## 1. Introduction

The dominant statistical mode of tropical Atlantic interannual–decadal atmosphere–ocean variability is an anomalous meridional SST gradient across the mean intertropical convergence zone (ITCZ) latitude and a cross-gradient atmospheric boundary layer flow toward the anomalously warmer hemisphere (Nobre and Shukla 1996; Chang et al. 1997). Hastenrath and Greischar (1993) proposed that this boundary layer flow is driven by the anomalous meridional SST gradient through its hydrostatic effect on sea level pressure (Lindzen and Nigam 1987). The cross-gradient flow implies a shift of the ITCZ and associated convection toward the anomalously warmer hemisphere (e.g., Hastenrath and Heller 1977). This basic picture—linking anomalous SST, winds, and convection—has been supported in subsequent observational and modeling studies (e.g., Ruiz-Barradas et al. 1999; Chang et al. 2000; Chiang et al. 2002).

Two interpretations had been proposed for how this mode (hereafter the meridional mode, also known as the gradient or interhemispheric mode) arises. Nobre and Shukla (1996) argued from observational analysis that trade wind variations in the north tropical Atlantic precede the tropical basinwide anomalies that in turn produced SST anomalies there (and hence an anomalous SST gradient across the mean ITCZ latitude), implying that the mode is externally forced. Two sources of external forcing have been identified: ENSO and the North Atlantic Oscillation (NAO; e.g., Curtis and Hastenrath 1995; Nobre and Shukla 1996; Xie and Tanimoto 1998; Giannini et al. 2000). They directly perturb the boreal winter north tropical Atlantic (NTA) trade wind strength, changing the underlying SST through latent heat fluxes. The resulting NTA SST perturbation peaks in boreal spring, creating the near-equatorial meridional SST gradient.

Chang et al. (1997) proposed an alternative explanation for the origins of the meridional mode, pointing to a positive feedback [sometimes known as wind–evaporation–SST (WES) feedback (e.g., Xie 1999)] between the SST gradient and the cross-gradient flow. The cross-gradient flow reduced the strength of the trade winds in the anomalously warmer hemisphere and increased the trades in the cooler hemisphere, and this reinforced the SST gradient through wind speed impact on anomalous evaporation. Idealized modeling studies (e.g., Chang et al. 1997; Xie 1999) have shown that this feedback can qualitatively reproduce the observed behavior, though it was found that under realistic coupling strength the model behavior was not self-sustaining, and that external forcing was required to sustain the variability. Subsequent modeling and observational studies show this feedback is relatively weak and limited to the deep Tropics (Chang et al. 2000; Chiang et al. 2002; Czaja et al. 2002), further underlining the need for external forcing. A more recent idealized model study by Kushnir et al. (2002) incorporating the deep Tropics–limited WES feedback suggests, however, that the feedback is essential to the decadal nature of meridional mode variability. Our current knowledge of the meridional mode physics thus suggests that it is externally forced through trade wind variations, with limited WES feedback in the deep Tropics.

An outstanding issue with the meridional mode is the issue of (out of phase) coherence between the SST anomalies on either side of the anomalous SST gradient. Houghton and Tourre 1992) showed that the dipolelike nature of the anomalies derived from a straight empirical orthogonal function (EOF) analysis of tropical Atlantic SST anomalies do not survive under varimax rotation; instead, the two SST lobes appear to vary independently of each other, a result supported by subsequent observational analysis (Mehta and Delworth 1995; Enfield et al. 1999), and also the modeling study by Dommenget and Latif (2000). On the other hand, Xie and Tanimoto (1998) argue using observational and modeling evidence for anticoherence between the two lobes at decadal time scales and part of a coherent pan-Atlantic decadal oscillation characterized by zonal bands of SST and wind anomalies with alternate polarities from the tropical South Atlantic to Greenland. The “dipole” issue is intimately tied to the physical interpretation of the mode, since independent behavior between the northern and southern SST lobes suggests an externally forced mechanism with little to no positive feedback in the Tropics, whereas dipole behavior between the northern and southern lobes implies a stronger role for positive feedback. At present, the prevailing evidence suggests that the SST lobes act essentially independently, in agreement with the prevailing physical interpretation; although we emphasize that this issue is not yet fully resolved in large part because of inadequate length of data, but also in part because simple models hint at dipolelike anomalous SST structures even with WES feedback limited to the deep Tropics (Kushnir et al. 2002).

A similar tropical Pacific mode of variability (hereafter the meridional mode, following Servain et al. 1999) has heretofore not been identified, in large part because the El Niño–Southern Oscillation (ENSO) dominates the variability there (Wallace et al. 1998). There are, however, compelling reasons for its existence. Each basin possesses similar mean states: namely, a cold tongue over the equatorial oceans (weighted to the east) and an ITCZ at its northern edge that follows an annual march lagging by a quarter cycle the seasonal cycle of the ITCZ over land (Mitchell and Wallace 1992). An idealized tropical coupled general circulation model study (Xie and Saito 2001) shows that in the absence of thermocline–SST feedback central to the ENSO physics, interannual variations in the north–south oceanic ITCZ position are present, suggesting that this variability is intrinsic to the ITCZ/cold tongue climate. Indeed, a mode in the Pacific with spatial pattern reminiscent of the Atlantic meridional mode was identified in a recent coupled model study with realistic configuration (Yukimoto et al. 2000). Finally, recent detailed observational and general circulation model (GCM) analyses by Vimont and collaborators (Vimont et al. 2001, 2003b) shows that a leading mode of wintertime atmospheric variability in the North Pacific [the North Pacific Oscillation (NPO); Rogers 1981] forces variations of the north tropical Pacific (NTP) SST through its control of the subtropical trades there, a mechanism similar to the external forcing of the tropical Atlantic meridional mode by the NAO.

These similarities motivate us to search for analogous tropical meridional modes in the two basins. We will show that once the ENSO influence is linearly removed from the SST and wind data we analyze, the dominant mode of atmosphere–ocean variability extracted separately from the tropical Pacific and tropical Atlantic basin resembles each other in spatial and temporal characteristics (section 2). The Atlantic pattern is the meridional mode pattern, whose interpretation is fairly well established from previous studies. A nonlinear component of ENSO does, however, project strongly on the derived Pacific mode. We show that by additionally filtering out the strong ENSO years from our analysis, the derived Pacific mode stays essentially the same, indicating that this mode does not depend on ENSO for its existence (section 3). We then show that the Pacific and Atlantic modes share the same physical interpretation, and on that basis we argue that the Pacific mode is the analogue to the Atlantic meridional mode pattern (section 4). We reinforce our interpretation by showing the similarity between the two basins using a different (composite) analysis (section 5). A summary and discussion of the potential implications of our results is given in section 6.

## 2. Data, method, and results

We use monthly mean SST and 10-m winds from the National Centers for Environmental Prediction–National Center for Atmospheric Research (NCEP–NCAR) reanalysis (Kalnay et al. 1996, hereafter simply the reanalysis) spanning January 1948–December 2001. For each field, we spatially averaged over six adjoining (two latitude by three longitude) grid points, computed detrended monthly mean anomalies, and applied a 3-month running mean. The spatial averaging does not unduly impact the result, as the extracted dominant mode possesses comparatively large space scales; similarly for the temporal smoothing. Since ENSO is not the variability of interest, we subtract the linear least squares fit to a commonly used ENSO index [cold tongue index (CTI), SST anomalies averaged over 6°S–6°N and 180°– 90°W] from all fields prior to analysis.

We apply maximum covariance analysis (MCA; also known as singular value decomposition, Bretherton et al. 1992) to the cross-covariance matrix between SST and both components of 10-m winds to extract the leading mode of coupled variability. Two domains that possess qualitatively similar mean ITCZ/cold tongue climates are separately and independently analyzed: 32°N– 21°S and 175°E–95°W for the Pacific, and 32°N–21°S and 74°W to the West African coastline for the Atlantic. The meridional offset in both domains is in keeping with the northward bias of the mean ITCZ position. The squared covariance fractions for the leading modes are each 53% for the Pacific and the Atlantic, explaining the majority of the total squared covariance. The normalized root-mean-square covariance, which measures the strength of the relation between two fields (in this case SST and 10-m winds in the Pacific or Atlantic) is 0.16 for the Pacific and 0.17 for the Atlantic, indicating that the SST and wind fields are closely coupled. We found that the leading MCA mode calculated in this method (in which the CTI has been removed via linear regression) is identical to the second MCA mode of the original data (in which the CTI was *not* removed via linear regression). For consistency with the analysis in the Atlantic basin, we show results from the original method.

