Clim Dyn (2011) 37:1205–1215 DOI 10.1007/s00382-010-0912-1

The role of mean state on changes in El Nin˜o’s flavor Jung Choi • Soon-Il An • Jong-Seong Kug Sang-Wook Yeh

•

Received: 7 June 2010 / Accepted: 10 September 2010 / Published online: 24 September 2010 Ó Springer-Verlag 2010

Abstract Recently, many studies have argued for the existence of two types of El Nin˜o phenomena based on different spatial distributions: the conventional El Nin˜o [or Eastern Pacific (EP) El Nin˜o], and the Central Pacific (CP) El Nin˜o. Here, we investigate the decadal modulation of CP El Nin˜o occurrences using a long-term coupled general circulation model simulation, focusing, in particular, on the role of climate state in the regime change between more and fewer CP El Nin˜o events. The higher occurrence regime of the CP El Nin˜o coincides with the lower occurrence regime of EP El Nin˜o, and vice versa. The climate states associated with these two opposite regimes resemble the leading principal component analysis (PCA) modes of tropical Pacific decadal variability, indicating that decadal change in climate state may lead to regime change in terms of two different types of El Nin˜o. In particular, the higher occurrence regime of CP El Nin˜o is associated with a strong zonal gradient of mean surface temperature in the equatorial Pacific, along with a strong equatorial Trade wind over the area east of the dateline. In addition, the oceanic variables—the mixed layer depth and the thermocline depth—show values indicating increased depth over the western-to-central Pacific. The aforementioned climate states obviously intensify zonal J. Choi  S.-I. An (&) Department of Atmospheric Sciences/Global Environmental Laboratory, Yonsei University, Seoul 120-742, Korea e-mail: [email protected] J.-S. Kug Korea Ocean Research and Development Institute, Ansan, Korea S.-W. Yeh Department of Environmental Marine Science, Hanyang University, Ansan, Korea

advective feedback, which promotes increased generation of the CP El Nin˜o. Frequent CP El Nin˜o occurrences are not fully described by oceanic subsurface dynamics, and dynamical or thermodynamical processes in the ocean mixed layer and air–sea interaction are important contributors to the generation of the CP El Nin˜o. Furthermore, the atmospheric response with respect to the SSTA tends to move toward the west, which leads to a weak air–sea coupling over the eastern Pacific. These features could be regarded as evidence that the climate state can provide a selection mechanism of the El Nin˜o type.

1 Introduction The El Nin˜o-Southern Oscillation (ENSO), characterized by interannual variation of the sea surface temperature (SST) over the eastern-to-central tropical Pacific, has great global impacts and teleconnections (Ropelewski and Halpert 1987, 1996; Trenberth and Carbon 2000). ENSO has been considered as a natural basin-wide oscillatory mode of the tropical Pacific ocean–atmosphere system; its oscillation mechanism has been described earlier by the delayed oscillator (Battisti and Hirst 1989; Schopf and Suarez 1988) and, more recently, by the recharge oscillator (Jin 1997a, b). In particular, these both emphasize thermocline feedback as a major feedback process driving the oscillatory feature of ENSO. At the same time, zonal advective feedback has been proposed as an oscillatory mechanism by Picaut et al. (1997). The zonal advective process has gained more attention recently, as El Nin˜o events with SST variability concentrated over the central Pacific occur more frequently. In the past two decades, research on ENSO has advanced rapidly (Neelin et al. 1998; Wallace et al. 1998).

123

1206

J. Choi et al.: The role of mean state on changes in El Nin˜o’s flavor

Specifically, many previous studies have suggested the coexistence of various coupled modes in intermediate coupled models (Zebiak 1984; Dewitte et al. 2007) and general circulation models (Philander et al. 1992), as well as in observations (Jin et al. 2003; Kang et al. 2004). An and Jin (2001) argued that each type of feedback—thermocline and zonal advective—destabilizes a different type of ENSO mode. Bejarano and Jin (2008) proved the coexistence of various equatorial coupled modes of ENSO by solving the eigenvalue problem of the intermediate tropical Pacific coupled model. Thus, at least two natural modes associated with ENSO may exist, each with its own characteristics and potential underlying mechanism. Recent studies have argued for the existence of two distinct types of El Nin˜o based on different spatial distributions. One is the conventional El Nin˜o, which is characterized by the maximum loading of sea surface temperature anomaly (SSTA) in the eastern Pacific. The other type places the center of the SSTA more toward the central Pacific. This new type of El Nin˜o has been variously termed ‘‘Central Pacific El Nin˜o’’ (Kao and Yu 2009; Yeh et al. 2009), ‘‘Warm Pool El Nin˜o’’ (Kug et al. 2009, 2010), ‘‘Dateline El Nin˜o’’ (Larkin and Harrison 2005a, b), and ‘‘El Nin˜o Modoki’’ (Ashok et al. 2007; Weng et al. 2007). Despite the terminological inconsistency, these basically describe the same phenomenon. Several characteristics of this new-type El Nin˜o differ from those of the conventional El Nin˜o; Larkin and Harrison (2005a, b) and Ashok et al. (2007) found that global impacts induced by the new-type El Nin˜o are also different from those coming from the conventional El Nin˜o. Moreover, Kug et al. (2009, 2010) suggested that zonal advective feedback plays a crucial role in the development and decay of the SST anomaly associated with the new-type El Nin˜o, while thermocline feedback is known to be a key process of the conventional El Nin˜o. It is known that the amplitude and frequency of the conventional El Nin˜os are modulated with a decadal time scale (Timmermann 2003; Rodgers et al. 2004), whereas the decadal modulation of the new-type El Nin˜o has not yet been fully investigated. Yeh et al. (2009) and Ashok and Yamagata (2009) have suggested that the new-type El Nin˜o will increase in frequency relative to the conventional El Nin˜o in future climate scenario. This implies that climate changes may lead to changes in the type of ENSO. In addition, Kug et al. (2010) discussed a possible interaction between new-type El Nin˜o and climate mean state, namely that the occurrence of the new-type El Nin˜o can contribute to accumulative warming of the mean state, owing to the warm-phase dominance of the new type of El Nin˜o. Nevertheless, the interaction between the new-type El Nin˜o and climate state is still not fully understood.

