The sensitivity of climate response to the wintertime Niño3.4 sea surface temperature anomalies of 1855–2002


  • Ivana Herceg Bulić

    Corresponding author
    1. Andija Mohorovičić Geophysical Institute, Faculty of Science, University of Zagreb, Horvatovac 95, 10000 Zagreb, Croatia
    • Andrija Mohorovičić Geophysical Institute, University of Zagreb, Horvatovac 95, 10000 Zagreb, Croatia.
    Search for more papers by this author


The sensitivity of winter atmosphere to ENSO forcing is analysed from an ensemble of integrations forced with observed monthly sea-surface temperatures (SSTs) for the period 1854–2002 with particular interest on the North Atlantic European (NAE) region. A categorisation into warm and cold events, according to the strength of Niño3.4 index, is performed providing main features of simulated atmospheric response to ENSO forcing. A symmetry in a spatial pattern of the atmospheric response to warm and cold phase is found, but with stronger amplitude for warm events. An additional experiment with SST forcing restricted to the tropical Pacific confirmed that atmospheric variability over the Northern Hemisphere is considerably affected by ENSO, even over the distant NAE region.

Signals calculated over the Niño3.4, Pacific-North American (PNA) and NAE regions are analysed as functions of the amplitude of the winter (JFM) SST anomalies in the Niño3.4 region. Both the precipitation and large-scale circulation signals over the Niño3.4 and PNA regions gradually increase with the intensity of SST forcing. However, an asymmetry in amplitude of atmospheric response respecting the polarity of the tropical SST anomalies is found. The tropical and PNA signals associated with warm ENSO events are found to be stronger than the signals related to the cold events. A saturation effect is noticed for the cold events. No such evident characteristics are obtained for the NAE region, but the results imply discernible ENSO impact on atmospheric signals over that region.

Using a correlation coefficient approach, it is shown that the JFM Niño3.4 SST anomaly affects not only the atmospheric response in the simultaneous season, but also during the next few seasons. Significant correlations between the strength of ENSO and the signals are found for the Niño3.4 and PNA regions. Although with smaller absolute values, statistically significant correlations are also found for the NAE region. Copyright © 2010 Royal Meteorological Society

1. Introduction

Extra-tropical climate is characterised by considerable inter-annual variability. The large internal (non-linear) variability of the mid-latitude atmosphere is its dominant feature. In contrast, the internal variability in the tropics is relatively weak because the externally forced variability exceeds the level of the internal variability. Chaotic behaviour of the atmosphere is now commonly accepted (Lorenz, 1963; Palmer, 1993). Molteni et al. (1993) and Palmer (1993) demonstrated how external forcing can influence the statistical properties of atmospheric circulation and presented a simple paradigm for the interactions between an extra-tropical system and a tropical system. They emphasised that the role of the SST forcing is to bias the probability distribution function of atmospheric states.

The connection between ocean surface temperatures and climate is particularly pronounced in the tropics and has been widely studied (e.g. Bjerknes, 1969; Graham and Barnett, 1987; Lindzen and Nigam, 1987; Neelin and Held, 1987). It has been found that sea surface temperatures (SSTs) play an important role in forcing atmospheric convection. Thus, deep cumulus clouds and heavy rains typically occur over the warmest surface waters that are associated with the warm (El Niño) phase of El Niño Southern Oscillation (ENSO). This heavy rainfall migrates together with warm waters resulting in an eastward shift in rainfall along the equator. La Niña, the cold phase of the ENSO phenomenon, is characterised by colder than normal tropical Pacific surface water and suppressed deep convection and rainfall.

However, the impacts of tropical Pacific SSTs are not restricted only to tropical climates. Large-scale convection over warm tropical oceans provides an important portion of the energy that drives the general circulation of the atmosphere. A considerable amount of both observational and modelling evidence of tropical SST variability and mid-latitude teleconnections (especially for the Northern Hemisphere during the winter) is available (see Trenberth et al., 1988 for a review).