The spatial structure of the leading MCA mode is depicted by regression maps of SST and winds onto the normalized SST expansion coefficients for the Pacific (Fig. 1a) and Atlantic (Fig. 1b). The Atlantic analysis identifies the meridional mode as the leading one. The Pacific pattern strongly resembles the Atlantic pattern, with anomalously warm SST in the NTP region producing an anomalous meridional SST gradient. Southerly atmospheric flow occurs over the strongest anomalous SST gradient in both patterns. The divergence of the anomalous surface winds (not shown) gives a dipole with increased convergence to the north of the mean ITCZ position (around 5°N for the Atlantic, 8°N for the Pacific) and increased divergence to the south, indicating an anomalous northward displacement of the ITCZ. A simultaneous regression on a satellite precipitation dataset 1979–2001 (Xie and Arkin 1997) (Figs. 1c,d) shows anomalous rainfall generally consistent with the anomalous northward displacement of the ITCZ, although the Pacific precipitation pattern resembles more a strengthening of the climatological mean rainfall—more precipitation in the mean ITCZ latitudes north of the cold tongue, and less precipitation over the cold tongue region itself. We will address this issue in section 3.

The temporal characteristics of the leading modes are also similar. The month-by-month variance of the SST expansion coefficient (Fig. 2, top row) peaks in boreal spring, following the late winter/early spring peak variance in the winds. A lag correlation between the expansion coefficients (Fig. 2, second row) shows the wind leading SST by a month in both basins, implying that the winds drive SST variability. The symmetry around zero lag implies, however, that a significant coupled component is also captured. This symmetry is not an artifact of the initial temporal smoothing on the analyses fields; the symmetry in the lagged correlations seen in Fig. 2 results when the MCA is repeated without first applying a 3-month running mean to the data. A noticeable difference is the long persistence in the Pacific lag correlation relative to the Atlantic, which we will address in the next section. Both SST expansion coefficients vary across many time scales (Fig. 2, bottom two rows) including interannual and decadal variations. In the present manuscript we do not argue for any preferred time scale of the variability. We note that the correlation between the Pacific and Atlantic SST expansion coefficients is weak (*r* = 0.18), indicating that the modes vary independently in each basin (recall, the analysis is conducted separately for each basin).

How robust are these results? The leading mode we obtain is insensitive to reasonable variations in domain size (e.g., from 120°E to 75°W in the Pacific, and from 32°S to 32°N in either basin). We repeated the analysis for subsets of the data periods, in particular separating the pre- and postsatellite data periods of 1948–79, and 1979–2001, as well as excluding the first 10 yr of data (1948–57) when a different set of observing times was used in the upper-air data going into the reanalysis (Kistler et al. 2001). All three subsets produced comparable results. The same analysis using an independent tropical Pacific wind stress product 1961–2001 (Legler and O'Brien 1988) and a reduced space optimal analysis SST product (Kaplan et al. 1998) gives essentially the same leading Pacific pattern. Similarly, using an Atlantic wind stress and SST product (Servain 1991) 1964– 88 gives the same leading Atlantic pattern. Insofar as these datasets are independent of the reanalysis, they suggest that our results are robust. We also repeated the MCA after removing ENSO-related variability from the fields by linearly regressing out the leading complex empirical orthogonal function of the global SST anomaly field (instead of the CTI) that has been shown to represent ENSO and its teleconnected response (Enfield and Mestas-Nunez 1999). Our results remain essentially the same.

## 3. ENSO nonlinearity and the Pacific meridional mode

Since ENSO so dominates in the Pacific variability, the question arises as to whether linear removal of CTI from our analysis fields removes ENSO sufficiently so that the dominant pattern arising from what remains is in fact independent of ENSO. It turns out that ENSO appears to project nonlinearly on the Pacific meridional mode. Figure 3 shows a scatterplot of an interannual ENSO index—CTI averaged over November–December–January (NDJ) and normalized (hereafter ctiNDJ)— against March–April–May (MAM)-averaged values of the Pacific meridional mode SST expansion coefficients (also normalized) following the NDJ months of the interannual ENSO index. We choose MAM for the Pacific meridional mode index as that is the season when the behavior is most pronounced. Visual examination of the scatterplot suggests that extreme (positive and negative) ENSO events tend to favor low values of the Pacific MAM meridional mode index.

In order to see if the extreme ENSO years affected our MCA result, we repeated the analysis but using only those years with low or no ENSO activity. We divided the 53 yr of the analysis into three bins based on ctiNDJ. We use the NDJ period rather than the MAM period as NDJ is the season when ENSO peaks; however, the results remain the same whether we use NDJ or MAM, with 9 yr in each of the extreme “high” and “low” index years (these years are indicated by light shading in Fig. 5), and 35 “neutral” years. We defined the year to be from the previous September through the following August in order to include the crucial boreal winter (for ENSO) and spring (for the meridional mode) seasons in the same year. The result of the MCA analysis with only neutral years (Figs. 4a,e,f) essentially reproduces the same mode 1 as the previous MCA analysis using all 53 years: the SST expansion coefficients from this analysis (middle curve in Fig. 5, the coefficients are in the nonshaded region of the graph) are correlated at *r* = 0.99 with the same time period from the MCA analysis with all data included (top curve in Fig. 5).

There are slight but noticeable differences between the original Pacific MCA 1 pattern and the one derived using only the neutral years: the spatial pattern in the neutral case (Fig. 4a) has slightly greater amplitude in the northern tropical Pacific features relative to the equatorial features compared to the original (Fig. 1a), and the peaks in the variance by the month (Fig. 4e) for the SST (in boreal spring) and wind expansion (in boreal winter) coefficients are more sharply defined than the original (Fig. 2, top left). The regression map of precipitation (note that this is computed using only 12 years of data, so interpretation of precipitation is speculative) shows a substantially different precipitation response than in Fig. 1—it shows primarily a reduction in the southern half of the mean ITCZ latitude of ∼8°N, and only a suggestion of increased precipitation to the north of 10°N. This pattern implies a northward shift in the position of maximum ITCZ rainfall (making the interpretation more consistent with the ITCZ response for the Atlantic meridional mode), although in this case it is achieved largely by reducing rainfall in the southern half of the climatological Pacific ITCZ. The most striking change, however, is the marked decrease in the long-range (greater than 5 months) persistence in the lag correlation between the SST and wind expansion coefficients (cf. Fig. 4f to Fig. 2, second row, left). Thus, with the extreme ENSO years removed, the temporal properties of this Pacific MCA mode 1 fall in line with the Atlantic meridional mode.

The persistence seen in the original Pacific MCA mode 1 arises from nonlinear ENSO behavior. To demonstrate this, we repeated the MCA analysis using only the 18 extreme ENSO years that were omitted in the neutral MCA analysis. The results for mode 1 (Fig. 4b) show similar SST and wind spatial pattern as before, but with a stronger equatorial emphasis (in particular, a stronger zonal SSTA gradient and anomalous equatorial easterlies). The regression of the expansion coefficients from this MCA mode on the available precipitation data (Fig. 4d—note that there are only 8 yr in this regression) resembles the same from the MCA analysis using all years (Fig. 1c). This indicates that the extreme ENSO years basically determine the “all years” regression for precipitation. The wind and SST expansion coefficients (Fig. 4g) possess more uniform variance across all months. The most pronounced difference between the temporal structure of the “neutral” and “strong” ENSO years is seen in the strong lagged correlation between the wind and SST time series out to several months (Fig. 4h). This explains the strong persistent lagged correlation in the original MCA time series. The combined SST expansion coefficients from the neutral ENSO years analysis (second curve in Fig. 5, nonshaded regions) and the extreme ENSO years (second curve in Fig. 5, shaded regions) are nearly identical to the SST expansion coefficient from the entire record (Fig. 5, first curve), indicating that the original MCA was not unduly affected by the inclusion of the extreme ENSO years and ENSO nonlinearity.

*negative*meridional mode index, as seen in Fig. 3.

So, while nonlinear ENSO projects strongly onto the Pacific meridional mode, *the existence of this mode does not depend on ENSO.* Once the nonlinear ENSO influence is largely removed, the resulting Pacific meridional mode shares similar characteristics to the Atlantic meridional mode in both spatial and temporal characteristics. These similarities suggest, therefore, that the meridional modes have similar physical interpretation in each basin—an interpretation we explore in the next section.