Hereafter, the terms ‘‘Central Pacific (CP) El Nin˜o’’ and ‘‘Eastern Pacific (EP) El Nin˜o’’ will be used for ‘‘new-type El Nin˜o’’ and ‘‘conventional El Nin˜o,’’ respectively. In this study, we focus on the decadal modulation of the CP El Nin˜o occurrences and examine the role of climate state changes on El Nin˜o behavior, especially with regard to the location of occurrences. Section 2 describes the data and model. Section 3 presents a model for decadal modulation of El Nin˜o’s flavor. Section 4 discusses the effects of climate change on ENSO. A summary and discussion are given in Sect. 5.

123

2 Data Here, we utilize a long-term simulation of GFDL coupled GCM CM2.1 (Delworth et al. 2006; Wittenberg et al. 2006), specifically a 500-year preindustrial run, which is available in the CMIP3 archives (https://esg.llnl.gov: 8443/home/publicHomePage.do). A detailed description of the model simulation may be found in Wittenberg et al. (2006). These data were also used in Kug et al. (2010). The model data is useful in compensating for the limited number of observations available. To determine the two types of El Nin˜o events, we follow the steps suggested in Kug et al. (2010): specifically, we ˜ O3 and NIN ˜ O4 indices (hereafter, use the modified NIN ˜ O3m and NIN ˜ O4m, respectively) because of a slight NIN westward extension of the cold tongue in this model, as well as in most general circulation models (GCMs), as ˜ O3m (NIN ˜ O4m) is compared with observations. NIN defined by the region 170°W–110°W, 5°S–5°N (140°E– ˜ O3m or NIN ˜ O4m 170°W, 5°S–5°N). SSTA over the NIN region during ND(0)J(1) in excess of 0.5°C was considered an El Nin˜o event. ND(0)J(1) indicates the average over November of the El Nin˜o year through the following January. A CP El Nin˜o event was defined by the condition ˜ O4m SSTA exceeds the NIN ˜ O3m SSTA, that the NIN while the opposite condition defined an EP El Nin˜o event. Based on these definitions, a total of 121 events were classified as CP El Nin˜o events and 84 as EP El Nin˜o events (see details in Kug et al. 2010). To check whether such simple definition is reasonable, we increased the threshold used for the criteria of the El Nin˜o type. That is, ˜ O4m and El Nin˜os, of which difference between NIN ˜ O3m SSTA is less than 0.25°C, are putted aside as a NIN mixed-type El Nin˜o. Kug et al. (2009), in their analysis of the observation, also described this mixed type of El Nino. These mixed type El Nin˜os account for about 30% of total El Nin˜o events. However, the excluding the mixed type El Nino does not lead to a significant change in the results. Therefore, the simple definition used in this study is valid.

J. Choi et al.: The role of mean state on changes in El Nin˜o’s flavor Fig. 1 Development and decay of a, c the EP El Nin˜o composite and b, d the CP El Nin˜o composite of SSTA [(K), upper panels] and 17°C isotherm depth anomaly [(m), lower panels] along the equator (averaged over 5°S–5°N)

1207

(a)

(b)

(c)

(d)

Following the classification of the two types of El Nin˜o, we constructed a composite analysis for each type. Figure 1 shows a time-longitude cross-section of SSTA and 17°C isotherm depth (a proxy for thermocline depth) anomalies along the equator, depicting the temporal evolution of each type of El Nin˜o event. Note that we take the 17°C isotherm depth, rather than the 20°C isotherm depth, as a proxy for the thermocline depth to avoid the possible outcropping of the 20°C isotherm that might occur owing to the cold bias of this model. In the case of the EP El Nin˜o, positive SSTA coincides with anomalous deepening thermocline over the eastern Pacific. Therefore, SSTA in the EP El Nin˜o shows a standing type of oscillation with a slight eastward propagating feature, which resembles a typical strong El Nino’s pattern observed since 1980s (e.g., An and Wang 2000). On the other hand, CP El Nin˜o shows a westward-propagating tendency in the developing phase. By contrast, in the case of CP El Nin˜o, the centers of SSTA and of the thermocline depth anomaly do not match. The peak of the warm SSTA during the winter season appeared near 160°E; a shallow thermocline depth anomaly, however, appeared in the same region, indicating that surface warming is not induced by subsurface dynamics. Kao and Yu (2009) also argue that CP El Nin˜o is not as strongly linked to the thermocline variation, which involves only a

shallow layer (about 100 m) of subsurface ocean temperatures variations: on the contrary, they suggested that CP El Nin˜o may be influenced more strongly by atmospheric forcing.

