Global Modes of Sea Surface Temperature Variability in Relation to Regional Climate Indices

Monique Messié Monterey Bay Aquarium Research Institute, Moss Landing, California

Search for other papers by Monique Messié in
Current site
Google Scholar
PubMed
Close
and
Francisco Chavez Monterey Bay Aquarium Research Institute, Moss Landing, California

Search for other papers by Francisco Chavez in
Current site
Google Scholar
PubMed
Close
Full access

We are aware of a technical issue preventing figures and tables from showing in some newly published articles in the full-text HTML view.
While we are resolving the problem, please use the online PDF version of these articles to view figures and tables.

Abstract

A century-long EOF analysis of global sea surface temperature (SST) was carried out and the first six modes, independent by construction, were found to be associated with well-known regional climate phenomena: the El Niño–Southern Oscillation (ENSO), the Atlantic multidecadal oscillation (AMO), the Pacific decadal oscillation (PDO), the North Pacific Gyre Oscillation (NPGO), El Niño Modoki, and the Atlantic El Niño. Four of the six global modes are dominated by Pacific changes, the other two (M2 and M6) being associated with the AMO and Atlantic El Niño, respectively. The principal component time series of the ENSO (M1) and North Pacific (M3) modes are coherent at time scales >10 yr, and their interaction results in the traditional PDO pattern and the dominant mode of Pacific multidecadal variability. The M3 and PDO time series are well correlated, but the EOFs have different spatial patterns. The fourth mode (M4) has been strengthening since the 1950s and is related to the NPGO but also to El Niño Modoki, especially at the decadal scale. The fifth global mode (M5) is also spatially and temporally correlated to El Niño Modoki. The Pacific SST modes are further related to atmospheric forcing and the circulation of the North Pacific subpolar and subtropical gyres.

Corresponding author address: Monique Messié, Monterey Bay Aquarium Research Institute, 7700 Sandholdt Road, Moss Landing, CA 95039. E-mail: monique@mbari.org

Abstract

A century-long EOF analysis of global sea surface temperature (SST) was carried out and the first six modes, independent by construction, were found to be associated with well-known regional climate phenomena: the El Niño–Southern Oscillation (ENSO), the Atlantic multidecadal oscillation (AMO), the Pacific decadal oscillation (PDO), the North Pacific Gyre Oscillation (NPGO), El Niño Modoki, and the Atlantic El Niño. Four of the six global modes are dominated by Pacific changes, the other two (M2 and M6) being associated with the AMO and Atlantic El Niño, respectively. The principal component time series of the ENSO (M1) and North Pacific (M3) modes are coherent at time scales >10 yr, and their interaction results in the traditional PDO pattern and the dominant mode of Pacific multidecadal variability. The M3 and PDO time series are well correlated, but the EOFs have different spatial patterns. The fourth mode (M4) has been strengthening since the 1950s and is related to the NPGO but also to El Niño Modoki, especially at the decadal scale. The fifth global mode (M5) is also spatially and temporally correlated to El Niño Modoki. The Pacific SST modes are further related to atmospheric forcing and the circulation of the North Pacific subpolar and subtropical gyres.

Corresponding author address: Monique Messié, Monterey Bay Aquarium Research Institute, 7700 Sandholdt Road, Moss Landing, CA 95039. E-mail: monique@mbari.org

1. Introduction

Multiple modes of interannual-to-multidecadal variations in ocean conditions, each with characteristic spatial and temporal patterns, have been recognized from analysis of sea surface temperatures, now available globally for more than 100 years [see Deser et al. (2010) for a review]. The strongest mode of variability is the El Niño–Southern Oscillation (ENSO) phenomenon (Philander 1990; McPhaden et al. 2006), whose intensity is described by the multivariate ENSO index (MEI) (Wolter and Timlin 1993). Modes recognized from regional analysis include the Atlantic multidecadal oscillation (AMO; after Kerr 2000), the Pacific decadal oscillation (PDO; Mantua et al. 1997), and the North Pacific Gyre Oscillation (NPGO; Di Lorenzo et al. 2008). A recently identified central equatorial Pacific warming has been coined El Niño Modoki (Ashok et al. 2007); it has also been called “dateline El Niño” (Larkin and Harrison 2005), “warm pool El Niño” (Kug et al. 2009), “central Pacific El Niño” (Yeh et al. 2009; Kao and Yu 2009; Lee and McPhaden 2010; Yu and Kim 2011), and “type 2 variability” (Yu et al. 2010). The term “El Niño Modoki” is used throughout this paper.

These ocean phenomena are coupled to variations in the atmosphere and characterized by well-established climate, ecosystem, and socioeconomic impacts, with the interannual ENSO the most frequently discussed (Barber and Chavez 1983; Philander 1990; Chavez et al. 2002; McPhaden 2002; Chavez 2005; McPhaden et al. 2006). Longer term cycles are now being recognized, and the PDO was originally defined in an effort to understand multidecadal changes in salmon abundance in the northeast Pacific (Mantua et al. 1997). More recently it has been shown to play a role in anchovy and sardine transitions (Chavez et al. 2003), chlorophyll decadal variability (Martinez et al. 2009), and “regime shifts” (Hare and Mantua 2000; Overland et al. 2008). The NPGO has been associated with variations in nutrients, salinity, and chlorophyll in the California Current and Gulf of Alaska (Di Lorenzo et al. 2009). El Niño Modoki has been linked to monsoon failure and drought in India (Kumar et al. 2006) and to tropical cyclone frequency (Chen and Tam 2010).

The numerical indices that define these modes of climate variability are often computed over regional scales, and as a result their full geographical extent is unclear: it is not fully understood if they are independent or over what spatial scales they can be recognized. In this paper a global empirical orthogonal function (EOF) analysis of SST is used to investigate these relations. EOFs separate the spatiotemporal SST signal into a sum of orthogonal modes by maximizing the variance explained by each mode. This technique has been widely used regionally in a given ocean basin; many of the indices listed above are computed using EOFs (MEI, PDO, NPGO). Previous EOF studies have characterized SST variability in the World Ocean using a variety of filters, including time series normalization (Kawamura 1994), bandpassing (Enfield and Mestas-Nuñez 1999, 2000), mode rotation (Mestas-Nuñez and Enfield 1999), or separation of seasons (Yasunaka and Hanawa 2005). Here pre- and postprocessing are kept to a minimum so that the nonseasonal modes are as close as possible to the observed SST.

The primary goal of this paper is to first identify global modes of SST and then relate them to regional climate phenomena. A second focus is on the interrelationships of the Pacific-dominated modes. The paper is organized as follows. Section 2 describes the datasets, the regional indices, and the analytical methods. Section 3 presents the global, ocean basin and coupled ocean–atmosphere EOF analysis, and then relates the results to the regional indices. Section 4 explores the interaction between global modes 1 (ENSO) and 3 (North Pacific) and how they interact to generate the dominant mode of Pacific multidecadal variability (PDO). Section 5 describes the links between the NPGO, El Niño Modoki, and the fourth and fifth global EOF modes. Section 5 discusses a hypothesis regarding the mode impacts on circulation of subpolar and subtropical gyres. Concluding remarks are given in section 6.

2. Data and methods

a. SST, sea surface height, sea level pressure, and regional indices

The Extended Reconstruction Sea Surface Temperature (ERSST) version v3b product, available online (www.ncdc.noaa.gov/oa/climate/research/sst/ersstv3.php) and based on the International Comprehensive Ocean–Atmosphere Data Set (ICOADS) release 2.4, forms the basis of the analysis. The resolution is 2° × 2° monthly, and data prior to 1875 and poleward of 65° latitude were not considered. The ERSST v3 product is described in Smith et al. (2008), v3b is identical but without satellite data (their use introduced a small residual cold bias because of cloud-contaminated data). A visual comparison with the higher-resolution Met Office Hadley Centre Sea Ice and Sea Surface Temperature (HadISST) dataset used in Deser et al. (2010) indicates that both datasets give essentially the same time series and patterns (see also Smith and Reynolds 2003). Available data varies widely as a function of time and area (Deser et al. 2010, their Fig. 3), being sparse in earlier years and in the Southern Ocean. As a result the earliest part of the record, and most particularly the Southern Hemisphere, should be considered with caution.

Sea level pressure (SLP) is the Hadley Centre’s mean SLP dataset (HadSLP2) updated with the near-real-time product (HadSLP2r) (Allan and Ansell 2006). Monthly data from 1910 to 2009 with a resolution of 5° × 5° were used. Sea level anomalies (SLAs) are the Segment Sol multimissions d’Altimétrie, d’Orbitographie et de localisation précise/Data Unification and Altimeter Combination System (SSALTO/DUACS) merged reference product DT-2010 (version 3.0.0) provided by AVISO (France). Weekly (averaged monthly) data from October 1992 to December 2009 with a resolution of 1/3° × 1/3° were used. The area corresponding to each pixel decreases poleward in degree grids, so to avoid a bias toward high latitudes all datasets were regridded to a 100 km × 100 km grid using linear interpolation.

Regional numerical indices were downloaded via the Internet or computed (Table 1). They include the MEI, AMO, PDO, NPGO, El Niño Modoki index (EMI), Atlantic El Niño index (ATL3), and North Pacific Oscillation (NPO).

Table 1.