## 4. Interpretation

Following a previous lead (Mitchell and Wallace 1992), we examine the spatial characteristics of the seasonal cycle in both the Pacific and Atlantic basins. Figure 7 shows the deviation of the boreal spring (MAM) seasonal mean SST and surface winds from the annual mean. Recall that the peak variance in the meridional mode SST expansion coefficients occurs in this season. The resemblance in the deep Tropics to the corresponding meridional mode spatial patterns (cf. Fig. 1, top row) is striking. This observation has been previously noted for the Atlantic (Mitchell and Wallace 1992). A similar pattern results if we apply MCA to the annual cycle in tropical SST and winds (not shown). Note that the qualitative differences between the Pacific and Atlantic meridional mode spatial patterns are reproduced in the MAM deviation too: the more zonal equatorial flow and equatorially confined negative SST lobe in the Pacific as compared to the Atlantic. The point made here is that the meridional mode pattern emerges in different physical situations and constitutes an independent extraction. While it suggests that the physics of the meridional mode is related to the physics of the seasonal cycle, we do not necessarily interpret the former as a modulation of the latter.

How are the SST and surface wind patterns of the MCA mode 1 related? Previous modeling (Chang et al. 2000; Chiang et al. 2001; Sutton et al. 2000) and observational (Chiang et al. 2002) work have suggested that the SST-forced surface wind response is limited to the deep Tropics, suggesting the same for the Pacific MCA mode 1 pattern. In order to determine the SST-forced surface atmospheric circulation, we devised a simple atmospheric general circulation model (AGCM) experiment. We ran four separate ensemble simulations (eight members in each ensemble) using the National Center for Atmospheric Research Community Climate model, version 3.10 (CCM3.10; see Kiehl et al. 1998, for a description) AGCM at standard T42 resolution, using different SST boundary conditions for each ensemble. The initial conditions in each ensemble were taken as different Januaries from a long simulation of CCM3.10 forced by climatological SSTs. In the first two ensembles, we forced CCM3.10 with the positive and negative polarity of the Pacific meridional mode SST map (shown in Fig. 1a), respectively. Similarly, the last two ensembles were forced by the meridional mode SST map in the Atlantic (Fig. 1b). We averaged the precipitation and lowest model-level wind output for each set of eight ensemble runs, subtracted the mean of the “negative polarity” runs from the mean of the “positive polarity” runs, and divided the result by 2. The result (Fig. 8) shows the SST-forced response in the surface winds and precipitation. It clearly shows a deep tropical surface wind response similar to that derived from the MCA analysis (Fig. 1), as well as a meridionally displaced ITCZ. On the other hand, the northern subtropical wind anomalies present in the MCA mode 1 Pacific and Atlantic patterns are conspicuously absent from the model anomaly field. This suggests that the underlying SST do not force the subtropical wind anomalies.

We now interpret the anomalous trade winds to the north of the mean ITCZ coincident with the northern lobe of the meridional mode SST anomaly. Their absence in the AGCM simulations indicates that they are not forced by the underlying SST anomaly. This suggests that they are either spurious [this is doubtful, given that the Atlantic (Pacific) meridional mode expansions coefficients are significantly correlated to the observed NTA (NTP) surface wind anomalies] or they are forcing rather than responding to the underlying SST anomalies. This latter interpretation is supported through lead–lag analysis of zonal mean SST, wind, and heat flux anomalies, similar to the analysis done by Czaja et al. (2002) for the Atlantic. We created an annual index of wintertime atmospheric circulation anomalies associated with the MCA modes by averaging the December–January– February (DJF) MCA wind expansion coefficients, after which we lag regressed this index against anomalies of SST, winds, and surface net heat flux arranged by the month and zonally averaged across the domain of the MCA analyses (Fig. 9). Extreme ENSO years (as defined in the previous section) are not included in the Pacific analysis. Recall that December–March are the months of maximum variance in the MCA wind expansion coefficients, while March–May are the months of maximum variance of the SST expansion coefficients. In both basins, decreased northern subtropical trade winds in DJF coincide with a decrease in surface fluxes (primarily decreased *latent* heat flux) out of the ocean and an increase in the SST there. This result suggests that the trade wind anomalies are forcing the northern subtropical SST anomalies.

Figure 9 shows other interesting characteristics of the meridional mode evolution in both basins. In the Atlantic, the decay in the subtropical SST anomaly (around 20°N) past boreal spring is coincident with increased fluxes out of the ocean. The increased flux out of the ocean is not as clear in the Pacific, though there are very weak upward net heat flux anomalies north of about 20°N (this may help explain the enhanced persistence of SST anomalies in the Pacific well into boreal summer and fall). In the Atlantic, the meridional near-equatorial flow in this analysis is coincident with the establishment of northern tropical SST anomalies and the meridional SST gradient in the deep Tropics, consistent with our interpretation of this flow as an SST-forced response. Likewise for the Pacific anomalies, in boreal spring there is a clear tendency for a meridional wind to blow across the maximum SST gradient. In agreement with earlier work by Chang et al. (1997), there is a clear indication in the deep Tropics of increased heat fluxes into the ocean associated with the Atlantic cross-gradient anomalous surface flow, indicating a positive feedback between the tropical meridional wind response and the SST anomalies through boreal spring and into summer. This feedback is, however, not readily apparent in the Pacific, though we note that Vimont et al. (2003a) find a positive feedback between similar subtropical SST anomalies and surface winds in coupled simulations of the Commonwealth Scientific and Industrial Research Organisation (CSIRO) general circulation model.

Where do the northern trade wind anomalies originate? To explore this question, we regressed the DJF wind expansion coefficients on concurrent DJF sea level pressure (SLP) anomalies (Fig. 10). The patterns strongly resemble the NPO (Rogers 1981) and NAO (Hurrell et al. 2002) pattern for the Pacific and Atlantic cases, respectively, suggesting that the northern trade wind anomalies originate in large part from wintertime atmospheric variability over their respective northern basins. This interpretation for variability of the northern wintertime subtropical trades and their impact on the underlying SST through resulting latent heat flux anomalies has already been established for both the Pacific (e.g., Vimont et al. 2003a,b) and the Atlantic (e.g., Czaja et al. 2002; Xie and Tanimoto 1998), so we will not go into depth here. We stress that our analysis here does not preclude other sources of forcing on the meridional mode; indeed, ENSO is known to force the Atlantic meridional mode, but is filtered out of our analysis.

The interpretation of a purely unidirectional influence—forcing from the northern extratropics, and response in the Tropics—is likely too simplistic, given the possible feedback from the Tropics to the northern extratropical circulation. Our results showing the northern midlatitude atmospheric variability preceding the meridional mode event do suggest, however, that the tropical meridional mode response can be *initiated* by wintertime northern midlatitude events, and that the patterns of the wintertime forcing resemble intrinsic modes of the midlatitude variability. The issue of feedback cannot be resolved purely from analysis of observational data, and GCM studies will be required; indeed, a feedback mechanism has been proposed from a GCM study for the Atlantic case (Okumura et al. 2001); and recent work by Cassou et al. (2004) using the Action de Recherche Petite Echelle Grande Echelle (ARPEGE) atmospheric general circulation model also suggests forcing on the NAO by SST anomalies in the north tropical Atlantic. On the other hand, we note that our ensemble GCM simulations, while responding strongly in the Tropics to the meridional mode SST anomalies, has SLP anomalies that do not project on the NPO (NAO) pattern in the Pacific (Atlantic) (not shown). A detailed analysis of our model extratropical response to meridional mode SST anomalies is beyond the scope of our study; we simply note that there is possibility for Atlantic (Pacific) tropical feedback on the NAO (NPO), but this knowledge requires more comprehensive analysis, and taking into account the sensitivity of this connection across different atmospheric models.

The Southern Hemisphere does not appear to play much of a role in the mode 1 results of our MCA analysis. The SST and surface wind amplitude in the southern lobe is generally much smaller in both the mode 1 spatial patterns (Fig. 1) and also in the lag regression of zonally averaged SST and winds onto the MCA 1 DJF wind expansion coefficients (Fig. 9). Furthermore, the SLP regressed onto the DJF MCA 1 wind expansion coefficients does not project significantly on the southern Tropics (Fig. 10). The MCA 1 SST pattern in the Southern Hemisphere also appears to be different between the two basins (Fig. 1), although we noted a similar difference with the MAM deviation of SST from the annual mean (Fig. 7). While it does not preclude significant involvement by the Southern Hemisphere in our mode 1, establishing such involvement requires further investigation that is beyond the scope of our study. For our purposes, it is sufficient to note that both the Atlantic and Pacific MCA mode 1 have primarily Northern Hemisphere origins and with impact on the deep Tropics.