A subject of interest regarding the ENSO teleconnections is the asymmetric atmospheric response to the warm (El Niño) and cold (La Niña) phases of the phenomenon. Some observational as well as numerical studies have emphasised that the extra-tropical response of the atmosphere to the tropical Pacific SST anomalies during an El Niño is statistically more significant than the response during a La Niña (Pitcher et al., 1988; Chen and van den Dool, 1995, 1997a, b; Chen, 2004). On the other hand, by analysing NCEP (National Center for Environmental Prediction) data of the 500-hPa geopotential height, Mo et al. (1998) found less significant correlations between the ENSO signal and the Pacific North American (PNA) pattern during the El Niño winters than during the La Niña. Considering both the observations and model simulations, Sardeshmukh et al. (2000) suggested a substantial asymmetry in the remote response to El Niño and La Niña events. According to them, the response of the remote seasonal mean geopotential height to El Niño simulations is stronger, but also more variable, than in La Niña simulations. This result indicates that seasonal extra-tropical anomalies may not necessarily be more predictable during El Niño events than during La Niña events. Hoerling et al. (1997) found that even for an idealised climate system with equal but opposite tropical SST anomalies, the climate response in the AGCM is non-linear. More recently, Hoerling et al. (2001) found that the non-linearity in climate responses emerges as dominant for stronger ENSO events, while mainly linear responses are associated with weaker tropical Pacific SST forcing. Palmer (1988) suggested that during a La Niña situation the general circulation is forced towards a less barotropically stable state; so, the internal variability may be large enough to suppress a response to the external SST forcing. Atmospheric general circulation model (AGCM) experiments performed by Hoerling et al. (1997) indicated that the intrinsic non-linearity of the thermodynamic control on deep convection over the tropical Pacific and non-linearity in tropical rainfall may be responsible for the phase shifts in the extra-tropical teleconnection patterns.

Furthermore, it is reasonable to presume that climatic responses as well as their predictability are dependent on the magnitude of the SST forcing. Thus, it is logical to expect stronger atmospheric signals during stronger events, which has been confirmed by AGCM simulations (Geisler et al., 1985; Branković et al., 1994; Kumar et al., 1996; Kumar and Hoerling, 1998).

While the impact of tropical Pacific SSTs on the PNA region is well documented, the atmospheric response over the North Atlantic–European (NAE) sector is still under discussion. Although some of the results are inconsistent, most studies have indicated a weak, but noticeable, signal of ENSO phenomenon in its climate anomalies (Fraedrich, 1990; Fraedrich and Müller, 1992; Fraedrich, 1993). Furthermore, a recent study (Herceg Bulić and Branković, 2007; hereafter referred to as HBB, 2007) based on SPEEDY AGCM simulations demonstrated relatively weak, but still noticeable, NAE atmospheric response to equatorial Pacific SST anomalies.

Although the recent work of HBB (2007) has showed that SPEEDY is capable of simulating winter extra-tropical response, there are also some open questions. For example, it is reasonable to expect that atmospheric signals calculated over the PNA region would be significant. However, that may not be also valid for the NAE region. Furthermore, they based their conclusions on the composite maps that are made according to a classification associated with the strength of ENSO events. Still, such composite maps represent not only the effect of tropical Pacific but also the impact of other sea surfaces (i.e. Atlantic and Indian Ocean) that may also have a significant impact on the Northern Hemisphere atmosphere, particularly on the NAE region. Thus, although the composites are based on the ENSO categorisation, there is still a possibility that they are also influenced by some impact (or impacts) from the Atlantic and/or Indian Ocean while their characteristics are attributed to the impact of the tropical Pacific only. Fortunately, the use of numerical models enables performing experiments that are not possible in nature. Thus, by prescribing SST variability only in the tropical Pacific (with keeping SSTs constant at climatological value elsewhere) it is provided to estimate if composites of atmospheric response to El Niño and La Niña events are mainly controlled by tropical Pacific or not.

Furthermore, JFM SST anomalies in the tropical Pacific affect winter (simultaneous) Northern Hemisphere atmosphere. But, the interesting point is also if JFM ENSO forcing has a prolonged impact on the following seasons and to what extent.

The current study examines the impact of the magnitude of the winter tropical Pacific SST forcing on atmospheric signals simulated by Speedy, an AGCM of intermediate complexity. The analysis presented in this study is based on a 35-member ensemble of AGCM simulations forced with the monthly global SST variations during 1854–2002. Also a 35-member ensemble experiment with SST forcing restricted to the tropical Pacific is also performed. Such a long-time series of SST anomalies made is encouraging for investigating the atmospheric signal dependence on the polarity and intensity of ENSO events. In contrast to many other similar studies, the atmospheric signals not only for the Niño3.4 and PNA areas, but also for the NAE region, are examined. It is hoped that due to the relative large ensemble size and long time period, it is possible to detect a signal that is strong enough to be separated from the weather noise, even for the NAE region. The structure of the paper is as follows. The model data and experimental design are described in Section 2. Section 3 presents the results of atmospheric response and simulated ENSO signal for both the tropical and extra-tropical regions. Concluding remarks and discussion appear in Section 4.