Compilation of regional climate indices and their sources, SSH: sea surface height.

Table 1.

b. Removal of the centennial warming signal

The global-mean SST is characterized by a decreasing trend before 1910, increasing thereafter (0.66°C century−1, Fig. 1), so the analysis was restricted to the 1910–2009 period to simplify statistical treatment of the trend and to avoid periods of sparse data coverage. To determine the relations between the century-scale warming and interannual-to-multidecadal variations, the EOF analysis was carried out on global time series with 1) no trend removal, 2) removal of the global mean (Fig. 1 shaded colors; Zhang et al. 1997; Mantua et al. 1997; Chavez and Messié 2009), 3) removal of a linear fit to the global mean (Fig. 1 black line; Chavez et al. 2011), and 4) removal of the local linear trend at each pixel (Enfield and Mestas-Nuñez 1999; Yeh and Kirtman 2005; Deser et al. 2010). The first, method 1 (not shown), produced complicated patterns in the first two modes (M1 and M2), where the long-term trend was co-mingled with ENSO with the following modes showing relations to regional climate indices. A “secular trend” has been reported as a distinct mode (Enfield and Mestas-Nuñez 2000; Guan and Nigam 2008), but some of temporal and spatial characteristics of the trend include higher frequency variability, and the constraint of orthogonality in the EOF calculation impacts other modes. Methods 2–4 resulted in the same identifiable modes but each has different caveats. Zhang et al. (1997) and Mantua et al. (1997) first subtracted the time series of global-mean SST at each pixel (Fig. 1) to calculate the PDO index (method 2). The global-mean SST, however, contains signatures of natural variability [see Ting et al. (2009); Wang and Dong (2010) for the AMO imprint], so its removal may bias the results. Method 3 is an improvement of method 2 as it only removes a linear fit, but results in nonseparable third and fourth modes (M3 and M4) and fifth and sixth modes (M5 and M6) [following the “rule of thumb” test described by North et al. (1982)]. Moreover, both methods 2 and 3 remove the same signal at each pixel, assuming that global long-term forcing processes, such as global warming, have a homogeneous distribution when studies have shown that this is unlikely (Xie et al. 2010). Method 4 can produce local biases as a result of natural low-frequency variability [see Fig. 4a of Chavez et al. (2011) for a map showing pixel-by-pixel trends], but it is used here because this bias is mostly restricted to the North Atlantic and this method clearly separates the modes (see following section).

Fig. 1.
Fig. 1.

Time series of monthly globally averaged SST with the global seasonal cycle removed. There is a decreasing SST trend from 1875 to 1910 and thereafter the general trend is increasing. The SST time series were restricted to the 1910–2009 period prior to any analysis (variance, local trend, and seasonal cycle removal; EOF analysis; etc.).

Citation: Journal of Climate 24, 16; 10.1175/2011JCLI3941.1

c. EOF calculations

After removal of the local linear trends, the seasonal cycle (computed by averaging each month at each pixel over the 1910–2009 period) was removed to obtain monthly SST anomalies. The SST pixel anomaly time series were not normalized (so that regions with higher variance logically have higher weight on the analysis), smoothed, or filtered. EOFs were then computed globally, in each of the ocean basins and regionally (Fig. 2). Each mode is composed of a spatial pattern (the so-called EOF) and a principal component (PC) time series that represents the temporal evolution of the EOF pattern. A given mode can be reconstructed by multiplying the EOF (space) by its PC (time).

Fig. 2.
Fig. 2.

(a) Variance of SST anomalies for the 1910–2009 period. The linear trend and seasonal cycle at each pixel were removed prior to the analysis. (b) Time series spatial coherence estimated from SST anomalies. At each pixel, the corresponding time series was correlated with its neighbors on expanding concentric squares (within the same ocean). The size of the squares was increased by including the following neighbor; this procedure continued until a pixel with less than 50% variance explained by the initial time series was found. The corresponding rank (number of squares before finding such a pixel) is high when coherence is strong. Black lines mark the boundaries for the ocean basin analysis reported in Table 2.

Citation: Journal of Climate 24, 16; 10.1175/2011JCLI3941.1

The significance of the modes was evaluated by computing their sampling error as λ (2/N)1/2, where λ is a given eigenvalue and N the number of realizations, here 1200 (North et al. 1982). The result (Fig. 3a) indicates that the first six modes are clearly separated, hence statistically well defined. The data were also subsampled spatially and temporally, the EOFs computed on 200 random subsets (perturbed EOFs) and then compared with the original modes (master EOFs) using the coefficient of congruence, similar to a correlation coefficient but better suited to vector comparisons because it does not remove the respective means (Cheng et al. 1995). When subsetting the dataset temporally (retaining 50% of the original time series) more than 80% of the perturbed EOFs (modes 1–5, it is 52% for M6) give a coefficient of congruence with the corresponding master EOF higher than 0.9. A subset of the results (comparison of the perturbed EOFs with the master M4 and M5 EOFs) is shown in Fig. 3b. The third mode markers align along the M4 axis but remain clearly separated from the fourth mode markers, indicating that M3 and M4 are separated even if the perturbed EOFs incorporate part of M4 in their third mode. Similar projections on the other master EOFs (not shown) confirmed the separation of all modes. When subsetting the dataset spatially (retaining 50% or 20% of the original pixels), the perturbed EOFs reproduce the modes even better than the temporal subsetting (Figs. 3c,d).

Fig. 3.
Fig. 3.

Statistical tests performed on the modes resulting from the EOF analysis: (a) eigenvalue separation following the rule of thumb proposed by North et al. (1982), (b) projection of the first six modes of 200 subsamples, each representing 50% of the time series, on the M4 and M5 master EOFs, (c) as in (b) but keeping 50% of the pixels, (d) as in (c) but keeping 20% of the pixels. The projection of a given perturbed mode P (any marker) on a master mode M (here M4, x axis, and M5, y axis) is realized using the coefficient of congruence: , where pi and mi are the loadings of vectors P and M (Cheng et al. 1995).

Citation: Journal of Climate 24, 16; 10.1175/2011JCLI3941.1

A combined SLP/SST EOF analysis, as in Di Lorenzo et al. (2010, manuscript submitted to Geophys. Res. Lett.), was performed to identify coupled atmosphere–ocean modes. The SST and SLP datasets were merged into one “big map” after being identically processed (removal of local linear trends and seasonal cycle) and then normalized (domain-average standard deviation for SST; standard deviation at each latitude for SLP). The method used here includes longer time series and no smoothing.

d. Wavelet analysis

Wavelet analysis, cross wavelets, and wavelet coherence were performed on the PCs to further investigate their frequency and coherence. Wavelet analysis decomposes a signal into time–frequency space to highlight the dominant scales of variability and period. It can be seen as a time-varying Fourier analysis, describing the energy not only as a function of frequency but also of time (see Torrence and Compo 1998). Wavelet analysis has been widely used in oceanography, in particular to study El Niño, El Niño Modoki, or the PDO (Wang and Wang 1996; Torrence and Compo 1998; Torrence and Webster 1999; An and Wang 2000; Newman et al. 2003; Wang et al. 2009; Kao and Yu 2009). The most commonly used wavelet function is the complex Morlet wavelet, but here a real derivative of a Gaussian (DOG) wavelet is used (Torrence and Compo 1998). The localization in spectral space is not as good as what is obtained with complex functions; however, the DOG wavelet temporally locates shifts from positive to negative anomalies (and vice versa) for each period and resolves longer time periods. Cross wavelets extract the common power between two time series, and wavelet coherence finds correlations between the time series in time–frequency space (Torrence and Webster 1999; Grinsted et al. 2004).

3. Global analysis of nonseasonal SST variability

a. SST variance and spatial coherence

The intensity and spatial coherence of the SST anomalies is shown in Fig. 2. The variance of the time series highlights areas characterized by strong nonseasonal variability, in particular the central/eastern equatorial Pacific, the North Pacific along the Kuroshio Extension, and the North Atlantic along the Gulf Stream (Fig. 2a). This map is complementary to Fig. 4 in Deser et al. (2010), which highlights similar regions, computed from shorter, higher-resolution time series. The time series spatial coherence map (how time series are similar to their neighbors, Fig. 2b) indicates that the tropics are generally more coherent than high-latitude regions. The most coherent region is the equatorial Pacific, with ranks higher than 8 (i.e., 800 km) over a large area.

b. Global SST modes

EOFs calculated globally yield six modes that together explain 38.1% of the variance of SST anomalies (Fig. 4: spatial patterns or EOFs and Fig. 5: associated time series or PCs). Highly variable regions (Fig. 2a) dominate the analysis, given the lack of normalization, and the six modes have strongest signatures in the equatorial band and the Northern Hemisphere. The general lack of coherence in SST anomalies (Fig. 2b) may help explain why the variance explained by M1, associated with the strongly coherent central/eastern equatorial Pacific region, is high but the variance explained by the following modes is low (<6%).

Fig. 4.
Fig. 4.

Global EOF spatial patterns of the first six SST modes calculated for the 1910–2009 period.

Citation: Journal of Climate 24, 16; 10.1175/2011JCLI3941.1

Fig. 5.
Fig. 5.