3 Decadal modulation of El Nin˜o’s flavor In order to investigate the decadal modulation of occurrence of the two types of El Nin˜o events, we counted the occurrences of each type of El Nin˜o event over a 20-year sliding period (Fig. 2). Solid and dashed lines indicate occurrence frequencies of CP El Nin˜o and EP El Nin˜o, respectively. Note that ‘‘frequency’’ indicates the number of events over the 20-year period. These sliding frequencies fluctuate on decadal-to-interdecadal time scales. We also examined the same data with sliding window sizes of 15 and 25 years, and concluded that results were not significantly modified by the window size. The average frequencies of the CP and EP El Nin˜o are 4.90 and 3.33 for the 20-year period, respectively. Overall, the CP El Nin˜o occurs more frequently than the EP El Nin˜o in this simulation; this model, the occurrence ratio differs from the observed occurrence. Since the 1850s, EP El Nin˜o has occurred 32 times, while CP El Nin˜o has occurred 7 times

123

1208

J. Choi et al.: The role of mean state on changes in El Nin˜o’s flavor

Fig. 2 Twenty-year sliding frequency of CP El Nin˜o (solid line) and EP El Nin˜o (dashed line). 20-year sliding frequency is defined by counting each El Nin˜o event during the 20-year period

(Yeh et al. 2009). In this model, the frequencies of both CP and EP El Nin˜o exhibit decadal modulation, with respective standard deviations of 2.15 and 1.36. Hence the decadal modulation intensity of CP El Nin˜o is larger than that of EP El Nin˜o. The correlation coefficient between the CP and EP El Nin˜o frequencies was found to be -0.65; this is statistically significant at the 99% confidence level, as determined using a critical value table for Pearson’s correlation coefficients. The negative correlation indicates an inverse relation between the number of CP El Nin˜o occurrences and that of EP El Nin˜o on decadal time scales. Since the characteristics of El Nin˜o are highly dependent on the background state (Ye and Hsieh 2006; An and Wang 2000; Fedorov and Philander 2000), one may infer that slowly varying tropical climate states control the selection between the CP El Nin˜o and the EP El Nin˜o (An and Jin 2000; An et al. 2006). In this study, we define two regimes based on the CP El Nin˜o frequency index. The one is the higher-frequency regime of CP El Nin˜o, referring the period when the normalized CP El Nin˜o frequency indices are greater than 1.0 (hereafter High-CP period). The other is the lowerfrequency regime, referring the period when the normalized CP El Nin˜o frequency indices are smaller than -1.0 (hereafter Low-CP period). Composites for High-CP (LowCP) periods indicate the average over the 20-year duration near a point of time which is above a threshold (High-CP and Low-CP for 1.0 and -1.0, respectively). The composite for a 20-year duration is sufficient to eliminate interannual signals, such as ENSO. Figure 3a, b show anomalies in the composite pattern’s deviation from climatology, indicating the mean state changes associated with High-CP and Low-CP periods, respectively. As seen in Fig. 3a, the centers for warming and cooling are located near 160°E and 100°W at the equator, respectively. Moreover, anomalous warming of mean SST extends to the northeast in the Northern Hemisphere and to the southeast in the Southern Hemisphere. This feature indicates that the cold tongue region is colder, while the warm pool region is warmer. Therefore, CP El Nin˜o events tend to occur more frequently when the zonal gradient of mean SST is stronger

123

Fig. 3 Composite map of SST anomaly (°C) associated with a HighCP and b Low-CP periods

over the tropical Pacific. We also checked the mean state of SST associated with Low-CP periods (Fig. 3b), that is, periods when the EP El Nin˜o occurs relatively more frequently. As expected, the zonal gradient of mean SST for Low-CP periods is weaker. The magnitude of the anomaly is slightly larger in Low-CP periods than in High-CP periods. In addition, we performed the same composite analysis based on the EP El Nin˜o frequency index for several variables—SST, MLD, 17°C isotherm depth, zonal wind stress, and precipitation. The composite map for Low-EP (High-EP) periods is almost identical with that for High-CP (Low-CP) periods. This finding illustrates the inverse relationship between the occurrence frequency of the CP El Nin˜o and that of the EP El Nin˜o, which is presumably related to the change in mean state. Because of this significant inverse relationship, hereafter we will only

J. Choi et al.: The role of mean state on changes in El Nin˜o’s flavor

consider the mean state associated with the CP El Nin˜o frequency. Figure 4a shows equatorially-averaged (5°S–5°N) composite anomaly maps of the zonal wind stress (solid line) and precipitation (dashed line) associated with High-CP periods. This is also consistent with the mean SST changes (Fig. 3a). The anomalous zonal wind stress is negative from 170°E to 100°W over the tropical Pacific in High-CP periods, reflecting a strengthening of the easterlies which might be associated with an increase in the zonal surface temperature gradient. By contrast, there is an anomalous westerly in the western Pacific. This resembles surface atmospheric circulation driven by a Gill-type model (Gill 1980) such that the anomalous convective heating over the warm SST anomaly region (150–160°E) leads the surface convergence toward the main heating region, and therefore westerly anomalies in the western Pacific. Over the western Pacific, the center for climatological precipitation is located between 120°E and 140°E (the maximum magnitude reaches 11 mm per day, not shown). The magnitude of climatological precipitation decreases sharply to the east and is about 7 mm per day near 150°E. In High-CP periods, anomalous high precipitation appears near 150°E, with a magnitude of 0.75 mm per day. Thus, the anomalous precipitation in a High-CP period is not sufficient to move the center of climatological precipitation, owing to its small magnitude. In addition to the SST composite (Fig. 3a), we also show composite anomaly maps of other oceanic variables: mixed layer depth in Fig. 4b and thermocline depth in Fig. 4c. The shaded area in the figures represents the region, where is statistically distinguished from the total climatological mean, with a 99% confidence level by Student’s t test. Mixed layer depth (MLD) is defined as the depth at which the temperature differs by 0.5°C from that at the surface (Monterey and Levitus 1997). The composite map for Low-CP periods is essentially out of phase with that of High-CP periods (see Fig. 3a, b); therefore, we only describe the climate state associated with High-CP periods. In a High-CP period, both MLD and thermocline depth are deepening over the western-to-central tropical Pacific. The center of MLD deepening is located near 165°W at 3°N. This feature is consistent with an increasing subsurface temperature (Fig. 4d); both of these features seem to be induced by a deepening of thermocline depth (Fig. 4c). The thermocline depth associated with High-CP periods is also deepening over the western Pacific, owing to mass convergence caused by surface wind stress. The location of the thermocline maximum is to the west of the MLD maximum, since thermocline depth is directly affected by the upwelling, whereas the MLD is affected by mixing which is induced by both current shear and surface warming. In addition, the decadal variability of thermocline depth is maximized near 10°S. This characteristic has

