1. Introduction

Numerous studies have been done on the linkage between the anomalous tropical Pacific sea surface temperature (SST) and the interannual variability of the extratropical circulation (Trenberth et al. 1998). The extratropical response to the El Niño forcing is a “Pacific– North American (PNA)” teleconnection pattern (Wallace and Gutzler 1981). This signal accounts for a significant part of the variance of interannual variability in the midlatitude North Pacific and North America, and is a dominant source of skill for seasonal forecasts (Zwiers 1987; Derome et al. 2001).

Diagnostic studies of the El Niño–Southern Oscillation (ENSO) often illustrate the associated global climate anomalies as the difference between warm and cold events (e.g., van Loon and Rogers 1981; Pan and Oort 1983; Kiladis and Diaz 1989), or as linear correlations or regressions between ENSO indices and the global circulation anomalies (e.g., Horel and Wallace 1981). It is assumed implicitly in these approaches that the climate response is mainly linear. In fact, the PNA response is rather well explained by linear wave propagation theory (Hoskins and Karoly 1981; Simmons 1982).

Some atmospheric anomalies associated with ENSO also exhibit appreciable nonlinearity. Hoerling et al. (1997) found that the North American surface temperature anomalies during El Niño and La Niña are almost in quadrature, and there is a phase shift of about 35° longitude in the upper-atmospheric response between the warm and cold events. The existence of nonlinearity is supported by other studies (e.g., Livezey et al. 1997; Montroy et al. 1998).

The nonlinear component of the response is less well understood. An anomalous SST in the tropical Pacific induces anomalous diabatic heating by changing the moisture supply and local circulation. It is this anomalous diabatic heating that acts as a forcing to the global atmosphere. Even if the SST anomalies were symmetric (equal in amplitude and opposite in sign), the diabatic heating anomalies would be asymmetric. Convection and precipitation in oceanic regions depend on the total underlying value of SST (e.g., Gadgil et al. 1984). Across the tropical Pacific, the climatological SST in the east is much cooler than the western warm pool. Hoerling et al. (1997) found that the maximum tropical rainfall anomalies are located east of the date line during warm events, but west of the date line during cold events, and they attribute the nonlinearity of the extratropical response to this phase shift of the tropical deep convection and diabatic heating.

There are other possible mechanisms that may lead to the nonlinearity of the extratropical response to tropical SST anomalies. For example, the direct response to tropical forcing changes the midlatitude flow, which modifies the wave energy propagation from the Tropics. There is no consistent result on the sensitivity of the response to the background flow. The barotropic experiments by Hoerling et al. (1997) show that the extratropical response to the tropical Pacific SST anomalies does not depend on whether the model was linearized about the zonally varying flow of El Niño or La Niña states. Other authors, for example, Hall and Derome (2000) and Ting and Sardeshmukh (1993), found that small variations in the basic state can make important differences in the response. Another possible mechanism for the nonlinearity is related to the transient eddy feedback, that is, the forced low-frequency flow pattern in the midlatitude flow—the direct response to tropical forcing—modulating the storm track activity, which then feeds back onto the direct response. The anomalous vorticity fluxes by the midlatitude transient eddies are found to reinforce the low-frequency flow component, whereas their thermal fluxes are dissipative (Lau 1988; Held et al. 1989; Sheng et al. 1998; Lin and Derome 1995).

In this study, we investigate the dependence of the extratropical atmospheric response on the amplitude of the tropical forcing. By fixing the diabatic forcing anomaly pattern, the effect of the phase shift of the tropical deep convection and diabatic heating is minimized. In this case, if the extratropical response still exhibits nonlinearity, it must be largely ascribed to the nonlinear midlatitude dynamics. Using a dry primitive equation model Hall and Derome (2000) found an asymmetry between the responses to tropical heating and cooling for forcing perturbations that are equal and opposite. Here, by using the same model with a different configuration of forcing, a systematic study of this problem is conducted. A large set of ensemble experiments is performed, the leading patterns of the extratropical response are identified, and their amplitudes as a function of the forcing are analyzed.

A brief description of the model and the experimental design are given in section 2. In section 3, a composite analysis is first presented for the observed extreme El Niño and La Niña winters, the model results with El Niño and La Niña forcings are then presented, and comparisons are made. In section 4, two leading modes of response are identified, and the dependence of their amplitudes on the forcing amplitude is discussed. In section 5, a set of linear experiments is performed to understand the mechanisms behind the asymmetry and nonlinearity in the extratropical response to the amplitude of tropical forcing. In section 6, the sensitivity of the response to the longitude of the forcing is discussed. Section 7 gives a summary and discussion.