Time series (1910–2009) of the principal components (PCs) associated with the first six global EOF modes. Time series of well-correlated regional indices are superimposed. The NPGO index (Table 1) has only been calculated from 1950 forward, is negatively correlated with M4, and has been inverted. The correlation between the EMI and M4 increases to 0.54 when computed over the same period as the NPGO and drops to 0.15 before. The correlation between the EMI and M5 remains high for the entire time series (0.51 before 1950, 0.62 after). The correlation between M6 and ATL3 increases to 0.56 when ATL3 is detrended over the 1910–2009 period.

Citation: Journal of Climate 24, 16; 10.1175/2011JCLI3941.1

The leading mode in the SST anomalies (M1, 18.1%) explains greater than three times more variance than any other mode. The EOF displays the well-known ENSO pattern (Fig. 4a), with most of the variance in the central/eastern equatorial Pacific (anomalies as high as 3°C, variance explained as high as 80%). The corresponding PC (Fig. 5a) is dominated by interannual variability and highly correlated with the MEI (r = 0.95). El Niño, anomalous warming in the eastern Pacific, occurs when M1 is much greater than 0 (maximum in 1997/98). At the same time the western tropical Indian Ocean warms to a lesser extent and the western Pacific cools (Fig. 4a). La Niña is the counterpart (M1 much less than 0). This ENSO pattern has been described previously using EOFs with varying data processing methods, different datasets, periods, and spatial regions (Weare et al. 1976; Kawamura 1994; Deser and Blackmon 1995; Tourre and White 1995; Enfield and Mestas-Nuñez 1999, 2000; Yasunaka and Hanawa 2005; Kao and Yu 2009; Deser et al. 2010). The results are very similar, illustrating the strength and coherence of the ENSO signal.

The second mode (M2, 5.3%) is strongest in the northern Atlantic (Fig. 4b). The associated PC displays the lowest frequency of the first six modes and is well correlated with the AMO (Fig. 5b, r = 0.80). The EOF also resembles the AMO pattern determined from regression of SST anomalies on the AMO index (Deser et al. 2010, their Fig. 11a) or from seasonal Atlantic EOFs (Guan and Nigam 2009). Global analysis conducted by Enfield and Mestas-Nuñez (1999, 2000) and Mestas-Nuñez and Enfield (1999) also identified a mode related to Atlantic multidecadal variability, but their spatial pattern is different, with more emphasis on the equatorial and North Pacific, whereas the present EOF is clearly dominated by the North Atlantic. The relatively strong positive phase in recent years, visible in Fig. 5b, is likely to be an artifact of the local detrending of the time series (Ting et al. 2009). Removing the global linear trend (black line in Fig. 1) at each pixel instead of the local trend yields a second mode whose PC is closer to the “real” AMO, that is, a very moderate warm phase in recent years (Ting et al. 2009; Wang and Dong 2010).

The PC of the third mode (M3, 4.7%) is most strongly correlated with the PDO (Fig. 5c, r = 0.56). The EOF pattern (Fig. 4c), however, differs from the traditional PDO, which has an ENSO character but is stronger at higher latitudes and weaker in the tropics (Mantua et al. 1997). The traditional PDO–ENSO similarity is not surprising, given the influence of ENSO on North Pacific SST and the PDO (Schneider and Cornuelle 2005). Here M3 is independent of M1 (ENSO). The resulting pattern is a tripole with the western North Pacific (~40°N) and the equatorial Pacific in phase with each other and out of phase with the northeast Pacific. This spatial pattern is similar to the intrinsic North Pacific mode identified by An and Wang (2005), the “ENSO leftover mode” identified by Vimont (2005), and the second non-ENSO rotated mode (R2) of Mestas-Nuñez and Enfield (1999); however, this last study lacks the eastern equatorial Pacific negative anomaly. M3 is also likely the North Pacific mode described in Deser and Blackmon (1995) and the coupled mode described by Latif and Barnett (1994, 1996) and Barnett et al. (1999b).

The fourth mode (M4, 4.0%) is correlated with both the NPGO and EMI time series (Fig. 5d) and exhibits an increase in variance with time. Highest loadings are observed in the North Pacific where M4 resembles the second EOF of SST for this region [the Victoria mode, described by Bond et al. (2003)] and the NPGO SST pattern, although the latter is stronger at midlatitudes (Di Lorenzo et al. 2008). M4 is strongest in the Gulf of Alaska, as is the sea-level-based NPGO (Di Lorenzo et al. 2008). M4 SST anomalies extend to the U.S./Canada coast and in a horseshoe pattern down to the central equatorial Pacific where the spatial distribution is similar to El Niño Modoki [warming in the central Pacific flanked by cooler anomalies to the west and east, Ashok et al. (2007)]. This mode could also be associated spatially and temporally to the fifth non-ENSO rotated mode (R5) of Mestas-Nuñez and Enfield (1999).

The fifth mode (M5, 3.3%) is also strongly correlated with the EMI (Fig. 5e, r = 0.57). The highest loadings are observed in the tropical Pacific and bear the imprint of El Niño Modoki (Fig. 4e). There is a concurrent warming in the southern Pacific Ocean (south of 40°S around 150°W), a region where a record warming was observed during the 2009/10 El Niño Modoki (Lee et al. 2010). The pattern is retained without 2009 in the analysis, indicating a recurring coupling between El Niño Modoki and warming in the South Pacific.

The sixth mode (M6, 2.8%) is characterized by a warming in the equatorial Atlantic (Fig. 4f), which resembles El Niño (Philander 1986) and shows up as the first mode of equatorial Atlantic variability (Deser et al. 2010, their Fig. 8a). This Atlantic El Niño mode has been described previously (e.g., Ruiz-Barradas et al. 2000; Chang et al. 2006) and is captured by the ATL3 index (Zebiak 1993) correlated with the M6 PC (Fig. 5f). It is temporally and spatially consistent with the sixth non-ENSO rotated mode (R6) of Mestas-Nuñez and Enfield (1999). The time series of the six modes can be found online (www.mbari.org/bog/GlobalModes/Indices.htm).

c. Scales of variability in the dominant modes

The wavelet analysis highlights the dominant scales of variability and period (Fig. 6, see the caption for an interpretation of the wavelets). The global power spectra show that all PC time series except M1 exhibit significant low-frequency variability (periods > 15 yr). The results obtained here are qualitatively similar to those obtained from the classical Morlet wavelet [e.g., the ENSO analysis in Torrence and Compo (1998)].

Fig. 6.
Fig. 6.

Wavelet analysis of the PC time series associated with the first six global EOF modes: (left) wavelet power spectrum and (right) corresponding global power spectrum (calculated as time-averaged wavelet power spectrum). The time series were normalized to unit variance prior to analysis so that the power is relative (value of 1 = white noise level). The wavelet function is DOG, m = 2 (classical “Mexican hat,” see Torrence and Compo 1998). Color patches, wider as the period increases, describe anomalies (either positive or negative) in the time-varying wavelet power spectra. Troughs (white) correspond to shifts from positive to negative anomalies (or inversely), and color (power or energy) is a function of signal intensity. For each patch, the period corresponding to the most intense color roughly gives the oscillation period. (The DOG wavelet is not as precise as the Morlet wavelet in spectral localization, Torrence and Compo 1998.) As an example, for M1/ENSO (a) each patch in the 2–8 band represents El Niño or La Niña conditions, the oscillation period varying with time. Strong events such as the 1982–83 El Niño, 1987–88 and 1988–89 El Niño/La Niña, and 1997–98 and 1998/99 El Niño/La Niña exhibit a strong signal at short time scales (minimum 2 yr here). The black contours (wavelet power spectrum) and blue dashed line (global power spectrum) represent the 95% significance level for a red noise background (lag 1 autocorrelation = 0.72). The cone of influence (COI), representing possible edge effects, is superimposed on the wavelets plots. Wavelet software was provided by C. Torrence and G. Compo (http://paos.colorado.edu/research/wavelets/).

Citation: Journal of Climate 24, 16; 10.1175/2011JCLI3941.1

The M1 wavelet analysis is consistent with ENSO having a period of 2–8 yr (Trenberth 1976), with the red patches in this time scale associated with El Niño or La Niña conditions. Evident are fluctuations in the number, strength, and period of El Niño/La Niña events following the 1976/77 regime shift from 2–4 to 4–6 yr (An and Wang 2000). There are events with large amplitude after ~1965 (the oscillation period increasing with time) and fewer events between 1920 and 1960 (Torrence and Compo 1998). Studies have described the lower frequency modulation of ENSO (Torrence and Webster 1999; Yeh and Kirtman 2005; Wang et al. 2009), and lower frequency variations are apparent in the M1 wavelet power spectrum but remain below the red-noise background (Fig. 6a).