## 5. Composite analysis

As a final step, we composite NCEP reanalyses monthly anomaly SST, surface wind, and surface pressure fields based on the high and low years of the DJF-averaged wind expansion coefficients of the meridional modes. For the Pacific, we begin by excluding strong ENSO years (based on the same criteria as in section 3) from the analysis. Composites are taken around the nine largest positive and negative values of the DJF-averaged wind expansion coefficient (taken from the 35 ENSO neutral years). In the Atlantic, composites are taken around the nine largest positive and negative values of the DJF-averaged wind expansion coefficient (taken from the full 53 years available). For each basin, we compute means of the high and low years for each 3-month interval from the previous September–November through June–August, take their difference, and divide by 2. A two-sided *t* test shows significance of the difference in the means. The important thing to note here is that we use the “full” anomaly data without linear removal of the CTI in this analysis, as we want to show that the Pacific meridional mode pattern comes out even in the absence of this filter.

The composites (Fig. 11 for the Pacific, Fig. 12 for the Atlantic) show the evolution of the Pacific and Atlantic meridional modes for SST and surface wind anomalies (Figs. 11 and 12, left) and sea level pressure anomalies (Figs. 11 and 12, right). All that we had said before is apparent in these composites: the northern midlatitude circulation anomalies in boreal winter led to altered trade winds in the northern tropical region, and anomalous SST therein that maximizes in boreal spring. The SST-driven meridional circulation is coincident with the maximum meridional gradient in SST anomalies; and the northern tropical SST anomalies begin to decay by boreal summer. The composites also hint at similarities not mentioned before, in particular the equatorial southerlies in the western portion of the Pacific and Atlantic basins in June–August (JJA). Though there are also notable differences in the evolution of the meridional modes in the two basins—in particular, the southeastern tropical ocean response differs between the basins—the overall similarity between the spatial structures and physical mechanisms in the Pacific (Fig. 11) and Atlantic (Fig. 12) basins is striking.

The composites for DJF, MAM, and JJA remain essentially the same if we use the MAM SST expansion coefficients instead of the DJF wind coefficients to create the composites. The Pacific composite for the preceding September–October–November (SON) period using the MAM SST expansion coefficients, however, does change in this fashion: while the SON SST composite *pattern* resembles the Pacific SON pattern in Fig. 12, the peak SST amplitudes in the pattern increases by around 0.2 K in the central Pacific warm patch, and these SST differences survive the significance test. In other words, there is a suggestion of a precursor event in the SON equatorial Pacific SST prior to the peak MAM period. We think this difference comes about because the SST expansion coefficients preferentially picks years with stronger ENSO events (while our composite excludes strong ENSO years, it does not remove all years with ENSO events, obviously), more so that the wind expansion coefficients and the “precursor” SST in the central Pacific are part of the nonlinear ENSO signature [recall that the spatial pattern of MCA mode 1 in Fig. 4b, which was constructed using only extreme ENSO years, shows a sizable SST amplitude in the central equatorial Pacific that is absent in the MCA mode 1 using neutral ENSO years (Fig. 4a)]. Given this result, we leave open the possibility of a tropical precursor in our interpretation of the Pacific and Atlantic meridional modes. However, establishing the existence of precursors requires further elucidation that is beyond the scope of this paper (we note also another hint of an Atlantic meridional mode precursor in the Fig. 12 SON SST anomaly map in the western midlatitude Atlantic, although the area represented is very small). Our purpose is to point out the similarities between the Pacific and the Atlantic, and to this end we believe the composites reinforce this comparison.

## 6. Summary and discussion

The tropical Atlantic has a well-documented “meridional” mode of interannual–decadal atmosphere–ocean variability, distinct from the “zonal” ENSO-like mode. Several factors motivated us to look for an analogous mode of variability in the tropical Pacific, which has heretofore not been discovered: the fact that the two basins share similar ITCZ/cold tongue mean state climates and possess qualitatively similar midlatitude wintertime variability (NAO and NPO); also note the suggestion by idealized model studies that this variability *should* exist as part of the intrinsic variability of the ITCZ/cold tongue climate. To proceed, we employed MCA analysis to find coupled modes of variability between SST and surface wind in climatically equivalent regions of both tropical basins. The MCA had been employed successfully in previous studies (e.g., Chang et al. 1997) to extract the Atlantic meridional mode. Because we were not interested in ENSO (though ENSO dominates tropical Pacific variability), we removed ENSO from all fields prior to the analysis through linear regression onto CTI.

The dominant mode in the Atlantic was indeed the meridional mode, and the Pacific mode 1 resembled the Atlantic meridional mode in both spatial and temporal characteristics. A notable exception, however, was the substantially larger temporal persistence in the Pacific mode 1 expansion coefficients relative to the Atlantic. We realized, however, that the ENSO nonlinearity (the part of ENSO evolution not taken out by linear regression on the CTI) was projecting on the Pacific meridional mode during extreme ENSO events. By repeating the MCA without the extreme ENSO years, we showed that the original Pacific MCA mode 1 pattern derived was in fact robust, and that furthermore without the extreme ENSO years, the temporal persistence of the Pacific MCA mode reduced to the same levels as that for the Atlantic meridional mode.

We tentatively labeled the Pacific MCA 1 the Pacific meridional mode and proceeded to show that it had the same physical interpretation as the Atlantic meridional mode. In particular, we showed that (i) the deviation of boreal spring climate from the annual mean climate in both basins had spatial patterns that resembled the meridional mode in the respective basins; (ii) that the near-equatorial surface flow across the meridional SST gradient was coupled to the SST gradient; (iii) the anomalous trades in the northern tropical region drove the SST variability there through modulation of latent heat flux, thus giving rise to the near-equatorial meridional SST gradient in boreal spring; (iv) it did not appear, based on a ensemble GCM experiment, that the tropical response fed back onto the trade wind forcing. We additionally showed evidence that the anomalous trades were associated with NAO and NPO wintertime midlatitude atmospheric variability in the Atlantic and Pacific, respectively, though this had been previously established.

What are the potential implications of this study? First, we note a pleasing symmetry that the NAO and NPO are leading modes of wintertime atmospheric variability (Hsu and Wallace 1985) that can drive the meridional modes in their respective basins. This leads us to postulate that the meridional mode is an effective conduit for extratropical atmospheric influence on the Tropics. We think this is the consequence of two particular features of the atmosphere–ocean climate: the ability of the wintertime midlatitude variability to leave an anomalous “footprint” on surface ocean temperature (e.g., Vimont et al. 2001), and of the tropical atmosphere to be sensitive to relatively small (meridional) SST gradients (e.g., Chiang et al. 2002). If this is indeed correct, it means that the midlatitude atmospheric influence has to be given serious consideration in tropical marine climate variability. In particular, should the meridional mode event forced by the midlatitudes be picked up by the Bjerknes feedback in that basin, it could imply significant midlatitude control of ENSO and Atlantic Niño events. Indeed, such scenarios have already been proposed and explored (Servain et al. 1999; Vimont et al. 2003b) with potentially important implications for tropical climate prediction.

There is no intrinsic time scale associated with the meridional mode as described here, beyond what is implied by the ocean surface mixed-layer thermodynamics and the adjustment time of the atmosphere. The meridional mode physics should be equally applicable to mean state climate changes and to climate variability. In particular, if something causes (say) the northern Atlantic trade wind intensity to change, the tropical Atlantic climate and in particular the mean position of the Atlantic ITCZ should change accordingly. This scenario has been proposed to explain the apparent southward shift in the mean position of the tropical Atlantic ITCZ during the Last Glacial Maximum (LGM) 21 000 yr ago (Chiang et al. 2003). The mechanism for the northern trade wind increase during LGM is through perturbation of the North Atlantic stationary wave circulation brought about by the presence of the Laurentide ice sheet over North America.

It could still be argued that the Pacific mode we obtain is a result of statistical artifact—perhaps there is still nonlinearity in ENSO that has not been accounted for. While in principle we acknowledge that possibility, there is one compelling and general argument we can make against it: that the striking similarity of the Pacific mode to the well-established Atlantic meridional mode in both spatial and temporal characteristics, and in their interpretation, makes it unlikely that our result arises from statistical artifact. Our interpretation—that they arise from analogous processes given similar mean ITCZ/cold tongue climates—is simpler and more general.

## Acknowledgments

Special thanks to J. M. Wallace for his advice and encouragement. We also thank M. Biasutti, A. Czaja, C. Deser, S. Hastenrath, Y. Kushnir, and S.-P. Xie for useful discussions and comments. This research is funded by the NOAA Postdoctoral Program in Climate and Global Change, administered by the University Corporation for Atmospheric Research, and by the Joint Institute for the Study of the Atmosphere and Ocean (JISAO) under NOAA Cooperative Agreement NA17RJ1232. The authors contributed equally to this work.