2. The model and experimental design

The model used in this study is a primitive equation dry atmospheric model as described in detail in Hall (2000) and Hall et al. (2001a,b). It has a global domain with horizontal resolution of T21 and five vertical levels. An important feature of this model is that it uses a time-averaged forcing calculated empirically from observed daily data. By computing the dynamical terms of the model, together with a linear damping, with daily global analyses and averaging in time, the residual term for each time tendency equation is obtained as the forcing. This forcing, thus, includes all processes that are not resolved by the model's dynamics such as diabatic heating (including latent heat release related to the transient eddies) and the deviation of dissipative processes from linear damping. The model has no orography, so the forcing also mimics the time mean orographic forcing. Due to its simplicity in physics, it is referred to as a “simple general circulation model (SGCM).” As shown in Hall (2000), this model is able to reproduce remarkably realistic stationary planetary waves and the broad climatological characteristics of the transients are in general agreement with observations. This model was also used to do seasonal predictions, and was found to be similar in skill to a more complex GCM (Derome et al. 2003, manuscript submitted to J. Climate).

The daily data required to calculate the forcing are taken from the National Centers for Environmental Prediction–National Center for Atmospheric Research (NCEP–NCAR) reanalyses (Kalnay et al. 1996). To represent the interannual variability a forcing field is calculated separately for each of the 51 winters [December–January–February (DJF)] from 1948/49 to 1998/99. When the model is run with a different time-independent forcing for each winter it is able to simulate the structure, the interannual variability, and trend of the observed North Atlantic Oscillation (NAO) and Arctic Oscillation (AO) (Peterson et al. 2002; Lin et al. 2002; Greatbatch et al. 2003).

To obtain the forcing anomaly associated with the El Niño–Southern Oscillation, an EOF analysis is first performed on DJF seasonal mean SST over the tropical Pacific. The SSTs are also taken from the NCEP–NCAR reanalyses and have a 51-winter coverage. El Niño's signal is identified as the first mode of the EOF analysis for the SST in the tropical Pacific (40°S–40°N, 120°E– 90°W). The structure of the first EOF, a typical El Niño SST pattern (not shown), explains about 30% of the variance in that area. The interannual variability of El Niño is represented by the principal component (PC) of the leading mode. When results are compared with those obtained with the Global Sea Ice Coverage and Sea Surface Temperature version 2.2 (GISST2.2) dataset (Parker et al. 1995a,b) during the overlapping 47 winters, the agreement is very good. Next, a forcing pattern is obtained as a linear regression of the model forcing fields upon the PC of the first EOF of SST, and, thus, represents a linear fit of the forcing to the El Niño SST pattern. Shown in Fig. 1 is the vertically averaged thermal forcing. Heating is seen in the central equatorial Pacific with a vertically averaged maximum near 2°C day⁻¹, corresponding to one standard deviation of the El Niño SST pattern, with cooling bands to its north and south. It should be noted that the diabatic heating is the sum of this forcing and the linear temperature damping. We have looked at the corresponding diabatic heating (not shown), and found that it has almost the same distribution as Fig. 1 in the tropical Pacific. The high-amplitude positive temperature forcing to the west of Australia and over Southeast Asia in Fig. 1 is opposed by strong thermal damping, resulting in no net diabatic heating west of Australia and only weak heating over Southeast Asia.

With the obtained forcing pattern fixed in space, atmospheric response experiments can be performed for its positive and negative phases with different amplitudes. We choose the forcing amplitudes to correspond to the 51 values of the PC of El Niño's SST mode (PC1). In this way, they are distributed in a reasonable range, and each forcing represents a linear fit of the model forcing to the El Niño signal in a winter between 1948/ 49 and 1998/99. In the discussion to follow, we will refer to the PC of the El Niño SST mode as the forcing amplitude. The extreme values of PC1 in the winters of 1982/83 (2.57) and 1997/98 (2.61) create a gap with the PC1 value (1.41) of the third strongest El Niño winter (1957/58). Here, the PC values are normalized with their standard deviation. In order to better see the relationship of the response to the distribution of the forcing amplitude, we insert 10 equally spaced PC1 values to fill the gap. Therefore, a total of 61 forcing amplitudes are used in the numerical experiment.