The M2 is strongly multidecadal, dominated by variations above ~20 yr, and changed sign in the mid-1920s, early 1960s, and mid-1990s (troughs in Fig. 6b), as seen directly from the time series (Fig. 5b). The M3 (Fig. 6c) and M1 (Fig. 6a) wavelet power spectra present similarities, particularly at multidecadal time scales. M4 is characterized by an increase in energy with time, in particular since 1990 at the decadal scale (Fig. 6d).

d. Ocean basin analysis

EOF analysis was performed for each of the major ocean basins to hone in on regional impacts (Table 2). The M1, M3, M4, and M5 dominate variability in the Pacific, while M2 and M6 dominate in the Atlantic (Fig. 4). M1 and M3 show up as the first and second modes in the Pacific basin analysis (M4 explains 84.2% of the third mode, not shown). M1 dominates equatorial Pacific variability (also impacted by M5) and to a lesser extent the North and South Pacific; M3 and M4 are most active in the North Pacific, less in the equatorial Pacific and even less in the South Pacific, suggesting a fundamental relationship with the North Pacific. As expected the first and second modes of variability in the North Pacific are the PDO [Mantua et al. (1997); r = 0.86, not unity given differences in methodology) and the Victoria mode/NPGO (Bond et al. 2003; Di Lorenzo et al. 2008; r = 0.55 with the NPGO). The first mode (PDO) is a combination of M1 (36%) and M3 (~52%).

Table 2.

Percentage of variance of the first two EOF modes calculated regionally or by ocean basin explained by each of the first six global EOF modes. North Pacific/Atlantic: north of 20°N, equatorial region: 20°S–20°N, South Pacific: south of 20°S (see Fig. 2). Bold values are greater than 25% variance explained.

Table 2.

In the Indian Ocean the first mode is linked to ENSO [see also the corresponding spatial pattern in Deser et al. (2010), their Fig. 7a, left]; when computed over the equatorial Indian only the variance explained by M1 reaches 41.0% when M1 leads by 3 months.

In the Atlantic, the first mode is dominated by M2; the second by M6. This order is reversed in the equatorial Atlantic where M6 dominates. Consistent with studies of the Atlantic El Niño (e.g., Ruiz-Barradas et al. 2000; Chang et al. 2006), no link was found between M6 and M1 (ENSO). No strong connections between the Pacific and the Atlantic were evident (little or no signature of M2 and M6 in the Pacific nor M1, M3, M4, and M5 in the Atlantic).

e. Atmospheric modes coupled to M1, M3, and M4

The first mode of the combined SST–SLP EOF analysis (not shown) yields a pattern associated with ENSO (Di Lorenzo et al. 2010, manuscript submitted to Geophys. Res. Lett.). The relation between El Niño and its atmospheric counterpart, the Southern Oscillation, has been described repeatedly (e.g., Philander 1990; Neelin et al. 1998). It involves an east–west seesaw SLP and changes in the strength of the Pacific trades. ENSO can influence higher latitudes through atmospheric teleconnections, and a link to the Aleutian low has been established (Alexander et al. 2002; Vimont 2005).

In the second coupled mode described in Di Lorenzo et al. (2010, manuscript submitted to Geophys. Res. Lett.), the SLP pattern is associated with the North Pacific Oscillation (NPO) and the SST pattern with the NPGO. Their Fig. 1 is very similar to the M4 spatial pattern in the Pacific (Fig. 4d). The relation between the NPO and NPGO has previously been reported (Di Lorenzo et al. 2008; Chhak et al. 2009). The NPO is characterized by concurrent changes in the Aleutian low and North Pacific high, which strengthen (weaken) (but do not move) in phase, increasing (decreasing) the gradients between them (Walker and Bliss 1932).

The third coupled mode is linked to M3 by the spatial SST pattern (Fig. 7b) and a PC that is well correlated with the M3 PC (r = 0.65). The corresponding SLP pattern (Fig. 7a) shows an intensification and southward shift of the Aleutian low and a weakening of the North Pacific high: negative anomalies in the North Pacific, strongest south of the average Aleutian low position (~50°N; e.g., Rodionov et al. 2005). Performing EOFs on the SLP field alone yields very similar EOFs, and lag correlations between the corresponding PC (third mode of SLP) and M3 are highest when SLP leads by a month (not shown), consistent with atmospheric forcing of M3 (An and Wang 2005). The SLP pattern is similar to one described by Mestas-Nuñez and Enfield (1999) and is associated with their R2 SST mode.

Fig. 7.
Fig. 7.

Third EOF mode of coupled SLP–SST anomalies for the Pacific and Indian Oceans: (a) SLP and (b) SST, obtained by performing an EOF analysis on the joined SLP–SST fields [following Di Lorenzo et al. (2010), manuscript submitted to Geophys. Res. Lett.].

Citation: Journal of Climate 24, 16; 10.1175/2011JCLI3941.1

4. M1, M3, and the PDO

a. The PDO as a combination of M1 and M3

The M1 and M3 spectral signatures present similarities at low frequencies (Figs. 6a,c), highlighted by the cross-wavelet analysis (Fig. 8a). PCs are independent by construction (orthogonal, i.e., null correlation overall) but can be correlated at a given frequency or during specific periods. There is common power and strong coherence between the M1 and M3 PCs at decadal-to-multidecadal time scales over the entire period. Hence, M1 and M3 combine to generate the characteristic multidecadal variability associated with the PDO. The interaction is strongest in the North Pacific where they explain 88% of the first mode (by definition, the PDO; Table 2). M1 explains 29% and M3 explains 32% of the traditional PDO index. In both cases, M3 contributes a greater proportion of the variance.

Fig. 8.
Fig. 8.

Cross-wavelet analysis of the (a) M1 and M3 PC time series and (b) M4 and M5 PC time series: (left) cross-wavelet power spectrum (black contours: 8, 16, 32, 48, and 64) superimposed on wavelet coherence (colors, only plotted when the common power is ≥2) and (right) global cross-power spectrum averaged over time. A coherence close to 1 (−1) indicates that the time series are strongly correlated (anticorrelated) at a given time and frequency. COI as in Fig. 6. Shifts visible in the time series and/or reported in the literature are highlighted (gray dashed bars): (a) around 1924–25, 1945, 1976–77, and 1997–98 for M1 and M3; (b) 1971 for M4 and M5. The same cross-wavelet analysis performed between other PCs showed no consistent coherence between the time series. Cross-wavelet and wavelet coherence software was provided by A. Grinsted (www.pol.ac.uk/home/research/waveletcoherence/).

Citation: Journal of Climate 24, 16; 10.1175/2011JCLI3941.1

Moreover, the spatial pattern of the traditional PDO is a combination of the M1 and M3 patterns (Figs. 4a,c): the negative anomaly at the equator in M3 balances the positive anomaly in ENSO at low frequencies, leading to a pattern similar to M1 but with a weaker signature along the equator. The PDO multidecadal spatial pattern is evident in the first global EOF of SST when only periodicities >8 yr are considered (Fig. 9). Mestas-Nuñez and Enfield (1999) arrive at this pattern directly (their R2) because they remove the high-frequency ENSO variability, leaving the lower frequency component, prior to analysis.

Fig. 9.
Fig. 9.

First global EOF mode of SST anomalies calculated when periodicities shorter than 8 yr have been removed with a Fourier transform prior to analysis: (top) PC and (bottom) EOF. (Note that the color bar is different from that in Fig. 4.) The time series of the PDO index is plotted with the PC time series (r = 0.48), which is also significantly correlated with M1 (r = 0.43) and M3 (r = 0.42) but not with M2, M4, M5, or M6 (r < 0.1). Gray bars around 1924/25, 1945, 1976/77, and 1997/98 are as in Fig. 8a.

Citation: Journal of Climate 24, 16; 10.1175/2011JCLI3941.1

The PDO has been described either as a reddened response to ENSO (Newman et al. 2003; Deser et al. 2004; An et al. 2007; Shakun and Shaman 2009), an independent mode centered in the North Pacific (Latif and Barnett 1996; Barnett et al. 1999b), or a sum of several modes (Schneider and Cornuelle 2005; Deser et al. 2010). Shakun and Shaman (2009) argued that the North and South Pacific present similar multidecadal variability and concluded that ENSO was the primary driver. Strong similarities in the influence of ENSO on the North and South Pacific are observed (36% and 49%, respectively), but the influence of M3 is restricted to the North Pacific (Table 2). The PDO is described here not as a single dynamical mode but as a combination of the ENSO equatorial Pacific M1 mode and the North Pacific M3 mode (intrinsic North Pacific variability), following Schneider and Cornuelle (2005).

b. Regime shifts

When applied to PCs (representative of large-scale variability), wavelet analysis can also be used to identify “shifts” in a dataset, defined as “sudden,” “high amplitude,” “low frequency,” and “large spatial scale” (Lees et al. 2006), hence corresponding to troughs spanning high and low frequency scales (abrupt and long-lasting change). Shifts can be identified in the M1 and M3 cross-wavelet analysis around 1925, 1947, 1976, and 1998 (Fig. 8a, gray dashed lines). These shifts are also evident in the low frequency EOF (Fig. 9) and in the M1 and M3 wavelet power spectra (Figs. 6a,c). The 1976/77 shift has been documented for ENSO (Fedorov and Philander 2000; An and Wang 2000; Deser et al. 2004); however, the 1925 and 1947 shifts are not as clear (Deser et al. 2004), although a change in ENSO variance has been reported for the 1920s (Torrence and Compo 1998; Torrence and Webster 1999; Yeh and Kirtman 2005). These fluctuations are consistent with PDO transitions around 1925 to a warm regime, in the mid-1940s to a cool one, and in 1976/77 to warm once again (Zhang et al. 1997; Mantua et al. 1997; Francis et al. 1998; Mantua and Hare 2002; Deser et al. 2004; Miller et al. 1994; Breaker 2007). A cool transition has been reported around 1998 (Chavez et al. 2003; Peterson and Schwing 2003; Overland et al. 2008; Chavez et al. 2011). The positive/negative phases not only coincide with warmer/cooler temperatures but also correlate with profound changes in ecosystem structure (e.g., sardine/anchovy dominance, Chavez et al. 2003) and, hence, correspond to so-called regime shifts (deYoung et al. 2004; Bakun 2005; Lees et al. 2006; Overland et al. 2008; and references above).

