to follow the ice concentration by 2 years. The sea ice anomalies in the Labrador Sea seemed to be advected southeastward, dominating the decadal fluctuation of SST anomaly in the south of Greenland. This implies that the quasi-decadal cycle of SST anomalies in the Labrador Sea region is connected to the low frequency variability of Arctic sea ice.

The sea ice anomalies from the Arctic are also believed to play a role in the formation of “The Great Salinity Anomaly” (GSA) (Dickson et al. 1988) and the“Lesser Great Salinity Anomaly” (LGSA) (Lozier et al. 1995). The GSA and LGSA were characterized as large, near-surface fresh, cold water. These two events arrived in the Labrador Sea in around 1968–71 and 1978–84, respectively. Both events roughly coincide with the two minima in the time series of CEOF mode 2 + 3 in the Labrador Sea (Fig. 9) in around 1971 and 1983–85, respectively. There appears to be a correspondence between the low salinity anomalies and the cold anomalies of SST. The relationship implies a strong GSA also appeared in 1957–58. There is little observational evidence from this time period, but the model results from Häkkinen (1993, her Fig. 7b) showed that there were strong salinity minima, which were highly surface trapped within the upper 100 m, in 1959, 1964–65, and 1971 in the Greenland Sea. The appearance of the salinity minima in 1959 in the model results roughly coincides with that deduced from the CEOF. The timing of the other salinity minima of the model also corresponds with that from CEOF. Thus, the temperature minimum in 1957–58 might indicate a previously unrecognized salinity event. If the GSA events are cyclic events with an approximate 14-yr period, as seen both in an observation and in the CEOF time series, another GSA is soon to be expected to arrive at the end of this millennium.

The propagation speed of CEOF mode 2 + 3 is estimated approximately as 1.3 cm s−1 and the scale of anomalies along the west–east path roughly as 2000 km (∼20°). This speed is comparable to that from Weaver and Sarachik (1991) although their model experiments produced only warm anomalies (their speed was about 1.2 to 2.3 cm s−1). Sutton and Allen (1997) showed a series of warm and cold anomalies that propagated downstream along the Gulf Stream and NAC with an average velocity of 1.7 cm s−1, alongpath length scale 2000 km, and amplitude in the range 0.5–1.0°C. They were advected from the storm formation region off Cape Hatteras in the subtropical gyre, which is not included in our analysis, all the way to the southern tip of Greenland in the subpolar gyre. The advection along the Gulf Stream is not seen in the CEOF mode 2 + 3; however it occurs from the Labrador Sea region. Hansen and Bezdek (1996) estimated the propagation speed of their SST anomalies using autocorrelation analysis as 2.3 cm s−1. The order of the speeds is considerably slower than the near-surface currents in the core of Gulf Stream. This suggests the existence of a deep signal below the mixed layer. Observational data found evidence for decadal patterns of variability a few hundred meters below the mixed layer (Antonov 1993; Levitus et al. 1994; Molinari et al. 1997). The advective role of SST anomalies below the mixed layer by mean currents seems to be very important in determining the evolution of the SST anomalies at this timescale. Grötzner et al. (1998) explained the physical mechanism of the quasi-decadal oscillations as the delayed feedback of the Gulf Stream, which is carried by the mean currents and the modification of the gyral strength through anomalous wind stress curl. One possible scenario is that the quasi-decadal oscillation mode in the North Atlantic may be explained in terms of the atmosphere–ocean interaction and the advection of temperature anomalies by mean currents.

Greatbatch and Peterson (1996, hereafter GP) ran a simple ocean model under mixed boundary conditions and found a 17-year self-sustained oscillation. Alternating positive and negative surface salinity anomalies propagated from the western boundary of an ocean domain (their Fig. 10b) and then westward along the northern boundary. The propagation pattern is similar to that represented by our mode 2 + 3 (Fig. 8). Greatbatch and Peterson explained the quasi-decadal self-sustained oscillation in the context of the adjustment process of the thermohaline overturning circulation (TOC), associated with the low frequency form of viscous, baroclinic Kelvin wave propagation (Winton 1996). An ordinary baroclinic Kelvin wave propagates much faster than interdecadal timescales. The realization of the viscous baroclinic Kelvin waves is attributed to weak or nonexistent stratification at high latitudes, which is due to the presence of deep, convective mixing. The weak stratification slows down wave propagation and gives rise to an interdecadal baroclinic Kelvin waves. In response to changes in the strength of the TOC the ocean adjusts through the east–west pressure difference across the basin, which induces the Kelvin wave propagation along the northern and western boundaries and the alternating anomalies advecting eastwards as part of the closed circle. The cause of the oscillation is considered to be originated from perturbations to the western boundary current arising from the southward boundary wave propagation due to the TOC.