For each amplitude of the forcing pattern, an ensemble of 30 integrations of 120 days is performed starting from different initial conditions. The initial conditions are randomly chosen from winter days over the 51 yr. The analysis is done over the final 90 days in order to allow time for the model to adjust to the forcing. The ensemble average of the atmospheric anomaly field represents a forced signal by the anomalous forcing field.

To interpret the model response, a linear perturbation model is used that is based on the same primitive equation model, with the approach as described in Hall and Derome (2000). Once a basic state is chosen, it is substituted into the SGCM equations and the forcing is taken to be that which will keep the basic state at equilibrium. A time independent small-amplitude anomalous forcing is then injected at t = 0, and the evolution of the forced response is followed in time over a period that is sufficiently short to keep the integration linear. The anomalous forcing has the same structure as the linear fit of the model forcing to the El Niño signal, as discussed above. For the linear experiments, the anomalous forcing amplitude is scaled by 10⁻⁴ times the anomalous forcing corresponding to one standard deviation of the SST PC1, and the response is then scaled back for display purposes.

3. Composite study

To examine the Northern Hemisphere circulation patterns associated with warm and cold events, a composite study is performed for the observations. The selection of warm and cold events is based on the PC of the first EOF of winter SST in the tropical Pacific as discussed in section 2. Winters (DJF) with a PC exceeding plus one standard deviation of the PC variability are chosen as warm events, while those with a PC smaller than minus one standard deviation are chosen as cold cases. With this criterion, six warm events and seven cold events are selected in the 51 winters from 1948/49 to 1998/99. The six El Niño winters are 1957/58, 1972/ 73, 1982/83, 1986/87, 1991/92, and 1997/98. The seven La Niña winters are 1949/50, 1955/56, 1970/71, 1973/ 74, 1975/76, 1988/89, and 1998/99. They are consistent with those identified in other diagnostic studies (e.g., Kiladis and Diaz 1989; Hoerling et al. 1997).

Figures 2a and 2b depict the composites of observed DJF seasonal mean 500-hPa geopotential height anomalies for warm and cold events. The composite anomalies are computed relative to a climatology of winters that have “near normal” tropical Pacific wintertime SSTs, that is, the 13 winters of extreme phases of ENSO are excluded. Following Hoerling et al. (1997), the linear component is represented as the difference between the composites of warm and cold events (Fig. 2c), while the nonlinear component is estimated as their summation (Fig. 2d). In general a positive phase of PNA is associated with the warm events and a negative phase of PNA is associated with the cold events. This linear feature is clear in Fig. 2c. Figure 2b, however, is not just a sign reverse of Fig. 2a. Over the North Pacific, the positive height anomaly during the cold events is located about 30° west of the negative height anomaly during the warm events, which confirms the observation in Hoerling et al. (1997). This phase shift is also observed for the anomaly center over North America. Further downstream, from the east coast of North America to western Europe, a band of positive height anomaly appears in both warm and cold events. The negative height anomaly to its northeast also does not switch sign from warm to cold cases. The nonlinearity is seen in Fig. 2d. Two important features are observed: 1) a negative center in the eastern Pacific reflects the phase shift of the PNA pattern, and 2) a pattern from the western North Atlantic extends to Asia, reminiscent of a positive phase of NAO for both warm and cold cases.

Using a long record of observational data, Pozo-Vazquez et al. (2001) made a composite analysis of Northern Hemisphere sea level pressure (SLP) and temperature during extreme El Niño and La Niña winters during 1873–1995. They found that the climate anomalies during an extreme El Niño and La Niña are asymmetric. The asymmetry occurs mainly in the North Atlantic region, where no statistically significant anomaly patterns were found during El Niño winters, while during La Niña winters a statistically significant anomaly pattern resembling the positive phase of the NAO was observed. This feature is not obvious in our composite study (Figs. 2a,b), probably due to the shorter record and the fact that the El Niño composite is biased by two very strong El Niño events (1982/83 and 1997/98).