1209

Fig. 4 Composite map associated with High-CP periods for a precipitation [dashed line (mm/day)] and zonal wind stress [solid line (10-3 Pa)], b mixed layer depth (m), c 17°C isotherm depth (m), and d vertical profile of equatorially-averaged (5°S–5°N) subsurface temperature. Upper (lower) thick solid line in d indicates climatology of mixed layer (17°C isotherm depth)

appeared in other CGCMs (Schneider 2000; Cibot et al. 2005). Cibot et al. (2005) suggested that decadal thermocline variability over the western off-equatorial Pacific (Southern Hemisphere) arises primarily from Ekman pumping anomalies in this region. Figure 4d shows the vertical structure of the equatoriallyaveraged (5°S–5°N) subsurface temperature associated with High-CP periods. The upper and lower thick solid lines indicate the climatology of MLD and thermocline depth respectively. We did not depict the MLD and thermocline depth anomaly associated with High-CP periods due to the small magnitudes of changes; these magnitudes are less

123

1210

J. Choi et al.: The role of mean state on changes in El Nin˜o’s flavor

than 3 m at the equator (see Fig. 4b, c). In the western-tocentral Pacific, the maximum core of the anomaly appears near the bottom of the MLD, which reaches 0.35°C, as is consistent with the deepening of MLD and thermocline depth in this region. In the eastern Pacific, the minimum anomaly is located near the thermocline. These subsurface changes are consistent with the strong zonal gradient of surface mean temperature. In order to identify decadal changes in oceanic vertical structure at the equator, we applied an empirical orthogonal function (EOF) analysis (a.k.a. principal component analysis) to the 20-year moving averaged equatorially-averaged (5°S–5°N) subsurface temperature (hereafter, 20 yearSubT). EOF analysis can capture the internal variability of subsurface temperature in this simulation. Many previous studies have determined decadal fluctuations using EOF analysis (Yeh and Kirtman 2004; Cibot et al. 2005; Choi et al. 2009). The first EOF mode of 20 year-SubT accounts for 59.0% of total variance; Fig. 5 shows the spatial pattern of the first EOF mode. The first EOF mode is almost identical to the composite map of subsurface temperature associated with High-CP periods (Fig. 4d). The pattern correlation between the composite map of High-CP periods (Fig. 4d) and the pattern of the first EOF mode is 0.97 over the domain depicted in Fig. 5 (120°E–80°W, from the surface to 250.6 m depth). The pattern of the first EOF mode is correlated with the composite map of Low-CP period with the correlation coefficient of -0.95. Next, we analyzed the periodicity of the principal component (PC) time series (Fig. 6, dashed line). To check the dominant time scale of the first mode, we applied spectral analysis to the PC time series. The first mode has its maximum spectral peaks at 12, 13.5, 29, and 83-year periods. In essence, the first mode indicates a decadal-tointerdecadal mode, generated intrinsically by the mechanism of this model and also possibly existing in nature.

Wittenberg (2009) noted that there is interdecadal-tointercentennial modulation of ENSO behavior in this simulation, indicating that the time scale of the first mode seems to be related to ENSO modulation. The temporal correlation coefficient between the CP El Nin˜o frequency index (Fig. 6, solid line) and the first PC time series is 0.64, and it is statistically significant at the 99% confidence level. Thus, the climate state associated with CP El Nin˜o occurrences seems to be related to internal interdecadal oscillation in this simulation. We have confirmed, therefore, that the oceanic variables (in Fig. 4b–d) are physically linked with the mean state of SST (in Fig. 3a), associated with High-CP periods, and that these climate state changes are induced by internal variability over the tropical Pacific in this model. On the other hand, an inconsistency with the results of Yeh et al. (2009) is apparent. Based on future scenario CGCM experiments, Yeh et al. (2009) argued that shallow thermocline depth over the central Pacific would lead to more frequent occurrences of CP El Nin˜o. Under conditions of shallow thermocline depth, variability of thermocline depth has an increased effect on variability of SST. This contradiction might be resolved as follows: The mean thermocline depth of multi model ensemble in the future climate change scenario run (Special Report for Emission Scenario A1B climate change projection) tends to be more than 20 m shallower as compared with the twentiethcentury climate change simulation (control run) over the equatorial central Pacific (see Fig. 4 in Yeh et al. 2009); this difference is observed over 10% of climatology relative to the control run. Thus, anthropogenic forcing creates a change in ocean mean state sufficient to induce frequent CP El Nin˜o events. On the other hand, the maximum magnitude of thermocline depth due to the internal oscillation reaches 3 m in this model. It follows that decadal variability of thermocline change in a CGCM could not be enough to effectively generate CP El Nin˜o events, owing to its relatively small magnitude. Nevertheless, the possible role of subsurface dynamics should be investigated in the future. In this study, the difference between the thermocline depth and MLD is smaller at the equator, except in the far western Pacific, during High-CP periods. Thus vertical stratification of the central tropical Pacific becomes more unstable in High-CP periods, even though the thermocline depth is deepening. This indicates that the entrainment of MLD induced by variability of thermocline depth is increased, leading to an intensification of SST variability over the central Pacific in High-CP periods. This effect is due to intensified vertical temperature advection of mean subsurface temperature by anomalous upwelling  [w0oozT ; ‘‘surface layer feedback’’ in Neelin et al. (1998)] over this region, caused, in turn, by an increase in the mean

Fig. 5 Pattern of the first EOF mode of 20-year moving averaged equatorially-averaged (5°S–5°N) subsurface temperature. Upper (lower) thick solid line indicates the climatology of mixed layer depth (17°C isotherm depth)

123

J. Choi et al.: The role of mean state on changes in El Nin˜o’s flavor

1211

Fig. 6 PC time series for the first EOF mode of 20-year moving averaged equatoriallyaveraged (5°S–5°N) subsurface temperature (dashed line) and CP El Nin˜o frequency index (solid line)

vertical temperature gradient. It seems, however, that the frequent CP El Nin˜o occurrences in High-CP periods are not fully described by the subsurface dynamics. Therefore, we suggest that the interaction between ocean surface and atmosphere plays a crucial role in the frequent generation of CP El Nin˜o events. The details are described in Sect. 4.