The present study examines interdecadal changes in North Atlantic SST anomalies. The important role of the heat flux forcing is suggested at the both multi- and quasi-decadal timescales. The quasi-decadal variation also implies the importance of including other parameters such as salinity, ice, and freshwater anomalies. The ocean’s advective nature may play a key role to determine the timescales in interdecadal variability. Further modeling work coupled to ice and atmosphere would help in understanding the interaction occurring at the climate timescales.

Acknowledgments

This work was supported by National Oceanic and Atmospheric Administration Grant NA57FL0122 and by Office of Naval Research, Secretary of the Navy, Grant N00014-94-1-0369, which provides the base support for the research conducted by Professor James J. O’Brien, Director, the Center for Ocean–Atmospheric Prediction Studies at The Florida State University.

REFERENCES

  • Antonov, J. I., 1993: Linear trends of temperature intermediate and deep layers of the North Atlantic and the North Pacific Oceans:1957–1981. J. Climate,6, 1928–1942.

  • Barnett, T. P., 1983: Interaction of the monsoon and Pacific trade wind system at interannual time scales. Part I: The equatorial zone. Mon. Wea. Rev.,111, 756–773.

  • Cayan, D. R., 1992: Latent and sensible heat flux anomalies over the north oceans: Driving the sea surface temperature. J. Phys. Oceanogr.,22, 859–881.

  • Delworth, T., S. Manabe, and R. J. Stouffer, 1993: Interdecadal variations of the thermohaline circulation in a coupled-atmosphere model. J. Climate,6, 1993–2011.

  • ——, ——, and ——, 1997: Multidecadal climate variability in the Greenland Sea and surrounding regions: A coupled model simulation. Geophys. Res. Lett.,24, 257–260.

  • Deser, C., and M. L. Blackmon, 1993: Surface climate variations over the North Atlantic Ocean during winter: 1900–1989. J. Climate,6, 1743–1353.

  • Dickson, R. R., J. Meincke, S.-A. Malmberg, and A. J. Lee, 1988: The “Great Salinity Anomaly” in the northern North Atlantic 1968–1982. Progress in Oceanography, Vol. 20, Pergamon, 103–151.

  • Greatbatch, R. J., and S. Zhang, 1995: An interdecadal oscillation in an idealized ocean basin forced by constant heat flux. J. Climate,8, 81–91.

  • ——, and K. A. Peterson, 1996: Interdecadal variability and oceanic thermohaline adjustment. J. Geophys. Res.,101 (C9), 20 467–20 482.

  • Griffies, S. M., and E. Tziperman, 1995: A linear thermohaline oscillator driven by stochastic atmospheric forcing. J. Climate,8, 2440–2453.

  • Grötzner, A., M. Latif, and T. P. Barnett, 1998: A decadal climate cycle in the North Atlantic Ocean as simulated by the ECHO coupled GCM. J. Climate,11, 831–847.

  • Häkkinen, S., 1993: An Arctic source for the Great Salinity Anomaly:A simulation of the Arctic Ice-Ocean System for 1955–1975. J. Geophys. Res.,98 (C9), 16 397–16 410.

  • Hansen, D. V., and H. F. Bezdek, 1996: On the nature of decadal anomalies in North Atlantic sea surface temperature. J. Geophys. Res.,101 (C4), 8749–8758.

  • Horel, J., 1984: Complex principal component analysis: Theory and examples. J. Climate Appl. Meteor.,23, 1660–1673.

  • Houghton, R. W., 1996: Surface quasi-decadal fluctuations in the North Atlantic. J. Climate,9, 1363–1373.

  • Kushnir, Y., 1994: Interdecadal variations in North Atlantic sea surface temperature and associated atmospheric conditions. J. Climate,7, 141–157.

  • Latif, M., and T. P. Barnett, 1994: Cause of decadal climate variability over the North Pacific and North America. Science,266, 634–637.

  • Levitus, S., J. I. Antonov, and T. P. Boyer, 1994: Interannual variability of temperature at a depth of 125 meters in the North Atlantic Ocean. Science,266, 96–99.

  • Lozier, M. S., W. B. Owens, and R. G. Curry, 1995: The climatology of the North Atlantic. Progress in Oceanography, Vol. 36, Pergamon, 1–44.

  • Molinari, R. L., D. A. Mayer, J. F. Festa, and H. F. Bezdek, 1997: Multiyear variability in the near-surface temperature structure of the midlatitude western North Atlantic Ocean. J. Geophys. Res.,102 (C2), 3267–3278.

  • Sharp, R. J., 1996: Complex empirical orthogonal function analysis in two dimensional data sets: Theory and examples. Thesis, The Florida State University, 35 pp. [Available as Honors 1519, from Strozier Library Subbasement Archives, The Florida State University, Tallahassee, FL 32306-1005.].