Our model results are shown in Fig. 3, which presents the composites of ensemble mean DJF 500-hPa geopotential height anomalies for warm and cold events as well as their difference and summation. It should be noted that the average of the forcing amplitudes of the six warm cases is almost the same as that of the seven cold cases. Two 2000-day integrations with opposite signs of forcing, corresponding to 1.5 standard deviation of PC 1 of SST, give almost the same result as the composite study. Comparing Figs. 3a and 3b, the asymmetry occurs mainly in the North Atlantic where the positive height anomaly in the La Niña case is much stronger than the negative height anomaly in the El Niño case. We also see a phase shift in the response over the North Pacific and North America. The shift, however, is about 10°, much smaller than that of the observations. The linear part of the response in the model (Fig. 3c) shows a clear wave train PNA structure. The two features cited earlier about the observed nonlinear component are also found in the model response (Fig. 3d): that is, a negative center in the eastern Pacific and a pattern from the western North Atlantic extending to Asia. Note that the contour interval in Fig. 3d is 10 m instead of 20 m as used in Figs. 3a,b,c. The agreement between the model results and the observations, however, is not as good as for the linear part of the response.

The result shown in Fig. 3 can also be compared with Fig. 10 of Hoerling et al. (1997), where the NCEP atmospheric model is used. They specified symmetric SST anomalies for El Niño and La Niña simulations, thus, allowing the tropical deep convection, determined by the model, to show a phase shift between the warm and cold events. They attribute the nonlinearity of the extratropical response to the phase shift of the tropical deep convection. Comparing their Fig. 10d with our Fig. 3d, the similarity of the nonlinear components in the two models is striking. We emphasize that in our simulations there is no phase shift in the diabatic forcing. Thus, our result suggests that the extratropical atmospheric dynamics is an important contributing factor to the nonlinear response to tropical forcing. The mechanism will be explained in section 5. The weaker nonlinearity in the North Pacific area in Fig. 3d, compared to Fig. 10d of Hoerling et al., may result from the lack of phase shift of the tropical deep convection in our experiment. The sensitivity of the response to longitude of the forcing will be discussed in section 6.

The fact that the observed phase shift of the response over the North Pacific and North America is greater than in our simulations indicates that other factors such as the phase shift of the tropical deep convection are also contributing.

4. Response patterns

An EOF analysis is performed on the 61 ensemble mean 500-hPa geopotential height fields in order to identify the structure of the atmospheric response to a fixed tropical forcing. Figure 4 shows the distribution of the first two modes, as obtained by regressing the ensemble mean 500-hPa height to the corresponding PCs. The magnitudes on the maps represent anomalous height responses corresponding to one standard deviation of the PCs. As can be seen, the first mode, which represents 76% of the variance of the forced interannual variability, is a PNA-type wave train. The second mode, which accounts for 10% of the variance, is a wave train from the western North Atlantic to Asia. Comparison with Fig. 3 shows that these two modes have very similar patterns to the linear and nonlinear components of the response, respectively.

Shown in Fig. 5 is the dependence of the response amplitude to the forcing strength, by plotting the PCs of the response EOFs as functions of the forcing amplitude. As seen in Fig. 5a, the amplitude of the PNA response (EOF1) is a linear function of the forcing amplitude. The correlation between the response and the forcing is 0.95. The relationship between the amplitude of EOF2 and the forcing strength is approximately parabolic (Fig. 5b). A strong forcing anomaly of either positive or negative value excites the same pattern and sign of response.

5. Linear experiments

We now use a linear perturbation model to investigate the mechanism(s) behind the nonlinearity of the extratropical response to the tropical forcing. Because we have a geographically fixed diabatic forcing for both El Niño and La Niña, the nonlinearity comes from the extratropical atmospheric dynamics. Two aspects are of particular interest, which are the changes in the basic-state flow associated with the direct response to the tropical forcing and the feedback of the transient eddies.

The linear experiments are performed with the SGCM following the approach of Hall and Derome (2000). A description of the linear experiment was provided in section 2. In the following the perturbation forcing will take two forms: 1) a forcing anomaly that is linearly related to an El Niño or a La Niña, as in the nonlinear experiments, and 2) a forcing anomaly associated with the transient eddies—the latter being diagnosed from a separate equilibrium nonlinear experiment. In the latter case, the transient eddy forcing appears in all predictive equations in the SGCM, so that both the vorticity and thermal eddy forcing are included.

Figures 6a and 6b show the day-7 and day-15 500-hPa height perturbation of the linear integration when using the model climatology as a basic state, with an anomalous positive tropical forcing. It is seen that the linear model is able to simulate the development of the PNA wave train as discussed in Hall and Derome (2000). Rossby wave energy propagation is clear because the downstream perturbation amplifies when the integration time increases.