## REFERENCES

Bretherton, C. S., C. Smith, and J. M. Wallace, 1992: An intercomparison of methods for finding coupled patterns in climate data.

,*J. Climate***5****,**541–560.Cassou, C., L. Terray, J. W. Hurrell, and C. Deser, 2004: North Atlantic winter climate regimes: Spatial asymmetry, stationarity with time, and oceanic forcing.

,*J. Climate***17****,**1055–1068.Chang, P., L. Ji, and H. Li, 1997: A decadal climate variation in the tropical Atlantic Ocean from thermodynamic air–sea interactions.

,*Nature***385****,**516–518.Chang, P., R. Saravanan, L. Ji, and G. C. Hegerl, 2000: The effect of local sea surface temperatures on the atmospheric circulation over the tropical Atlantic sector.

,*J. Climate***13****,**2195–2216.Chiang, J. C. H., S. E. Zebiak, and M. A. Cane, 2001: Relative roles of elevated heating and surface temperature gradients in driving anomalous surface winds over tropical oceans.

,*J. Atmos. Sci***58****,**1371–1394.Chiang, J. C. H., Y. Kushnir, and A. Giannini, 2002: Deconstructing Atlantic ITCZ variability: Influence of the local cross-equatorial SST gradient, and remote forcing from the eastern equatorial Pacific.

*J. Geophys. Res.,***107,**4004, doi:10.1029/2000JD000307.Chiang, J. C. H., M. Biasutti, and D. S. Battisti, 2003: Sensitivity of the Atlantic Intertropical Convergence Zone to Last Glacial Maximum boundary conditions.

*Paleoceanography,***18,**1094, doi:10.1029/ 2003PA000916.Curtis, S., and S. Hastenrath, 1995: Forcing of anomalous sea-surface temperature evolution in the tropical Atlantic during Pacific warm events.

,*J. Geophys. Res***100C****,**15835–15847.Czaja, A., P. van der Vaart, and J. Marshall, 2002: A diagnostic study of the role of remote forcing in tropical Atlantic variability.

,*J. Climate***15****,**3280–3290.Dommenget, D., and M. Latif, 2000: Interannual to decadal variability in the tropical Atlantic.

,*J. Climate***13****,**777–792.Enfield, D. B., and A. M. Mestas-Nunez, 1999: Multiscale variabilities in global sea surface temperatures and their relationships with tropospheric climate patterns.

,*J. Climate***12****,**2719–2733.Enfield, D. B., A. M. Mestas-Nunez, D. A. Mayer, and L. Cid-Serrano, 1999: How ubiquitous is the dipole relationship in tropical Atlantic sea surface temperatures?

,*J. Geophys. Res***104C****,**7841–7848.Giannini, A., Y. Kushnir, and M. A. Cane, 2000: Interannual variability of Caribbean rainfall, ENSO, and the Atlantic Ocean.

,*J. Climate***13****,**297–311.Hastenrath, S., and L. Heller, 1977: Dynamics of climatic hazards in northeast Brazil.

,*Quart. J. Roy. Meteor. Soc***103****,**77–92.Hastenrath, S., and L. Greischar, 1993: Circulation mechanisms related to northeast Brazil rainfall anomalies.

,*J. Geophys. Res***98D****,**5093–5102.Hoerling, M. P., A. Kumar, and M. Zhong, 1997: El Niño, La Niña, and the nonlinearity of their teleconnections.

,*J. Climate***10****,**1769–1786.Houghton, R. W., and Y. M. Tourre, 1992: Characteristic low-frequency sea surface temperature fluctuations in the tropical Atlantic.

,*J. Climate***5****,**765–771.Hsu, H. H., and J. M. Wallace, 1985: Vertical structure of wintertime teleconnection patterns.

,*J. Atmos. Sci***42****,**1693–1710.Hurrell, J. W., Y. Kushnir, G. Ottersen, and M. Visbeck, 2002: An overview of the North Atlantic Oscillation.

*The North Atlantic Oscillation: Climatic Significance and Environmental Impact, Geophys. Monogr.,*No. 134, Amer. Geophys. Union, 1–35.Kalnay, E., and Coauthors, 1996: The NCEP/NCAR 40-Year Reanalysis Project.

,*Bull. Amer. Meteor. Soc***77****,**437–471.Kaplan, A., M. A. Cane, Y. Kushnir, A. C. Clement, M. B. Blumenthal, and B. Rajagopalan, 1998: Analyses of global sea surface temperature 1856–1991.

,*J. Geophys. Res***103C****,**18567–18589.Kiehl, J. T., J. J. Hack, G. B. Bonan, B. A. Boville, D. L. Williamson, and P. J. Rasch, 1998: The National Center for Atmospheric Research Community Climate Model: CCM3.

,*J. Climate***11****,**1131–1149.Kistler, R., and Coauthors, 2001: The NCEP–NCAR 50-year reanalysis: Monthly means CD-ROM and documentation.

,*Bull. Amer. Meteor. Soc***82****,**247–267.Kushnir, Y., R. Seager, J. Miller, and J. C. H. Chiang, 2002: A simple coupled model of tropical Atlantic decadal climate variability.

*Geophys. Res. Lett.,***29,**2133, doi:10.1029/2002GL015874.Legler, D. M., and J. J. O'Brien, 1988: Tropical Pacific wind stress analysis for TOGA, IOC time series of ocean measurements. IOC Tech. Series 33, Vol. 4, UNESCO, 11–17.

Lindzen, R. S., and S. Nigam, 1987: On the role of sea-surface temperature gradients in forcing low-level winds and convergence in the Tropics.

,*J. Atmos. Sci***44****,**2440–2458.Mehta, V. M., and T. Delworth, 1995: Decadal variability of the tropical Atlantic Ocean surface temperature in shipboard measurements and in a global ocean–atmosphere model.

,*J. Climate***8****,**172–190.Mitchell, T. P., and J. M. Wallace, 1992: The annual cycle in equatorial convection and sea surface temperature.

,*J. Climate***5****,**1140–1156.Nobre, P., and J. Shukla, 1996: Variations of sea surface temperature, wind stress, and rainfall over the tropical Atlantic and South America.

,*J. Climate***9****,**2464–2479.Okumura, Y., S-P. Xie, A. Numaguti, and Y. Tanimoto, 2001: Tropical Atlantic air–sea interaction and its influence on the NAO.

,*Geophys. Res. Lett***28****,**1507–1510.Rogers, J. C., 1981: The North Pacific Oscillation.

,*J. Climatol***1****,**39–57.Ruiz-Barradas, A., J. A. Carton, and S. Nigam, 1999: Structure of interannual-to-decadal climate variability in the tropical Atlantic sector.

,*J. Climate***12****,**1–43.Servain, J., 1991: Simple climatic indices for the tropical Atlantic Ocean and some applications.

,*J. Geophys. Res***96****,**15137–15146.Servain, J., I. Wainer, J. P. McCreary, and A. Dessier, 1999: Relationship between the equatorial and meridional modes of climatic variability in the tropical Atlantic.

,*Geophys. Res. Lett***26****,**485–488.Sutton, R. T., S. P. Jewson, and D. P. Rowell, 2000: The elements of climate variability in the tropical Atlantic region.

,*J. Climate***13****,**3261–3284.Vimont, D. J., D. S. Battisti, and A. C. Hirst, 2001: Footprinting: A seasonal connection between the Tropics and mid-latitudes.

,*Geophys. Res. Lett***28****,**3923–3926.Vimont, D. J., D. S. Battisti, and A. C. Hirst, 2003a: The seasonal footprinting mechanism in the CSIRO general circulation models.

,*J. Climate***16****,**2653–2667.Vimont, D. J., J. M. Wallace, and D. S. Battisti, 2003b: The seasonal footprinting mechanism in the Pacific: Implications for ENSO.

,*J. Climate***16****,**2668–2675.Wallace, J. M., E. M. Rasmusson, T. P. Mitchell, V. E. Kousky, E. S. Sarachik, and H. von Storch, 1998: The structure and evolution of ENSO-related climate variability in the tropical Pacific: Lessons from TOGA.

,*J. Geophys. Res***103C****,**14241–14259.Xie, P. P., and P. A. Arkin, 1997: Global precipitation: A 17-year monthly analysis based on gauge observations, satellite estimates, and numerical model outputs.

,*Bull. Amer. Meteor. Soc***78****,**2539–2558.Xie, S-P., 1999: A dynamic ocean–atmosphere model of the tropical Atalantic decadal variability.