There is no signature of the 1988/89 regime shift (Hare and Mantua 2000; Overland et al. 2008) in the M1 and M3 cross-wavelet power spectrum (Fig. 8a), although it could represent a change in slope rather than sign in the time series (Chavez et al. 2003). There is a shift in M3 (visible at periods < 15 yr only, Fig. 6c), suggesting that changes in M3 that are independent of M1 are weaker and restricted to the North Pacific, while the larger shifts visible in Figs. 8a, 9 have wider impacts and are driven by the combined effects of M1 and M3.

c. The M1 + M3 variability in the North Pacific

M1 and M3 are statistically independent; however, lagged correlations between them are weak but significant (r ~ 0.27, p < 0.001 when M1 leads M3 by 6 months to a year). This lag and the M3 spatial pattern in the North Pacific are consistent with anomalies generated by ENSO the previous winter (Fig. 4a) resurfacing a year later (“reemergence mechanism,” Alexander et al. 2002). The pattern is also consistent with a coincident negative anomaly in the equatorial band [see Fig. 10 in Alexander et al. (2002), observed DJ(1/2) is close to Fig. 4c]. This lag and equatorial negative anomaly are part of the independent ENSO leftover mode identified by Vimont (2005). This mode corresponds to an “overshoot” of the equatorial thermocline after the El Niño peak and persists from spring to summer (Vimont 2005). This equatorial anomaly is much reduced in the coupled SLP–SST mode (Fig. 7b) compared to the “pure” SST mode (Fig. 4c), which indicates that its origin is not related to the Aleutian low variability (Fig. 7a).

ENSO (M1) impacts the strength and location of the Aleutian low (Bjerknes 1969) via heat flux anomalies associated with the atmospheric bridge (Alexander et al. 2002; Vimont 2005). This impact is then thought to generate decadal-to-multidecadal variability in the North Pacific (M3), described in more detail below. The largest anomalies in SST and ocean circulation associated with M3 are in the region of the Kuroshio–Oyashio Extension (KOE), yet the largest changes in sea level pressure associated with the SST variability are located in the central/eastern Pacific (Fig. 7a). The relationship between the atmosphere and ocean has been reported to be via anomalies in wind stress curl near 160°W (strongest signal in both SLP and SST, Fig. 7) that generate westward propagating Rossby waves (Deser et al. 1999; Miller and Schneider 2000; Seager et al. 2001; Schneider et al. 2002; Qiu 2003; Qiu and Chen 2010). The theoretical time scale for the KOE to adjust to the wind-anomaly-generated Rossby waves is about 3–5 yr, leading to decadal-scale variability. However, the lag estimated here between the independent SLP and SST modes is only one month. An and Wang (2005) obtained the same 1-month lag for their North Pacific mode and proposed that “immediate” atmospheric impact on SST can be achieved through changes in surface heat flux (although weaker, there is an atmospheric signal above the KOE, Fig. 7a). This direct atmospheric forcing is strongest above the central and eastern North Pacific (Miller and Schneider 2000). As the KOE adjusts to the delayed Rossby wave forcing, positive feedback to the atmosphere (Seager et al. 2001; Schneider et al. 2002) would cause the SST and SLP modes to evolve mostly in phase.

The analysis suggests that changes in M3 may be initiated by ENSO via the Aleutian low and delayed teleconnections (Bjerknes 1969; Alexander et al. 2002). The Aleutian low perturbations lead to a lagged KOE/gyre circulation response (see section 6), and positive feedback in the KOE region expands the ENSO interannual forcing to longer time scales (Miller et al. 2004) by initiating a longer-term resonance in M3 that leads to the observed multidecadal variability in the PDO (or the combined M1 + M3 mode, Table 2 and Fig. 9). In support, tropical SST variability and independent processes along the KOE have been both previously linked to the interannual-to-multidecadal scales of the PDO (Nakamura et al. 1997; Schneider and Cornuelle 2005; Kwon et al. 2010). Clearly every ENSO cannot trigger this response, and the maintenance of the M3 state may be a function of the frequency and intensity of ENSO. M1 also contains significant variability at the decadal-to-multidecadal time scale (Fig. 6a). It is not clear if this lower frequency variability is reflected in M3 or if the M3 lower frequency resonance drives this scale of variability in M1 (Barnett et al. 1999a; Pierce et al. 2000).

5. M4, M5, the NPGO, and El Niño Modoki

a. Atmospheric coupling and mechanisms

Reports on El Niño Modoki and the NPGO have been growing steadily over the last decade; however, their mechanisms, forcing, and continued evolution remain poorly understood. It has only been recently established that these two modes are related (Di Lorenzo et al. 2010; Chavez et al. 2011): here both are associated with M4. M5, on the other hand, is only associated with El Niño Modoki (section 3b). When the two PCs are added, the correlation with the EMI increases to 0.82 for the 1960–2009 period (compared to 0.60 and 0.65 for M4 and M5, respectively). The coupled ocean–atmosphere analysis indicates that M4 is associated with the NPO, while no coupled SLP–SST pattern emerged that could be identified with M5. This suggests that part of the El Niño Modoki variability is linked to M4 (and the northeast Pacific and the NPGO, possibly via atmospheric teleconnections), while part of the variability is independent (M5).

Conflicting scenarios have been proposed for the relation between North Pacific variability and El Niño Modoki. On one hand, Yu et al. (2010) found that El Niño Modoki first develops in the northeastern subtropical Pacific, then extends southwestward toward the central equatorial Pacific, consistent with the M4 pattern and an origin of M4 in the northeast Pacific. Yu and Kim (2011) show that NPO extratropical SLP variations (linked to M4) are capable of exciting El Niño Modoki via surface heat flux anomalies that subsist after the NPO peak (seasonal footprinting mechanism, Vimont et al. 2001). The highest correlation between the NPO and EMI indices indeed occurs when the NPO leads by a few months to a year (not shown). On the other hand, Di Lorenzo et al. (2010) suggest that atmospheric teleconnections generated by El Niño Modoki variability on the NPO are low passed by the ocean, leading to the NPGO variability (similar to ENSO forcing of the PDO via the atmospheric bridge).

Other studies have focused on the local forcing and character of El Niño Modoki. Kao and Yu (2009) found that that the onset, development, and decay of El Niño Modoki anomalies (surface and subsurface) remain mostly confined to the central Pacific. Bejarano and Jin (2008) and Kug et al. (2009) found that El Niño Modoki develops because of zonal advection by anomalous currents and vertical advection of an anomalous temperature gradient, which suggests a phenomenon confined to the tropics. This local character of El Niño Modoki is consistent with highest loadings in the tropical Pacific and the lack of an SLP link to M5, making remote forcing via the atmospheric bridge unlikely.

b. Temporal variability and evolution

A cross-wavelet analysis shows coherence between M4 and M5 at longer periods that starts around 1950 and a shift around 1970 (Fig. 8b). This shift is missed by the shorter NPGO time series but is evident in the EMI time series (wavelet analysis not shown). Prior to 1950 the coherence between M4 and M5 is absent or weak (Fig. 8b). It is not clear if this is the result of sparse data coverage in the equatorial Pacific prior to ~1960 (Deser et al. 2010); however, an analysis of only the 1910–60 period (not shown) does produce a robust M5 (equatorial Pacific) but not M4 (northeast Pacific where data coverage was >70%). A hypothesis, that should be considered with caution given the available data, is that a new and strengthening mode of climate and ocean variability (M4/NPGO) with significant implications to climate and ecosystems has become apparent in the second half of the twentieth century, consistent with increasing frequency and persistence of El Niño Modoki events in recent decades (Ashok et al. 2007; Kug et al. 2009; Yeh et al. 2009; Yu et al. 2010; Lee and McPhaden 2010) and increased variance of the NPGO/Victoria mode since the early 1990s (Bond et al. 2003; Di Lorenzo et al. 2008).

Cross-wavelet coherence (not shown) indicates that the M4 and EMI time series are coherent starting in the 1950s and mostly at time scales ≳8 yr and that coherence is increasing. While it is not clear whether the appearance and strengthening of M4 is responsible for the increased variance and intensity of El Niño Modoki or if the latter generated this new mode, this study does suggest that enhanced El Niño Modoki activity is linked to new or increased low-frequency variability in the North Pacific (M4) after ~1960. Cross-wavelets/coherence between M5 and EMI (not shown) show coherence at interannual periods during the entire time series but at low frequencies only after ~1960, presumably linked to the strengthening M4 (Fig. 8b). This implies that weak El Niño Modoki activity (associated with M5) has probably been present prior to 1960 but that the low-frequency variability and recent increase of variance in El Niño Modoki is linked to M4.