  • Shriver, J. F., M. A. Johnson, and J. J. O’Brien, 1991: Analysis of remotely forced oceanic Rossby waves off California. J. Geophys. Res.,96 (C1), 749–757.

  • Sutton, R. T., and M. R. Allen, 1997: Decadal predictability of North Atlantic sea surface temperature and climate. Nature,388, 563–567.

  • Talley, L. D., 1984: Meridional heat transport in the Pacific Ocean. J. Phys. Oceanogr.,14, 231–241.

  • Weaver, A. J., and E. S. Sarachik, 1991: Evidence for decadal variability in an ocean general circulation model: An advective mechanism. Atmos.–Ocean,29, 197–231.

  • Winton, M., 1996: The role of horizontal boundaries in parameter sensitivity and decadal-scale variability of coarse-resolution ocean general circulation models. J. Phys. Oceanogr.,26, 289–304.

  • Woodruff, S. D., R. J. Slutz, R. L. Jenne, and P. M. Steurer, 1987: A comprehensive ocean–atmosphere data set. Bull. Amer. Meteor. Soc.,68, 1239–1250.

Fig. 1.
Fig. 1.

(a) January SST from climatology. Isobath intervals are 2°C for positive values and 1°C for negative values. (b) Eigenvalue spectrum (in percent of explained variance) of SST anomaly.

Citation: Journal of Physical Oceanography 29, 12; 10.1175/1520-0485(1999)029<3133:MAQDVO>2.0.CO;2

Fig. 2.
Fig. 2.

(a) Spatial amplitude function obtained from the CEOF mode 1 multiplied by 10. (b) SSTA trend of averaged warm event (1950–64) minus cold event (1970–84). The difference fields are based upon the CEOF reconstructions. (c) Spatial phase function obtained from CEOF mode 1. Only the phase from 0° to 80° is shown in order to highlight the region of interest.

Citation: Journal of Physical Oceanography 29, 12; 10.1175/1520-0485(1999)029<3133:MAQDVO>2.0.CO;2

Fig. 3.
Fig. 3.

(a) Temporal functions obtained from CEOF mode 1. Solid line represents the time series of the real part and dotted line is the imaginary part. (b) Dotted line is the temporal phase function obtained from the CEOF mode 1. Solid lines is the best linear fit to the functions; a is a y intersect and b is the slope of the best fit function; y = a + bx represents the linear function. The x axis is a time axis from 1949 to 1990 and the y axis is an argument of the phase functions in degrees. Reciprocal of the slope is a period of one cycle.

Citation: Journal of Physical Oceanography 29, 12; 10.1175/1520-0485(1999)029<3133:MAQDVO>2.0.CO;2

Fig. 4.
Fig. 4.

Chronological sequence of January reconstructed SST anomalies obtained from the CEOF mode 1 (4-yr low-pass filtered). The values are multiplied by 10. Units are in degrees Celsius.

Citation: Journal of Physical Oceanography 29, 12; 10.1175/1520-0485(1999)029<3133:MAQDVO>2.0.CO;2

Fig. 5.
Fig. 5.

Time–latitude plot of SSTA obtained from the CEOF mode 1 and zonally averaged across the North Atlantic between 20° and 80°N. The values are multiplied by 10. Units are in degrees Celsius.

Citation: Journal of Physical Oceanography 29, 12; 10.1175/1520-0485(1999)029<3133:MAQDVO>2.0.CO;2

Fig. 6.
Fig. 6.

Spatial amplitude functions obtained from CEOF mode 2 (a) and mode 3 (b). Spatial phase functions obtained from CEOF mode 2 (c) and mode 3 (d).

Citation: Journal of Physical Oceanography 29, 12; 10.1175/1520-0485(1999)029<3133:MAQDVO>2.0.CO;2

Fig. 7.
Fig. 7.

As in Fig. 3a except for CEOF mode 2 (a) and mode 3 (b). As in Fig. 3b except for CEOF mode 2 (c) and mode 3 (d).

Citation: Journal of Physical Oceanography 29, 12; 10.1175/1520-0485(1999)029<3133:MAQDVO>2.0.CO;2

Fig. 8.
Fig. 8.

Chronological sequence of January reconstructed SST anomalies obtained from CEOF mode 2 + 3 (4-yr low-pass filtered). The values are multiplied by 10. Units are in degrees Celsius.

Citation: Journal of Physical Oceanography 29, 12; 10.1175/1520-0485(1999)029<3133:MAQDVO>2.0.CO;2

Fig. 9.
Fig. 9.

Solid line is the SSTA time series obtained from CEOF mode 2 + 3 averaged over the Labrador Sea region from 48° to 60°N, 50° to 60°W. Superimposed dotted line is the SSTA time series from Houghton (1996).

Citation: Journal of Physical Oceanography 29, 12; 10.1175/1520-0485(1999)029<3133:MAQDVO>2.0.CO;2

Save