,*J. Climate***12****,**64–70.Xie, S-P., and Y. Tanimoto, 1998: A pan-Atlantic decadal climate oscillation.

,*Geophys. Res. Lett***25****,**2185–2188.Xie, S-P., and K. Saito, 2001: Formation and variability of a northerly ITCZ in a hybrid coupled AGCM: Continental forcing and oceanic–atmospheric feedback.

,*J. Climate***14****,**1262–1276.Yukimoto, S., M. Endoh, Y. Kitamura, A. Kitoh, T. Motoi, and A. Noda, 2000: ENSO-like interdecadal variability in the Pacific Ocean as simulated in a coupled general circulation model.

,*J. Geophys. Res***105C****,**13945–13963.

Temporal properties of the leading MCA mode: (left) Pacific and (right) Atlantic. (top row) Variance computed by the month for the SST (light shading) and wind (dark shading) expansion coefficients. (second row) Lagged correlation between the SST and wind expansion coefficients. The simultaneous correlation is the black bar, and dark (light) shaded bars are for winds leading (lagging) SST. For reference, correlations above 0.27 are significant at the 95% confidence level assuming a decorrelation time scale of 1 yr (54 independent samples). (third and fourth rows) SST expansion coefficients for the Pacific and Atlantic MCA mode 1

Citation: Journal of Climate 17, 21; 10.1175/JCLI4953.1

Temporal properties of the leading MCA mode: (left) Pacific and (right) Atlantic. (top row) Variance computed by the month for the SST (light shading) and wind (dark shading) expansion coefficients. (second row) Lagged correlation between the SST and wind expansion coefficients. The simultaneous correlation is the black bar, and dark (light) shaded bars are for winds leading (lagging) SST. For reference, correlations above 0.27 are significant at the 95% confidence level assuming a decorrelation time scale of 1 yr (54 independent samples). (third and fourth rows) SST expansion coefficients for the Pacific and Atlantic MCA mode 1

Citation: Journal of Climate 17, 21; 10.1175/JCLI4953.1

Temporal properties of the leading MCA mode: (left) Pacific and (right) Atlantic. (top row) Variance computed by the month for the SST (light shading) and wind (dark shading) expansion coefficients. (second row) Lagged correlation between the SST and wind expansion coefficients. The simultaneous correlation is the black bar, and dark (light) shaded bars are for winds leading (lagging) SST. For reference, correlations above 0.27 are significant at the 95% confidence level assuming a decorrelation time scale of 1 yr (54 independent samples). (third and fourth rows) SST expansion coefficients for the Pacific and Atlantic MCA mode 1

Citation: Journal of Climate 17, 21; 10.1175/JCLI4953.1

MAM-averaged values of the Pacific MCA 1 SST expansion coefficients vs CTI averaged over NDJ (ctiNDJ). The reverse U shape of the scatterplot suggests that extreme ENSO events (both positive and negative) favor low values of the MAM Pacific meridional mode

Citation: Journal of Climate 17, 21; 10.1175/JCLI4953.1

MAM-averaged values of the Pacific MCA 1 SST expansion coefficients vs CTI averaged over NDJ (ctiNDJ). The reverse U shape of the scatterplot suggests that extreme ENSO events (both positive and negative) favor low values of the MAM Pacific meridional mode

Citation: Journal of Climate 17, 21; 10.1175/JCLI4953.1

MAM-averaged values of the Pacific MCA 1 SST expansion coefficients vs CTI averaged over NDJ (ctiNDJ). The reverse U shape of the scatterplot suggests that extreme ENSO events (both positive and negative) favor low values of the MAM Pacific meridional mode

Citation: Journal of Climate 17, 21; 10.1175/JCLI4953.1

Results of the MCA on the tropical Pacific domain, using only (left column) ENSO “neutral” years and (right column) only extreme ENSO years to highlight the influence of ENSO nonlinearity on the meridional mode. (top row) SST and wind fields regressed onto the SST expansion coefficient for (a) neutral and (b) extreme ENSO years. The contour interval is 0.1°C (std dev)^{−1}. Wind vectors are plotted where the geometric sum of their correlation coefficients exceeds the 95% confidence level. The reference wind vector is 0.5 m s^{−1}. Shading denotes SST regions where the correlation coefficients exceed the 95% confidence level. (middle row) Precipitation regressed onto the SST expansion coefficient for (c) neutral and (d) extreme ENSO years. Contour interval: (c) 0.2 mm day^{−1} (std dev)^{−1}, (d) 0.4 mm day^{−1} (std dev)^{−1}. (a)–(d) Solid contours denote positive anomalies, dashed contours denote negative anomalies, and the zero contour has been omitted. (bottom row) (e) and (g) Variance computed by the month for the SST (light gray) and wind (dark gray) expansion coefficients, similar to Fig. 2, top row. (f) and (h) Lagged correlation between the SST and wind expansion coefficients, similar to Fig. 2, second row. Dark (light) gray indicates that the wind expansion coefficient leads (lags) the SST expansion coefficient. Lagged correlations greater than 0.34 in (f) [greater than 0.46 in (h)] exceed the 95% confidence level

Citation: Journal of Climate 17, 21; 10.1175/JCLI4953.1

Results of the MCA on the tropical Pacific domain, using only (left column) ENSO “neutral” years and (right column) only extreme ENSO years to highlight the influence of ENSO nonlinearity on the meridional mode. (top row) SST and wind fields regressed onto the SST expansion coefficient for (a) neutral and (b) extreme ENSO years. The contour interval is 0.1°C (std dev)^{−1}. Wind vectors are plotted where the geometric sum of their correlation coefficients exceeds the 95% confidence level. The reference wind vector is 0.5 m s^{−1}. Shading denotes SST regions where the correlation coefficients exceed the 95% confidence level. (middle row) Precipitation regressed onto the SST expansion coefficient for (c) neutral and (d) extreme ENSO years. Contour interval: (c) 0.2 mm day^{−1} (std dev)^{−1}, (d) 0.4 mm day^{−1} (std dev)^{−1}. (a)–(d) Solid contours denote positive anomalies, dashed contours denote negative anomalies, and the zero contour has been omitted. (bottom row) (e) and (g) Variance computed by the month for the SST (light gray) and wind (dark gray) expansion coefficients, similar to Fig. 2, top row. (f) and (h) Lagged correlation between the SST and wind expansion coefficients, similar to Fig. 2, second row. Dark (light) gray indicates that the wind expansion coefficient leads (lags) the SST expansion coefficient. Lagged correlations greater than 0.34 in (f) [greater than 0.46 in (h)] exceed the 95% confidence level

Citation: Journal of Climate 17, 21; 10.1175/JCLI4953.1

Results of the MCA on the tropical Pacific domain, using only (left column) ENSO “neutral” years and (right column) only extreme ENSO years to highlight the influence of ENSO nonlinearity on the meridional mode. (top row) SST and wind fields regressed onto the SST expansion coefficient for (a) neutral and (b) extreme ENSO years. The contour interval is 0.1°C (std dev)^{−1}. Wind vectors are plotted where the geometric sum of their correlation coefficients exceeds the 95% confidence level. The reference wind vector is 0.5 m s^{−1}. Shading denotes SST regions where the correlation coefficients exceed the 95% confidence level. (middle row) Precipitation regressed onto the SST expansion coefficient for (c) neutral and (d) extreme ENSO years. Contour interval: (c) 0.2 mm day^{−1} (std dev)^{−1}, (d) 0.4 mm day^{−1} (std dev)^{−1}. (a)–(d) Solid contours denote positive anomalies, dashed contours denote negative anomalies, and the zero contour has been omitted. (bottom row) (e) and (g) Variance computed by the month for the SST (light gray) and wind (dark gray) expansion coefficients, similar to Fig. 2, top row. (f) and (h) Lagged correlation between the SST and wind expansion coefficients, similar to Fig. 2, second row. Dark (light) gray indicates that the wind expansion coefficient leads (lags) the SST expansion coefficient. Lagged correlations greater than 0.34 in (f) [greater than 0.46 in (h)] exceed the 95% confidence level

Citation: Journal of Climate 17, 21; 10.1175/JCLI4953.1

Mode 1 SST expansion coefficients from the Pacific MCA of the entire record (top curve) for the ENSO neutral years only (middle curve, within the nonshaded regions) and for the extreme ENSO years only (middle curve, within the shaded regions), plotted together with the CTI (bottom curve). The months corresponding to the extreme ENSO years are denoted by the shading. Note that the expansion coefficient for the entire record (top curve) is nearly identical to the expansion coefficients for the neutral and extreme ENSO years combined (middle curve)