6. Large-scale decadal circulation (M1 + M3 and M4)

Decadal-to-multidecadal changes in the circulation of the North Pacific subpolar and subtropical gyres have been extensively reported (Lagerloef 1995; Latif and Barnett 1996; Miller et al. 1998; Deser et al. 1999; Seager et al. 2001; Schneider et al. 2002; Di Lorenzo et al. 2009; Ceballos et al. 2009; Chhak et al. 2009; Kwon et al. 2010). This section focuses on changes in the location and intensity of the Aleutian low and North Pacific high (such as those associated with M1 + M3 and M4) and the adjustments these induce via Sverdrup balance in the strength and position of the North Pacific subpolar and subtropical gyres (Fig. 10). These gyres converge in the KOE where decadal-to-multidecadal changes in transport and the north/south location of the subpolar/subtropical gyre boundary have been linked to the atmospheric changes discussed above (Latif and Barnett 1994; Barnett et al. 1999b; Schneider et al. 2002; Qiu 2003; Schneider and Cornuelle 2005; Ceballos et al. 2009; Qiu and Chen 2010).

Fig. 10.
Fig. 10.

Conceptual model of the decadal changes in North Pacific circulation associated with (left) M1 + M3 (PDO) and (right) M4 (NPGO). When M1 + M3 are >0, the subpolar gyre intensifies and the subtropical gyres of the North and South Pacific weaken. As a consequence, the KOE and subtropical-subpolar gyre boundary are driven southward. When M4 > 0, both subtropical and subpolar gyres weaken. Over the past decade conditions have been like (b) and (d). This combined case implies an average subpolar gyre and greatly enhanced transport in the subtropical gyres. Background colors are the mode shown in Fig. 9 for M1 + M3 and Fig. 4d for M4. Background arrows are from Sverdrup et al. (1942). Bold arrows show the enhanced transport, thin arrows show the weakened transport, dashed arrows show the role largely unknown.

Citation: Journal of Climate 24, 16; 10.1175/2011JCLI3941.1

During the positive phase of the combined M1 + M3 (PDO) multidecadal variability (Fig. 10a), the intensification and southward displacement of the Aleutian low reportedly causes the subpolar gyre to expand and strengthen and the subtropical gyre to weaken (Chhak et al. 2009). The Oyashio intensifies (Nakamura et al. 1997), the Kuroshio weakens, and the location of the Kuroshio bifurcation is driven southward (Qiu and Chen 2010) together with the subtropical–subpolar gyre boundary (Seager et al. 2001; Schneider et al. 2002). Shifts in the position and changes in strength of the Kuroshio and Oyashio contribute to anomalous cooling in the frontal region (Seager et al. 2001). Since one weakens while the other strengthens, it is likely that the total transport of the KOE (sum of Oyashio and Kuroshio) does not change significantly with M1 + M3, but its location shifts north or south (Schneider et al. 2002). A time series at a given location (i.e., Deser et al. 1999) would then show large changes in KOE transport on the M1 + M3 time scale, but this observation might reflect changes in location rather than strength. A stronger Alaskan Current and weaker California Current are consistent with positive SST anomalies in the eastern North Pacific, leading the dipole pattern observed in M3 (Fig. 4c) and the combined M1 + M3 decadal EOFs (Fig. 10a). The opposite occurs during the negative phase when the subtropical gyre strengthens and the subpolar gyre weakens (Fig. 10b).

During the negative phase of M4 (NPO and NPGO > 0), the Aleutian low and North Pacific high both strengthen and the westerlies intensify, leading to an intensification of both the subpolar and subtropical gyres (Di Lorenzo et al. 2008; Chhak et al. 2009) (Fig. 10d). M4 (NPGO) has been associated with changes in KOE transport but not in position (Ceballos et al. 2009). The opposite occurs when M4 > 0 (Fig. 10c).

Changes in circulation of the North Pacific gyres would then be the result of the different possible permutations shown in Fig. 10. The sum of M1 + M3 and M4 effects on gyre circulation is consistent with observed multidecadal changes in equatorial divergence over the past 50 years (McPhaden and Zhang 2002). When M1 + M3 and M4 are both positive (Figs. 10a,c), the two effects would counteract each other in the subpolar gyre, resulting in weak changes, while they would be superimposed in the subtropical gyre, consistent with an enhanced weakening of the gyre, the overturning circulation, and equatorial divergence that began in the late 1970s and peaked in the early 1990s (McPhaden and Zhang 2002; Fig. 4).

The KOE responds differently to the two modes: changes in position for M1 + M3/PDO and strength for M4/NPGO. As a result observed changes in the KOE depend on the coherence or lack thereof in the phases of each of these (Fig. 10). This has made it difficult to pinpoint the individual effects of M1 + M3 and M4 on the circulation of the North Pacific gyres (Qiu and Chen 2010). The recently observed decadal-to-multidecadal KOE variability is reported to show concurrent changes in position and strength: a southern position coinciding with a weak Kuroshio extension jet (Qiu and Chen 2010). This is likely the result of M1, M3, and M4 being in phase during the past two decades (r = 0.40 for M3 and M4 during the “altimetric” period 1992–2009).

7. Conclusions

This contribution reports on a global analysis of SST that clearly captures well-known regional climate phenomena in the first six EOF modes (Figs. 4 and 5). Local trends were removed prior to the EOF analysis to focus on interannual-to-multidecadal variations. The analysis suggests greater variability in the Northern Hemisphere; however, data density in the Southern Hemisphere is sparse, especially in the first half of the twentieth century. The synthesis portion of the paper focuses on the Northern Hemisphere and, in particular, the North Pacific, which has been the subject of considerable recent work.

Four of the first six modes (M1, M3, M4, and M5) display large variations in the equatorial and North Pacific (Fig. 4). These modes capture ENSO, the PDO, the NPGO, and El Niño Modoki. The PDO is shown to result from the interaction between the M1 (ENSO) and M3 (North Pacific) modes, the latest being associated with the intensification and southward migration of the Aleutian low. The M4 mode is linked to the North Pacific Oscillation and is associated with the NPGO and El Niño Modoki at decadal and longer periods after 1960. A hypothesis put forward here is that a new and strengthening mode of variability (M4/NPGO) has become apparent since the 1950s, related to more common and intense episodes of El Niño Modoki. Yeh et al. (2009) have linked the intensification of El Niño Modoki to anthropogenically induced warming and have predicted these events will continue to intensify. This may drive a positive feedback loop that intensifies the NPGO (Di Lorenzo et al. 2010). The appearance and intensification of M4 may then also be associated with a warmer world. The M5 mode is associated with the interannual time scale of El Niño Modoki throughout the 1910–2009 time series, suggesting that this phenomenon has been present prior to the 1950s, albeit with weaker intensity. The relation between the NPO and ENSO (M1) via the seasonal footprinting mechanism (Vimont et al. 2003; Alexander et al. 2010) suggests additional links between the modes, and further studies will be required to determine the direction and nature of the relationships between atmospheric variability and the four (M1, M3, M4, M5) modes of Pacific variability.

The M1, M3, and M4 modes are associated with changes in the circulation in the subpolar and subtropical gyres (Fig. 10) and, hence, the transfer of heat between the tropics and higher latitudes. The modes are also associated with dramatic changes in the productivity and composition of ocean ecosystems. These impacts may be changing as a result of the present day warming trend via increases in the frequency and intensity of the modes as described above. A shift around 1997/98 is evident in many of the modes of SST variability. It is not clear if the coherence in the modes over the past few decades (M1, M3, and M4, in particular) is the result of pure chance or if it is driven by a common and large unidentified source. Regardless, there are indications that the 1998 shift may be larger than previous similar ones (Chavez et al. 2011). The next 20 years will bring additional surprises together with increased understanding of ocean–atmosphere interactions and their variability and change.

Acknowledgments

Shannon Boedecker and Annette Gough assisted with Fig. 10, and John Ryan provided guidance on the EOF calculations. This work was supported by NASA (NNA07CN11A and NNX09AU39G) and the David and Lucile Packard Foundation. We thank the reviewers for constructive comments, which greatly improved the manuscript.

REFERENCES

  • Alexander, M. A., I. Bladé, M. Newman, J. R. Lanzante, N. C. Lau, and J. D. Scott, 2002: The atmospheric bridge: The influence of ENSO teleconnections on air–sea interaction over the global oceans. J. Climate, 15, 22052231.

    • Search Google Scholar
    • Export Citation
  • Alexander, M. A., D. J. Vimont, P. Chang, and J. D. Scott, 2010: The impact of extratropical atmospheric variability on ENSO: Testing the seasonal footprinting mechanism using coupled model experiments. J. Climate, 23, 28852901.

    • Search Google Scholar
    • Export Citation
  • Allan, R., and T. Ansell, 2006: A new globally complete monthly historical gridded mean sea level pressure dataset (HadSLP2): 1850–2004. J. Climate, 19, 58165842.

    • Search Google Scholar
    • Export Citation
  • An, S. I., and B. Wang, 2000: Interdecadal change of the structure of the ENSO mode and its impact on the ENSO frequency. J. Climate, 13, 20442055.

    • Search Google Scholar
    • Export Citation
  • An, S. I., and B. Wang, 2005: The forced and intrinsic low-frequency modes in the North Pacific. J. Climate, 18, 876885.

  • An, S. I., J. S. Kug, A. Timmermann, I. S. Kang, and O. Timm, 2007: The influence of ENSO on the generation of decadal variability in the North Pacific. J. Climate, 20, 667680.