In order to assess whether a changed basic state can significantly alter the forced solution, the modeled El Niño and La Niña states are used as the basic state of the linear model. These states are obtained by two long integrations of the nonlinear model with positive and negative tropical forcing corresponding to a 1.5 standard deviation of the SST PC1. The 500-hPa geopotential height anomalies of these two states (not shown) are almost identical to Figs. 3a and 3b. Figures 7a and 7b show the day-7 and day-15 500-hPa height perturbation of the linear integration when a positive (El Niño) forcing and El Niño basic state are used. A positive-phase PNA wave train originating from the tropical Pacific can be seen with some propagation but no unexpected amplification from day 7 to day 15 (when compared to Fig. 6). On day 15 the perturbation over the North Atlantic is weaker than that obtained when using the model climatology as a basic state (Fig. 6), indicating that El Niño's basic state is less favorable for the development of disturbances downstream to the North Atlantic and Europe. Figure 8 shows the 500-hPa height perturbation when a negative (La Niña) forcing and La Niña basic state are used. On day 7 (Fig. 8a), the perturbation is almost antisymmetric (same amplitude and opposite sign) to its El Niño counterpart (Fig. 7a), with a negative phase of the PNA wave train. By day 15, however, an intense positive height anomaly has developed in the North Atlantic (Fig. 8b), which is clearly not observed when an El Niño or climatological basic state is used.

In the previous section we found that the Atlantic response mode is nonlinear to the amplitude of the tropical forcing anomaly, such that a strong forcing anomaly of either positive or negative value excites the same pattern and sign of response. This can be explained by the linear experiments. We will refer to the linear solution with respect to the model climatological basic state as the direct, or first order, response to the forcing anomaly, which is illustrated in Fig. 6. Now with a modified basic state, the linear solution can be viewed as the sum of the direct response and a secondary part, which is related to the change of basic state. The secondary response for an El Niño on day 15 is calculated as Fig. 7b minus Fig. 6b, while that for a La Niña on day 15 is Fig. 8b plus Fig. 6b. They are illustrated in Figs. 9a and 9b, respectively. In general they show a wave train pattern in the North Atlantic similar to the second EOF of the nonlinear model response (Fig. 4b), and have the same sign for both El Niño and La Niña forcing. The secondary response for El Niño forcing is appreciably weaker in amplitude than that for La Niña forcing, and the centers over the eastern United States and the North Atlantic are somewhat displaced, suggesting that it may not be well developed in a 15-day period for El Niño forcing. It can be concluded that the sensitivity of the response to the basic state is responsible for a significant part of the asymmetry and nonlinearity of the response in the North Atlantic region as observed in the composite analysis and the EOF analysis of the response.

To assess the contribution by the transient eddies, another two linear integrations were performed. This time, the transient flux convergence anomalies from the equilibrium model simulations for El Niño and La Niña cases were added to the anomalous forcing. Figure 10 shows the result for El Niño forcing. Again, El Niño's basic state was used, therefore, the only difference with the experiment leading to Fig. 7 is that here the forcing perturbation includes the transient flux convergence anomaly. Comparing with Fig. 7, we see that the transient eddy feedback reinforces the response, in agreement with previous studies (e.g., Lau 1988; Held et al. 1989; Sheng et al. 1998; Lin and Derome 1997), leading to an Aleutian low that is more than double the amplitude of the response without transient contributions. The result for La Niña forcing is shown is Fig. 11. On both day 7 and day 15, we see that, by comparison to Fig. 8, the transient eddy feedback not only reinforces the positive height anomaly in the North Pacific, but also moves it westward about 20° longitude. On day 15 (Fig. 11b), the strong positive height anomaly in the North Atlantic that we saw in Fig. 8b remains in place and is not much affected by the transient eddies.

The forced solutions by the anomalous forcing and eddy flux convergence anomaly at day 15 (Figs. 10b and 11b) compare well with the model equilibrium response (Figs. 3a and 3b). This suggests that these two forcings are responsible for most of the equilibrium response.

6. Sensitivity to the longitude of the forcing

In order to study the dependence of the response on the longitudinal location of the tropical forcing, two additional 2000-day integrations were conducted—one for warm events and one for cold events. The forcing anomaly corresponding to the SST PC1 whose thermal forcing is shown in Fig. 1 is shifted 30° westward, so that the maximum forcing is now located to the west of the date line. The two long integrations were performed using the same shifted forcing pattern, but their amplitudes are +1.5 and −1.5 standard deviation of the SST PC1.