2. Experimental design

2.1. Model and experiment

The model used in this study is a simplified atmospheric general circulation model, nicknamed SPEEDY (from ‘Simplified Parameterizations, primitivE-Equation Dynamics’), with 8 vertical levels and a triangular truncation of horizontal spectral fields at total wave number 30 (T30-L8). An earlier version of the model with 5 vertical layers is described in details in Molteni (2003) together with its climatology and variability. SPEEDY is based on a spectral dynamical core developed at the GFDL (Geophysical Fluid Dynamics Laboratory, University of Princeton, Princeton, New Jersey, USA; see Held and Suarez, 1994). It is a hydrostatic, σ-coordinate, spectral transform model in the vorticity-divergence form described by Bourke (1974), with a semi-implicit treatment of gravity waves. The parameterised processes include the short- and long-wave radiation, large-scale condensation, convection, surface fluxes of momentum, heat and moisture, and vertical diffusion. Land and ice temperature anomalies are determined by a simple one-layer thermodynamic model. The inclusion of the sea-ice thermodynamic model in SPEEDY amplifies the seasonal cycle of temperature (Herceg Bulić and Branković, 2006). One of the main features of SPEEDY is its computational efficiency.

Molteni (2003) discussed the SPEEDY response to tropical and extra-tropical SST anomalies for a former model version with 5 vertical layers. He showed that the model simulates atmospheric flow realistically, especially during the boreal winter. Furthermore, different ranges of atmospheric variability were reproduced in SPEEDY simulations, albeit with smaller amplitudes than those obtained from the observations. The 7-layer version of the model is significantly improved compared with the 5-layer version. Comparison of the 7-layer model response for the period 1952–2001 to the NCEP/NCAR reanalysis data (Kalnay et al., 1996) confirmed the ability of SPEEDY to simulate reasonably well the forced and internal components of the atmospheric inter-decadal variability and to estimate the response of the atmospheric circulation to the well documented SST trends in the tropical oceans (Bracco et al., 2004). Furthermore, SPEEDY was efficient in simulating various aspects of dynamical climatology (e.g. Molteni 2003; Bracco et al.2004; Hazeleger et al.2005; Kucharski et al.2006; HBB07; Kucharski et al.2009).

Here, the experiment is conducted by integrating SPEEDY over the period 1854–2002 with observed SST boundary conditions. The SST dataset is in the form of monthly averages provided by NOAA_ERSST_v2 data1 (Smith and Reynolds, 2004). Relatively large member ensembles may be needed to obtain a significant ENSO signal, especially for the extra-tropics where it may be suppressed by strong atmospheric internal variability even in the presence of a relatively strong tropical signal (Branković and Palmer, 1997). In order to obtain statistically representative results, a 35-member ensemble is generated. By the use of this ensemble, the ability to detect the ENSO signal is enhanced since the climate noise related to internal atmospheric variability is reduced by a factor 1/n, where n is the ensemble size (e.g. Branković and Molteni, 2004). The model integration of every member is performed with the same observed SSTs for the period of investigation, but with slightly different initial conditions. This procedure enables to isolate the ‘external’ variability caused by the SST forcing from the ‘internal’ variability due to internal atmospheric processes. For the purpose of analysing the results, the first year of the model integration is discarded due to drift of the model climate; so, the analytical results refer to the period 1855–2002.

2.2. The measure of signal

An analysis of the atmospheric signal during the period from 1855–2002 is performed using the results from the ensemble of integrations. A simple approach to define signal is adopted in this study (Branković and Molteni, 2004). The signal is defined as the mean squared anomaly of all ensemble means Āj with respect to the climatological mean Ā, and reflects the impact of the external forcing:

equation image(1)


equation image(2)

is the ensemble mean for year j, and Aij is the i-th ensemble member in the j-th year.

3. Results

3.1. SST forcing and Niño3.4 index

Since the goal of this study is to examine the dependence of atmospheric signals on the equatorial Pacific winter SST amplitude, the Niño3.4 SST index is employed (Trenberth, 1997). It is derived as an area average of the SST anomaly (SSTA) over the Niño3.4 region (5°N–5°S, 170°W–120°W). As a concise way of summarising the strength of the tropical Pacific SST forcing during 1855–2002, a graphical representation is illustrated in Figure 1. The JFM (January, February, March) mean of the Niño3.4 SST index is shown in Figure 1. Each bar in the graph corresponds to the JFM mean, and the individual events have been ranked from the warmest events (on the left hand side) to the coldest events (on the right hand side). It is easily seen that warm SST anomalies acquire larger peak amplitudes than cold events (i.e. extremely warm ENSO events are stronger than extremely cold events). However, the cold events quantitatively dominate the warm events during the investigated period (there are 92 events with negative Niño3.4 SST index and 55 events with positive Niño3.4 SST index).

Figure 1.