4 Effects of the mean state on ENSO In this section, we further investigate dynamical causal relationships between the climate mean state and decadal modulation of two types of ENSO. 4.1 Heat budget analysis in the ocean mixed layer In this subsection, we investigate changes in oceanic dynamics during High-CP and Low-CP periods. First, in order to calculate the SST tendency, a budget analysis of mixed-layer temperatures over the equatorial Pacific is performed, as in Kug et al. (2010). Here, we focus, in particular, on SST tendencies due to two feedback mechanisms: thermocline and zonal advective feedback. The terms ‘‘thermocline feedback’’ and ‘‘zonal advective feedback’’ are defined as follows: 0 Thermocline feedback: woT oz Zonal advective feedback: u0oT ox where u is the oceanic zonal current, and T is the temperature averaged over the mixed layer (fixed at 50 m). Vertical velocity (w) is calculated at the bottom of the mixed layer (55 m). An upper bar () represents climatology, while a prime symbol (0 ) indicates anomaly, i.e., deviation from the climatological mean. We also calculated these two quantities using a long-term varying mean state (20-year moving averaged value) instead of the 500-year mean; the results were not significantly affected, because the magnitudes of the mean state changes are relatively small, as compared to the climatology. Figure 7 represents the equatorially-averaged (5°S–5°N) linear regression coefficient of each feedback with respect to the normalized total SSTA tendency. The linear

regression coefficient is the covariance between each feedback and SSTA tendency divided by the variance of the SSTA tendency at each grid point. Therefore, the regression coefficient in Fig. 7 represents the contribution of each feedback (K/month) to the unit SSTA tendency (K/ month). The solid and dashed lines, respectively, represent the regression associated with High-CP and Low-CP periods. The contribution of thermocline feedback to total SSTA tendency is maximized between 120°W and 110°W, owing to the shallow mean thermocline over the eastern Pacific. In the case of Low-CP periods (dashed line), the magnitude reaches 0.5, larger than that in High-CP periods. This suggests that thermocline feedback plays a dominant role in frequent generation of EP-type ENSO. The magnitude of the zonal advective feedback over the central Pacific in High-CP periods is larger than that in Low-CP period. We applied the bootstrap method for the statistical significance test. First, we randomly select N data from the 500-year simulation run. Where N is a total number of cases used for the calculation of the regression coefficient for both High-CP and Low-CP period. Then we calculate the linear regression coefficient at each grid point from the selected N data. By repeating this process 1,000 times, we obtain 1,000 values for the regression coefficient and the probability distribution function for the regression coefficient is constructed. The 2.5 and 97.5% ranking from the probability distribution function are indicating the 95% confidence level. As a result, these changes are statistically significant at 95% confidence level by the bootstrap method. This figure implies that the increased frequency of CP-type ENSO is related to strong zonal advective feedback. These results are consistent with those of Kug et al. (2009, 2010), who argued that the thermocline and zonal advective feedbacks are key processes for induction of the EP El Nin˜o and CP El Nin˜o, respectively. However, we note that the center of SSTA for the CP El Nin˜o is located west of the dateline, while the contribution of zonal advective feedback is stronger to the east of the dateline. This inconsistency may be due to the effect induced by EPtype ENSO in High-CP periods, which is incorporated into this calculation.

123

1212

J. Choi et al.: The role of mean state on changes in El Nin˜o’s flavor

Fig. 7 Linear regression of a thermocline feedback and b zonal advective feedback averaged over the equatorial band (5°S–5°N) with respect to the local SSTA tendency (dT0 / dt). Regressions for High-CP and Low-CP periods are indicated by the solid and dashed lines, respectively. Units are [(K/month)/(K/month)]

Fig. 8 Equatorially-averaged (5°S–5°N) linear regression coefficient of precipitation anomaly with respect to the local SST anomaly (mm/ day/K). The regressions for High-CP and Low-CP periods are indicated by the solid and dashed lines, respectively

4.2 Changes in air–sea coupling strength In this subsection, we explore the mechanism by which the climate state modifies El Nin˜o’s flavor through air–sea interaction. We investigate changes in air–sea coupling strength as a function of changes in climate state, since air– sea coupling strength is a key component of ENSO stability. Atmospheric heating is a major contributing cause of atmospheric circulation in the tropics, which is related to the mean SST (cf. Ham et al. 2007). That is, climate states corresponding to increased warming in the western Pacific (High-CP periods, in the terminology of this study) imply location of the maximum surface wind anomaly to the west of the maximum SSTA. Through this process, the local airsea coupling strength becomes stronger (weaker) over the western (eastern) Pacific. To verify the previous argument, we calculate the regression coefficient of the equatorially-averaged (5°S– 5°N) atmospheric variable anomaly with respect to the SSTA at each grid point. In Fig. 8, the solid (dashed) line represents the regression coefficient of the precipitation anomaly with respect to the in situ SSTA associated with High-CP (Low-CP) periods. Also we performed the bootstrap method as previously described to check the