Citation: Journal of Climate 17, 21; 10.1175/JCLI4953.1

Mode 1 SST expansion coefficients from the Pacific MCA of the entire record (top curve) for the ENSO neutral years only (middle curve, within the nonshaded regions) and for the extreme ENSO years only (middle curve, within the shaded regions), plotted together with the CTI (bottom curve). The months corresponding to the extreme ENSO years are denoted by the shading. Note that the expansion coefficient for the entire record (top curve) is nearly identical to the expansion coefficients for the neutral and extreme ENSO years combined (middle curve)

Citation: Journal of Climate 17, 21; 10.1175/JCLI4953.1

Mode 1 SST expansion coefficients from the Pacific MCA of the entire record (top curve) for the ENSO neutral years only (middle curve, within the nonshaded regions) and for the extreme ENSO years only (middle curve, within the shaded regions), plotted together with the CTI (bottom curve). The months corresponding to the extreme ENSO years are denoted by the shading. Note that the expansion coefficient for the entire record (top curve) is nearly identical to the expansion coefficients for the neutral and extreme ENSO years combined (middle curve)

Citation: Journal of Climate 17, 21; 10.1175/JCLI4953.1

A demonstration of the how “nonlinear ENSO” projects on the meridional mode. (a) The CTI-weighted mean of the nine high ctiNDJ years, and (b) the same for the nine low ctiNDJ years. Note that in (a) and (b) the mean SST and winds are divided by the sum of the ctiNDJ, which is negative for the “low” years, so that the polarity of (b) is opposite what would be expected from a negative ENSO event. (c) An estimate of “linear” ENSO, the sum of (a) and (b). (d) The nonlinear component of ENSO estimated as the difference between (a) and (b). (a)–(d) SST is contoured [contour interval (a) and (b) 0.4 K, in (c) 0.8 K, and (d) 0.2 K] and vectors denote 10-m winds. Solid contours denote positive SST anomalies, dashed contours denote negative SST anomalies, and the zero contour has been omitted

Citation: Journal of Climate 17, 21; 10.1175/JCLI4953.1

A demonstration of the how “nonlinear ENSO” projects on the meridional mode. (a) The CTI-weighted mean of the nine high ctiNDJ years, and (b) the same for the nine low ctiNDJ years. Note that in (a) and (b) the mean SST and winds are divided by the sum of the ctiNDJ, which is negative for the “low” years, so that the polarity of (b) is opposite what would be expected from a negative ENSO event. (c) An estimate of “linear” ENSO, the sum of (a) and (b). (d) The nonlinear component of ENSO estimated as the difference between (a) and (b). (a)–(d) SST is contoured [contour interval (a) and (b) 0.4 K, in (c) 0.8 K, and (d) 0.2 K] and vectors denote 10-m winds. Solid contours denote positive SST anomalies, dashed contours denote negative SST anomalies, and the zero contour has been omitted

Citation: Journal of Climate 17, 21; 10.1175/JCLI4953.1

A demonstration of the how “nonlinear ENSO” projects on the meridional mode. (a) The CTI-weighted mean of the nine high ctiNDJ years, and (b) the same for the nine low ctiNDJ years. Note that in (a) and (b) the mean SST and winds are divided by the sum of the ctiNDJ, which is negative for the “low” years, so that the polarity of (b) is opposite what would be expected from a negative ENSO event. (c) An estimate of “linear” ENSO, the sum of (a) and (b). (d) The nonlinear component of ENSO estimated as the difference between (a) and (b). (a)–(d) SST is contoured [contour interval (a) and (b) 0.4 K, in (c) 0.8 K, and (d) 0.2 K] and vectors denote 10-m winds. Solid contours denote positive SST anomalies, dashed contours denote negative SST anomalies, and the zero contour has been omitted

Citation: Journal of Climate 17, 21; 10.1175/JCLI4953.1

Deviations of the Mar–May climatological SST and winds from the annual mean, showing their resemblance to the meridional mode patterns. To facilitate comparison with Fig. 1, the sign of every field has been reversed, and the color scale offset so that white represents values of SST between −1.5 and −0.5 K. Note also that the SST color scale relative to the magnitude of the wind vector is the same as in Fig. 1. Hence, the magnitude of the wind vector can be directly compared to the strength of the SST gradient as indicated by the colors. The zero contour has been added as a solid line. Winds are plotted where the amplitude of the vector anomaly exceeds 1 m s^{−1}

Citation: Journal of Climate 17, 21; 10.1175/JCLI4953.1

Deviations of the Mar–May climatological SST and winds from the annual mean, showing their resemblance to the meridional mode patterns. To facilitate comparison with Fig. 1, the sign of every field has been reversed, and the color scale offset so that white represents values of SST between −1.5 and −0.5 K. Note also that the SST color scale relative to the magnitude of the wind vector is the same as in Fig. 1. Hence, the magnitude of the wind vector can be directly compared to the strength of the SST gradient as indicated by the colors. The zero contour has been added as a solid line. Winds are plotted where the amplitude of the vector anomaly exceeds 1 m s^{−1}

Citation: Journal of Climate 17, 21; 10.1175/JCLI4953.1

Deviations of the Mar–May climatological SST and winds from the annual mean, showing their resemblance to the meridional mode patterns. To facilitate comparison with Fig. 1, the sign of every field has been reversed, and the color scale offset so that white represents values of SST between −1.5 and −0.5 K. Note also that the SST color scale relative to the magnitude of the wind vector is the same as in Fig. 1. Hence, the magnitude of the wind vector can be directly compared to the strength of the SST gradient as indicated by the colors. The zero contour has been added as a solid line. Winds are plotted where the amplitude of the vector anomaly exceeds 1 m s^{−1}

Citation: Journal of Climate 17, 21; 10.1175/JCLI4953.1

Results of the CCM3 model experiments with imposed SST anomalies from the leading MCA mode, showing the SST-forced component of the surface wind and precipitation associated with the meridional mode. Shown is one-half of the difference between the mean of an ensemble of eight simulations with the polarity of SST anomalies in Fig. 1, and the same but with SST anomalies of the opposite polarity. Vectors denote 995-mb winds (reference vector is 2 m s^{−1}) and gray shading denotes precipitation anomalies in mm day^{−1}. Winds are plotted only where significant at the 95% confidence level (based on a multivariate *t* test for both components of the wind). In general, shaded regions (precipitation) exceed the 95% confidence level (based on a two-tailed *t* test)

Citation: Journal of Climate 17, 21; 10.1175/JCLI4953.1

Results of the CCM3 model experiments with imposed SST anomalies from the leading MCA mode, showing the SST-forced component of the surface wind and precipitation associated with the meridional mode. Shown is one-half of the difference between the mean of an ensemble of eight simulations with the polarity of SST anomalies in Fig. 1, and the same but with SST anomalies of the opposite polarity. Vectors denote 995-mb winds (reference vector is 2 m s^{−1}) and gray shading denotes precipitation anomalies in mm day^{−1}. Winds are plotted only where significant at the 95% confidence level (based on a multivariate *t* test for both components of the wind). In general, shaded regions (precipitation) exceed the 95% confidence level (based on a two-tailed *t* test)

Citation: Journal of Climate 17, 21; 10.1175/JCLI4953.1

Results of the CCM3 model experiments with imposed SST anomalies from the leading MCA mode, showing the SST-forced component of the surface wind and precipitation associated with the meridional mode. Shown is one-half of the difference between the mean of an ensemble of eight simulations with the polarity of SST anomalies in Fig. 1, and the same but with SST anomalies of the opposite polarity. Vectors denote 995-mb winds (reference vector is 2 m s^{−1}) and gray shading denotes precipitation anomalies in mm day^{−1}. Winds are plotted only where significant at the 95% confidence level (based on a multivariate *t* test for both components of the wind). In general, shaded regions (precipitation) exceed the 95% confidence level (based on a two-tailed *t* test)

Citation: Journal of Climate 17, 21; 10.1175/JCLI4953.1

Lead–lag regression of monthly mean, zonally averaged anomalies onto the normalized DJF wind expansion coefficient of the (top) Pacific and (bottom) Atlantic MCA leading mode (see Fig. 1) from Jul prior to the wind DJF through the Jan following. This follows a technique used by Czaja et al. (2002). Data are zonally averaged from 165° to 115°W in the Pacific, and from 60°W to 0° in the Atlantic. A 3-month running mean is applied to the data, and the cold tongue index is linearly removed prior to analysis; also, extreme ENSO years (as defined in section 3) are not used in the Pacific analysis. Colors indicate SST anomalies, vectors are 10-m winds, and contours denote net surface heat flux (contour interval is 2 W m^{−2}). The surface wind vectors are plotted where the magnitude of either component of the correlation vector exceeds 0.27 (the 95% confidence level). Solid contours denote downward heat flux anomalies (into the ocean), dashed contours denote upward heat flux anomalies (out of the ocean), and the zero contour has been omitted. In general, SST amplitudes greater than about 0.075°C (std dev)^{−1} and heat flux amplitudes greater than about 3 W m^{−2} (std dev)^{−1} exceed the 95% confidence level