Niño3.4 SST anomaly index for the period 1854–2002. Each bar corresponds to a JFM (January, February, March) mean of the index, and these are ranked from the strongest warm events on the left hand side to the strongest cold events on the right hand side. Units are °C

For the purposes of analysing ENSO impact on the extra-tropical atmospheric response, the years of interest under the study are categorised into warm (El Niño) and cold (La Niña) years. If standard deviation of JFM Niño3.4 index in Figure 1 is denoted as σ, then the years associated with a Niño3.4 index larger than 1.0σ are defined as warm ENSO years, while the years associated with Niño3.4 index smaller than − 1.0σ are defined as cold ENSO years. The composites of JFM SSTA for warm and cold ENSO events described above are shown in Figure 2. Anomalies show SST ENSO pattern with cold (warm) tongue in tropical Pacific for La Niña (El Niño) events. The amplitude is similar for both composites, but an asymmetry in spatial structure between the cold (Figure 2(a)) and warm composite (Figure 2(b)) is evident. For the cold events in Figure 2(a), the area associated with negative SSTA smaller than 1.5 °C is between 135°W and 165°W and is flanked with positive anomalies at the north and south. However, for the warm composite (Figure 2(b)) area associated with positive SSTA amplitudes is narrower than its counterpart for the cold composite, while the area bounded with maximal amplitudes (greater than the 1.5 °C) is shifted toward the west and placed between 115°W and 155°W. Similarly, as found for the cold composite, the warm tongue is flanked with the anomalies of the opposite sign at the northern and southern side.

Figure 2.

The JFM sea surface temperature (SST) anomalies for: (a) cold ENSO events; and (b) warm ENSO events. Contours are for every 0.3 °C. Negative values are dashed. This figure is available in colour online at

3.2. Extra-tropical response

The influence of the tropical Pacific on the Northern Hemisphere atmosphere is presented by composites of atmospheric circulation anomalies (i.e. geopotential height anomalies at 200-hPa, Z200), with anomalies defined regarding the climatology for the 1855–2002 time period (Figure 3). The composites are based on the categorisation according to the strength of the JFM Niño3.4 index as described above.

Figure 3.

The composites of 200-hPa geopotential height anomalies for the Northern Hemisphere for: (a) JFM cold composite; (b) JFM warm composite; (c) JJA cold composite; and (d) JJA warm composite. Contours are for every 0.5 dam. Negative values are dashed. This figure is available in colour online at

Winter (JFM) and summer (JJA) composites of Z200 are shown in Figure 3. For the both seasons, a remarkable similarity between the cold (Figure 3(a)) and warm (Figure 3(b)) composite is evident. The spatial structure of the anomalies is almost the same for the cold and warm composite, but with the anomalies of reversed sign. The strongest atmospheric response is found over the PNA region composed from three centres of actions (two with the same sign of anomalies; one over the northeastern Pacific and the other over the Gulf of Mexico, while the third centre with the anomalies of opposite sign is placed over the northern United States and Canada). The NAE region, which is of particular interest for this study, is also characterised with a detectible atmospheric response to tropical Pacific forcing. Albeit considerably weaker than the PNA region, the NAE region is also affected by ENSO showing a response that is associated with positive (negative) Z200 for cold (warm) ENSO events. The essential features for both the PNA and NAE region is stronger response to warm than to cold events and its remarkable linearity. Although the amplitude of SST forcing in the tropical Pacific is the same for both cold and warm ENSO events (cf. Figure 2(a) and (b)), the amplitudes of Z200 anomalies for the warm composite (Figure 3(b)) are considerably greater than for the cold composite (Figure 3(a)) for all of three action centres over the PNA as well as for the action centre over the NAE region. The linearity of atmospheric response to ENSO forcing is noticeable as no shifting between the position of action centres for warm and cold composites. In other words, although the substantial shift in cold and warm SST forcing is found in Figure 2 (and also in the tropical Pacific precipitation; not shown), the action centres associated with atmospheric response to warm ENSO events (Figure 2(b)) have the same position as the action centres associated with cold ENSO forcing (Figure 2(a)).

Composites for JJA season (Figure 3(c) and (d)) resembles, to a great extent, those obtained for the JFM season, but with much smaller amplitudes. This is an indication of prolonged impact of winter ENSO forcing, not only on the PNA but also on the NAE region.

Precipitation anomalies (not shown) correspond very well with Z200 anomalies. Generally, the largest precipitation anomalies coincide with the largest Z200 anomalies and negative (positive) precipitation anomalies are associated with positive (negative) Z200 anomalies. As found for Z200, the strongest precipitation response is over the PNA region. However, precipitation over the NAE region is also affected by ENSO with wetter conditions during La Niña events and drier conditions associated with El Niño events. Main features of simulated Z200 response are also noticed for precipitation (stronger response to warm events with strong linearity between the cold and warm ENSO events).