123

significance for changes in air–sea coupling strength (Figs. 8, 9), and as a result, those changes are statistically significant at 95% confidence level. The regression coefficient indicates the sensitivity of precipitation anomaly [mm/day] with respect to the change of unit SSTA [K]. In general, the magnitude of the regression coefficient increases toward the west, and the maximum loading of precipitation is located to the west of the dateline. This is because the convection anomaly appears to be linearly proportional to the SST anomaly in the western Pacific. However, since negative SST anomalies have no decreased effect on convective activity in the cold tongue region (Hoerling et al. 1997; Kang and Kug 2002), a strongly nonlinear relationship exists between convective activities and local SST anomaly in the eastern Pacific. Furthermore, the magnitude to the west of the dateline in High-CP periods (solid line) is greater than that in Low-CP periods (dashed line). On the other hand, to the east of the dateline, the magnitude associated with Low-CP periods is larger than that associated with High-CP periods. This means that the sensitivity of precipitation anomalies to the local SSTA is intensified when the total SST increases. Thus, western Pacific warming (associated with High-CP periods) leads to more active convection than in the case of eastern Pacific warming, when the same SSTA are applied to the two regions. Figure 9a, b display the linear regression coefficients of ˜ O3m the zonal wind stress anomaly with respect to the NIN ˜ O4m SSTA indices, respectively. The general and NIN behavior is almost identical to that of the regression coefficients for precipitation, because of their relationship. The action center of zonal wind stress anomaly with respect to ˜ O3m SSTA is broad and located near 160°E, while the NIN the activity of zonal wind stress anomaly with respect to ˜ O4m SSTA has a sharp peak over the western the NIN Pacific. This indicates that stronger zonal wind stress is ˜ O4m region than in the needed to induce SSTA in the NIN ˜ O3m region. In other words, westward (eastward) NIN shifts of the maximum zonal wind stress anomaly could induce a relatively weak (strong) SSTA over the central (eastern) Pacific even absent any change in the magnitude

J. Choi et al.: The role of mean state on changes in El Nin˜o’s flavor

1213

Fig. 9 Equatorially-averaged (5°S–5°N) linear regression coefficient of zonal wind stress anomaly (Pa) with respect to the ˜ O3m SSTA and a NIN ˜ O4m SSTA indices. The b NIN regressions for High-CP and Low-CP periods are indicated by the solid and dashed lines, respectively. Units are (Pa/K)

of wind stress. These facts are consistent with Kang and Kug (2002), who argued that the amplitude of SSTA oscillations (such as ENSO) increases as wind stress shifts to the east. Furthermore, the zonal wind stress with respect to both SSTA indices over western Pacific increases in High-CP periods. This implies that climate states corresponding to High-CP periods provide strong air-sea coupling over the western Pacific, while the coupling strength over the eastern Pacific is relatively weak.

5 Summary and discussion In this study, we examined the modulation characteristics of two types of El Nin˜o using a 500-year preindustrial simulation of the GFDL CM2.1 coupled GCM, and followed the steps suggested in Kug et al. (2010) to determine the two types of El Nin˜o events. We used the terms ‘‘Central Pacific (CP) El Nin˜o’’ and ‘‘Eastern Pacific (EP) El Nin˜o’’ instead of ‘‘new-type El Nin˜o’’ and ‘‘conventional El Nin˜o,’’ respectively. The frequency index (number of occurrences) of both types of El Nin˜o is defined by the number of events in a sliding 20-year period. There is a strong negative correlation between the two indices, indicating a possible relationship between the mean state and El Nin˜o’s flavor. In order to investigate this relationship, climate states are classified into two periods based on the CP El Nin˜o frequency index: High-CP and Low-CP periods, referring respectively to decades in which CP El Nin˜o occurred more frequently and less frequently. The characteristics of the climate state associated with High-CP periods are essentially represented by a strong zonal gradient of the mean SST. Regarding the mean atmospheric variable, the equatorial Trade wind (easterly) is intensified to the east of the dateline, whereas it is reduced to the west of the dateline. The change in the oceanic variables is also consistent with the change in mean SST. Both the mixed layer depth and the thermocline depth are deepening, primarily over the central to western Pacific. These oceanic changes are due to internal variability over the tropical Pacific in this model.

Climate states for High-CP periods promote zonal advective feedback, which generates SST anomaly tendencies over the central Pacific; by contrast, climate states for Low-CP periods promote thermocline feedback, inducing SSTA tendencies over the eastern Pacific, as in Kug et al. (2009, 2010). Moreover, Kug et al. (2009, 2010) pointed out that zonal advective feedback has a greater effect on the generation of CP El Nin˜o than does thermocline feedback. Furthermore, air-sea coupling strength is increased (decreased) over the western (eastern) Pacific in High-CP periods, leading to intensification (suppression) of SSTA over the western (eastern) Pacific. Thus, the climate state in High-CP periods can provide favorable conditions for the development of CP El Nin˜o and the suppression of EP El Nin˜o, especially through changes in mean surface temperature gradient and local air–sea coupling strength. Interestingly, the composite pattern of mean state with respect to annual cycle amplitude or ENSO amplitude shown in An et al. (2010) is similar to our composite, implying that the mean state, ENSO, and annual cycle are closely related to each other. Also, Sun and Yu (2009) analyzed the decadal modulation of ENSO intensity using observational dataset, and showed that the mean state changes are related to the ENSO intensity. The mean SSTs associated with the weak and strong ENSO decades in Sun and Yu (2009) (see their Figs. 6 and 9) are similar to those associated with the High- and Low-CP periods, respectively. This may be because the strong ENSO period account for mostly EP El Nin˜o, which has larger amplitude than CP El Nin˜o’s, and the weak ENSO period mostly account for CP El Nin˜o. Aforementioned studies again imply that climate changes in the tropics could lead to the effective modification of ENSO behaviors (e.g., amplitude or spatial pattern) internally in the tropics without external forcing. Analysis of the annual cycle amplitude and ENSO amplitude over the eastern Pacific is out of our scope; hence, we only described the relationship between climate changes and El Nin˜o’s flavor. We intend to examine these relationships in the future. The origin of decadal-to-interdecadal climate state changes over the tropical Pacific remains an open question.