Figure 12 presents the 500-hPa height response for these two experiments. In the North Pacific–North American region, we again see that the response centers of the negative forcing (Fig. 12b) are located to the west of their counterparts of the positive forcing (Fig. 12a). Comparing with Fig. 3a, the response to the shifted positive forcing (Fig. 12a) does not show much phase shift. The North Pacific negative center is only shifted about 10° west, but the other three centers downstream remain in almost the same longitudes. The response is, thus, not sensitive to the longitude of the forcing for El Niño. This is consistent with previous studies (e.g., Ting and Yu 1998; Hoerling and Kumar 2002). On the other hand, for the case of negative forcing (Fig. 12b), when compared with Fig. 3b, both the North Pacific and North American centers of the responses are moved westward with the former by about 30°. The response in the North Atlantic, however, is almost unchanged.

As discussed in Hoerling et al. (1997), the tropical Pacific negative precipitation anomaly during La Niña is located about 30° west of the positive precipitation anomaly during El Niño. Our experiments confirm that this phase difference in diabatic forcing anomaly increases the phase shift of the response in the North Pacific and North American region, which is evident when comparing Figs. 3a and 12b.

7. Summary and discussion

The extratropical response to a geographically fixed anomalous tropical forcing was studied. A primitive equations dry atmospheric model was used where the forcing is calculated from observational data. The anomalous forcing represents a linear fit to the ENSO signal.

It was found that some features of the midlatitude response to the geographically fixed forcing anomaly are linear, while others are nonlinear in the amplitude of the forcing. An EOF analysis of the atmospheric response reveals that the first EOF is a PNA pattern and its amplitude is a linear function of the forcing amplitude. The second response EOF shows a wave train structure extending from the western North Atlantic to Asia. The amplitude of this mode has a nearly parabolic relationship with the amplitude of the tropical forcing.

Our model results compare well with observations and other model results, except that the phase shift of the PNA pattern between El Niño and La Niña winters is smaller. In the real atmosphere, the displacement of the PNA centers between El Niño and La Niña cases can be enhanced by a phase shift of the tropical deep convection. Hoerling et al. (2001) found that the phase shift in the Pacific center of the PNA between warm and cold events is a robust feature in several different GCMs, where the forcing is through the SST, which gives a phase shift in equatorial precipitation and, thus, in the diabatic forcing. This has also been shown in our experiment when the forcing anomaly is shifted as discussed in section 6. The response elsewhere, especially over the North Atlantic, however, is not sentitive to the shift of forcing anomaly.

Linear experiments indicate that the PNA pattern is a direct response to the tropical forcing. The response pattern in the North Atlantic as observed in the full model integration is strongly affected by the nonlinearity associated with the time mean flow. In essence, the sensitivity of the response to the basic state is responsible for the asymmetry and nonlinearity of the response in the North Atlantic region. The feedback of transient eddies reinforces the response over the North Pacific, and is also responsible for the westward phase shift of the response in La Niña winters.

A possible link between the North Atlantic and European climate to ENSO has important implications for seasonal forecasts in that region. Several diagnostic studies have attempted to associate the circulation changes in the North Atlantic with ENSO (van Loon and Madden 1981; Fraedrich et al. 1992; Fraedrich 1994). Pozo-Vazquez et al. (2001) did a composite study of Northern Hemisphere SLP and temperature in Europe during extreme El Niño and La Niña winters during 1873–1995. In the North Atlantic area, no statistically significant anomaly patterns were found to be associated with El Niño, while during La Niña winters a statistically significant anomaly pattern resembling the positive phase of the NAO was found. This finding can be understood in the light of our model result. As seen in Figs 3a and 3b, the response in the North Atlantic to La Niña forcing is much stronger than that to El Niño forcing. This asymmetry is caused by the sensitivity of the response to changes in the basic state as demonstrated by the linear experiments.

As seen in our linear experiments, the changed basic flow associated with the tropical forcing has a major influence on the North Atlantic/European response to further tropical forcing. A La Niña–type basic state yields a stronger response in the North Atlantic to the tropical Pacific forcing than does an El Niño–type basic state. This suggests that seasonal predictions in the North Atlantic and Europe may benefit from the tropical signal when the background state is of a La Niña type.