    • Search Google Scholar
    • Export Citation
  • Ashok, K., S. K. Behera, S. A. Rao, H. Weng, and T. Yamagata, 2007: El Niño Modoki and its possible teleconnection. J. Geophys. Res., 112, C11007, doi:10.1029/2006JC003798.

    • Search Google Scholar
    • Export Citation
  • Bakun, A., 2005: Regime shifts. The Sea, A. R. Robinson and K. Brink, Eds., The Global Coastal Ocean: Multiscale Interdisciplinary Processes, Vol. 13, Harvard University Press, 971–1018.

    • Search Google Scholar
    • Export Citation
  • Barber, R. T., and F. P. Chavez, 1983: Biological consequences of El Niño. Science, 222, 12031210.

  • Barnett, T. P., D. W. Pierce, M. Latif, D. Dommenget, and R. Saravanan, 1999a: Interdecadal interactions between the tropics and midlatitudes in the Pacific basin. Geophys. Res. Lett., 26, 615618.

    • Search Google Scholar
    • Export Citation
  • Barnett, T. P., D. W. Pierce, R. Saravanan, N. Schneider, D. Dommenget, and M. Latif, 1999b: Origins of the midlatitude Pacific decadal variability. Geophys. Res. Lett., 26, 14531456.

    • Search Google Scholar
    • Export Citation
  • Bejarano, L., and F.-F. Jin, 2008: Coexistence of equatorial coupled modes of ENSO. J. Climate, 21, 30513067.

  • Bjerknes, J., 1969: Atmospheric teleconnections from the equatorial Pacific. Mon. Wea. Rev., 97, 163172.

  • Bond, N. A., J. E. Overland, M. Spillane, and P. Stabeno, 2003: Recent shifts in the state of the North Pacific. Geophys. Res. Lett., 30, 2183, doi:10.1029/2003GL018597.

    • Search Google Scholar
    • Export Citation
  • Breaker, L. C., 2007: A closer look at regime shifts based on coastal observations along the eastern boundary of the North Pacific. Cont. Shelf Res., 27, 22502277.

    • Search Google Scholar
    • Export Citation
  • Ceballos, L. I., E. Di Lorenzo, C. D. Hoyos, N. Schneider, and B. Taguchi, 2009: North Pacific Gyre Oscillation synchronizes climate fluctuations in the eastern and western boundary systems. J. Climate, 22, 51635174.

    • Search Google Scholar
    • Export Citation
  • Chang, P., and Coauthors, 2006: Climate fluctuations of tropical coupled systems—The role of ocean dynamics. J. Climate, 19, 51225174.

    • Search Google Scholar
    • Export Citation
  • Chavez, F. P., 2005: Biological consequences of interannual to multidecadal variability. The Sea, A. R. Robinson and K. Brink, Eds., The Global Coastal Ocean: Multiscale Interdisciplinary Processes, Vol. 13, Harvard University Press, 643–680.

    • Search Google Scholar
    • Export Citation
  • Chavez, F. P., and M. Messié, 2009: A comparison of eastern boundary upwelling ecosystems. Prog. Oceanogr., 83, 8096.

  • Chavez, F. P., and Coauthors, 2002: Biological and chemical consequences of the 1997–1998 El Niño in central California waters. Prog. Oceanogr., 54, 205232.

    • Search Google Scholar
    • Export Citation
  • Chavez, F. P., J. P. Ryan, S. E. Lluch-Cota, and C. M. Ñiquen, 2003: From anchovies to sardines and back: Multidecadal change in the Pacific Ocean. Science, 299, 217221.

    • Search Google Scholar
    • Export Citation
  • Chavez, F. P., M. Messié, and J. T. Pennington, 2011: Marine primary production in relation to climate variability and change. Annu. Rev. Mar. Sci., 3, 227260.

    • Search Google Scholar
    • Export Citation
  • Chen, G., and C.-Y. Tam, 2010: Different impacts of two kinds of Pacific Ocean warming on tropical cyclone frequency over the western North Pacific. Geophys. Res. Lett., 37, L01803, doi:10.1029/2009GL041708.

    • Search Google Scholar
    • Export Citation
  • Cheng, X., G. Nitsche, and J. M. Wallace, 1995: Robustness of low-frequency circulation patterns derived from EOF and rotated EOF analyses. J. Climate, 8, 17091720.

    • Search Google Scholar
    • Export Citation
  • Chhak, K. C., E. Di Lorenzo, N. Schneider, and P. F. Cummins, 2009: Forcing of low-frequency ocean variability in the northeast Pacific. J. Climate, 22, 12551276.

    • Search Google Scholar
    • Export Citation
  • Deser, C., and M. L. Blackmon, 1995: On the relationship between tropical and North Pacific sea surface temperature variations. J. Climate, 8, 16771680.

    • Search Google Scholar
    • Export Citation
  • Deser, C., M. A. Alexander, and M. S. Timlin, 1999: Evidence for a wind-driven intensification of the Kuroshio Current Extension from the 1970s to the 1980s. J. Climate, 12, 16971706.

    • Search Google Scholar
    • Export Citation
  • Deser, C., A. S. Phillips, and J. W. Hurrell, 2004: Pacific interdecadal climate variability: Linkages between the tropics and the North Pacific during boreal winter since 1900. J. Climate, 17, 31093124.

    • Search Google Scholar
    • Export Citation
  • Deser, C., M. A. Alexander, S. P. Xie, and A. S. Phillips, 2010: Sea surface temperature variability: Patterns and mechanisms. Annu. Rev. Mar. Sci., 2, 115143.

    • Search Google Scholar
    • Export Citation
  • deYoung, B., R. Harris, J. Alheit, G. Beaugrand, N. Mantua, and L. Shannon, 2004: Detecting regime shifts in the ocean: Data considerations. Prog. Oceanogr., 60, 143164.

    • Search Google Scholar
    • Export Citation
  • Di Lorenzo, E., and Coauthors, 2008: North Pacific Gyre Oscillation links ocean climate and ecosystem change. Geophys. Res. Lett., 35, L08607, doi:10.1029/2007GL032838.

    • Search Google Scholar
    • Export Citation
  • Di Lorenzo, E., and Coauthors, 2009: Nutrient and salinity decadal variations in the central and eastern North Pacific. Geophys. Res. Lett., 36, L14601, doi:10.1029/2009GL038261.

    • Search Google Scholar
    • Export Citation
  • Di Lorenzo, E., K. M. Cobb, J. C. Furtado, N. Schneider, B. Anderson, A. Bracco, M. A. Alexander, and D. Vimont, 2010: Central Pacific El Niño and decadal climate change in the North Pacific Ocean. Nat. Geosci., 3, 762765.

    • Search Google Scholar
    • Export Citation
  • Enfield, D. B., and A. M. Mestas-Nuñez, 1999: Multiscale variabilities in global sea surface temperatures and their relationships with tropospheric climate patterns. J. Climate, 12, 27192733.

    • Search Google Scholar
    • Export Citation
  • Enfield, D. B., and A. M. Mestas-Nuñez, 2000: Global modes of ENSO and non-ENSO sea surface temperature variability and their association with climate. El Niño and the Southern Oscillation: Multiscale Variability and Global and Regional Impacts, H. F. Diaz and V. Markgraf, Eds., Cambridge University Press, 89–112.

    • Search Google Scholar
    • Export Citation
  • Enfield, D. B., A. M. Mestas-Nuñez, and P. J. Trimble, 2001: The Atlantic multidecadal oscillation and its relation to rainfall and river flows in the continental U.S. Geophys. Res. Lett., 28, 20772080.

    • Search Google Scholar
    • Export Citation
  • Fedorov, A. V., and S. G. Philander, 2000: Is El Niño changing? Science, 288, 19972002.

  • Francis, R. C., S. R. Hare, A. B. Hollowed, and W. S. Wooster, 1998: Effects of interdecadal climate variability on the oceanic ecosystems of the NE Pacific. Fish. Oceanogr., 7, 121.

    • Search Google Scholar
    • Export Citation
  • Grinsted, A., J. Moore, and S. Jevrejeva, 2004: Application of the cross wavelet transform and wavelet coherence to geophysical time series. Nonlinear Processes Geophys., 11, 561566.

    • Search Google Scholar
    • Export Citation
  • Guan, B., and S. Nigam, 2008: Pacific sea surface temperatures in the twentieth century: An evolution-centric analysis of variability and trend. J. Climate, 21, 27902809.

    • Search Google Scholar
    • Export Citation
  • Guan, B., and S. Nigam, 2009: Analysis of Atlantic SST variability factoring interbasin links and the secular trend: Clarified structure of the Atlantic multidecadal oscillation. J. Climate, 22, 42284240.

    • Search Google Scholar
    • Export Citation
  • Hare, S. R., and N. J. Mantua, 2000: Empirical evidence for North Pacific regime shifts in 1977 and 1989. Prog. Oceanogr., 47, 103145.

    • Search Google Scholar
    • Export Citation
  • Kao, H.-Y., and J.-Y. Yu, 2009: Contrasting eastern-Pacific and central-Pacific types of ENSO. J. Climate, 22, 615632.

  • Kawamura, R., 1994: A rotated EOF analysis of global sea surface temperature variability with interannual and interdecadal scales. J. Phys. Oceanogr., 24, 707715.