3.3. Extra-tropical response to SST forcing restricted to tropical Pacific

The analysis in the previous section indicated atmospheric response to ENSO forcing, not only over the PNA, but also over the NAE region, although the latter is much weaker. Although the composites are created with respect to the cold and warm ENSO events, they still reflect the impact of oceans other than the Pacific, since SPEEDY simulations are forced with SSTs prescribed in all oceans. This, particularly, may be important for the NAE region because of the vicinity of the Atlantic and Indian Ocean that may considerably blur the atmospheric response to tropical Pacific response. Thus, an additional experiment (TroPac) with SST forcing only in the tropical Pacific is performed. This experiment is based on the ensemble of 35 SPEEDY integrations with SST variability prescribed only in tropical Pacific and climatological SST elsewhere. In this way, the impact of other SSTs except those associated with tropical Pacific is turned off.

The winter Z200 response simulated in the TroPac experiment is shown in Figure 4. Obviously, response pattern reflects to a great extent that obtained with global SST forcing (Figure 3(a) and (b)). The action centres have the same position as in Figure 3 indicating that the strong linearity of atmospheric response and stronger response to warm ENSO events is found. Generally, the Z200 amplitudes are greater in the TroPac experiment than those achieved in the experiment with global SST forcing. Particularly, the response over the NAE region is considerably stronger in the TroPac simulations. The absolute value of maximal Z200 anomalies over the NAE region is increased from 1.0 to 1.5 dam for the cold composite (cf. Figures 3(a) and 4(a)), while for the warm composite it is increased from 1.5 to 2.5 dam (cf. Figures 3(b) and 4(b)). Similar result is also obtained for precipitation (not shown). Thus, both precipitation and Z200 anomalies have greater amplitudes in simulations with SST forcing prescribed only in tropical Pacific revealing a discernible ENSO signature even in the relatively distant NAE region.

Figure 4.

The composites of 200-hPa geopotential height anomalies for the Northern Hemisphere simulated in TroPac experiment for: (a) JFM cold composite; and (b) warm composite. Contours are for every 0.5 dam. Negative values are dashed. This figure is available in colour online at

3.4. ENSO signal simulated by SPEEDY

The model results described in this section are also arranged according to the ranked amplitude of the SST anomaly in Figure 1. Those signals are calculated for each of the 12 overlapping 3-month seasons beginning with December, January and February (DJF) and ending with November, December and January (NDJ) according to Equation (1).

3.4.1. Precipitation signal

Figure 5. shows the winter (JFM) and summer (JJA) seasonal precipitation signal averaged over the Niño3.4 region (Fig. 5.a,b), PNA region (Fig. 5c,d) and NAE region (Fig. 5e,f). The signal is presented in relation to its climatological mean (i.e. each particular signal is divided by a corresponding climatological mean for considered time period). PNA domain is defined as the region (20°N–70°N; 180°W–60°W) while NAE domain is defined as the region (25°N–75°N; 15°W–60°E). According to the results described in the previous section, the atmospheric response is linear (i.e. with no shift in the centres of action for cold and warm composite), so it is acceptable to use the same averaging area for both cold and warm regional signals. Seasonal signals are arranged according to the ranked amplitude of the JFM Niño3.4 SST anomaly in Figure 1.

Figure 5.

JFM and JJA seasonal mean precipitation signal during the period 1854–2002 calculated over the Niño3.4, PNA and NAE region. The signal is divided by the seasonal climatological mean and arranged according to the ranking of the JFM Niño3.4 index in Figure 1

As expected, tropical precipitation signal is strongly associated with the JFM SSTA in the Niño3.4 region, particularly for JFM (Figure 5(a)) as well as for DJF and FMA seasons (not shown) since the shape of their distributions corresponds to that of the SST anomaly in Figure 1. The shapes of the MAM and AMJ distributions (not shown) qualitatively imply that the tropical atmosphere still ‘remembers’ the JFM SST anomaly, while the JJA season (Figure 5(b)) and those after that do not reflect the impact of SST anomaly in Figure 1.

Tropical precipitation signal is enhanced (suppressed) during warm (cold) events due to tropical convection that is enhanced (suppressed) by warm (cold) ocean surface temperatures. Comparison of Figure 1 with Figure 5(a) indicates the dependence of the precipitation signal on the amplitude of the SST forcing. However, the signal is not symmetric with respect to the polarity of ENSO phases. Thus, strong warm SST forcing generates larger precipitation signals than strong cold SST forcing. This is consistent with some previous findings (e.g. Hoerling et al., 1997; Kumar and Hoerling, 1998). Note that the rainfall signal demonstrates a saturation effect during cold events; the further increase of SST forcing during cold events with negative SST anomalies greater than 0.5 °C is accompanied only by a very small, if any, increase in the simulated atmospheric response. On the contrary, a saturation effect is not found for the signals associated with positive SST anomalies (the simulated rainfall signal intensifies with the increasing SST anomalies). The saturation effect for cold ENSO events has also been reported previously (e.g. Kumar and Hoerling, 1998).