Citation: Journal of Climate 17, 21; 10.1175/JCLI4953.1

Lead–lag regression of monthly mean, zonally averaged anomalies onto the normalized DJF wind expansion coefficient of the (top) Pacific and (bottom) Atlantic MCA leading mode (see Fig. 1) from Jul prior to the wind DJF through the Jan following. This follows a technique used by Czaja et al. (2002). Data are zonally averaged from 165° to 115°W in the Pacific, and from 60°W to 0° in the Atlantic. A 3-month running mean is applied to the data, and the cold tongue index is linearly removed prior to analysis; also, extreme ENSO years (as defined in section 3) are not used in the Pacific analysis. Colors indicate SST anomalies, vectors are 10-m winds, and contours denote net surface heat flux (contour interval is 2 W m^{−2}). The surface wind vectors are plotted where the magnitude of either component of the correlation vector exceeds 0.27 (the 95% confidence level). Solid contours denote downward heat flux anomalies (into the ocean), dashed contours denote upward heat flux anomalies (out of the ocean), and the zero contour has been omitted. In general, SST amplitudes greater than about 0.075°C (std dev)^{−1} and heat flux amplitudes greater than about 3 W m^{−2} (std dev)^{−1} exceed the 95% confidence level

Citation: Journal of Climate 17, 21; 10.1175/JCLI4953.1

Lead–lag regression of monthly mean, zonally averaged anomalies onto the normalized DJF wind expansion coefficient of the (top) Pacific and (bottom) Atlantic MCA leading mode (see Fig. 1) from Jul prior to the wind DJF through the Jan following. This follows a technique used by Czaja et al. (2002). Data are zonally averaged from 165° to 115°W in the Pacific, and from 60°W to 0° in the Atlantic. A 3-month running mean is applied to the data, and the cold tongue index is linearly removed prior to analysis; also, extreme ENSO years (as defined in section 3) are not used in the Pacific analysis. Colors indicate SST anomalies, vectors are 10-m winds, and contours denote net surface heat flux (contour interval is 2 W m^{−2}). The surface wind vectors are plotted where the magnitude of either component of the correlation vector exceeds 0.27 (the 95% confidence level). Solid contours denote downward heat flux anomalies (into the ocean), dashed contours denote upward heat flux anomalies (out of the ocean), and the zero contour has been omitted. In general, SST amplitudes greater than about 0.075°C (std dev)^{−1} and heat flux amplitudes greater than about 3 W m^{−2} (std dev)^{−1} exceed the 95% confidence level

Citation: Journal of Climate 17, 21; 10.1175/JCLI4953.1

Simultaneous regression maps of DJF sea level pressure onto the standardized DJF wind expansion coefficients for (a) the Pacific MCA mode 1 and (b) the Atlantic MCA mode 1, suggesting that the origins of the north tropical Pacific and Atlantic trade wind forcing originates from the NPO and NAO, respectively. Contour interval is 0.4 mb (std dev)^{−1}. Light shading denotes regions where the correlation coefficient exceeds the 95% confidence level

Citation: Journal of Climate 17, 21; 10.1175/JCLI4953.1

Simultaneous regression maps of DJF sea level pressure onto the standardized DJF wind expansion coefficients for (a) the Pacific MCA mode 1 and (b) the Atlantic MCA mode 1, suggesting that the origins of the north tropical Pacific and Atlantic trade wind forcing originates from the NPO and NAO, respectively. Contour interval is 0.4 mb (std dev)^{−1}. Light shading denotes regions where the correlation coefficient exceeds the 95% confidence level

Citation: Journal of Climate 17, 21; 10.1175/JCLI4953.1

Simultaneous regression maps of DJF sea level pressure onto the standardized DJF wind expansion coefficients for (a) the Pacific MCA mode 1 and (b) the Atlantic MCA mode 1, suggesting that the origins of the north tropical Pacific and Atlantic trade wind forcing originates from the NPO and NAO, respectively. Contour interval is 0.4 mb (std dev)^{−1}. Light shading denotes regions where the correlation coefficient exceeds the 95% confidence level

Citation: Journal of Climate 17, 21; 10.1175/JCLI4953.1

Composite analysis of the Pacific meridional mode based on positive and negative values of the DJF wind expansion coefficients of MCA mode 1. Only neutral ENSO years (as described in section 3) are used. The composites are in 3-month averages from Sep–Nov through the following Jun–Aug (thus, the first row indicates variability that occurs 6 months before the peak of the MAM SST expansion coefficient, and so forth). The maps are generated by subtracting the mean of the “negative” years from the mean of the “positive” years, and dividing by 2. (left column) SST (shaded) and surface wind (reference vector 1 m s^{−1}) differences. Regions where the SST is statistically significant at the 95% level are enclosed by the black contour line. For graphical purposes only, wind contours are plotted only where the F value exceeds the 80% level based on a multivariate *t* test. (right column) Sea level pressure (contour interval 0.4 mb) differences. Significant regions (based on 95% confidence of a two-sided *t* test) are lightly shaded. Solid contours denote positive anomalies, dashed contours denote negative anomalies, and the zero contour has been omitted

Citation: Journal of Climate 17, 21; 10.1175/JCLI4953.1

Composite analysis of the Pacific meridional mode based on positive and negative values of the DJF wind expansion coefficients of MCA mode 1. Only neutral ENSO years (as described in section 3) are used. The composites are in 3-month averages from Sep–Nov through the following Jun–Aug (thus, the first row indicates variability that occurs 6 months before the peak of the MAM SST expansion coefficient, and so forth). The maps are generated by subtracting the mean of the “negative” years from the mean of the “positive” years, and dividing by 2. (left column) SST (shaded) and surface wind (reference vector 1 m s^{−1}) differences. Regions where the SST is statistically significant at the 95% level are enclosed by the black contour line. For graphical purposes only, wind contours are plotted only where the F value exceeds the 80% level based on a multivariate *t* test. (right column) Sea level pressure (contour interval 0.4 mb) differences. Significant regions (based on 95% confidence of a two-sided *t* test) are lightly shaded. Solid contours denote positive anomalies, dashed contours denote negative anomalies, and the zero contour has been omitted

Citation: Journal of Climate 17, 21; 10.1175/JCLI4953.1

Composite analysis of the Pacific meridional mode based on positive and negative values of the DJF wind expansion coefficients of MCA mode 1. Only neutral ENSO years (as described in section 3) are used. The composites are in 3-month averages from Sep–Nov through the following Jun–Aug (thus, the first row indicates variability that occurs 6 months before the peak of the MAM SST expansion coefficient, and so forth). The maps are generated by subtracting the mean of the “negative” years from the mean of the “positive” years, and dividing by 2. (left column) SST (shaded) and surface wind (reference vector 1 m s^{−1}) differences. Regions where the SST is statistically significant at the 95% level are enclosed by the black contour line. For graphical purposes only, wind contours are plotted only where the F value exceeds the 80% level based on a multivariate *t* test. (right column) Sea level pressure (contour interval 0.4 mb) differences. Significant regions (based on 95% confidence of a two-sided *t* test) are lightly shaded. Solid contours denote positive anomalies, dashed contours denote negative anomalies, and the zero contour has been omitted

Citation: Journal of Climate 17, 21; 10.1175/JCLI4953.1

Same as in Fig. 11 but for the Atlantic meridional mode. Unlike the Pacific composite, the positive and negative composite years here are taken from the full 53 yr of the available data

Citation: Journal of Climate 17, 21; 10.1175/JCLI4953.1

Same as in Fig. 11 but for the Atlantic meridional mode. Unlike the Pacific composite, the positive and negative composite years here are taken from the full 53 yr of the available data

Citation: Journal of Climate 17, 21; 10.1175/JCLI4953.1

Same as in Fig. 11 but for the Atlantic meridional mode. Unlike the Pacific composite, the positive and negative composite years here are taken from the full 53 yr of the available data

Citation: Journal of Climate 17, 21; 10.1175/JCLI4953.1

^{*}

National Oceanic and Atmospheric Administration Contribution Number 1010.