123

1214

J. Choi et al.: The role of mean state on changes in El Nin˜o’s flavor Acknowledgment This work was funded by Grant RACS_20102601 from the Korea Meteorological Administration Research and Development Program.

References

Fig. 10 Equatorially-averaged (5°S–5°N) composite map for the normalized skewness of SSTA, of which climatological mean has been removed. Solid (dashed) line indicates High-CP (Low-CP) periods

Two explanations are possible: first, such changes could be described by an internal process of the tropical Pacific without invoking any extratropical influences (Timmermann and Jin 2002; Rodgers et al. 2004; An 2009; Choi et al. 2009; Sun and Yu 2009). This might be attributable to two-way feedback between climate state and ENSO: while ENSO induces tropical Pacific decadal variability through a nonlinear rectification effect, the climate state provides favorable conditions for the specific type of ENSO. Figure 10 shows the equatorial averaged (5°S–5°N) skewness for SSTA associated with both High-CP and Low-CP periods. Statistically, skewness, i.e. the normalized third moment, is a way of measuring the nonlinearity of ENSO (Burgers and Stephenson 1999; An and Jin 2004; An 2009). Dashed and solid lines are those calculated for Low-CP and High-CP periods respectively. The skewness in Low-CP periods tends to increase toward the eastern Pacific, whereas that in High-CP periods shows a smooth positive curve over the central-to-eastern Pacific, indicating that the nonlinear rectification effect depends on the type of ENSO. Secondly, decadal variability with respect to the tropical Pacific could be caused by external impacts, such as midlatitude signal or atmospheric noise (Gu and Philander 1997; Kleeman et al. 1999; Nonaka et al. 2002). To investigate these external effects, we calculated the leadlag correlation between the CP El Nin˜o frequency index and the 20-year moving-averaged heat content over the entire Pacific (not shown here). The lead-lag correlation did not show a significant signal from mid-latitude to the tropics. Therefore, at least under this model, there is no significant influence from the extra-tropics that generates the decadal-to-interdecadal tropical fluctuation. This result, however, may be model-dependent; thus, more comprehensive analysis should be performed in the future.

123

An S-I (2009) A review of interdecadal changes in the nonlinearity of the El Nin˜o-Southern Oscillation. Theor Appl Climatol 97:29–40 An S-I, Jin F-F (2000) An eigen analysis of the interdecadal changes in the structure and frequency of ENSO mode. Geophys Res Lett 27:2573–2576 An S-I, Jin F-F (2001) Collective role of thermocline and zonal advective feedbacks in the ENSO mode. J Clim 14:3421–3432 An S-I, Jin F-F (2004) Nonlinearity and asymmetry of ENSO. J Clim 17:2399–2412 An S-I, Wang B (2000) Interdecadal change of the structure of the ENSO mode and its impact on the ENSO frequency. J Clim 13:2044–2055 An S-I, Ye Z, Hsieh W (2006) Changes in the leading ENSO modes associated with the late 1970s climate shift: role of surface zonal current. Geophys Res Lett 33:L14609 An S-I, Ham Y-G, Kug J-S, Timmermann A, Choi J, Kang I-S (2010) The inverse effect of annual mean state and annual cycle changes on ENSO. J Clim 23:1095–1110 Ashok K, Yamagata T (2009) The El Nin˜o with a difference. Nature 461:481–484 Ashok K, Behera SK, Rao SA, Weng H, Yamagata T (2007) El Nin˜o Modoki and its teleconnection. J Geophys Res 112. doi: 10.1029/2006JC003798 Battisti DS, Hirst AC (1989) Interannual variability in the tropical atmosphere/ocean system: influence of the basic state and ocean geometry. J Atmos Sci 46:1687–1712 Bejarano L, Jin F-F (2008) Coexistence of equatorial coupled modes of ENSO. J Clim 21:3051–3067 Burgers G, Stephenson DB (1999) The ‘‘normality’’ of El Nin˜o. Geophys Res Lett 26:1027–1030 Choi J, An S-I, Dewitte B, Hsieh WW (2009) Interactive feedback between the tropical Pacific decadal oscillation and ENSO in a coupled general circulation model. J Clim 22:6597–6611 Cibot C, Maisonnave E, Terray L, Dewitte B (2005) Mechanisms of tropical Pacific interannual-to-decadal variability in the ARPEGE/ ORCA global coupled model. Clim Dyn 24:823–842 Delworth TL et al (2006) GFDL’s CM2 global coupled climate models. Part I: formulation and simulation characteristics. J Clim 19:634–674 Dewitte B, Cibot C, Pe´rigaud C, An S-I, Terray L (2007) Interaction between near-annual and ENSO modes in a CGCM simulation: role of the equatorial background mean state. J Clim 20:1035–1052 Fedorov AV, Philander SGH (2000) Is El Nin˜o changing? Science 288:1997–2002 Gill AE (1980) Some simple solutions for heat induced tropical circulation. Quart J R Meteor Soc 106:447–462 Gu D, Philander SGH (1997) Interdecadal climate fluctuations that depend on exchanges between the tropics and extratropics. Science 275:805–807 Ham Y-G, Kug J-S, Kang I-S (2007) Role of moist energy advection in formulating anomalous walker circulation associated with El Nin˜o. J Geophys Res 112:D24105. doi:10.1029/2007JD008744 Hoerling MP, Kumar A, Zhong M (1997) El Nin˜o, La Nin˜a, and the nonlinearity of their teleconnections. J Clim 10:1769–1786 Jin F-F (1997a) An equatorial ocean recharge paradigm for ENSO. Part I: conceptual model. J Atmos Sci 54:811–829