Figure 5(c) displays the JFM precipitation signal averaged over the PNA domain. Although the amplitude is smaller compared to the tropical signal, the similarity between the tropical and PNA precipitation signal is clearly evident (cf. Figure 5(a) and (c)). The particular similarity is also noticed for DJF as well as for FMA season (not shown). Again, strong signals appear during warm rather than cold events. Thus, the PNA signal also manifests an asymmetric response with respect to the polarity of the SST anomaly and exhibits a saturation effect for cold events. The PNA signals for MAM, AMJ, and MJJ season (not shown) still indicate on weak, but possible influence of winter tropical SST anomaly on the seasonal precipitation over the PNA region. However, from the JJA season (Figure 5(d)) further (not shown) there is no indication of ENSO impact.

The manifestation of the teleconnections between the tropics and PNA region demonstrated above is not surprising since the impact of tropical Pacific on PNA is well documented in available literature. Still, ENSO influence on the Europe is still contradictory and a topic of discussions. Some previous results of the SPEEDY's simulations (HBB, 2007) have shown some striking similarities between the model response over the PNA and NAE. These findings encouraged the analysing of NAE signal. Figure 5(c) shows JFM seasonal precipitation signal averaged over the NAE region. Although it appears a little bit noisy, it resembles, to a certain extent, the shape of signal distributions obtained for the Niño3.4 and PNA region indicating that it is also possibly affected with ENSO. For the FMA and MAM seasons (not shown) the qualitative resemblance is also found.

3.4.2. Z200 signal

Figure 6 shows the equatorial, PNA and NAE 200-hPa geopotential height (Z200) signal, which provides a picture of large-scale circulation. The Z200 signal is presented in relation to its climatological mean (the same as is done for the precipitation signal in Figure 5). The JFM equatorial atmospheric signal (Figure 6(a)) increases with the amplitude of the Niño3.4 SST anomaly and mirrors the distribution of rainfall signal in Figure 5(a). This is also valid for the DJF and FMA season (not shown) and partly for the MAM, MJJ, and AMJ season. Again, the signal is not symmetric with respect to extreme ENSO phases; thus, larger circulation responses are found for warm SST events. For the JJA (Figure 6(b)) and the following seasons (not shown) there is no resemblance with the distribution of the JFM season.

Figure 6.

The same as Figure 5, except for the signal of the 200-hPa geopotential height

The shape of JFM PNA Z200 signal distribution (Figure 6(c)) is similar to that over the Niño3.4 region (Figure 6(a)), suggesting a tropical–extra-tropical teleconnection. It also reflects the distribution of the JFM precipitation signal over the same region (Figure 5(c)). The PNA Z200 signal demonstrates the same characteristics as the tropical signal (i.e. asymmetric response and saturation effect for cold events). Distribution of JJA Z200 signal (Figure 6(d)) slightly reveals a similarity with the shape of the JFM distribution what is contrary to the precipitation signal for JJA season in Figure 5(d) which does not have any similarity with the shape of JFM signal in Figure 5(c). This result indicates that SST anomalies in the Niño3.4 region generated Rossby waves that have affected large circulation signal over the PNA region (and also JJA composites in Figure 3(c) and (d)), but not to the extent required to produce a significant impact on the precipitation signal.

Over the NAE region, the shape of JFM Z200 signal resembles, to some extent, the shape of JFM signal for the Niño3.4 and PNA region, although with some very strong signals appearing for cold ENSO years which cannot be associated with ENSO forcing. Still, there is a weak qualitative indication of ENSO impact on the Z200 signal over the NAE region. On the other hand, JJA Z200 signal (Figure 6(f)) looks rather noisy and it is hard to find any correlation with winter SSTA in this stage of analysis.

3.5. Correlations between the Niño3.4 indices and simulated signals

The results presented above for the tropical and PNA regions have implied the existence of a connection between the JFM tropical SST forcing and the simulated precipitation and Z200 signals. Therefore, the correlation coefficients between the JFM Niño3.4 indices and the simulated seasonal signals are calculated.

The correlation coefficients between the JFM Niño3.4 indices and the seasonal tropical precipitation as well as Z200 signals are shown in Figure 7. In this way, the correlation between the JFM Niño3.4 indices in Figure 1 and the simulated precipitation (Z200) signal of any particular season can be examined. The correlation coefficients are calculated for all events (r, presented by the solid line), and also separately for years associated with positive Niño3.4 indices (rpos, the dashed line) and also for years associated with negative Niño3.4 indices (rneg, the dotted line). The circles indicate seasons with correlation coefficients for which the t-statistic is significant at the 95% confidence level, and only those seasons will be considered further.

Figure 7.