    • Search Google Scholar
    • Export Citation
  • Kerr, R. A., 2000: A North Atlantic climate pacemaker for the centuries. Science, 288, 19841985.

  • Kug, J.-S., F.-F. Jin, and S.-I. An, 2009: Two types of El Niño events: Cold tongue El Niño and warm pool El Niño. J. Climate, 22, 14991515.

    • Search Google Scholar
    • Export Citation
  • Kumar, K. K., B. Rajagopalan, M. Hoerling, G. Bates, and M. Cane, 2006: Unraveling the mystery of Indian monsoon failure during El Niño. Science, 314, 115119.

    • Search Google Scholar
    • Export Citation
  • Kwon, Y.-O., M. A. Alexander, N. A. Bond, C. Frankignoul, H. Nakamura, B. Qiu, and L. Thompson, 2010: Role of the Gulf Stream and Kuroshio–Oyashio systems in large-scale atmosphere–ocean interaction: A review. J. Climate, 23, 32493281.

    • Search Google Scholar
    • Export Citation
  • Lagerloef, G. S. E., 1995: Interdecadal variations in the Alaska gyre. J. Phys. Oceanogr., 25, 22422258.

  • Larkin, N. K., and D. E. Harrison, 2005: Global seasonal temperature and precipitation anomalies during El Niño autumn and winter. Geophys. Res. Lett., 32, L16705, doi:10.1029/2005GL022860.

    • Search Google Scholar
    • Export Citation
  • Latif, M., and T. P. Barnett, 1994: Causes of decadal climate variability over the North Pacific and North America. Science, 266, 634637.

    • Search Google Scholar
    • Export Citation
  • Latif, M., and T. P. Barnett, 1996: Decadal climate variability over the North Pacific and North America: Dynamics and predictability. J. Climate, 9, 24072423.

    • Search Google Scholar
    • Export Citation
  • Lee, T., and M. J. McPhaden, 2010: Increasing intensity of El Niño in the central-equatorial Pacific. Geophys. Res. Lett., 37, L14603, doi:10.1029/2010GL044007.

    • Search Google Scholar
    • Export Citation
  • Lee, T., and Coauthors, 2010: Record warming in the South Pacific and western Antarctica associated with the strong central-Pacific El Niño in 2009–10. Geophys. Res. Lett., 37, L19704, doi:10.1029/2010GL044865.

    • Search Google Scholar
    • Export Citation
  • Lees, K., S. Pitois, C. Scott, C. Frid, and S. Mackinson, 2006: Characterizing regime shifts in the marine environment. Fish Fish., 7, 104127.

    • Search Google Scholar
    • Export Citation
  • Mantua, N. J., and S. R. Hare, 2002: The Pacific decadal oscillation. J. Oceanogr., 58, 3544.

  • Mantua, N. J., S. R. Hare, Y. Zhang, J. M. Wallace, and R. C. Francis, 1997: A Pacific interdecadal climate oscillation with impacts on salmon production. Bull. Amer. Meteor. Soc., 78, 10691079.

    • Search Google Scholar
    • Export Citation
  • Martinez, E., D. Antoine, F. d’Ortenzio, and B. Gentili, 2009: Climate-driven basin-scale decadal oscillations of oceanic phytoplankton. Science, 326, 12531256.

    • Search Google Scholar
    • Export Citation
  • McPhaden, M. J., 2002: El Niño and La Niña: Causes and global consequences. The Earth System: Physical and Chemical Dimensions of Global Environmental Change, M. C. MacCracken and J. S. Perry, Eds., Vol. 1, Encyclopedia of Global Environmental Change, John Wiley and Sons, 353–370.

    • Search Google Scholar
    • Export Citation
  • McPhaden, M. J., and D. Zhang, 2002: Slowdown of the meridional overturning circulation in the upper Pacific Ocean. Nature, 415, 603608.

    • Search Google Scholar
    • Export Citation
  • McPhaden, M. J., S. E. Zebiak, and M. H. Glantz, 2006: ENSO as an integrating concept in earth science. Science, 314, 1740.

  • Mestas-Nuñez, A. M., and D. B. Enfield, 1999: Rotated global modes of non-ENSO sea surface temperature variability. J. Climate, 12, 27342746.

    • Search Google Scholar
    • Export Citation
  • Miller, A. J., and N. Schneider, 2000: Interdecadal climate regime dynamics in the North Pacific Ocean: Theories, observations and ecosystem impacts. Prog. Oceanogr., 47, 355379.

    • Search Google Scholar
    • Export Citation
  • Miller, A. J., D. R. Cayan, T. P. Barnett, N. E. Graham, and J. M. Oberhuber, 1994: The 1976-77 climate shift of the Pacific Ocean. Oceanography, 7, 2126.

    • Search Google Scholar
    • Export Citation
  • Miller, A. J., D. R. Cayan, and W. B. White, 1998: A westward-intensified decadal change in the North Pacific thermocline and gyre-scale circulation. J. Climate, 11, 31123127.

    • Search Google Scholar
    • Export Citation
  • Miller, A. J., F. Chai, S. Chiba, J. R. Moisan, and D. J. Neilson, 2004: Decadal-scale climate and ecosystem interactions in the North Pacific Ocean. J. Oceanogr., 60, 163188.

    • Search Google Scholar
    • Export Citation
  • Nakamura, H., G. Lin, and T. Yamagata, 1997: Decadal climate variability in the North Pacific during the recent decades. Bull. Amer. Meteor. Soc., 78, 22152225.

    • Search Google Scholar
    • Export Citation
  • Neelin, J. D., D. S. Battisti, A. C. Hirst, F.-F. Jin, Y. Wakata, T. Yamagata, and S. E. Zebiak, 1998: ENSO theory. J. Geophys. Res., 103, 14 26114 290.

    • Search Google Scholar
    • Export Citation
  • Newman, M., G. P. Compo, and M. A. Alexander, 2003: ENSO-forced variability of the Pacific decadal oscillation. J. Climate, 16, 38533857.

    • Search Google Scholar
    • Export Citation
  • North, G. R., T. L. Bell, R. F. Cahalan, and F. J. Moeng, 1982: Sampling errors in the estimation of empirical orthogonal functions. Mon. Wea. Rev., 110, 699706.

    • Search Google Scholar
    • Export Citation
  • Overland, J., S. Rodionov, S. Minobe, and N. Bond, 2008: North Pacific regime shifts: Definitions, issues and recent transitions. Prog. Oceanogr., 77, 92102.

    • Search Google Scholar
    • Export Citation
  • Peterson, W. T., and F. B. Schwing, 2003: A new climate regime in northeast Pacific ecosystems. Geophys. Res. Lett., 30, 1896, doi:10.1029/2003GL017528.

    • Search Google Scholar
    • Export Citation
  • Philander, S. G. H., 1986: Unusual conditions in the tropical Atlantic Ocean in 1984. Nature, 322, 236238.

  • Philander, S. G. H., 1990: El Niño, La Niña, and the Southern Oscillation. Academic Press, 289 pp.

  • Pierce, D. W., T. P. Barnett, and M. Latif, 2000: Connections between the Pacific Ocean tropics and midlatitudes on decadal timescales. J. Climate, 13, 11731194.

    • Search Google Scholar
    • Export Citation
  • Qiu, B., 2003: Kuroshio Extension variability and forcing of the Pacific decadal oscillations: Responses and potential feedback. J. Phys. Oceanogr., 33, 24652483.

    • Search Google Scholar
    • Export Citation
  • Qiu, B., and S. Chen, 2010: Eddy-mean flow interaction in the decadally modulating Kuroshio Extension system. Deep-Sea Res. II, 57, 10981110.

    • Search Google Scholar
    • Export Citation
  • Rodionov, S. N., J. E. Overland, and N. A. Bond, 2005: The Aleutian low and winter climatic conditions in the Bering Sea. Part I: Classification. J. Climate, 18, 160177.

    • Search Google Scholar
    • Export Citation
  • Ruiz-Barradas, A., J. A. Carton, and S. Nigam, 2000: Structure of interannual-to-decadal climate variability in the tropical Atlantic sector. J. Climate, 13, 32853297.

    • Search Google Scholar
    • Export Citation
  • Schneider, N., and B. D. Cornuelle, 2005: The forcing of the Pacific decadal oscillation. J. Climate, 18, 43554373.

  • Schneider, N., A. J. Miller, and D. W. Pierce, 2002: Anatomy of North Pacific decadal variability. J. Climate, 15, 586605.

  • Seager, R., Y. Kushnir, N. H. Naik, M. A. Cane, and J. Miller, 2001: Wind-driven shifts in the latitude of the Kuroshio–Oyashio Extension and generation of SST anomalies on decadal timescales. J. Climate, 14, 42494265.

    • Search Google Scholar
    • Export Citation
  • Shakun, J. D., and J. Shaman, 2009: Tropical origins of North and South Pacific decadal variability. Geophys. Res. Lett., 36, L19711, doi:10.1029/2009GL040313.

    • Search Google Scholar
    • Export Citation
  • Smith, T. M., and R. W. Reynolds, 2003: Extended reconstruction of global sea surface temperatures based on COADS data (1854–1997). J. Climate, 16, 14951510.

    • Search Google Scholar
    • Export Citation
  • Smith