J. Choi et al.: The role of mean state on changes in El Nin˜o’s flavor

1215

Jin F-F (1997b) An equatorial ocean recharge paradigm for ENSO. Part II: a stripped-down coupled model. J Atmos Sci 54:830–847 Jin F-F, Kug J-S, An S-I, Kang I-S (2003) A near-annual coupled ocean-atmosphere mode in the equatorial Pacific ocean. Geophy Res Lett 30:1080 Kang I-S, Kug J-S (2002) El Nin˜o and La Nin˜a sea surface temperature anomalies: asymmetry characteristics associated with their wind stress anomalies. J Geophys Res 107:4372. doi: 10.1029/2001JD000393 Kang I-S, Kug J-S, An S-I, Jin F-F (2004) A near-annual Pacific Ocean basin mode. J Clim 17:2478–2488 Kao HY, Yu JY (2009) Contrasting Eastern-Pacific and CentralPacific types of ENSO. J Clim 22:615–632 Kleeman R, McCreary JP, Klinger BA (1999) A mechanism for generating ENSO decadal variability. Geophys Res Lett 26:1743–1746 Kug J-S, Jin F-F, An S-I (2009) Two-types of El Nin˜o events: cold tongue El Nin˜o and warm pool El Nin˜o. J Clim 22:1499–1515 Kug J-S, Choi J, An S-I, Jin F-F, Wittenberg AT (2010) Warm pool and cold tongue El Nin˜o events as simulated by the GFDL 2.1 coupled GCM. J Clim 23:1226–1239 Larkin NK, Harrison DE (2005a) On the definition of El Nin˜o and associated seasonal average US weather anomalies. Geophys Res Lett 32:L13705. doi: 10.1029/2005GL022738 Larkin NK, Harrison DE (2005b) Global seasonal temperature and precipitation anomalies during El Nin˜o autumn and winter. Geophys Res Lett 32:L16705. doi:10.1029/2005GL022860 Monterey GI, Levitus S (1997) Seasonal variability of mixed layer depth for the world ocean. NOAA Atlas NESDIS 14, US Government Printing Office, pp 5 Neelin JD, Battisti DS, Hirst AC, Jin F-F, Wakata Y, Yamagata T, Zebiak SE (1998) ENSO theory. J Geophys Res Oceans 103:14261–14290 Nonaka M, Xie SP, McCreary JP (2002) Decadal variations in the subtropical cells and equatorial Pacific SST. Geophys Res Lett 29:1116. doi:10.1029/2001GL013717 Philander SGH, Pacanowski RC, Lau N-C, Nath MJ (1992) Simulation of ENSO with a global atmospheric GCM coupled to a high-resolution, tropical Pacific Ocean GCM. J Clim 5:308–329 Picaut J, Masia F, du Penhoat Y (1997) An advective–reflective conceptual model for the oscillatory nature of the ENSO. Science 277:663–666 Rodgers KB, Friederichs P, Latif M (2004) Tropical Pacific decadal variability and its relation to decadal modulations of ENSO. J Clim 17:3761–3774

Ropelewski CF, Halpert MS (1987) Global and regional scale precipiation patterns associated with the El Nin˜o/Southern Oscillation. Mon Weather Rev 115:1606–1626 Ropelewski CF, Halpert MS (1996) Quantifying Southern Oscillation–precipitation relationship. J Clim 9:1043–1059 Schneider N (2000) A decadal spiciness mode in the tropics. Geophys Res Lett 27:257–260 Schopf PS, Suarez MJ (1988) Vacillations in a coupled ocean– atmosphere model. J Atmos Sci 45:549–566 Sun F, Yu J-Y (2009) A 10–15-year modulation cycle of ENSO intensity. J Clim 22:1718–1735 Timmermann A (2003) Decadal ENSO amplitude modulations: a nonlinear paradigm. Glob Planet Change 37:135–156 Timmermann A, Jin F-F (2002) A nonlinear mechanism for decadal El Nin˜o amplitude changes. Geophys Res Lett 29:1003. doi: 10.1029/2001GL013369 Trenberth KE, Carbon JM (2000) The Southern Oscillation revisited: sea level pressures, surface temperatures and precipitation. J Clim 13:4358–4365 Wallace JM, Rasmusson EM, Mitchell TP, Kousky VE, Sarachik ES, von Storch H (1998) On the structure and evolution of ENSOrelated climate variability in the tropical Pacific: lessons from TOGA. J Geophy Res 103:14169–14240 Weng H, Ashok K, Behera SK, Rao A, Yamagata T (2007) Impacts of recent El Nin˜o Modoki on dry/wet conditions in the Pacific rim during boreal summer. Clim Dyn 29. doi: 10.1007/s00382-0070234-0 Wittenberg AT (2009) Are historical records sufficient to constrain ENSO simulations? Geophy Res Lett 36:L12702. doi:10.1029/ 2009GL038710 Wittenberg AT, Rosati A, Lau N-C, Ploshay JJ (2006) GFDL’s CM2 global coupled climate models. Part III: tropical Pacific climate and ENSO. J Clim 19:698–722 Ye Z, Hsieh WW (2006) The influence of climate regime shift on ENSO. Clim Dyn 26:823–833 Yeh S-W, Kirtman B (2004) Tropical Pacific decadal variability and ENSO amplitude modulation in a CGCM. J Geophys Res 109:C11009. doi:10.1029/2004JC002442 Yeh S-W, Kug J-S, Dewitte B, Kwon M-H, Kirtman B, Jin F-F (2009) Recent changes in El Nin˜o and its projection under global warming. Nature 461:511–515 Zebiak SE (1984) Tropical atmospheric-ocean interaction and the El Nin˜o/Southern Oscillation phenomenon. Ph.D. Thesis, MIT, pp 261

123