Correlation coefficients calculated between the JFM Niño3.4 index, and (a) seasonal mean Niño3.4 total precipitation signal, and (b) seasonal mean Niño3.4 200-hPa geopotential height signal. The solid black line represents coefficients calculated for all years; the dashed line represents coefficients calculated for the years associated with the warm events; and the dotted black line represents coefficients calculated for the years associated with the cold events. The circles on the lines indicate those seasons for which the t-statistics are significant at the 95% confidence level

Generally, Figure 7(a) shows that the correlations between the Niño3.4 indices and tropical precipitation as well as Z200 signals acheives greater values when cold and warm events are considered separately than for r-values calculated for all events. Furthermore, the rpos-values are greater than the absolute value of the rneg-values, likely due to the aforementioned saturation effect of the cold events. Although both the precipitation and Z200 signals exhibit the strongest correlations with the JFM Niño3.4 indices for the simultaneous season, the JFM Niño3.4 SST anomalies substantially affect atmosphere also in a few following seasons. Generally, the correlations with the JFM Niño3.4 indices (presented as the rpos- and rneg-values) are stronger for precipitation signals than for Z200 signals.

The PNA precipitation and Z200 signals are also significantly correlated with the JFM SST anomaly in Niño3.4 region (Figure 8). This connection is the strongest for the JFM season, but is also found for a few following seasons. In contrast to the results for the tropical signal, the PNA Z200 signals exhibit stronger correlations than the PNA precipitation signals (for both positive and negative events). However, as found for the Niño3.4 signal, it is likely that JFM warm events have longer impact on the atmosphere for the Z200 signals than for precipitation signals. It is probably associated with the ENSO impact which may be seen in Z200 signal, but is not strong enough to affect precipitation signal.

Figure 8.

The same as Figure 7, except for the PNA region

The results for the NAE region slightly resemble characteristics seen in the previous sections for the tropical and PNA regions. However, correlation coefficients indicate that connection between the JFM SST anomalies in the equatorial Pacific and the NAE signals is statistically significant (Figure 9). This result is in agreement with some previous findings that suggested a teleconnection between ENSO and the remote atmospheric responses over the NAE region (e.g. Fraedrich and Müller, 1992; HBB, 2007). The strongest correlation is found again for the simultaneous JFM season, but the duration of the period with significant correlations is shorter than that for the Niño3.4 and PNA regions. Surprisingly, a relatively long statistically significant period is found for the correlation of NAE precipitation with the negative SSTA, starting with the FMA season and ending with the JAS season. However, this result should be considered doubtful, since it does not seem reasonable to expect such long impact of a negative JFM Pacific SST anomaly on a region as remote as the NAE. The cause of this result is out of the scope of this study; so, it is considered as a coincidence, rather than giving some speculative explanations.

Figure 9.

The same as Figure 8, except for the NAE region

4. Summary and discussion

In this study, a large ensemble of very long integrations is used to examine the dependence of the atmospheric response on the amplitude and polarity of JFM equatorial Pacific forcing. For this purpose, a model of intermediate complexity was employed to create a 35-member ensemble of integrations for the period from 1855–2002. Such a large number of very long AGCM simulations enabled detection of the ENSO signal in the extra-tropics, where atmospheric internal variability is relatively strong (HBB, 2007). In addition, an ensemble of SPEEDY integrations with SST forcing prescribed only in tropical Pacific is created in order to estimate if features of composites based on the strength of the Niño3.4 index could be attributed mainly to the impact of tropical Pacific. The results are focused on the Niño3.4 region and also on the Pacific North American and North Atlantic European regions for the purpose of examining the teleconnection between the tropical Pacific and Northern Hemisphere mid-latitudes.

Warm and cold composites of atmospheric response (Z200 and precipitation anomalies) based on the categorisation in accordance with the strength of the Niño3.4 index is used to analyse atmospheric response to El Niño and La Niña events. It is found that El Niño signature in the Northern Hemisphere mid-latitude atmosphere is more robust than that of La Niña. Composite maps reveal similar structure of atmospheric response to warm and cold ENSO events, but with opposite sign of the corresponding anomalies. Some previous studies indicated that model responses to cold and warm ENSO forcing have a substantial shift in the location of the action centres. For example, Hoerling et al. reported a 30° longitudinal shift between action centres of the 500-hPa height anomalies over the PNA region, while Peng and Kumar (2005) found a 15° shift in the 200-hPa height anomaly pattern over the same region. Such a shift between the atmospheric responses to ENSO events of opposite polarity may be a consequence of investigation period as well as decadal variability (Hoerling et al.1997; Hoerling et al.2001; DeWeaver and Nigam 2002). However, results presented here indicate the strong linearity of the response with no shift in the spatial pattern of the cold and warm composite for both the PNA and NAE region.

Although the major ENSO impact on the extra-tropical Northern Hemisphere is found over the PNA region, the NAE region is also considerably affected by ENSO phenomena. Some striking similarities between the atmospheric response over the PNA and NAE are found (i.e. the linearity of atmospheric response to El Niña and La Niña events, stronger response to warm ENSO phase than to the cold ENSO phase).

Results shown for the model integrations with SST forcing prescribed only in tropical Pacific (TroPac experiment) are very similar to those obtained for global forcing indicating that atmospheric response presented by composites in this study may be attributed to the impact of tropical Pacific. In other words, tropical Pacific has substantial influence on the winter atmospheric variability of the Northern Hemisphere. Furthermore, composites made for TroPac experiment reveal stronger atmospheric response than composites based on integrations with global SST forcing, particularly pronounced for the NAE region (cf. Figure 3(a) and (b); and Figure 4(a) and (b)). This may be an indication of some additional processes (related to the other parts of world's oceans than tropical Pacific) that actually attenuate atmospheric response to ENSO forcing. The Atlantic Ocean and/or Indian Ocean may be a source of such processes due their relative vicinity to the NAE region, but an extensive study is required to examine this presumption.

The tropical rainfall signal, as well as the tropical Z200 signal, is shown to increase with the intensity of tropical Pacific SST forcing. It is found that the amplitudes of the simulated signals for warm events exceed those for cold events. This is consistent with the composite maps presented in this study as well as with some previous studies such as observational analyses of Hoerling et al. (1997) and the modelling results of Kumar and Hoerling (1998). Besides, the results presented here also show a saturation effect for the cold events. Kumar and Hoerling (1998) have explained this effect with the fact that Pacific rainfall is climatologically sparse over the cold tongue region, and thus, a moderate cold event is itself sufficient to entirely suppress convection there.

The analysis of the atmospheric signals for the PNA region reveals similar characteristics as those calculated for the Niño3.4 region, albeit with weaker amplitudes. Thus, increasing magnitudes of the atmospheric signals with positive tropical Pacific SST anomalies and saturation effect for cold ENSO events are also found for both the precipitation and Z200 PNA signals, highlighting the teleconnection between the tropical Pacific and Northern Hemisphere mid-latitudes. Atmospheric signal for the NAE region does not indicate such clear dependence on the Niño3.4 index. Still, distribution of Z200 and precipitation signals calculated for the NAE region imply a possible physical connection with ENSO forcing. The analysis of the correlation coefficients between the JFM Niño3.4 indices and the simulated precipitation and Z200 signals shows that the JFM Pacific SST anomaly significantly affects not only the atmospheric response in simultaneous season (JFM) but also in some following seasons. Over the Niño3.4 region, the JFM SST anomaly is significantly correlated with the precipitation from the DJF season through the AMJ season (Figure 8(a)). The duration of this correlation is the same for both cold and warm SST forcing. However, a positive JFM SST anomaly generates a period with a statistically significant correlation with the Z200 signal that has a longer duration than that of a negative JFM SST anomaly (Figure 8(b); note seven successive seasons are with significant rpos, while only four successive seasons are with rneg). This may be an indication that sensitivity of large-scale circulation is somehow longer for warm JFM forcing than for cold JFM forcing. Particularly, both characteristics (asymmetric response to cold and warm SST forcing and saturation effect for cold SST forcing) are found for the both Z200 and precipitation signal, while the longer period of correlation with the warm Pacific SST anomaly is found only for Z200 signal.

Over the PNA region (Figure 8), the correlation coefficients show similar characteristics to those found for the Niño3.4 region highlighting tropical–extra-tropical teleconnection. An intriguing result is found for the NAE region (Figure 9), where a significant correlation is also obtained (although with smaller r-values than for the Niño3.4 and PNA regions). Together with composite maps in Figures 3 and 4, this is an indication of discernible ENSO effect on the atmosphere over the NAE region. Furthermore, according to Figure 9, JFM SST anomalies in tropical Pacific may affect atmospheric variability over the NAE region not only in simultaneous (winter) season but also in some following seasons.

Although ENSO impact on the NAE region is relatively weak, enhanced statistical significance in this study is obtained from a relatively large number of very long SPEEDY integrations. Since stronger atmospheric response and greater values of correlation coefficients are found for warm ENSO events (Figures 3 and 4), it is reasonable to presume that stronger El Niño events would also generate considerably stronger atmospheric response. According to some modelling studies (e.g. Timmermann et al.1999; Collins 2000; Guilyardi et al.2009), it is possible that El Niño events in future climate will have larger amplitudes than in the current climate due to warming caused by anthropogenic influence. Basing on the above discussion and results presented here it is possible that in such a future climate with stronger warm ENSO events teleconnections between the tropical Pacific and North Atlantic and Europe would be also intensified.


This work was supported by the Ministry of Science, Education and Sports of the Republic of Croatia under the Project No. 119-1193086-1323. Additional support was received through the ECOOP project funded by the European Commission's Sixth Framework Programme (Contract 36355).